Modern Magnetic Resonance, Vol. 1: Applications in Chemistry, Biological and Marine Sciences 1402038941, 9781402038945

Citation preview

Modern Magnetic Resonance Part 1

Part 1: Applications in Chemistry, Biological and Marine Sciences Part 2: Applications in Medical and Pharmaceutical Sciences Part 3: Applications in Materials Science and Food Science

Modern Magnetic Resonance Part 1: Applications in Chemistry, Biological and Marine Sciences Graham A. Webb (Ed.) Royal Society of Chemistry, London, UK

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-1-4020-3894-5 (HB) ISBN 978-1-4020-3910-2 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

www.springer.com

Printed on acid-free paper

First published in 2006 Reprinted with corrections in 2008

All rights reserved.
C 2008 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

V

Contents

List of Section Editors ........................................................................................................... Preface ............................................................................................................................... Color Plate Section................................................................................................................

XI XIII XV

APPLICATIONS IN CHEMISTRY Foreword ............................................................................................................................ Abbreviations ......................................................................................................................

3 5

Amyloids

9

Kinetics of Amyloid Fibril Formation of Human Calcitonin, Miya Kamihira, Hazime Saitô, and Akira Naito.............................................................................................................. Polymorphism of Alzheimer’s Aβ Amyloid Fibrils, Oleg N. Antzutkin ........................................... Chemical Shifts and Spin-Couplings 13 C, 15 N, 1 H, 2 H,

and 17 O NMR Chemical Shift NMR for Hydrogen Bonds, Shigeki Kuroki............ ......... NMR Chemical Shift Map, Isao Ando and Tetsuo Asakura............................................................ NMR Chemical Shifts Based on Band Theory, Hiromichi Kurosu and Takeshi Yamanobe...................... Modeling NMR Chemical Shifts, Julio C. Facelli ....................................................................... Ab Initio Calculation of NMR Shielding Constants, Peter B. Karadakov ......................................... Crystal Structure Refinement Using Chemical Shifts, Ulrich Sternberg, Raiker Witter, and Anne S. Ulrich The Theory of Nuclear Spin–Spin Couplings, Hiroyuki Fukui ....................................................... Fibrous Proteins Investigation of Collagen Dynamics by Solid-State NMR Spectroscopy, Göran Zernia and Daniel Huster Solid-State NMR Studies of Elastin and Elastin Peptides, Kristin K. Kumashiro............................... Structural Analysis of Silk Fibroins using NMR, Tetsuo Asakura and Yasumoto Nakazawa .................. Field Gradient NMR NMR Diffusometry, William S. Price ...................................................................................... Field Gradient NMR of Liquid Crystals, István Furó and Sergey V. Dvinskikh .................................... Field Gradient NMR for Polymer Systems with Cavities, Yuji Yamane and Sunmi Kim........................ NMR Measurements Using Field Gradients and Spatial Information, Shingo Matsukawa ................... Theory and Application of NMR Diffusion Studies, Torsten Brand, Eurico J. Cabrita, and Stefan Berger Host–Guest Chemistry Solid-State NMR in Host–Guest Chemistry, John A. Ripmeester and Christopher I. Ratcliffe ................ Imaging Mapping of Flow and Acceleration with NMR Microscopy Techniques, Elke Kossel, Bogdan Buhai, and Rainer Kimmich ....................................................................................................... Industrial Application of In situ NMR Imaging Experiments to Steel-Making Process, Koji Saito ...... Biomedical NMR Spectroscopy and Imaging, Toshiro Inubushi and Sigehiro Morikawa .......................

11 19 29 31 37 43 53 63 71 79 85 87 93 101 107 109 117 123 129 135 145 147 155 157 163 173

VI

Contents

Electron Spin Resonance Imaging in Polymer Research, Shulamith Schlick .............. ..................... NMR Imaging: Monitoring of Swelling of Environmental Sensitive Hydrogels, Karl-Friedrich Arndt, Manfred Knörgen, Sven Richter, and Thomas Schmidt ................................................................... Inorganic Materials and Catalysis Exploiting 1 H→29 Si Cross-Polarization Features for Structural Characterization of Inorganic Materials, Piotr Tekely .................................................................................... Solid State NMR Characterization of Solid Surface of Heterogeneous Catalysts, Feng Deng, Jun Yang, and Chaohui Ye .............................................................................................................. Isotope Labeling Recent Developments in Stable-Isotope-Aided Methods for Protein NMR Spectroscopy, Shin-ya Ohki and Masatsune Kainosho................................................................................................... Structural Glycobiology by Stable-isotope-assisted NMR Spectroscopy, Yoshiki Yamaguchi and Koichi Kato.................................................................................................................... Lipid Bilayer and Bicelle Development and Application of Bicelles for Use in Biological NMR and Other Biophysical Studies, Charles R. Sanders ............................................................................... Nuclear Magnetic Resonance of Oriented Bilayer Systems, Akira Naito, Shuichi Toraya, and Katsuyuki Nishimura ........................................................................................................ Solid-State Deuterium NMR Spectroscopy of Membranes, Michael F. Brown, Silvia Lope-Piedrafita, Gray V. Martinez, and Horia I. Petrache ................................................................................. Solid State 19 F-NMR Analysis of Oriented Biomembranes, Anne S. Ulrich ..................................... Membrane-Associated Peptides Solid-State NMR Studies of the Interactions and Structure of Antimicrobial Peptides in Model Membranes, Marise Ouellet and Michèle Auger............................................... Anisotropic Chemical Shift Perturbation Induced by Ions in Conducting Channels, Jun Hu, Eduard Chekmenev, and Timothy A. Cross............................................................................... NMR Studies of Ion-Transporting Biological Channels, James F. Hinton ....................................... Membrane Proteins Site-Directed NMR Studies on Membrane Proteins, Hazime Saitô................................................ Structure of Membrane-Binding Proteins Revealed by Solid-State NMR, Satoru Tuzi, Naoko Uekama, Masashi Okada, and Hitoshi Yagisawa .................................................................................. Solid-State NMR of Membrane-Active Proteins and Peptides, John D. Gehman and Frances Separovic .. Magnetic Resonance Spectroscopic Studies of the Integral Membrane Protein Phospholamban, Gary A. Lorigan............................................................................................................... NMR Studies of the Interactions Between Ligands and Membrane-Embedded Receptors: New Methods for Drug Discovery, David A. Middleton............................................... Photosynthetic Antennae and Reaction Centers, H.J.M. de Groot ................................................ Insight into Membrane Protein Structure from High-Resolution NMR, Philip L. Yeagle and Arlene Albert. New Developments Fast Multidimensional NMR: New Ways to Explore Evolution Space, Ray Freeman and Eriks Kupþe...... High-Sensitivity NMR Probe Systems, P.J.M. van Bentum and A.P.M. Kentgens ............................... CRAMPS, B.C. Gerstein and H. Kimura ..................................................................................... Mobile NMR, B. Blümich and F. Casanova............................................................................... Rheo-NMR, Paul T Callaghan................................................................................................ Analytical Aspects of Solid-State NMR Spectroscopy, Cecil Dybowski............................................

179 187 195 197 205 213 215 223 231 233 241 249 261 269 271 279 285 289 291 299 305 313 319 327 335 345 347 353 363 373 383 389

Contents 3 H NMR and Its Application, John R. Jones and Shui-Yu Lu......................................................... On-line SEC–NMR, Tatsuki Kitayama and Koichi Ute...................................................................

NOE and Chemical Exchange The Nuclear Overhauser Effect, Mike P Williamson.................................................................... Solute–Solvent Interactions Examined by the Nuclear Overhauser Effect, J.T. Gerig ....................... Chemical Exchange, Alex D . Bain........................................................................................... NQR & ESR Separated Detection of H-Transfer Motions in Multi-H-Bonded Systems Studied by Combined 1 H NMR and 35 Cl NQR Measurements, Ryuichi Ikeda................................................ EPR: Principles, Bruce R. McGarvey........................................................................................ Zero Field NMR: NMR and NQR in Zero Magnetic Field, David B. Zax ............................................ Organo Metallic Chemistry Organoboron Chemistry, Bernd Wrackmeyer ............................................................................ Organogermanium Chemistry, Bernd Wrackmeyer ..................................................................... Organotin Chemistry, Bernd Wrackmeyer ................................................................................ Paramagnetic Effects and 13 C High-Resolution Solid-State NMR of Paramagnetic Compounds Under Very Fast Magic Angle Spinning, Yoshitaka Ishii and Nalinda P. Wickramasinghe............................ Paramagnetic Effects of Dioxygen in Solution NMR—Studies of Membrane Immersion Depth, Protein Topology, and Protein Interactions, R.S. Prosser and F. Evanics ..............

395 399 407 409 413 421 429 431 439 445 453 455 459 461 465

1H

Protein Structure TROSY NMR for Studies of Large Biological Macromolecules in Solution, César Fernández, Gerhard Wider NMR Insight of Structural Stability and Folding of Calcium-Binding Lysozyme, Makoto Demura........ NMR Investigation of Calmodulin, Tapas K. Mal and Mitsuhiko Ikura ............................................ Analytical Framework for Protein Structure Determination by Solid-State NMR of Aligned Samples, Alexander A. Nevzorov and Stanley J. Opella ......................................... Determining Protein 3D Structure by Magic Angle Spinning NMR, Ovidiu C. Andronesi, Henrike Heise, and Marc Baldus ............................................................................................................. 19 F NMR Study of b-Type Haemoproteins, Yasuhiko Yamamoto, Satoshi Nagao, and Akihiro Suzuki....... Polymer Structure NMR in Dry or Swollen Temporary or Permanent Networks, Jean-Pierre Cohen Addad ....................... Crystalline Structure of Ethylene Copolymers and Its Relation to the Comonomer Content, Qun Chen......................................................................................... Isomorphism in Bacterially Synthesized Biodegradable Copolyesters, Naoko Yoshie and Yoshio Inoue... Two-Dimensional NMR Analysis of Stereoregularity of Polymers, A.S. Brar and Gurmeet Singh........... Quantitative Analysis of Conformations in Disordered Polymers by Solid-State Multiple-Quantum NMR, Hironori Kaji.................................................................. Polymer Microstructure: The Conformational Connection to NMR, Alan E. Tonelli........................... Solid-State NMR Characterization of Polymer Interfaces, Peter A. Mirau....................................... The Structure of Polymer Networks, Andrew K. Whittaker ........................................................... 1 H CRAMPS NMR of Polypeptides in the Solid State, Akira Shoji................................................. Polymer Dynamics Dynamics of Amorphous Polymers, Fumitaka Horii................................................................... Molecular Motions of Crystalline Polymers by Solid-State MAS NMR, Toshikazu Miyoshi...................

467 475 485 487 497 503 517 527 531 539 541 545 551 557 563 567 575 583 591 605 607 615

VII

VIII Contents

Dynamics in Polypeptides by Solid State 2 H NMR, Toshifumi Hiraoki........................................... Polymer Blends Polymer Blends, Atsushi Asano ............................................................................................ Configurational Entropy and Polymer Miscibility: New Experimental Insights From Solid-State NMR, Jeffery L. White ............................................................................... Quantum Information Processing Quantum Information Processing as Studied by Molecule-Based Pulsed ENDOR Spectroscopy, Robabeh Rahimi, Kazunobu Sato, Daisuke Shiomi, and Takeji Takui ................. Residual Dipolar Couplings and Nucleic Acids New Applications for Residual Dipolar Couplings: Extending the Range of NMR in Structural Biology, Rebecca S. Lipsitz and Nico Tjandra......................................................... Refinement of Nucleic Acid Structures with Residual Dipolar Coupling Restraints in Cartesian Coordinate Space, Nikolai B. Ulyanov, Zhihua Du, and Thomas L. James .......................... Conformational Analysis of DNA and RNA, Gota Kawai............................................................. Solid-State NMR Technique Analytical and Numerical Tools for Experiment Design in Solid-State NMR Spectroscopy, Niels Chr. Nielsen, Thomas Vosegaard, and Anders Malmendal....................................................... Homonuclear Shift-Correlation Experiment in Solids, K. Takegoshi ............................................. Two-Dimensional 17 O Multiple-Quantum Magic-Angle Spinning NMR of Organic Solids, Gang Wu..... A Family of PISEMA Experiments for Structural Studies of Biological Solids, Ayyalusamy Ramamoorthy and Kazutoshi Yamamoto ................................................................................................. Structural Constraints in Solids Rotational-Echo, Double-Resonance NMR, Terry Gullion ............................................................ REDOR in Multiple Spin System, Katsuyuki Nishimura and Akira Naito ........................................ Torsion Angle Determination by Solid-State NMR, Mei Hong...................................................... Secondary Structure Analysis of Proteins from Angle-Dependent Interactions, Toshimichi Fujiwara and Hideo Akutsu............................................................................................................ Telomeric DNA Comple xes Comparison of DNA-Binding Activities Between hTRF2 and hTRF1 with hTRF2 Mutants, Shingo Hanaoka, Aritaka Nagadoi, and Yoshifumi Nishimura........................................................

621 629 631 637 645 647 655 657 665 671 677 679 689 695 703 711 713 719 727 735 741 743

APPLICATIONS IN BIOLOGICAL SCIENCES Abbreviations ......................................................................................................................

755

Optimization of MRI Contrast for Pre-Clinical Studies at High Magnetic Field, Yu-Ting Kuo, Amy H. Herlihy The Application of In Vivo MRI and MRS in Phenomic Studies of Murine Models of Disease, Po-Wah So and Jimmy D. Bell................................................................... Experimental Models of Brain Disease: MRI Contrast Mechanisms for the Assessment of Pathophysiological Status, David L. Thomas, Louise van der Weerd, Mark F. Lythgoe, and John S. Thornton Experimental Models of Brain Disease: MRI Studies, Louise van der Weerd, David Thomas, John Thornton Mankin choy, and Mark F. Lythgoe ....................................................................................... Application of MRS in Cancer in Pre-clinical Models, Y.-L. Chung, M. Stubbs, and J.R. Griffiths ........... Experimental Cardiovascular MR in Small Animals, Jürgen E. Schneider and Stefan Neubauer.............

759 769 787 801 823 835

Contents

Application of Pharmacological MRI to Preclinical Drug Discovery and Development, Matthew D. Ireland and Steven C.R. Williams.................................... Application of MRI to Cell Tracking, Kishore Bhakoo, Catherine Chapon, Johanna Jackson, and William Jones

855 879

APPLICATIONS IN MARINE SCIENCE Foreword ............................................................................................................................. Abbreviations....................................................................................................................... Comprehensive Compositional Analysis of Fish Feed by Time Domain NMR, Emil Veliyulin, Karl Østerhus, Wolfgang Burk, Trond Singstad, and Tore Skjetne ....... Low Field NMR Studies of Atlantic Salmon (Salmo salar), Ida Grong Aursand, Emil Veliyulin, and Ulf Erikson............................................................................................................... Water Distribution and Mobility in Fish Products in Relation to Quality, Bo M. Jørgensen and Kristina N. Jensen ........................................................................................................... Proton NMR of Fish Oils and Lipids, R. Sacchi, M. Savarese, L. Falcigno, I. Giudicianni, and L. Paolillo .. Determination of Fatty Acid Composition and Oxidation in Fish Oils by High Resolution Nuclear Magnetic Resonance Spectroscopy, Rosario Zamora and Francisco J. Hidalgo ...................... Resonance Spectroscopy to Study Lipid Oxidation in Fish and Fish Products, E. Falch and M. Aursand ................................................................ Omega-3 Fatty Acid Content of Intact Muscle of Farmed Atlantic Salmon (Salmo Salar) Examined by 1 H MAS NMR Spectroscopy, M. Aursand, I.S. Gribbestad, and I. Martinez.................... HR MAS NMR Spectroscopy of Marine Microalgae, Part 1: Classification and Metabolite Composition from HR MAS 1 H NMR Spectra and Multivariate Analysis, Matilde Skogen Chauton and Trond Røvik Størseth......................................................................................................... HR MAS NMR Spectroscopy of Marine Microalgae, Part 2: 13 C and 13 C HR MAS NMR Analysis Used to Study Fatty Acid Composition and Polysaccharide Structure, Trond Røvik Størseth, Matilde S. Chauton, and Jostein Krane................................................................................... Post-mortem Studies of Fish Using Magnetic Resonance Imaging, Emil Veliyulin, Alv Borge, Trond Singstad, Ingrid Gribbestad, and Ulf Erikson.................................................................... How is the Fish Meat Affected by Technological Processes?, Loïc Foucat, Ragni Ofstad, and Jean-Pierre Renou...................................................................................................... Author Index Subject Index

893 895 897 905 915 919 925 933 941 947

953 959 967

IX

XI

List of Section Editors Subject

Name

Type of Editor

Graham A. Webb

Editor-In-Chief

Chemistry

Hazime Saitˆo Himeji Institute of Technology and QuLiS, Hiroshima University Japan e-mail: [email protected] Isao Ando Department of Chemistry and Materials Science Tokyo Institute of Technology, Ookayama, Meguro-ku Tokyo 152-0033 Japan e-mail: [email protected] Tetsuo Asakura Department of Biotechnology Tokyo University of Agriculture and Technology Koganei, Tokyo 184-8588 Japan e-mail: [email protected]

Section Editors

Biological Sciences

Jimmy D. Bell Molecular Imaging Group, MRC Clinical Sciences Centre Hammersmith Hospital Campus, Imperial College London London, W12 OHS UK e-mail: [email protected]

Section Editor

Marine Science

M. Aursand SINTEF Fisheries and Aquaculture Ltd N-7465 Trondheim Norway e-mail: [email protected]

Section Editor

Medical Science

Carolyn Mountford Institute for Magnetic Resonance Research, and Department of Magnetic Resonance in Medicine University of Sydney, PO Box 148, St Leopards, 1590, NSW Australia email: [email protected] Uwe Himmelreich, PhD Max-Planck-Institute for Neurological Research In vivo NMR Group, Gleueler Str 50, Cologne, D-50931 Germany email: [email protected] Deborah Edward Pageturner (Editing & Research Services)

Section Editor

Assistant Editor

Subject

Name

Type of Editor

Pharmaceutical Science

David Craik Institute for Molecular Bioscience University of Queensland Brisbane 4072, Queensland Australia e-mail: [email protected]

Section Editor

Materials Science

Marcel Utz University of Connecticut, 97 N Eagleville Rd Storrs CT 06269-3136 e-mail: [email protected]

Section Editor

Food Science

Peter Belton School of Chemical Sciences and Pharmacy University of East Anglia Norwich NR4 7TJ, UK e-mail: [email protected]

Section Editor

XIII

Preface

It is a great pleasure for me to Introduce the handbook of Modern Magnetic Resonance, MMR. The various techniques which comprise MMR derive essentially from three sources, all of which were produced by physicists. Today they are widely used by scientists working in many diverse areas such as chemistry, biology, materials, food, medicine and healthcare, pharmacy and marine studies. The first source of MMR studies is nuclear magnetic resonance, NMR. This provides details on the relative positions of nuclei, i.e. atoms, in a molecule. Consequently NMR provides structural information on samples which may be in the solid, liquid or gaseous state. Nuclear relaxation data yield dynamic information on the sample and the topology of the dynamic processes if the sample is undergoing a molecular change. Thus high and low resolution NMR studies provide information on all interesting aspects of molecular science. The protean nature of NMR is reflected in its many applications in chemistry, biology and physics which explore and characterize chemical reactions, molecular conformations, biochemical pathways and solid state materials, to name a few examples. Magnetic resonance imaging, MRI, is the second source of MMR data. MRI provides a three-dimensional image of a substance, and is consequently widely employed to assess materials both in vitro and in vivo. The importance of MRI studies in many areas of science and

medicine is shown by the recent award of the Nobel Prize to Lauterbur and Mansfield. The third source of MMR results is due to electron spin resonance, ESR. This is a technique for detecting unpaired electrons and their interactions with nuclear spins in a given sample. Thus ESR data are often used to complement the results of NMR experiments. Taken together NMR, MRI and ESR comprise the field of MMR, recent years have witnessed the fecundity of these techniques in many scientific areas. The present three volumes cover applications in most of these areas. Part 1 deals with Chemical Applications, Biological and Marine Sciences. Medical and Pharmaceutical Sciences are covered in Part 2. Part 3 provides examples of recent work in the Materials Science and Food Science. I wish to express my gratitude to all of the Section Editors and their many contributors for their hard work and dedication in the creation of MMR. My thanks also go to Emma Roberts and the production staff at Springer, London, for their assistance in the realization of these volumes. Royal Society of Chemistry Burlington House Piccadilly London, W1J OBA

G.A.WEBB February 2005

Please note that authors used either British or American English spelling depending on the language of their choice for individual papers.

Graham A. Webb (ed.), Modern Magnetic Resonance, XIII.
C 2008 Springer.

Part I

Color Plate Section

Part I

Plate 1. See also Figure 1 on page 21.

XVII

Part I

34.00 b3lyp/CSGT b3lyp/GIAO b3lyp/IGAIM

32.00

mpw1pw91/CSGT mpw1pw91/GIAO

Chemical Shieldings (ppm)

30.00

mpw1pw91/IGAIM olyp/CSGT olyp/GIAO

28.00

olyp/IGAIM Linear (b3lyp/CSGT) 26.00

Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) Linear (mpw1pw91/CSGT)

24.00

Linear (mpw1pw91/GIAO) Linear (mpw1pw91/IGAIM) 22.00

Linear (olyp/CSGT) Linear (olyp/GIAO) Linear (olyp/IGAIM)

20.00 0

1

2

3

4

5

6

7

8

9

10

Chemical Shifts (ppm)

Plate 2. See also Figure 2 on page 57. 250.00 b3lyp/CSGT b3lyp/GIAO b3lyp/IGAIM 200.00 "mpw1pw91/CSGT mpw1pw91/GIAO mpw1pw91/IGAIM Chemical Shieldings (ppm)

150.00 olyp/CSGT olyp/GIAO olyp/IGAIM 100.00 Linear (b3lyp/CSGT) Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) 50.00

Linear ("mpw1pw91/CSGT) Linear (mpw1pw91/GIAO) Linear (mpw1pw91/IGAIM)

0.00 -50

0

50

100

150

200

250

Linear (olyp/CSGT) Linear (olyp/GIAO) Linear (olyp/IGAIM)

-50.00 Chemical Shifts (ppm)

Plate 3. See also Figure 3 on page 58.

XVIII

b3lyp CSGT b3lyp GIAO b3lyp IGAIM

300

mpw1pw91 CSGT mpw1pw91 GIAO

Chemical Shieldings (ppm)

mpw1pw91 IGAIM 200 olyp CSGT olyp GIAO olyp IGAIM 100 Linear (b3lyp CSGT) Linear (b3lyp GIAO) Linear (b3lyp IGAIM) 0 0

50

100

150

200

250

300

350

400

450

Linear (mpw1pw91 CSGT) Linear (mpw1pw91 GIAO) Linear (mpw1pw91 IGAIM)

-100 Linear (olyp CSGT) Linear (olyp GIAO) Linear (olyp IGAIM) -200 Chemical Shifts (ppm)

Plate 4. See also Figure 4 on page 59. 400.00 b3lyp/CSGT b3lyp/GIAO 300.00 b3lyp/IGAIM mpw1pw91/CSGT 200.00 mpw1pw91/GIAO mpw1pw91/IGAIM Chamical Shielding (ppm)

100.00 olyp/CSGT olyp/GIAO 0.00 0

100

200

300

400

500

600

700

olyp/IGAIM Linear (b3lyp/CSGT)

-100.00 Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) -200.00 Linear (mpw1pw91/CSGT) Linear (mpw1pw91/GIAO) -300.00 Linear (mpw1pw91/IGAIM) Linear (olyp/CSGT) -400.00 Linear (olyp/GIAO) Linear (olyp/IGAIM) -500.00 Chemical Shifts (ppm)

Plate 5. See also Figure 5 on page 60.

XIX

Part I

400

Part I Plate 7. See also Figure 4 on page 77.

a c

b

Plate 6. See also Figure 2 on page 76.

B)

A)

C)

Plate 8. See also Figure 1 on page 88.

XX

Part I

Plate 9. See also Figure 4 on page 138.

Plate 10. See also Figure 3 on page 161.

XXI

Part I

Relative signal intensity(-)

100

80 initial 40min. 60 80min. 160min. 40

200min. 280min.

20

360min. 480min

0 1

21 11

41 31

61 51

81 71

91

From surface position

101 121 141 161 181 111 131 151 171 191

1step = 60micro meter

Relative signal intensity(-)

100

80

60

initial

560min.

80min.

640min.

160min.

720min.

240min.

800min.

320min.

880min.

400min.

1080min.

40

20

480min

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

0

From surface position

1step = 60micro meter

Plate 11. See also Figure 1 on page 164.

XXII

(b)

25˚C

(d)

(c)

(e)

(g)

375˚C

350˚C

400˚C

(f)

450˚C

425˚C

(h)

475˚C

Part I

(a)

(i)

500˚C

525˚C

Plate 12. See also Figure 7 on page 170.

Plate 13. See also Figure 8 on page 170.

Plate 14. See also Figure 2 on page 175.

XXIII

Part I Plate 15. See also Figure 3 on page 176.

Plate 16. See also Figure 5 on page 177.

ABS2H, 241 h

ABS2H, 834 h

0

0 3380

1 2 De pt h, 3 m m

3360 3340 4

d/

iel

cF

ti ne

3320

ag

G

3380

1 2 De pt h, 3 m m

3360 3340 4

g

M

Ma

Plate 17. See also Figure 4 on page 184.

XXIV

d/

iel

cF

ti ne

3320

G

Part I

ABS2H 30

%F

20

72 h 241 h 834 h

10

0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Depth, mm Plate 18. See also Figure 5 on page 184.

Plate 19. See also Figure 4 on page 190.

Plate 20. See also Figure 5 on page 190.

gel

5mm

20mm 20mm Plate 21. See also Figure 6 on page 191.

XXV

Part I

a

c

b

250 nm

- D2O/CH3OD-environment

- shrunk gel (barrier)

- collapsed skin

- swollen gel

Plate 22. See also Figure 7 on page 192.

Plate 23. See also Figure 8 on page 192.

Plate 24. See also Figure 9 on page 193.

XXVI

C N H

H

C N H

H C N C

H

C H N

N

+

C N H

H C N C

H

C H N

C H

N H

H

N

+

H

H C N C

H

C H N

C H

N H

Part I

a

N

+.....

N

+.....

N

+.....

C H

N H

b 13

C N

H

H

13

C

H

C H N

N

+

H

13

C

H

C H N

N

+

H

H C N C

H

C H N

C H

N H

C N

H

H C N

C H

N H

C N

H

H C N

C H

N H

c C H

15

N H

H C N C

H

C H N

H 2

N H

13

H

H

C

N 2

+

H

13

2

H

C N 15

N H

13

+

H

H

C

13

2H

N 2

+

H C N C

C

2

H

C N 15

13

+

Plate 25. See also Figure 1 on page 216.

XXVII

H

H

C

N 2

+.....

H

C 15N H H

13

C 15N

C H 15N C H 15N 13C H 15N H 13

13

2

H N

N H

C 15N H H

C H

N H

C N

C H 15N C H 15N 13C H 15N H 13

13

2

H N

H

C H N

H

H C N

C

+

N H

H C N

2H

C

C 15N H

N

15

C

C H

N H

C N

C H 15N C H 15N 13C H 15N H 13

13

2

H N

C H

H

C H N

H

H C N C

C

+

N H

C

2H

C N

d

N

15

H C N

C H

N H

e

C H

+.....

Part I

f 13 2

C 15N

H C

H

C H 15N 13

N H

g 13

C 15N

H

13

13

2

H

13

15

+

N

13 13

C

H

13C 2H 15N C H 15N 13C 2H 15N 2H

+

2

C 15N H

H 13

C

H

13

C 15

13

N

C

H

H 15

15

N H

13

i 13

13

C

N

H

C

N

H

13

C 15

N

15

13

N H

C

H

15

N N

H 13

C

2

13

H

2

13

C H 15N N 2H

H

C 15N

C

15

H

13

2

H

15

C H 2

C

N

N

H

N

C

C

N

H

H H

H C

H

N

N

C-term.

C H

Plate 25. (Continued) See also Figure 1 on page 217.

Plate 26. See also Figure 4 on page 221.

XXVIII

C

+.....

N

C H

C 15N

H 13

13

15

H

N H

H

N 2H

H

C

C

15

2

+

2

C H 15N 15

C 15N

C 15N

C

H H

H 15

N

C

H N-term.

15

13

13

C

13

H

C

+

H

13

H

N

H

N

N

2

13

N

13C 2H 15N C H 15N 13C 2H 15N 2H

h 13

H

15

C 15N H

H

C 2H

C 15N

H

13

13

N H

13

C 15N

H

C 2H 15N 15

2

13 2

N 2H 13C 15

C 2H

2

C 15N H

H

H 13C N

15

2

H 15

13 13

2

N

+.....

Part I

Plate 27. See also Figure 1 on page 234.

Plate 28. See also Figure 1 on page 250.

Plate 29. See also Figure 4 on page 253.

XXIX

Part I Plate 30. See also Figure 5 on page 254.

Plate 31. See also Figure 1 on page 262.

Plate 33. See also Figure 2 on page 280.

Plate 32. See also Figure 2 on page 265.

XXX

Part I

Plate 34. See also Figure 1 on page 307.

Plate 35. See also Figure 1 on page 314.

Plate 36. See also Figure 3 on page 324.

XXXI

Part I

N

N

13

O

φ2

φ3

N

N

φ1

F

C H3

O

F

F

F F

flexible, IC50 = 10 µM

NMR conformation

Constrained, IC50 = 100 nM

Plate 37. See also Figure 4 on page 325.

β α

Plate 38. See also Figure 1 on page 328.

Plate 39. See also Figure 4 on page 331.

XXXII

Part I

Plate 40. See also Figure 5 on page 332.

XXXIII

Part I Plate 43. See also Figure 7 on page 360.

Plate 41. See also Figure 2 on page 340.

Plate 42. See also Figure 5 on page 358.

RheoNMR Controller

Motor, gearbox, drive interface & drive adapter

Drive shaft

Plate 44. See also Figure 2 on page 384.

XXXIV

Cell Kit

Part I

a)

25

20

15

20 0 600

10 500 400 300 tim e ( 200 S) 100

t en

0

em lac isp s gap d ial ros rad ac

velocity (mm/s)

40

5

0

60

b)

50

time (s)

40 30 20 10 0

Plate 45. See also Figure 4 on page 385. Plate 47. See also Figure 3 on page 419.

90x

90-x

d-benzene in sheared PDMS

180y

t2

t1

t

rf

scan of complete cell

45-x 45-x

G 80

Sxx

Doi-Edwards

polymer in the gap

splitting /Hz

60 40

Sxy 20 τd=215

0

ms

-20

Szz

0.5 mm gap

Syy

-40 0

10

20

30

40

50

60

shear rate /s-1 -200 hz

0

f1

200 Hz

Plate 46. See also Figure 6 on page 387.

XXXV

70

80

90

100

Part I

∆σP [ppm] 3

2.5

2

1.5

1

0.5

0

-0.5 30

32

34

36

38

40

42

44

46

48

residue #

Plate 48. See also Figure 4 on page 480.

Plate 49. See also Figure 6 on page 482.

XXXVI

50

Part I

Plate 50. See also Figure 1 on page 498.

Plate 51. See also Figure 2 on page 499.

XXXVII

Part I Plate 52. See also Figure 3 on page 500.

Plate 54. See also Figure 3 on page 508.

Plate 53. See also Figure 1 on page 504.

XXXVIII

Part I

Plate 55. See also Figure 4 on page 509.

XXXIX

Part I Plate 56. See also Figure 5 on page 510.

XL

Part I

Plate 57. See also Figure 2 on page 522.

XLI

Part I Plate 58. See also Figure 3 on page 523.

Plate 59. See also Figure 4 on page 524.

XLII

Part I

Plate 61. See also Figure 1 on page 659.

Plate 60. See also Figure 2 on page 529.

Plate 62. See also Figure 1 on page 680.

XLIII

Part I Plate 63. See also Figure 1 on page 744.

Plate 64. See also Figure 2 on page 744.

XLIV

Part I

Plate 65. See also Figure 3 on page 745.

Group I: Mismatch disappeared at 180 mins

Group II: Mismatch persisted at 180 mins

(A) CBF ADC 30m ADC 180m

CBF (ml/g/min)

(B) LH

30m

(C)

180m

LH

ADC (x10-3 mm2/s) 30m 60m 90m 120m 180m TTC

Plate 66. See also Figure 2 on page 806.

XLV

30m

180m

Part I

T2 +rCBVpeak s (0.40-0.85) A

0.90

B

0.75

0.60

0.45

0.30

0.15

0.00

Plate 67. See also Figure 3 on page 809.

T2

T1

MTR

Cortex EAE

Cortex normal

Plate 68. See also Figure 4 on page 812.

XLVI

Enh

Part I

Plate 70. See also Figure 2 on page 837.

Plate 69. See also Figure 1 on page 836.

-15

a

b

15

c

-15

a’

15

b’

Plate 71. See also Figure 9 on page 843.

XLVII

d

Part I Plate 72. See also Figure 12 on page 848.

Plate 73. See also Figure 13 on page 849.

XLVIII

Part I

Plate 74. See also Figure 14 on page 850.

XLIX

Part I Plate 75. See also Figure 1 on page 866.

L

Part I

-11.8

-11.3

-9.3

-8.3

-7.3

-6.3

-5.3

-4.3

-3.3

-12.72

-0.3

+4.2

+0.7

+5.2

+0.7

-1.3

+1.6

+2.7 8 7.5

8 7.5

7 6.5 6

7 6.5 6

5.5 5

5.5 5

4.5

4.5

Plate 76. See also Figure 3 on page 868.

LI

QUINELRANE-INDUCED LOCOMOTION

Part I

90 Quinelorane 30ug/kg locomotion Quinelorane 30ug/kg locomotion interpolated

70

Quinelorane 30ug/kg locomotion interpolated minus saline locomotion interpolated Saline locomotion Saline locomotion interpolated

50

30

10 0

10

20

30

40

50

-10

-30

-50

MINUTES POST-INJECTION

Plate 77. See also Figure 4 on page 869.

-12.72

-11.8

-11.3

-9.3

-8.3

-7.3

-6.3

-5.3

-4.3

-3.3

+0.7

-1.3

-0.3

+0.7

+2.7

+1.6 8

8 7.5

7.5 7 6.5 6

7 6.5 6

5.5

+4.2

+5.2

5.5 5

5 4.5

4.5

Plate 78. See also Figure 5 on page 870.

LII

60

Part I

-12.72

-11.8

-11.3

-9.3

-8.3

-7.3

-6.3

-5.3

-4.3

-3.3

+0.7

-1.3

+1.6

+2.7

-0.3

+0.7

8 7.5

8 7.5

7 6.5

7 6.5 6

6

5.5 5

5.5

+4.2

+5.2

5 4.5

4.5

Plate 79. See also Figure 6 on page 871.

LIII

120 100 80 Adjusted Locomotion 60 Quinelorane Pharmacokinetics 40 BOLD signal in nucleus accumbens

20 0 0

5

10

15

20

25

30

35

40

45

50

55

Plate 80. See also Figure 7 on page 872.

D-T correlation

fat + water

2.6

2.6

2.4

2.4

2.2

2.2

2.0 1.8

water

1.6 1.4 1.2 1.0

Log10(T) (s)

Part I

140

30

25

20

2 1.8

15

1.6

10

1.4 5

1.2 1

a)

-1.5

-1 -0.5 0 0.5 Log10(D) (10-9 m2/s) b)

Plate 81. See also Figure 8 on page 912.

LIV

0

Part I

Part I, Section 1: Applications in Chemistry

3

Foreword to Application in Chemistry

Magnetic resonance has continued to be an emerging technique, to be applied to almost all fields of pure and applied sciences, including chemistry, physics, biology, materials science, medicine, etc. during past 60 years since its discovery. The applications in chemistry of this volume covers advanced studies on chemical aspect of magnetic resonance spectroscopy and imaging dealing with the state-of-the-art developments of new techniques together with those of basic concepts and techniques, consisting of 93 articles which are grouped to 25 chapters. They are alphabetically arranged for convenience of readers: amyloids, chemical shifts and spin coupling constants, fibrous proteins, field gradient NMR, host-guest chemistry,

imaging, inorganic materials and catalysis, lipid bilayers and bicelles, membrane-associated peptides, membrane proteins, new developments, NOE and chemical exchange, NQR and ESR, organometallic chemistry, paramagnetic effects, protein structures, polymer structure, polymer dynamics, polymer blends, quantum information processing, residual dipolar couplings and nucleic acids, solid state NMR techniques, structural constraints in solids, and telomeric DNA complexes. The section editors are grateful to contributors to this section for their fine contributions. Tetsuo Asakura, Hazime Saitˆo and Isao Ando

5

Abbreviations

AFM: atomic force microscopy

DFT: density functional theory

AHT: average Hamiltonian theory

DIPSHIFT: dipolar chemical shift

Bicelle: bilayered micelles

DNMR: dynamic NMR

BPPLED: bipolar pulse longitudinal eddy current

DNP: dynamic nuclear polarization

BPT: bond polarization theory

DOQSY: double quantum spectroscopy

CC: coupled cluster

DOR: double rotation

CD: circular dichroism

DOSY: diffusion-ordered NMR spectroscopy

CHF: coupled Hartree-Fock

DPMAS: direct polarization magic angle spinning

CNDO: complete neglect of differential overlap

DQ: double quantum

CP-MAS: cross polarization-magic angle spinning CODEX: centerband-only detection of exchange

DQDRAW: double quantum, dipolar recovery with windowless sequence

COSY: correlated spectroscopy

DRAW: dipolar recovery with windowless sequence

CPMG: Carr-Purcell-Meiboom-Gill

DSO: diamagnetic spin orbital

CRAMPS: combined rotation and multiple pulse spectroscopy

EFG: electric field gradient

CRINEPT: cross-correlated relaxation-enhanced polarization transfer

EHT: effective Hamiltonian theory

CRIPT: cross-correlated relaxation induced polarization transfer

EEHT: exact effective Hamiltonian theory

ENDOR: electron nuclear double resonance EPSI: echo planar spectroscopic imaging

CS: chemical shift

EPR: electron paramagnetic resonance

CSA: chemical shift anisotropy

ESRI: electron spin resonance imaging

CSI: chemical shift imaging

Et-NOESY: exchange transferred nuclear Overhauser effect spectroscopy

CTDQFD: constant-time double-quantum filter CTOCD: continuous transformation of the current density

EXSY: exchange spectroscopy FDR: frequency selective dipolar recoupling

DARR: dipolar-assisted rotational resonance

FOV: field of view

DAS: dynamic angle spinning

FPT: finite perturbation theory

DEPT: distortionless enhancement by polarization transfer

FC: Fermi contact

DD-MAS: dipolar decoupled-magic angle spinning

GE-HMQC: gradient enhanced-heternuclear multiple quantum coherence

DECORDER: direction exchange with correlation for orientation-distribution evaluation and reconstruction

GIAO-CHF: gauge-independent atomic-orbitals coupled Hartree-Fock

DFS: double frequency sweep

GC: gas chromatograph

6 Abbreviations

Chemistry

HETCOR: heteronuclear correlation HMBC: heteronuclear multiple bond correlation HMQC: heteronuclear multiple quantum correlation HOHAHA: homonuclear Hartmann Hahn HSQC: heteronuclear single quantum correlation HPDEC: high power decoupling

ONIOM: Our own n-layered integrated molecular Orbital + molecular mechanics ONP: optical nuclear polarization PET: positron emission tomography PFG: pulsed field-gradient PGSE: pulsed gradient-field spin echo

IGLO: individual gauge for localized orbitals

Photo-CIDNP: photochemically induced dynamic nuclear polarization

INADEQUATE: incredible natural abundance double quatum transfer experiment

PISA: polarity index slant angle

INDO: intermediate neglect of differential overlap

PISEMA: polarization inversion spin exchange at magic angle

INEPT: insensitive nuclei enhancement by polarization transfer

PM5: parametric method 5

LC-NMR: liquid chromatography-NMR LDA: local density approximation LDBS: locally dense basis set LORG: localized orbitals local origin

PSO: paramagnetic spin orbital QC: quantum computation QEDOR: quadrupole echo double resonance QED: quantum electrodynamics

LG-CP: Lee Goldburg-cross polarization

QCPMG: quadrupolar Carr-Purcell-Meiboom-Gill refocusing pulse

MAOSS: magic angle oriented sample spinning

QIP: quantum information processing

MAS: magic angle spinning

QM/MM: quantum mechanics/molecular mechanics

MCSCF: multi-configurational self-consistent field

QRI: quantum resonance interferometry

MD: molecular dynamics

PFG: pulse field gradient

MI: molecular imaging

RACO: relayed anisotropy correlation

MOVS: magnetically oriented vesicle systems

RDC: residual dipolar coupling

MQMAS: multiple quantum magic angle spinning

REAPDOR: rotational echo adiabatic passage double resonance

MREV-8: Mansfield-Rhim-Elleman-Vaughan 8 cycle MRI: magnetic resonance imaging MRFM: magnetic resonance force microscopy MSREDOR: multi spin REDOR NMR: nuclear magnetic resonance NMR-MOUSE: NMR-mobile universal surface explorer NOE: nuclear Overhauser enhancement NOESY: nuclear overhauser and exchange spectroscopy

REDOR: rotational echo double resonance RFDR: radio frequency driven resonance RMSD: root mean-square deviation ROCSA: recoupling of chemical shift anisotropy ROCSA-LG: recoupling of chemical shift anisotropyLee Goldburg ROE: rotating frame Overhauser experiment RR: rotational resonance SAIL: stereo-array-isotope-labelling

NQR: nuclear quadrupole resonance

SASS: switching angle sample spinning

ODF: order-director fluctuation

scBCH: semi-continuous Baker-Campbell-Hausdorff

Abbreviations

SDC: superdense coding

STO: Slater-type orbital

SEC-NMR: size exclusion chromatography-NMR

TB MO: tight-binding molecular-orbital

SEDOR: spin echo double resonance

TOCSY: total correlation spectroscopy

SELFIDOQ: separated-local-field double quantum

TORQUE: T one rho quenching

SFAM: simultaneous frequency amplitude modulation

TRAPDOR: transfer of populations in double resonance

SOPPA: second order polarization propagator approximation

TPPM: two pulse phase modulation

SOS: sum-over-states method SQ: single quantum

TROSY: transverse relaxation optimized spectroscopy VFMAS: very fast magic angle spinning

SQUID: superconducting quantum interference device

water LOGSY: water-ligand observation by gradient spectroscopy

SSNMR: solid state NMR

WISE: wide-line separation

STD: saturation transfer difference spectroscopy

XRD: x-ray diffraction

STRAFI: stray field magnetic resonance imaging

ZQ: zero-quantum

7

Part I

Amyloids

11

Miya Kamihira1 , Hazime Saitˆo1 , and Akira Naito2 1 Department

of Life Science, Himeji Institute of Technology, Harima Science Garden City, Kamigori, Hyogo 678-1297, Japan; and 2 Department of Engineering, Yokohama National University, Hodogaya, Yokohama 240-8501, Japan

Introduction Amyloid fibril formation is one of the common phenomena associated with many serious diseases such as Alzheimer’s disease, Parkinson’s, bovine spongiform encephalopathy (BSE), scrapie, and so on. Independent of the constituent polypeptides, the amyloid fibrils exhibit highly organized filamentous structures which are typi˚ in diameter, as revealed by electron mically ∼100 A croscopy and atomic force microscopy. Mechanism of the amyloid fibril formation has been extensively studied by various spectroscopic techniques, related to misfolding of proteins. Especially, solid-state NMR spectroscopy has made a great contribution to determine the structures of the fibrils from several peptides/proteins at the molecular level. For example, Alzheimer’s β-amyloid peptides, which consist of 40–42 amino acid residues, have gained insights into the three-dimensional (3D) structures in the fibrils as a double-layered cross-β structure with parallel β-sheets by accumulating the local and spatial conformational restraints [1–3]. Also, an 11-residue fragment of human transthyretin (TTR) in its fibrillar form which in vivo is allied with familial amyloid polyneuropathy and senile systemic amyloidosis, was revealed the complete 3D structures of the extended β-strand conformation, by establishing dihedral angles of the backbone and 13 C– 15 N distances [4,5]. These results indicate that solid-state NMR spectroscopy is a powerful tool to determine the non-crystal, non-soluble, fibrillar structures. In this chapter, a solid-state NMR application on the kinetics analyses of the amyloid fibril formation is described. Human calcitonin (hCT) is a thyroid hormone which regulates the mineral metabolism in the bones [6– 8]. hCT contains 32 amino acid residues and its sequence is CGNLSTCMLGTYTQDFNKFHTFPQTAIGVGAPNH2 with a disulfide bond between Cys1 and Cys7 and a C-terminus amide. In a high concentrated solution, however, it is known to form the amyloid fibrils, which are ˚ in diameter [9,10]. typically 80 A

Properties of Fibril Formation of hCT For concentrations above 15 mg/ml hCT, the solution changes in time into a turbid gel as the end fibrillated state, Graham A. Webb (ed.), Modern Magnetic Resonance, 11–17.
C 2008 Springer.

while for the concentrations below and around 1 mg/ml, the equilibrium state consists of a clear solution containing punctuate aggregates which may precipitate [11]. Long fibrils were observed from the gel (pH 3.3, 80 mg/ml) and short fibril aggregates were seen in the precipitates from a diluted solution (1.5 mg/ml, pH 7.5) [12]. Turbidity measurement showed absorption of the hCT solution increased gradually after a lag time which was dependent on the hCT concentrations [11]. As mentioned later, from the results, the kinetics of hCT fibrillation was proposed to be a double nucleation mechanism [11,13–15]. The fibrillation is also temperature-dependent and an apparent activation enthalpy for the reaction was obtained to be 20 kcal/mol at 10 mg/ml (pH 7.4) [11]. Also the rate of the fibril formation was found to be largely pH-dependent and in acidic solution it is much slower than that in neutral pH [11,12]. Solution NMR studies on hCT (80 mg/ml, pH 2.9) showed a gradual broadening of the peptide peaks, followed by a rapid broadening and subsequent disappearance of the NMR signals [16]. The phenomenon was not seen simultaneously and the peaks from the residues in the N-terminal (Cys1 -Cys7 ) and in the central (Met8 -Pro23 ) regions broadened and disappeared faster than those in the C-terminal region (Gln24 -Pro32 ). Furthermore the peaks of Cys1 , Leu4,9 , Met8 , Tyr12 , Asp15 , and Phe16,19,22 disappeared faster than the others [16]. These results together with hydrogen–deuterium exchange of amide protons indicate that the amphiphilicity of hCT in the central region might cause a formation of α-helical bundles leading to the fibril formation [16].

Conformational Changes of hCT To determine the fibrillation process further, the solidstate NMR methods were applied. For this purpose the conformation-dependent 13 C chemical shifts are efficient means to determine the secondary structures around the 13 C sites straightaway [17–21], especially in the case where the state changes as elapsed time. 13 C CP-MAS spectra of the hCT fibrils formed in 15 mM acetic acid solution (80 mg/ml) showed much narrower signals than those before dissolved in the solution (lyophilized powder) (Figure 1), suggesting that the fibril is conformationally more homogeneous than the lyophilized powder. Also

Part I

Kinetics of Amyloid Fibril Formation of Human Calcitonin

12 Part I

Chemistry

Part I

A

B

ppm 100

150

50

0

C Relative Intensity (%)

100 80 60

α-helix random coil β-sheet

40 20 0

A

B sample

Fig. 1. 13 C CP-MAS spectra of lyophilized powder of hCT (Ciba-Geigy, Japan) (A) and the hCT fibrils obtained after 48 h from dissolution in 15 mM acetic acid solution (80 mg/ml) (B). The spectra were recorded on a Chemagnetics CMX 400 NMR spectrometer at the resonance frequency of 100.6 MHz (13 C). Insets show deconvoluted spectra of the carbonyl resonances using Peak Fit (SPSS Inc., Chicago, USA). The deconvoluted signals were assigned to be β-sheet (lower frequency than 172.2 ppm; black bars), random coil (172.2–174.5 ppm; open bars), and α-helix (higher frequency than 174.5 ppm; hatched bars) respectively (C).

it is noted that the peak positions have shifted. Deconvolution of the carbonyl signals clearly indicated that the β-sheet conformation gained largely during the fibril formation (Figure 1C). The signals which could be assigned to Thr Cβ marked by arrows were also shifted from 65.7

(assigned to random coil) to 67.8 ppm (β-sheet) [18] (Figure 1A and B). Local conformations were examined using site-specifically 13 C labeled hCTs [12]. DD-MAS (single 90o pulse excitation with a proton decoupling under magic angle spinning) and CP-MAS (cross-polarization with a proton decoupling under magic angle spinning) spectra were recorded, since in the CP-MAS spectra a fibril component was detected, while in the DD-MAS spectrum the solution component was mainly observed, especially for the carbonyl carbons in view of the long spin–lattice relaxation time compared with the recycle delay. During the fibrillation at pH 3.3, it was clarified that conformational transitions occur from an α-helix (in the solution) to a βsheet structure (in the fibril), and from a random coil to a β-sheet structure in the central region, around Gly10 and Phe22 , respectively [12]. The C-terminus region (around Ala26 and Ala31 ) also changed the conformation partially from a random coil to a β-sheet structure [12]. Further the existence of polymorphs of the fibrils was clearly shown in molecular level, depending on the pH (3.3, 4.1, and 7.5) in the solution [12,22]. It is suggested that at pH 7.5 hCT forms the antiparallel β-sheet by a favorable electrostatic interaction between Asp15 (−) and Lys18 (+), in addition to the hydrophobic interaction among the amphiphilic helices [12]. The fibrils at pH 3.3 may be a mixture of antiparallel and parallel β-sheet structures, because no attractive ionic interaction to fix the unique direction for the molecular association is present in this case to result in the presence of two conformations up to the C-terminus region [12]. Whereas the β-sheet formed at pH 4.1 is shorter than the others, suggesting probable ionic interactions of the side chain of Asp15 with the amino group of the N-terminus (Cys1 ), rather than with the side chains of Lys18 (+) or His20 (+) [22]. Accordingly, it was demonstrated that the charged residues existing in hCT (amide nitrogen, Asp15 , His18 , and Lys20 ) play a central role for determination of the molecular alignment of the hCT monomers. Indeed absence of the negative charge at Asp15 (mutated to Asn15 : D15N-hCT) did not make any differences in the local conformations of the fibrils from neutral and acidic solution [23].

Kinetic Analysis of hCT Fibrillation When hCT solution was dissolved in 15 mM acetic acid solution (80 mg/ml, pH 3.3), it became a turbid, viscous gel in 2–3 days. Time course of the 13 C DD- and CP-MAS NMR spectra accumulated repeatedly, showed gradual increase of the CP-MAS signals, synchronously with decrease of the DD-MAS signals, after a certain time (Figure 2A). Although a MAS frequency of 4000 Hz was applied to the sample in a 5-mm o.d. rotor in a 400 MHz spectrometer, the observed increases in the CP-MAS signals were corresponding to the increases in turbidity and

Kinetics of Amyloid Fibril Formation

Kinetic Analysis of hCT Fibrillation 13

Ellipticity (m°)

Part I

Wave length (nm) Fig. 2. Time course of 13 C DD- (A) and CP-MAS (B) NMR spectra of hCT monomers and fibrils, respectively [pH 3.3, 80 mg/ml (23.4 mM)] at 20 ◦ C. The number of accumulations for the DD- and CP-MAS signals was 1000 and 2000, respectively. Magic angle spinning frequency of 4000 Hz was applied. Stacked CD spectra measured on an AVIV model 62DS using quartz cuvettes with path length of 0.02 cm (B). Sample concentration was 0.2 mg/ml (58.5 µM) in 20 mM phosphate buffer (pH 7.5). Temperature was controlled to 25 ◦ C. The time when hCT was dissolved was regarded as 0.

Fig. 3. A plot of [1-13 C]Gly10 peak heights in 13 C DD- (open circle) and CP-MAS (closed circle) of hCT (pH 3.3, 80 mg/ml) against the elapsed time (A). The time of dissolution was taken as 0. Acquisition was started 6 h after dissolution. The intensity of the CP-MAS signals was normalized as that observed at 119 h after dissolution as unity (B). The line in (B) shows the best fit to Equation (7) representing the two-step reaction mechanism.

proposed that a formation of micelles which corresponds to the α-helical bundle [16], are reversibly formed from monomers with the same aggregation number n 0 (An 0 ), as shown in Reaction (1) (Figure 4A and B), n 0 A(monomers) ↔ An 0 (micelle).

viscosity by visual observation of the rest of the same solution located outside the magnet. The changes of the peak intensities in the DD- and CP-MAS spectra (Figures 2B and 3A) show a two-step reaction process: for the case of [1-13 C]Gly10 -hCT, the first and the second step may occur at ∼60 and 60–118 h, respectively. The chained changes in the DD- and CP-MAS spectra and the presence of the lag time suggest that this hCT fibrillation process could be explained by the two-step autocatalytic reaction mechanism, in which the first reaction is a homogeneous nucleation step and the second one is a heterogeneous fibrillation process to elongate and to mature the fibrils. Here, the components of hCT molecules observed by the DD- and CP-MAS experiments are defined as A and B forms, respectively. For an early stage, it is

(1)

Here, the hCT molecules in the monomer and the micelle states are supposed to give the same DD-MAS NMR signals as the other signals were not appeared. We consider the case where the total hCT concentration of A form ([AT ] = n 0 [An 0 ] + c∗ ) is always much higher than the critical micellar concentration c∗ , then [An 0 ] can be expressed as [AT ]/n 0 . Under these conditions, the first reaction step is given by, k1

An 0 −→ Bn 0 ,

(2)

where k1 is the rate constant of Reaction (2) and Bn 0 is the nucleus of fibril consisting of n 0 number of hCT. If f is defined as the fraction of the B form (fibril) in the system,

14 Part I

Chemistry

Part I

Fig. 4. Schematic representation of a proposed model for the fibril formation. hCT monomers in solution (A) make a homogeneous association to form α-helical bundles (micelles) (B) and simultaneously they change conformations to form β-sheet (C) which could be nuclei of the fibril for heterogeneous fibrillation process to grow the fibril (D).

the kinetic equation of Reaction (2) can be given by
 df = k1 (1 − f ). (3) dt 1 The second autocatalytic fibrillation process can be given by k2

A + Bn −→ Bn + 1,

of hCT in the solution. Although micelles are formed, individual hCT molecules (A form) could also react with Bn in the second autocatalytic step. As a consequence of the Reaction (4), [AT ] = a(1 − f ) can be used as [A], and [B] increases stepwise after a certain delay time. The relevant differential equation is given by


(4)

where k2 is the rate constant of the Reaction (4) and Bn and Bn+1 are the elongated fibrils with n and n + 1 numbers of hCT molecules, respectively. In this process, each hCT molecule in Bn is assumed to act as catalytic sites to accelerate the change from A to B forms. Thus [Bn ] can be replaced by [B] = n 0 [Bn 0 ] + (n 0 + 1)[Bn0 + 1] + · · · + n[Bn ] + · · · = af in the kinetic equation where a is the initial concentration

df dt

 = k2 a f (1 − f ).

(5)

2

Then the overall kinetic equation for the two-step autocatalytic reaction may be expressed as df = dt




df dt




+ 1

df dt

 = k1 (1 − f ) + k2 a f (1 − f ). 2

(6) The differential equation can be integrated to provide f =

ρ {exp [(1 + ρ) kt] − 1} , 1 + ρ exp [(1 + ρ) kt]

(7)

Kinetics of Amyloid Fibril Formation

Kinetic Analysis of hCT Fibrillation 15

Part I

Table 1: Kinetic parameters for the fibril formation of hCTs in various pH solution

Sample pH

Method (observed signal)

Concentration (mM)

k1 (s−1 )

k2 (s−1 M−1 )

ak2 (s−1 )

hCT pH 3.3∗ pH 4.1 pH 7.5∗ pH 7.5∗

NMR (Gly10 C=O) NMR (Gly10 C=O) CD (205 nm) CD (205 nm)

23.4 23.4 0.0585 0.439

2.71(±0.11) × 10−8 3.86(±1.79) × 10−6 2.79(±0.04) × 10−6 6.44(±0.29) × 10−8

9.01(±0.81) × 10−4 5.89(±2.94) × 10−4 2.29(±0.14) 2.78(±0.19)

2.11(±0.19) × 10−5 1.38(±0.69) × 10−5 1.34(±0.08) × 10−4 1.22(±0.08) × 10−3

DFNKF† pH 7.5

NMR (Phe2(16) C=O)

23.4

1.02(±0.35) × 10−7

7.28(±1.54) × 10−3

1.70(±0.36) × 10−4

† Taken

from Ref. [12]. from Ref. [22].

under the boundary condition of t = 0 and f = 0, where ρ = k1 /k represents the dimensionless value to describe the ratio of k1 to k and k = ak2 is an effective rate constant of Reaction (4). After the peak height observed in the CPMAS spectra was normalized as that observed at 119 h after dissolution as unity (Figure 3B), fitting of the data to the Equation (7) yielded the rate constants, k1 and k2 , separately (Table 1). Almost the same values were obtained from increase of the intensities in the methyl signals in the 13 C CP-MAS signals as well [12] or from analysis of the decrease of the DD-MAS signals (data not shown). A proposed fibrillation process is illustrated in Figure 4. Similarly the rate constants for the fibrillation at pH 4.1 were obtained [22]. The fibrillation of hCT at pH 7.5 was examined using CD spectroscopy instead of NMR at low peptide concentrations (0.2 and 1.5 mg/ml), because the solution becomes a gel in a short time. Decrease of the intensity was observed as elapsed time (Figure 2B) and the same reaction mechanism was applied to it too (Table 1) [12]. Although the effective rate constant, ak2 , is affected by different initial concentrations, it is considered that the reaction rates should be compared as the rate constants, k1 and k2 . This fact was justified by observing that the comparable k2 values were determined in the two different initial concentrations (at pH 7.5; Table 1). The most striking feature was that in the case of fibril formation of hCT the k1 values were a couple of orders smaller than the k2 and ak2 values (Table 1). This suggests that the first homogeneous nucleation process is much slower than the second heterogeneous fibrillation step. Simulation of the Equation (7) reveals that if k1 ≥ k2 , the lag time disappears and as k1 becomes longer the lag time increases gradually (Figure 5A). Basically, ak2 (k) and the ratio of k1 and k(ρ) determine the reaction effectively. On the other hand, it became clear that if k2 (k) increases by one order, the reaction attains to f = 1 by ∼10 times faster (Figure 5B), while 10 times larger k1 does not provide such big differences (Figure 5A). Accordingly,

1.0

0.6

fraction of fibril (f )

∗ Taken

0.2 0

1.0

0.6

0.2 0 0

2000

6000

10000

t Fig. 5. A computational simulation of kinetics of the two-step autocatalytic reaction. The plot (A) shows at k = 10−2 and ρ is varied for 10 (open circle), 1 (closed square), 10−1 (open diamond), 10−2 (×), and 10−3 (+) respectively, representing the effect of k1 under fixed k2 on the plot. Whereas the plot (B) shows the variation of k and ρ at the same time to demonstrate the effect of k2 under fixed k1 : closed circle; 10−1 , open square; 10−2 , closed diamond; 10−3 , cross; 10−4 .

16 Part I

Chemistry

Part I

reflecting the large difference in the lag time, clear difference in the k2 values appeared among the samples at pH 3.3 (and 4.1) and 7.5 (Table 1). Thus it is important to determine the rate constants for the first and the second reaction steps separately. The separation of k1 from k2 is also important to discuss the factor of fibrillation mechanism in the first step separately from that in the second step.

Mechanism of Fibril Formation Recently many models have been proposed for the mechanism of amyloid fibril formations from several peptides/proteins [24]. The double nucleation mechanism explains the fibril formation starts with a homogeneous nucleation step from hCT monomers and afterward fibrillation continues with development of new fibrils from existing ones [11,13–15]. Formation of peptide micelles above a certain critical peptide concentration has been proposed in the nucleated polymerization model in which fibrils nucleate within these micelles or on existing nuclei (seeds) heterogeneously, following fibrils grow by irreversible binding of monomers to the fibrils ends [25– 27]. Then the nucleated conformational conversion model describes that structurally fluid oligomeric complexes accumulate into nuclei or associate with existing ones where conformational conversion takes place as a ratedetermining step [28]. The autocatalytic reaction mechanism we proposed, however, explains the conformational changes occurred together and the rate-limiting step that is characteristic to the amyloid fibril formation clearly. In the solution state, there also exist several different models. It has been proposed for the amyloidosis of β2 microglobulin that there should be a monomeric amyloidogenic intermediate from a native monomer to assemble each other to form an early assembly intermediate, following it changes to a nucleus where the monomeric intermediates make an interaction together to elongate the fibril [29,30]. On the contrary, in a mathematical model, a rapid, irreversible commitment occurs to form either stable monomer/dimer or unstable intermediate, only which associates cooperatively into a multimeric nucleus (filament) [31]. Further, elongation of the filament may occur via addition of the unstable intermediate and by end-toend association of the filaments [31]. We have also considered that many other fast reactions may exist during the process of a large fibril formation. However, the secondary structure and the chemical environments of the components observed in the DD- and CP-MAS spectra did not change throughout this process, and no additional peaks were observed (Figure 2A). These findings in 13 C NMR experiments imply that it is sufficient to consider the two-step reaction for the fibrillation kinetics of peptides.

Then what is the direct force to cause the molecular interaction among the monomers, to form a nucleus (at the first step) or to make an interaction of a monomer with a nucleus (at the second step)? Generally, it has been claimed that the hydrophobic and the electrostatic interaction might be necessary for the fibril formation since the “core region” which is essential to form a fibril, contains (a) cluster(s) of those amino acid residues. In the case of hCT, there exists only one negatively charged residue (Asp15 ) in it, together with 18 three positively charged group/side chains (NH+ 3 , Lys , and His20 ). And the results used D15N-hCT demonstrated that the negative charge at Asp15 does not increase the rate of fibrillation [22,23]. Instead larger positive net charges around Lys18 and His20 could cause decrease of the reaction rates, because the side chains of them locate on the same side of the β-strand which might destabilize the structure and disturb elongation of the fibrils [23]. On the other hand, a hCT fragment DFNKF (15–19) which is the shortest one to form a fibril in hCT [32] and is determined to be important for in vivo bioactivity too [33], gave 300 times smaller k2 values at pH 7.5 compared with those of hCT at pH 7.5 (Table 1) [22]. Further, the loss of aromatic rings in the central region was observed to cause the delay in the second step of the fibrillation (Kamihira et al., manuscript in preparation). These results could be a clue to the elucidation of the molecular association to lead to fibril formation.

Conclusion It was clearly demonstrated that the use of solid-state NMR spectroscopy is very efficient to determine the local conformational changes during the amyloid fibril formation of hCT. Especially the analysis of the signal intensities enabled to examine the kinetic property of hCT fibrillation as a two-step autocatalytic reaction. Further determination using this method could clarify the mechanism of amyloid fibril formations more in detail.

Acknowledgment We thank Dr. Atsuko Y. Nosaka for helpful discussions.

References 1. 2. 3. 4.

Tycko R. Curr. Opin. Chem. Biol. 2000;4:500. Tycko R. Methods Enzymol. 2001;339:390. Tycko R. Curr. Opin. Struct. Biol. 2004;14:96. Jaroniec CP, MacPhee CE, Bajaj VS, McMahon MT, Dobson CM, Griffin RG. Proc. Natl. Acad. Sci. U.S.A. 2004;101:711. 5. Jaroniec CP, MacPhee CE, Astrof NS, Dobson CM, Griffin RG. Proc. Natl. Acad. Sci. U.S.A. 2002;99:16748.

Kinetics of Amyloid Fibril Formation

21. Wishart DS, Sykes BD, Richards FM. J. Mol. Biol. 1991;222:311. 22. Naito A, Kamihira M, Inoue R, Saito H. Magn. Reson. Chem. 2004;42:247. 23. Kamihira M, Oshiro Y, Tuzi S, Nosaka YA, Saito H, Naito A. J. Biol. Chem. 2003;278:2859. 24. Zerovnik E. Eur. J. Biochem. 2002;269:3362. 25. Lomakin A, Chung DS, Benedek GB, Kirechner DA, Teplow DB. Proc. Natl. Acad. Sci. U.S.A. 1996;93:1125. 26. Lomakin A, Teplow DB, Kirschner DA, Benedek GB. Proc. Natl. Acad. Sci. U.S.A. 1997;94:7942. 27. Walsh DM, et al. J. Biol. Chem. 1999;274:25945. 28. Serio TR, Cashikar AG, Kowal AS, Sawicki GJ, Moslehi JJ, Serpell L, Arnsdorf MF, Lindquist SL. Science. 2000;289:1317. 29. McParland VJ, Kalverda AP, Homans SW, Radford SE. Nat. Struct. Biol. 2002;9:326. 30. Hoshino M, Katou H, Hagihara Y, Hasegawa K, Naiki H, Goto Y. Nat. Struct. Biol. 2002;9:332. 31. Pallitto MM, Murphy RM. Biophys. J. 2001;81:1805. 32. Reches M, Porat Y, Gazit E. J. Biol. Chem. 2002;277:35475. 33. Kazantzis A, Waldner M, Taylor JW, Kapurniotu A. Eur. J. Biochem. 2002;269:780.

Part I

6. Copp DH, Cameron EC, Cheney BA, Davidson AGF, Henze KG. Endocrinology. 1962;70:638. 7. Kumar MA, Foster GV, MacIntyre I. Lancet. 1963;2:480. 8. Austin LA, Heath HD. N. Engl. J. Med. 1981;304:269. 9. Sieber P, Riniker B, Brugger M, Kamber B, Rittel W. Helv. Chim. Acta. 1970;53:2135. 10. Bauer HH, Aebi U, Haner M, Hermann R, Muller M, Merkle HP. J. Struct. Biol. 1995;115:1. 11. Arvinte T, Cudd A, Drake AF. J. Biol. Chem. 1993;268:6415. 12. Kamihira M, Naito A, Tuzi S, Nosaka YA, Saito H. Protein Sci. 2000;9:867. 13. Ferrone FA, Hofrichter J, Eaton WA. J. Mol. Biol. 1985;183:591. 14. Ferrone FA, Hofrichter J, Sunshine HR, Eaton WA. Biophys. J. 1980;32:361. 15. Samuel RE, Salmon ED, Briehl RW. Nature. 1990;345: 833. 16. Kanaori K, Nosaka AY. Biochemistry. 1995;34:12138. 17. Saito H. Magn. Reson. Chem. 1986;24:835. 18. Saito H, Ando I. Annu. Rep. NMR Spectrosc. 1989;36:209. 19. Saito H, Tuzi S, Naito A. Annu. Rep. NMR Spectrosc. 1998;36:79. 20. Wishart DS, Sykes BD. Methods Enzymol. 1994;239:363.

References 17

19

Oleg N. Antzutkin Division of Chemistry, Lule˚a University of Technology, S-971 87 Lule˚a, Sweden

Abstract An overview of the strategy and experimental solid-state NMR, STEM, and AFM methods useful for obtaining structural constraints on Alzheimer’s amyloid-β peptide fibrils is presented. Polymorphism of amyloid fibrils and the relevance to neurotoxicity is discussed. Abbreviations: STEM, scanning transmission electron microscopy; AFM, atomic force microscopy. Alzheimer’s disease (AD) is a form of senile dementia, which affects ca. 40 million senior citizens worldwide [1]. AD is one of the most expensive diseases because an intense daycare of patients is needed over many years. For example, direct and indirect costs of AD and other forms of dementia in Sweden alone (ca. 160,000 patients, a half of them with AD) amount to more than 5400 million euro in 2004 [2]. To this day, a microscopic diagnosis of AD is made on the invariable presence of the two primary criteria: (i) the presence of extracellular senile amyloid plaques surrounded by dead or severely damaged nerve cells in certain regions of the cerebral cortex, such as the hippocampus, amygdala, and other regions of the brain important for memory, learning, and judgment and (ii) dense bundles of abnormal fibers, neurofibrillar tangles, formed by another normally occurring neuronal protein, tau-protein, in the cytoplasm of certain degenerating neurons [3]. In addition, the pathology of AD was found to involve marked decreases in acetylcholine, the chemical used by nerve cells to transmit signals. In 1984 Glenner and Wong [4,5] and later Masters et al. [6] found that amyloid isolated from blood vessels in the meninges and from Alzheimer’s amyloid plaques consist of ca. 90% of amyloid-β-peptides (Aβ), 39–43 amino acid residue peptides with the principal components Aβ(1−40) and Aβ(1−42) : DAEFRHDSGY10 EVHHQKLVFF20 AEDVGSNKGA30 IIGLMVGGVV40 IA [4–8]. Transmission electron microscopy of amyloid plaques revealed numerous unbranched Aβ amyloid fibrils with diameter 6–12 nm, surrounded by the amorphous aggregates, diffuse amyloid. Despite an imperfect correlation between amyloid deposits and dementia [9–11], recent reports from various research groups all tend to implicate Aβ, Graham A. Webb (ed.), Modern Magnetic Resonance, 19–27.
C 2008 Springer.

rather than tau-protein, as triggering a long cascade of biochemical reactions finally leading to neurodegeneration [12–14]. Early onsets of AD have been connected to the following genetic factors: (i) mutations in proteins Presenilin I and Presenilin II (putatively assigned to γ -secretases [12]) which lead to elevated plasma concentrations of Aβ(1−42) , a more hydrophobic and neurotoxic form of human Aβ; (ii) point mutations in human Aβ(1−40) (and Aβ(1−42) ), Aβ(1−40) (A21G) “Flemish” [15], Aβ(1−40) (E22Q) “Dutch” [16], Aβ(1−40) (E22K) “Italian” [17], Aβ(1−40) (E22G) “Arctic” [18], and Aβ(1−40) (D23N) “Iowa” [19]. Remarkably, these mutants aggregate faster and at lower critical concentrations compared with the wild type Aβ(1−40) [20]. Also, it has been found that the “Arctic” mutant found in Northern Swedish families with an early onset of AD (54–61 years), Aβ(1−40) (E22G), forms a larger relative amount of smaller aggregates and proto-fibrils [18] as well as the peptide shows a high degree of polymorphism of amyloid fibrils grown in TRIS buffer solution at pH 7.4 (two non-coiled and three coiled types of fibrils) [21,22]. The sequence of Aβ includes the first 28 residues (mainly hydrophilic) of the extracellular domain and 11– 15 residues (mainly hydrophobic) of the transmembrane region of a 695-residue amyloid precursor protein (APP), whose function is not fully understood [7]. It is believed that putatively incorrect processing [23] or abnormal posttranslational modifications of APP [24] give rise to the extracellular neurotoxic Aβ. Moreover, it has been observed in vitro that significant levels of peptide aggregation into a structural fibrillar form are always associated with significant Aβ-induced neurotoxicity [25–27]. The exact mechanism of the Aβ neurotoxicity is still unknown, though recent reports suggest that Aβ(1−40) and Aβ(1−42) amyloid fibrils with distinct different morphologies and different supramolecular structures show remarkably different toxicity in vitro for cell cultures of hippocampal neurons obtained from rat embryos [28,29]. Even small but putatively structured Aβ oligomers have been found toxic for nerve cell cultures both in vitro [13] and in vivo [14]. These make structural studies on Alzheimer’s amyloid fibers and oligomers significant neuropathologically. It has been reported in a fair number of studies that both synthetic Aβ and its fragments [30–46], as well as isolated Alzheimer’s senile plaque proteins [47,48],

Part I

Polymorphism of Alzheimer’s Aβ Amyloid Fibrils

20 Part I

Chemistry

Part I

˚ amyspontaneously assemble into the typical 60–120 A loid fibers that exibit a β-pleated sheet conformation and other properties consistent with native AD Aβ-peptides. Morphology and macroscopic features of the amyloid fibrils (shape, left or right handed twist, pitch, maximum and minimum heights) can be readily studied by AFM (see Figures 1A and B) or/and TEM (not shown). AFM is advantageous over TEM, because AFM allows very fine measurements of height and, therefore, all dimensions of coiled amyloid fibrils can be precisely elucidated (see Figure 1C and D). The latter is crucial in distinguishing of different types of fibril morphology, which may correlate with different neurotoxicity as discussed above. Polymorphism of Alzheimer’s Aβ(1−40) (two types of amyloid fibrils [28]) and Aβ(1−40) E22G (“Arctic” mutant, five different types of fibrils [21,22]) has been recently investigated in detail. Structural features of amyloid fibrils can be further investigated by scanning transmission electron microscopy (STEM) [28,49,50] (mass-per-length measurements, Figure 1E) using the methods developed in the earlier works [51–54]. By STEM mass-per-length of different polymorphs of amyloid fibrils can be measured by recording intensity of the dark-field STEM image across an amyloid fibril, correcting it on image background and further normalizing to intensity of the signal across a calibrant, tobacco mosaic virus (TMV) with mass-per-length 131 kDa/nm known from single crystal X-ray diffraction data. Additional information that Alzheimer’s amyloid fibrils adopt a cross-β-sheet structure with multiple folded or stacked β-sheet laminae, comes from X-ray diffraction data on oriented amyloid fibrils [32]. A cross-β-sheet structure is characterized by two typical reflections, which ˚ spacing (i.e. intermolecular hydrocorrespond to ca. 4.9 A ˚ spacing (interlamigen bonding) perpendicular to ∼10 A nae spacing filled with side groups of amino acid residues forming structures stabilized by van der Waals and electrostatic interactions). Therefore, mass-per-length statistics of amyloid fibrils can be readily recalculated into a number of β-sheet laminae folded and packed into fibrils, provided that molecular weight of peptide molecules is known (for example, MW(Aβ(1−40) ) ≈4329 a.u.). Fibrils with different morphology may be composed of different number of laminae: two, three or four, depending on the peptide sequence and on the preparation procedure (acidic or neutral pH, type of buffer, incubation with or without agitation [21,28,49]. For example, Alzheimer’s amyloid fibrils of Aβ(1−40) (R5G, Y10F, H13R), called as “rat” or “rodent” sequence, prepared at pH 7.25 in nonbuffered solution and without additional agitation, consist of three folded laminae (see statistics in Figure 1E). Human Aβ(1−40) fibrils prepared in vertical dialysis tubes in an unstirred bath of phosphate buffer solution at pH 7.45 also consist of three laminae. In contrast, a gentle circular agitation of 0.1–1.0 mM Aβ(1−40) solutions

(buffered or non-buffered, pH 7.4, room temperature, 1– 10 weeks of incubation) in polypropylene tubes gives rise to fibrils with either two or four β-sheet laminae and with remarkably lower neurotoxicity compared with the quiescent three-laminated fibrils [28]. Interestingly, Aβ(1−42) , which is more amyloidogenic peptide, and a shorter peptide, Aβ(10−35) , usually used as a convenient model system [55–58], also form amyloid fibrils with either two or four β-sheet laminae when prepared in non-buffered solutions at either pH 3.8 or 7.4 with additional agitation [49]. However, mass-per-length STEM measurements combined with X-ray diffraction data on oriented amyloid fibrils provide only information about a number of β-sheet laminae packed and propagating along the long axis of fibrils. Pertinent questions about more detailed supramolecular structure of fibrils and their properties are: (i) Which fragments of peptide molecules form parallel or antiparallel β-sheets? (ii) Are any fragments of molecules, which do not adopt a β-sheet secondary structure? (i.e. forming α-helices, turns or random coils) (iii)What are precise structural parameters of these non-β-sheet structural fragments? (iv) Do certain side groups of amino acid residues form salt bridges? (v) Which side groups are packed together forming hydrophobic clusters? (vi) How a single β-sheet lamina is folded, “upwards” or “downwards”, exposing side groups of different amino acid residues to the surrounding solution? (vii) What are these amino acid residues, which side groups form outer surfaces of amyloid fibrils and which may, therefore, induce neurotoxicity or can be accessed by “attacking” inhibitors or proteases of amyloid fibrils in vivo? (viii) How two (or three, or four) different β-sheet laminae are folded and packed together giving rise to a “ready” amyloid fibril? (ix) Where are binding sites either for metal ions that can stabilize certain structures of Aβ oligomers and fibrils (Cu2+ , Zn2+ , Fe3+ , and Al3+ found in Alzheimer’s amyloid plaques) or for highly soluble metal–ligand complexes that can easily penetrate brain-blood-barrier (for example, alumninum citrates, [Al3 (OH)(H–1 Cit)3 ]4– or [Al3 (OH)4 (H–1 Cit)3 ]7– )? (x) How different known pathologic point mutations in the peptide sequence affect the supramolecular structure, polymorphism, aggregation kinetics, and neurotoxicity of amyloid fibrils? (xi) Is the structure of aggregation intermediates (small oligomers, spherical bodies or protofilaments, which are also believed to be neurotoxic [13,14]) similar to structural fragments of amyloid fibrils or structural transitions do occur in the course of the aggregation cascade? (xii) How the assembly of amyloid fibrils proceeds, either by single molecules or by small oligomeric domains? Solid-state NMR spectroscopy combined with either selective or uniform 13 C, 15 N isotopic labeling of Alzheimer’s β-amyloid peptides were found as useful methods for answering on some of aforementioned questions about supramolecular structure of amyloid fibrils

Polymorphism of Alzheimer’s Aβ Amyloid Fibril

Polymorphism of Alzheimer’s Aβ Amyloid Fibril 21

Part I

Fig. 1. (A–D) Tapping mode AFM images of Aβ(1−40) preparations on mica. Aβ(1−40) was incubated at 50 µM in TRIS buffer (10 mM TRIS, 0.5 mM EDTA, 10 mM KCl and 0.01wt% NaN3 , pH 7.4 adjusted with NaOH) in plastic tubes without additional agitation for 8 days. Images were obtained by M. Hellberg and N. Norlin (MSc thesis, Lule˚a University of Technology, 2003). Heights (C) can be readily measured across the fibril (D). (E) Mass-per-length of Aβ(1−40) (R5G, Y10F, H13R), “rat”-sequence fibrils (incubated at pH 7.25 for 4 months without additional agitation) as determined by STEM. Fibrils appear as narrow structures of uniform intensity. Images were recorded at a dose of approximately 103 e/nm2 and a pixel size of 1 nm. White boxes show typical regions from which the integrated signal in a 100 nm segment of fibril and background film were measured; the TMV was used as a calibrant (seen as wide fibrillar structures). Scale bar is 50 nm. Mass-per-length values of 215 individual fibril segments pooled into (Continued overleaf )

22 Part I

Chemistry

Part I

(see Figure 1F) [21,59–61]. Recently, useful structural constraints on Alzheimer’s amyloid fibrils were obtained from solid-state NMR and new structural models for Aβ(10−35) and Aβ(1−40) fibril polymorphs with either two or four β-sheet laminae have been developed [49,62]. The model for Aβ(1−40) fibrils was based on: (i) a few tens of distance and torsion angle constraints on singly and doubly 13 C-labeled Aβ-peptides aggregated in fibrils, obtained from specific novel solid-state NMR experiments, 13 C-MQ-NMR [63,64], CT-fp-RFDR [65,66], 2D-exchange-MAS [67–69], 2Q-CSA [69,70] (see Figure 1F); (ii) from about 180 13 C and 15 N chemical shifts and line widths of resonance lines obtained from 2D correlation MAS NMR experiments on four Aβ(1−40) fibril samples with a small number (five, six, or seven) of selectively chosen and uniformly 13 C/15 N-labeled amino acid residues [28,62]; and (iii) from mass-per-length data for fibrils obtained from STEM [28,49]. For molecular structure determination local structural features of molecules can be elucidated by measuring either interspin distances or torsion angles in selectively 13 C- or 13 C/15 N- labeled parts of the peptide sequence. α-helix, β-sheet secondary structures can be estimated from chemical shifts and chemical shift anisotropies. Interspin distances can be measured using both homonuclear and heteronuclear dipole–dipole recoupling sequences, such as constant-time finite-pulse RFDR (fpRFDR-CT) and REDOR, respectively, and using other analogous methods. Peptide torsion angles, φ and ψ can be measured, for example, by correlating 13 C chemical shift tensors of carbonyl carbons in 13 C2 -labeled peptides (two consecutive amino acid residues) by means of either spin-diffusion (2D-13 C MAS exchange in Figure 1F) or by excitation of double-quantum coherences (2Q-13 CCSA, see Figure 1F) or in combination with fp-RFDRCT, which sets constraint on the interspin distance, rCC ,

and, therefore, on the peptide angle φ. Structural domains of aggregated peptides and the geometry of spin clusters can be tested by “spin-counting” techniques, such as 13 C multiple-quantum NMR spectroscopy [63,71,72]. These methods are particularly useful for testing supramolecular organization of singly 13 C-labeled Aβ-peptides that form either parallel or antiparallel β-sheet secondary structures. For an in-register parallel β-sheet organization 13 C spins form an infinite “in-register” cluster of spins with in˚ that are coupled by the homonuterspin distance of ∼5 A clear dipole–dipole interaction of ca. 70 Hz (see pink labels in Figure 1F). In this particular situation a number of coherent spin states can be excited by a specific pulse sequence, and the highest order of the excited multiplequantum coherence would be roughly the “count” of the number of coupled spins in the cluster. The aforementioned solid-state NMR methods were used in obtaining structural constraints and developing structural models for Alzheimer’s amyloid fibrils. The most important results are given below: (i) Y10 EVHHQKLVFFAEDV24 and A30 IIGLMVGGVV40 segments of Aβ(1−40) (fibrils prepared at pH 7.4) form parallel in-register β-sheets [64,66]; (ii) V24 GSNKGA30 fragment of Aβ(1−40) forms a “loop” in fibrils since (φ, ψ) angles deviate considerably from those in β-sheet structures [62,69]. However, this is not a true β-hairpin structure, since hydrogen bonding between these amino acid residues is intermolecular rather than intramolecular. Therefore, a “loop” is a “fold” of the Aβ(1−40) laminate [62]. (iii) The D1 AEFRHDSG9 segment is predominantly in a random coil conformation as was concluded from both 13 C chemical shifts and line widths of resonance lines in two-dimensional 13 C-13 C and 15 N-13 C MAS

Fig. 1. (Continued ) a histogram which was fitted with a single Gaussian curve to give an average mass-per-length of 26.15 kDa/nm with an s.d. of 3.29 kDa/nm. Images and the histogram were obtained from RD Leapman. (F) Putative model for a single Aβ(1−40) folded laminate with selective amino acid isotope labeling scheme shown in color (each label corresponds to a separate sample prepared for solid state NMR measurements): singly 13 C labeled carbonyl (red F4, V12, L17, F20, V24, L34 and V39) or methyl (pink A2, A21 and A30) sites in amino acid residues for 13 C fp-RFDR-CT (measurements of rcc interspin intermolecular distances) or 13 C MQ NMR for elucidation of either a parallel or an antiparallel β-sheet supramolecular structures; 15 N labeled sites (blue) for frequency selective 13 C{15 N}-REDOR measurements of salt bridges (black link D23-K28) between the negatively charged side chain carboxylate carbon of Asp23 and the positively charged side chain amino nitrogen of Lys28 (uniformly 13 C and 15 N labeled amino acid residues); doubly 13 C labeled samples at carbonyl carbons of two consecutive amino acid residues (orange links D23-V24, V24-G25, G25-S26, K28G29 and G29-A30) for φ and ψ peptide angle measurements using a combination of methods, 2Q-13 C-CSA, 2D-13 C-MAS exchange and 13 C fp-RFDR-CT NMR. 13 C and 15 N chemical shifts, line widths and sequential assignment were extracted from 2D 13 C-13 C and 15 N-13 C MAS NMR correlation spectra on Aβ(1−40) fibril samples with a few (five, six or seven) 15 N, 13 C uniformly labeled amino acid residues scattered across the peptide sequence. Binding of Cu+2 ions or Al-citrate complexes to Aβ(1−40) fibrils was tested by either Electron Paramagnetic Resonance or by 27 Al MAS NMR (after incubation, samples were dialyzed in 1 kDa cut-off dialysis tubes to remove unbound ions or complexes). (See also Plate 1 on page XVII in the Color Plate Section.)

Polymorphism of Alzheimer’s Aβ Amyloid Fibril

It is important to note that the central, mostly hydrophobic, region of Alzheimer’s amyloid peptides is of particular importance for the formation and stability of amyloid fibrils [75]. Therefore, it can be appreciated that all point mutations found so far in Aβ(1−40) which are associated with an early-onset of dementia are either at amino acid residues next to the central hydrophobic region of the peptide, LVFFA or those replacing negatively charged Glu22 or Asp23 residues on neutral (E22Q, D23N), hydrophobic (E22G), or positively charged (E22K) residues. All these mutations would change the net charge of the peptide, changing its solubility at neutral pH, make the “folding” region of the peptide molecule more flexible (as in the “Arctic” mutation, E22G) or enlarge the central hydrophobic region of the peptide, which will additionally stabilize β-sheet secondary structure in amyloid fibrils.

Figure 1F also shows that another NMR active isotope, such as 27 Al (I = 5/2, 100% natural abundance) can be useful in studies of binding of various biologically relevant soluble aluminum complexes (for example, aluminum citrate species, which may pass the brainblood-barrier [76]) to Aβ-oligomers and fibrils. It is well known in biochemistry and medicine that aluminum ions are highly toxic. A link between aluminum and AD has been extensively discussed since beginning of the 1980s when high concentrations of aluminum were detected for the first time in Alzheimer’s neurofibrillary tangles and later also in amyloid plaquies [77,78]. However, exact mechanism, binding sites for aluminum ions and Al complexes on Aβ-peptides and other important features of Al–Aβ-interaction are still unknown. For example, different Al-citrate complexes at low concentrations can either accelerate or retard the aggregation kinetics of Aβ(1−40) and also stabilize certain polymorphs of Aβ fibrils [79]. Binding of Cu(II) ions to Aβ(1−28) , Aβ(1−40), and Aβ(1−42) molecules [80,81] has been studied by another magnetic resonance method, electron spin resonance (ESR). Due to its high sensitivity, ESR is a very useful method in studies of metal-ion binding to Aβ. The effects of metal ions on Aβ(1−40) aggregation are currently widely discussed after the observation of co-localization of high concentrations of Al(III), Zn(II), Cu(II), and Fe(III) at the center of the core of Alzheimer’s amyloid plaques [82]. These metal ions accelerate Aβ aggregation kinetics, may stabilize amyloid fibrils and also increase neurotoxic effects of Aβ peptides [83,84]. It has been also suggested by Bush and co-workers that Cu(II) and Zn(II) may induce Aβ to form allosterically ordered oligomers that can penetrate lipid membranes [80]: Cu(II) ion initially coordinates His6, His13, His14, and Tyr10 in one Aβ molecule (see Figure 1F) but subsequently can coordinate two peptide molecules stabilizing a dimer and facilitating further aggregation of Aβ. Coordination of Cu(II) with His13 and His14 in two neighboring Aβ(1−40) molecules would facilitate propagation and stabilization of amyloid fibrils with in-register parallel β-sheet arrangement as found in all known polymorphs of Aβ(1−40) fibrils by recent solid state NMR measurements discussed above. Thus, studies of complexation of metal ions with Aβ are important in the search for the causes of and potential treatments for AD. In order to answer the questions formulated in this article more efforts must be directed towards determining the structure of different polymorphs of Aβ-fibrils and oligomers, the effects of point mutations, metal ions, and metal complexes on the aggregation kinetics of Aβpeptides, the search for potential inhibitors [85] and finally neurotoxicity tests on nerve cells cultures. Solidstate NMR has been already proven as a powerful tool in structural studies on other amyloidogenic peptides [86,87].

Part I

correlation spectra of uniformly labeled amino acid residues scattered across the peptide sequence [62]. (iv) Two short fragments of Aβ, Ac-Aβ(16−22) -NH2 (Ac-K16 LVFFAE22 -NH2 ) and Aβ(11−25) also form amyloid fibrils at pH 7.4. However, Aβ(16−22) or Aβ(11−25) molecules are organized in in-register anti-parallel β-sheets which are stabilized by electrostatic interactions (for example, Lys16 and Glu22) as well as by hydrophobic interactions between side-groups of amino acid residues in the central region of the peptide, LVFFA [68,73]. (v) Peptide molecules in Aβ(1−42) -fibrils form parallel in-register β-sheets as concluded from 13 C fpRFDR-CT and 13 C{15 N}-REDOR measurements on a single sample of Aβ(1−42) 13 C-labeled in Ala21 (13 CH3 ) and Leu34(13 CO) positions and 15 N-labeled in Val40 [49]. (vi) Aβ(10−35) (NH2 -Y10 EVHHQKLVFFAEDVGSNKGAIIGLM35 -NH2 ) fibrils also form a parallel in-register β-sheet structure as has been earlier suggested by Meredith, Lynn, and Botto on the basis of 2Q-DRAWS solid-state NMR measurements [55–58]. Our fp-RFDR-CT, REDOR, and 13 C-MQ NMR data have confirmed this conclusion [49]. However, we suggest that the Aβ(10−35) laminates are folded between V24 and A30 amino acid residues (similar to Aβ(1−40) fibrils) [49] instead of an extended β-sheet structure originally proposed by Lynn, Meredith, and co-workers [74]. The folded structure does not contradict with 2Q-DRAWS and other solid-state NMR measurements since molecules build two parallel in-register β-sheets, while it also fits well with fibril dimensions estimated from TEM and to STEM mass-per-length measurements consistent with only two or four laminates in fibrils prepared at pH 3.8 and 7.4, respectively [49].

Polymorphism of Alzheimer’s Aβ Amyloid Fibril 23

24 Part I

Chemistry

Part I

Acknowledgments O.N.A. acknowledges financial support from the Foundation to the memory of J.C. and Seth M. Kempe, the Swedish Foundation of International Cooperation in Research and Higher Education (STINT), the Swedish Research Council and the Swedish Alzheimer’s Fund. Collaboration on these projects with R. Tycko, R.D. Leapman, J.J. Balbach, Y. Ishii, A. Petkova, N.W. Rizzo, N.A. Oyler, D.J. Gordon, S.C. Meredith, J. Reed, F. Dyda, F. Delaglio in the U.S.A., with M. Lindberg, N. Almqvist, M. Hellberg, N. Norlin, P. Eriksson, G. Gr¨obner, A. Filippov, A. Lund in Sweden, I. T´oth in Hungary, and R. Dupree, M. Smith, and A. Kukol in the U.K. are greatly acknowledged.

References 1. Alzheimer A. On a distinctive disease of the cerebral cortex. Z. Psychiatry 1907;64:146–148. 2. http://www.alzheimerforeningen.nu. 3. Selkoe DJ. Amyloid protein and Alzheimer’s disease. Sci. Am. 1991;November:68–78. 4. Glenner GG, Wong CW. Alzheimer’s disease: initial report of the purification and characterization of a novel cerebrovascular amyloid protein. Biochem. Biophys. Res. Commun. 1984;120:885–889. 5. Glenner GG, Wong CW. Alzheimer’s disease and Down’s syndrome: sharing of a unique cerebrovascular amyloid fibril protein. Biochem. Biophys. Res. Commun. 1984;122:1131– 1135. 6. Masters CL, Multhaup G, Simms G, Pottgiesser J, Martins RN, Beyreuther K. Neuronal origin of a cerebral amyloid: neurofibrillary tangles of Alzheimer’s disease contain the same protein as the amyloid of plaque cores and blood vessels. EMBO J. 1985;4:2757–2763. 7. Kang J, Lemaire H-G, Unterbeck A, Salbaum JM, Masters CL, Grzeschik K-H, Multhaup G, Beyreuther K, M¨uller-Hill B. The precursor of Alzheimer’s disease amyloid A4 protein resembles a cell-surface receptor. Nature (London) 1987;325:733–736. 8. Shivers BD, Hilbich C, Multhaup G, Salbaum M, Beyreuther K, Seeburg PH. Alzheimer’s disease amiloidogenic glycoprotein: expression pattern in rat brain suggests a role in cell contact. EMBO J. 1988;7(5):1365–1370. 9. Marx J. Alzheimer’s debate boils over. Science 1992; 257: 1336–1338. 10. Dickson DW, Yen SH. Beta-amyloid deposition and paired filament formation: Which histopathological feature is more significant in Alzheimer’s disease? Neurobiol. Aging 1989;10:402–404. 11. Terry RD, Masliah E, Salmon DP, Butters N, deTeresa R, Hill R, Katzman R. Physical basis of cognitive alterations in Alzheimer’s disease: synapse loss is the major correlate of cognitive impairment. Ann. Neurol. 1991;30:572–580. 12. Selkoe DJ. Deciphering Alzheimer’s disease: molecular genetics and cell biology yield major clues. J. NIH Res. 1995;April 7:57–64.

13. Kayed R, Head E, Thompson JL, McIntire TM, Milton SC, Cotman CW, Glabe ChG. Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis. Science 2003;300:486–489. 14. Walsh DM, Klyubin I, Fadeeva JV, Cullen WK, Anwyl R, Wolfe MS, Rowan MJ, Selkoe DJ. Naturally secreted oligomers of amyloid β protein potently inhibit hippocampal long-term potentiation in vivo. Nature 2002;416:535–539. 15. Hendriks L, van Duijn CM, Cras P, Cruts M, Hul WV, van Harskamp F, Warren A, McInnis MG, Antonarakis SE, Martin J-J, Hofman A, Broeckhoven CV. Presenile dementia and cerebral haemorrhage linked to a mutation at codon 692 of the β-amyloid precursor protein gene. Nat. Genet. 1992;1:218–221. 16. Levy E, Carman MD, Fernandez-Madrid IJ, Power MD, Lieberburg I, van Duinen SG, Bots GthAM, Luyendijk W, Frangione B. Mutation of the Alzheimer’s disease amyloid gene in hereditary cerebral hemorrhage, Dutch type. Science 1990;248:1124–1126. 17. Tagliavini F, Rossi G, Padovani A, Magoni M, Andora G, Sgarzi M, Bizzi A, Savoiardo M, Carella F, Morbin M, Giaccone G, Bugiani O. A new βPP mutation related to hereditary cerebral haemorrhage. Alz. Report 1999;2:S28. 18. Nilsberth C, Westlind-Danielsson A, Eckman ChB, Condron MM, Axelman K, Forsell Ch, Stenh Ch, Luthman J, Teplow DB, Yonkin SG, N¨aslund J, Lannfelt L. The ‘Arctic’ APP mutation (E693G) causes Alzheimer’s disease by enhanced Aβ protofibril formation. Nature Neurosc. 2001;4(9):887– 893. 19. Grabowski TJ, Cho HS, Vonsattel JPG, Rebeck GW, Greenberg SM. Novel amyloid precursor protein mutation in an Iowa family with dementia and severe cerebral amyloid angiopathy. Ann. Neurol. 2001;49:697–705. 20. Kirkitadze MD, Condron MM, Teplow DB. Identification and characterization of key kinetic intermediates in amyloid βprotein fibrillogenesis. J. Mol. Biol. 2001;312:1103–1119. 21. Antzutkin ON. Amyloidosis of Alzheimer’s Aβ-peptides: solid-state nuclear magnetic resonance, electron paramagnetic resonance, transmission electron microscopy, scanning transmission electron microscopy and atomic force microscopy studies. Magn. Reson. Chem. 2004;42:231– 246. 22. Hellberg M, Norlin N, Antzutkin ON, Almqvist N. Morphology and Growth Kinetics of Alzheimer’s Amyloid β-peptides, Aβ(1−40) and the Arctic Mutation Aβ(1−40) (E22G): In Situ Atomic Force Microscopy Study. In: G Grateau, RA Kyle, M Skinner (Eds). Amyloid and Amyloidosis. CRC Press, Boca Raton, Florida, 2005, pp 408–410. 23. Sisodia SS, Koo EH, Beyreuther K, Unterbeck A, Price DL. Evidence that β-amyloid protein in Alzheimer’s disease is not derived by normal processing. Science 1990;248:492– 495. 24. Buxbaum JD, Gandy SE, Ciccetti P, Ehrlich ME, Czernik AJ, Fracasso RP, Ramabhadran TV, Unterbeck AJ, Greengard P. Processing of Alzheimer β/A4 amyloid precursor protein: modulation by agents that regulate protein phosphorylation. Proc. Natl. Acad. Sci. U.S.A. 1990;87:6003–6006. 25. Pike ChJ, Burdick D, Walencewicz AJ, Glabe GCh, Cotman CW. Neurodegeneration induced by β-amyloid peptides in vitro: the role of peptide assembly state. J. Neurosc. 1993;13(4):1676–1687.

Polymorphism of Alzheimer’s Aβ Amyloid Fibril

41. Barrow CJ, Yasuda A, Kenny PTM, Zagorski MG. Solution conformations and aggregational properties of synthetic amyloid β-peptides of Alzheimer’s disease. Analysis of circular dichroism spectra. J. Mol. Biol. 1992;225:1075– 1093. 42. Burdick D, Soreghan B, Kwon M, Kosmoski J, Knauer M, Henschen A, Yates J, Cotman C, Glabe C. Assembly and aggregation properties of synthetic Alzheimer’s A4/β amyloid peptide analogs. J. Biol. Chem. 1992;267:546– 554. 43. Malinchik SB, Inouye H, Szumowski KE, Kirschner DA. Structural analysis of Alzheimer’s β(1–40) amyloid: protofilament assembly of tubular fibrils. Biophys. J. 1998; 74:537–545. 44. Serpell LC, Blake CCF, Fraser PE. Molecular structure of a fibrillar Alzheimer’s Aβ fragment. Biochemistry 2000;39: 13269–13275. 45. Serpell LC, Smith JM. Direct visualization of the β-sheet structure of synthetic Alzheimer’s amyloid. J. Mol. Biol. 2000;299:225–231. 46. Lansbury PT, Costa PR, Griffiths JM, Simon EJ, Auger M, Halverson KJ, Kocisko DA, Hendsch ZS, Ashburn TT, Spencer RGS, Tidor B, Griffin RG. Structural model for the β-amyloid fibril based on interstrand alignment of an antiparallelsheet comprising a C-terminal peptide. Nat. Struct. Biol. 1995;2:990–998. 47. Kirschner DA, Abraham CR, Selkoe DJ. X-ray diffraction from intraneuronal paired helical filaments and extraneuronal amyloid fibers in Alzheimer disease indicated cross-β conformation. Proc. Nat. Acad. Sci. U.S.A. 1987;83:503– 507. 48. Roher AE, Lowenson JD, Clarke S, Woods AS, Cotter RJ, Gowing E, Ball MJ. β-Amyloid-(1–42) is a major component of cerebrovascular amyloid deposits: implications for the pathology of Alzheimer disease. Proc. Nat. Acad. Sci. U.S.A. 1993;90:10836–10840. 49. Antzutkin ON, Leapman RD, Rizzo NW, Balbach JJ, Tycko R. Supramolecular structural constraints on Alzheimers βamyloid fibrils from electron microscopy and solid state nuclear magnetic resonance. Biochemistry 2002;41:15436– 15450. 50. Goldsbury CS, Wirtz S, M¨uller SA, Sunderji S, Wicki P, Aebi U, Frey P. Studies on the in vitro assembly of Aβ(1−40) : Implications for the search for Aβ fibril formation inhibitors. J. Struc. Biol. 2000;130:217–231. 51. Engel A, Baumeister W, Saxton WO. Mass mapping of a protein complex with the scanning transmission electron microscope. Proc. Natl. Acad. Sci. U.S.A. 1982;79:4050– 4054. 52. Wall JS, Hainfeld JF. Mass mapping with the scanning transmission electron microscope. Ann. Rev. Biophys. Biophys. Chem. 1986;15:355–376. 53. M¨uller SA, Goldie KN, Burki R, Haring R, Engel A. Factors influencing the precision of quantitative scanning transmission electron microscopy. Ultramicroscopy 1992;46:317– 334. 54. Thomas D, Schultz P, Steven AC, Wall JS. Mass analysis of biological macromolecular complexes by STEM. Biol. Cell 1994;80:181–192. 55. Benzinger TLS, Gregory DM, Burkoth TS, Miller-Auer H, Lynn DG, Botto RE, Meredith SC. Propagating structure

Part I

26. Lorenzo A, Yankner B. β-amyloid neurotoxicity requires fibril formation and is inhibited by Congo red. Proc. Natl. Acad. Sci. U.S.A. 1994; 91: 12243–12247. 27. Howlett DR, Jennings KH, Lee DC, Clark MS, Brown F, Wetzel R, Wood SJ, Camilleri P, Roberts GW. Aggregation state and neurotoxic properties of Alzheimer beta-amyloid peptide. Neurodegeneration 1995;4:23–32. 28. Petkova AT, Leapman RD, Guo Z, Yau W-M, Mattson MP, Tycko R. Self-propagating, molecular-level polymorphism in Alzheimer’s β-amyloid fibrils. Science 2005;307:262– 265. 29. Seilheimer B, Bohrmann B, Bondolfi L, M¨uller F, St¨uber D, D¨obeli H. The toxicity of the Alzheimer’s β-amyloid peptide correlates with a distinct fiber morphology. J. Struct. Biol. 1997;119:59–71. 30. Castano ˜ EM, Ghiso J, Prelli F, Gorevic PD, Migheli A, Frangione B. In vitro formation of amyloid fibrils from two synthetic peptides of different lengths homologous to Alzheimer’s disease β-protein. Biochem. Biophys. Res. Commun. 1986;141:782–789. 31. Gorevic PC, Castano ˜ EM, Sarma R, Frangione B. Ten to fourteen residue peptides of Alzheimer’s disease protein are sufficient for amyloid fibril formation and its characteristic X-ray diffraction pattern. Biochem. Biophys. Res. Commun. 1987;147:854–862. 32. Kirschner DA, Inouye H, Duffy LK, Sinclair A, Lind M, Selkoe DJ. Synthetic peptide homologous to β protein from Alzheimer disease forms amyloid-like fibers in vitro. Proc. Nat. Acad. Sci. U.S.A. 1987;84:6953– 6957. 33. Halverson K, Fraser P, Kirschner DA, Lansbury PT Jr. Molecular determinants of amyloid deposition in Alzheimer’s disease: conformational studies of synthetic β-protein fragments. Biochemistry 1990;29:2639–2644. 34. Fraser PE, Duffy LK, O’Malley MB, Nguyen JT, Inouye H, Kirschner DA. Morphology and antibody recognition of synthetic beta-amyloid peptides. J. Neurosci. Res. 1991;28:474–485. 35. Fraser PE, Nguyen JT, Surewicz WK, Kirschner DA. pHdependent structural transitions of Alzheimer amyloid β/A4 peptides. Biophys. J. 1991;60:1190–1201. 36. Fraser PE, McLachlan DR, Surewicz WK, Mizzen CA, Snow AD, Nguyen JT, Kirschner DA. Conformation and fibrillogenesis of Alzheimer Aβ peptides with selected substitution of charged residues. J. Mol. Biol. 1994;244:64–73. 37. Hilbich C, Kisters-Woike B, Reed J, Masters CL, Beyreuthers K. Aggregation and secondary structure of synthetic amyloid βA4 peptides of Alzheimer’s disease. J. Mol. Biol. 1991;218:149–163. 38. Hilbich C, Kisters-Woike B, Reed J, Masters CL, Beyreuthers K. Human and rodent sequence analogs of Alzheimer’s amyloid βA4 share similar properties and can be solubilized in buffers of pH 7.4. Eur. J. Biochem. 1991;201:61– 69. 39. Hilbich C, Kisters-Woike B, Reed J, Masters CL, Beyreuthers K. Substitutions of hydrophobic amino acids reduce the amyloidogenicity of Alzheimer’s disease βA4 peptides. J. Mol. Biol. 1992;228:460–473. 40. Barrow CJ, Zagorski MG. Solution structures of β peptide and its constituent fragments: relation to amyloid deposition. Science 1991;253:179–182.

References 25

26 Part I

Chemistry

Part I 56.

57.

58.

59.

60. 61. 62.

63. 64.

65.

66.

67.

68.

69.

of Alzheimer’s β-amyloid(10−35) is parallel β-sheet with residues in exact register. Proc. Natl. Acad. Sci. U.S.A. 1998;95:13407–13412. Gregory DM, Benzinger TLS, Burkoth TS, Miller-Auer H, Lynn DG, Meredith SC, Botto RE. Dipolar recoupling NMR of biomolecular self-assemblies: determining inter- and intrastrand distances in fibrilized Alzheimer’s β-amyloid peptide. Solid State Nucl. Magn. Reson. 1998;13:149–166. Burkoth TS, Benzinger TLS, Urban V, Morgan DM, Gregory DM, Thiyagarajan P, Botto RE, Meredith SC, Lynn DG. Structure of the β-amyloid(10–35) fibril. J. Am. Chem. Soc. 2000;122:7883–7889. Benzinger TLS, Gregory DM, Burkoth TS, Miller-Auer H, Lynn DG, Botto RE, Meredith SC. Two-dimensional structure of β-amyloid(10–35) fibrils. Biochemistry 2000;39:3491– 3499. Antzutkin ON. Molecular structure determination: Applications in biology. In: MJ Duer (Ed). Solid State Nuclear Magnetic Resonance Spectroscopy, Principles and Application, Blackwell Science: Oxford, 2002, pp 280–390. Tycko R. Insight into amyloid folding problem from solidstate NMR. Biochemistry 2003;42(11):3151–3159. Tycko R. Application of Solid State NMR to the Structural Characterization of Amyloid Fibrils: Methods and Results. Prog. Nucl. Magn. Reson. Spectr. 2003; 42:53–68. Petkova AT, Ishii Y, Balbach JJ, Antzutkin ON, Leapman RD, Delaglio F, Tycko R. A structural model for Alzheimer’s β-amyloid fibrils based on experimental constraints from solid state NMR. Proc. Nat. Acad. Sci. U.S.A. 2002;99(2):16742–16747. Antzutkin ON, Tycko R. High-order multiple quantum excitation in 13 C nuclear magnetic resonance spectroscopy of organic solids. J. Chem. Phys. 1999;110(6):2749–2752. Antzutkin ON, Balbach JJ, Leapman RD, Rizzo NW, Reed J, Tycko R. Multiple quantum solid state NMR indicates a parallel, not antiparallel, organization of β-sheets in Alzheimer’s β-amyloid fibrils. Proc. Nat. Acad. Sci. U.S.A. 2000;97(24):13045–13050. Ishii Y, Balbach JJ, Tycko R. Measurement of dipolecoupled lineshapes in a many-spin system by constanttime two-dimensional solid state NMR with highspeed magic-angle-spinning. Chem. Phys. 2001;266:231– 236. Balbach JJ, Petkova AT, Oyler NA, Antzutkin ON, Gordon DJ, Meredith SC, Tycko R. Supramolecular structure in fulllength of Alzheimer’s β-amyloid fibrils: Evidence for a parallel β-sheet organization from solid state Nuclear Magnetic Resonance. Biophys. J. 2002;83(2):1205–1216. Weliky DP, Tycko R. Determination of peptide conformations by two-dimensional magic angle spinning NMR exchange spectroscopy with rotor synchronization. J. Am. Chem. Soc. 1996;118:8487–8488. Balbach JJ, Ishii Y, Antzutkin ON, Leapman RD, Rizzo NW, Dyda F, Reed J, Tycko R. Amyloid fibril formation by Aβ(16−22) , a seven-residue fragment of the Alzheimer’s βamyloid peptide, and structural characterization by solid state NMR. Biochemistry 2000;39(45):13748–13759. Antzutkin ON, Balbach JJ, Tycko R. Site-specific identification of non-β-strand conformations in Alzheimer’s β-amyloid fibrils by solid state NMR. Biophys. J. 2003;84(5):3326–3335.

70. Blanco FJ, Tycko R. Determination of peptide backbone dihedral angles in solid-state NMR by double quantum 13 C chemical shift anisotropy measurements. J. Magn. Reson. 2001;149:131–138. 71. Tycko R. Selection rules for multiple quantum NMR excitation in solids: derivation from time-reversal symmetry and comparisons with simulations and 13 C NMR experiments. J. Magn. Reson. 1999;139:302–307. 72. Oyler NA, Tycko R. Multiple quantum 13 C NMR spectroscopy in solids under high-speed magic-angle-spinning. J. Phys. Chem. B 2002;106:8382–8389. 73. Petkova AT, Buntkowsky G, Dyda F, Leapman RD, Yau WM, Tycko R. Solid state NMR reveals a pH-dependent antiparallel β-sheet registry in fibrils formed by a β-amyloid peptide. J. Mol. Biol. 2004;335:247–260. 74. Lynn DG, Meredith SC. Review: model peptides and the physicochemical approach to β-amyloids. J. Struc. Biol. 2000;130:153–173. 75. Tjernberg LO, N¨aslund J, Lindqvist F, Johansson J, Karlstr¨om AR, Thyberg J, Terenius L, Nordstedt C. Arrest of β-amyloid fibril formation by a pentapeptide ligand. J. Biol. Chem. 1996;271(15):8545–8548. 76. Slanina P, Falkeborn Y, Frech W, Cedergren A. Aluminium concentrations in the brain and bone of rats fed citric acid, aluminium citrate or aluminium hydroxide. Fd. Chem. Toxic. 1984;22:391–397. 77. Exley Ch. The aluminium-amyloid cascade hypothesis and alzheimer’s disease. In R Harris, F Fahrenholz (Eds). Alzheimer’s Disease: Cellular and Molecular Aspects of Amyloid beta, Series: Subcellular Biochemistry, Vol. 38, 225–234. 78. Exley Ch, Korchazhkina O. The association of aluminium and β amyloid in Alzheimer’s disease. In: Ch Exley (Ed). Aluminium and Alzheimer’s disease. The science that describes the link. Elsevier Science B. V., Amsterdam, The Netherlands, 2001, pp 421–433. 79. Antzutkin ON, Norlin N, Hellberg M, Eriksson P, Almqvist N, Leapman RD, Tycko R, Petkova AT, T´oth I, Howes AP, Dupree R. Binding of aluminium(III)-citrate complexes, [Al3 (H–1 Cit)3 (OH)]–4 and [Al3 (H–1 Cit)3 (OH)4 ]–7 , to alzheimer’s Aβ peptides: In situ atomic force, electron microscopy and solid state 13 C and 27 Al MAS NMR studies. Sixth Keele Meeting “Aluminium, Lithosphere to Biosphere (and Back)”, Buc¸aco, Portugal, February 26 to March 2, 2005. 80. Curtain CC, Ali FE, Volitakis I, Cherny RA, Norton RS, Beyreuther K, Barrow CJ, Masters CL, Bush AI, Barnham KJ. Alzheimer’s disease amyloid-β binds copper and zinc to generate an allosterically ordered membranepenetrating structure containing superoxide dismutaselike subunits. J. Biol. Chem. 2001;276(23):20466– 20473. 81. Curtain CC, Ali FE, Smith DG, Bush AI, Masters CL, Barnham KJ. Metal ions, pH, and cholesterol regulate the interactions of Alzheimer’s disease amyloid-β peptide with membrane lipid. J. Biol. Chem. 2003;278(5):2977– 2982. 82. Lovell MA, Robertson JD, Teesdale WJ, Campbell JL, Markesbery WR. Copper, iron and zinc in Alzheimer’s disease senile plaques. J. Neurol. Sci. 1998;158:47– 52.

Polymorphism of Alzheimer’s Aβ Amyloid Fibril

tides containing ester bonds at alternate positions. Biochemistry 2003;42:475–485. 86. Naito A, Kamihira M, Inoue R, Saitˆo H. Structural diversity of amyloid fibril formed in human calcitonin as revealed by site-directed 13 C solid-state NMR spectroscopy. Magn. Reson. Chem. 2004;42:247–257. 87. Jaroniec CP, MacPhee CE, Bajaj VS, McMahon MT, Dobson CM, Griffin RG. High-resolution molecular structure of a peptide in an amyloid fibril determined by magic angle spinning NMR spectroscopy. Proc. Natl. Acad. Sci. U.S.A. 2004;101:711.

Part I

83. Mantyh PW, Ghilardi JR, Rogers S, DeMaster E, Allen CJ, Stimson ER, Maggio JE. Aluminium, iron, and zinc ions promote aggregation of physiological concentrations of beta-amyloid peptide. J. Neurochem. 1993;61:1171–1174. 84. Atwood CS, Moir RD, Huang X, Scarpa RC, Bacarra NME, Romano DM, Hartshorn MA, Tanzi RE, Bush AI. Dramatic aggregation of Alzheimer Aβ by Cu(II) is induced by conditions representing physiological acidosis. J. Biol. Chem. 1998;273:12817–112826. 85. Gordon DJ, Meredith SC. Probing the role of backbone hydrogen bonding in β-amyloid fibrils with inhibitor pep-

References 27

Part I

Chemical Shifts and Spin-Couplings

31

and 17O NMR Chemical Shift NMR for Hydrogen Bonds Shigeki Kuroki

Department of Chemistry and Materials Science, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo 152-8552, Japan

Introduction Hydrogen bonding plays an important role in forming higher-order structures of peptides, polypeptides and proteins. Accordingly, the nature of the hydrogen bond has been widely studied by various spectroscopic methods. High-resolution NMR spectroscopy has been used as one of the most powerful methods for obtaining useful information on the details of the hydrogen-bonded structure. NMR chemical shifts are one of the most important parameters for providing information about molecular structure. Since the electronic structure around the carbonylcarbon and oxygen, amide-nitrogen and hydrogen atoms in peptides and polypeptides is greatly affected by the nature of the hydrogen bond, the NMR chemical shifts for these atoms are sensitive to the spatial arrangement of the nuclei comprising the hydrogen bond. Also the electritic gradient (eq), which determined by the NMR spectrum of quadrupdan nucleus, is greatly affected by the electronic structure around the hydrogen bond. Here, recent studies on the hydrogen-bonded structures of peptides and polypeptides in the solid state are presented through the observation of 13 C, 15 N, 1 H, 2 H, and 17 O NMR chemical shifts and theoretical calculations of nuclear shielding with a view to find a deeper understanding of the nature and inference of hydrogen bonds.

Hydrogen-bonded Structure and 13 C Chemical Shift [1−4] Hydrogen-Bond Length and the 13 C Isotropic Chemical Shift (δδ iso ) of the Carbonyl-Carbon in Several Amino Acids At first, the relationship between the isotropic 13 C chemical shifts of carbonyl-carbons and the hydrogen-bonded structure is discussed. Figure 1 shows the plot of the observed isotropic 13 C chemical shifts (δiso ) for the carbonylcarbon in Gly, L-Ala, L-Val, D,L-Leu, and L-Asp residues in peptides against the N · · · O hydrogen-bond length(RN ... O ). It is found that a decrease in RN...O leads to a higher frequency shift, and there exists approximately Graham A. Webb (ed.), Modern Magnetic Resonance, 31–35.
C 2008 Springer.

a linear relationship between the 13 C chemical shift and RN...O . It is noted that not only in oligopeptides (dimer or trimer) but also in polypeptides, the carbonyl-carbon chemical shifts give a similar hydrogen-bond length dependence. This suggests that the 13 C chemical shifts of the carbonyl-carbon taking the hydrogen-bond, which is formed between the amide >C=O and amide >N–H, are predominantly determined by the hydrogen-bond length. The slope of this linear relationship is quite characteristic of individual amino acid residues.

Hydrogen-Bond Length and the Principal Values(δ11 , δ22 and δ33 ) of the Carbonyl-Carbon It is expected that the principal values of the 13 C chemical shift tensors (δ11 , δ22 and δ33 , from higher to lower frequency) are, in principle, more sensitive as parameters for obtaining detailed information on hydrogen-bonding to be related with electronic structure, than will the isotropic 13 C chemical shift ( δiso = (δ11 + δ22 + δ33 ) /3 ). It is well known that δ11 is in the amide sp2 plane and lies along the direction normal to the C=O bond, the δ22 component lies almost along the amide C=O bond, and the δ33 component is aligned perpendicular to the amide sp2 plane. From the plots of the observed principal values of the 13 C chemical shift tensors such as δ11 , δ22 and δ33 for Gly, L-Ala, L-Val, D, L-Leu, and L-Asp residues in peptides against the N · · · O hydrogen-bond length (RN...O ), it is found that the experimental δ22 values are the most sensitive to RN...O , and the δ22 values move linearly to high frequency with a decrease of RN...O . The δ22 component lies almost along the amide C=O bond, so the δ22 values are the most sensitive to a changing in RN...O . The slope and intercept of the variation of the plot ofδ22 against RN...O varies depending on the amino acid residues. The δ11 and δ33 values are insensitive to a change in RN...O , but it seems that the δ11 and δ33 values move slightly to low and to high frequencies with a decrease in RN...O , respectively. Therefore, it can be said that the large high-frequency shift in δiso , with a decrease in RN...O , is predominantly governed by a decrease in δ22 . To understand the relationship between the 13 C chemical shifts and hydrogenbond length, some theoretical MO calculations on the

Part I

13 C, 15 N, 1 H, 2 H,

32 Part I

Chemistry

Part I Fig. 1. Plots of the observed isotropic 13 C chemical shifts (δiso ) for the carbonyl-carbon in Gly, L-Ala, L-Val, D, L-Leu, and LAsp residues in peptides against the N cdots O hydrogen-bond length(RN...O ).

13

C shielding tensors of model peptides have been carried out. From the calculations, it is found that δ22 is the most sensitive to a change of RN···O and moves linearly to high-frequency with a decrease in RN···O . Correspondingly, δ11 increases with a decrease in RN...O , whereas δ33 is insensitive to changes in RN...O . The results of the theoretical calculations agree well with the experimental results. Such an agreement indicates that the 13 C chemical shift changes originate predominately from the change of the electronic state of the amino carbonyl groups caused by the hydrogen-bond length variation. Further, it can be said that the amino acid residue dependence of the calculated tensor components is similar to the experimental one.

Hydrogen-bonded Structure and 15 N NMR Chemical Shift High-resolution 15 N NMR spectroscopy has been increasingly applied to the investigation of peptides, polypeptides and proteins in the solid state[5,6] . It is expected that a similar hydrogen-bond length dependence of the 13 C-carbonyl-carbon chemical shifts will aply to the amide nitrogen 15 N chemical shifts. But, there is no clear relationship between the observed 15 N chemical shifts of the Gly NH of peptides against the N· · ·O hydrogen-bond length (RN...O ).

Fig. 2. Plot of the observed 15 N chemical shifts of the glycine residue in X-Gly-Gly against the N–H bond length (RN–H ) associated with a hydrogen bond.

The plot of the observed 15 N chemical shifts of the glycine residue in X-Gly-Gly against the N–H bond length (RN–H ) associated with a hydrogen bond is shown as in Figure 2. It is found that there is a clear relationship between these parameters and the decrease of RN–H leads to a linear increase in shielding. Amide 15 N chemical shifts are closely related to the length of the N–H bond but are not related to RN···O distance. This implies that the 15 N chemical shift value gives useful information about the length of N–H bonds. It seems that the hydrogen bond angle ( N–H · · · O) is also related to the 15 N chemical shift. Theoretical calculations of 15 N chemical shift shows that a decrease of RN−H leads to an increase of the calculated 15 N isotropic shielding which agrees with the experimental results. Therefore, such a relationship suggests that the isotropic 15 N chemical shift value can be used in the estimation of RN−H . Combined with the carbonyl 13 C chemical shifts we can get very useful information about the hydrogen-bonded structure.

Hydrogen-bonded Structure and 1 H NMR Chemical Shift The chemical shift of a 1 H nucleus has been widely applied to many works on the hydrogen bonding studies of peptides and proteins in the solution state.[7,8] However, in the solution state, the 1 H chemical shifts of peptides

13 C, 15 N, 1 H, 2 H,

and 17 O NMR Chemical Shift NMR for Hydrogen Bonds

Hydrogen-bonded Structure and 17 O NMR Quadrupolar Coupling Constant and Chemical Shift[10−13] The oxygen atom is also one of the most important one forming hydrogen-bonded structures in peptides and polypeptides. Nevertheless, solid-state 17 O NMR studies of peptides and polypeptides have not been carried out due to the very weak sensitivity of the solid-state 17 O NMR measurements which arises from the fact that the 17 O nucleus has a very low natural abundance of 0.037 %, and that the 17 O nuclear spin quantum number (I) is 5/2, thus

Part I

with rotational isomers are often the averaged values for all isomers because of rapid inversion by rotation about bonds and further are strongly influenced by solvent, pH, etc. Therefore, it is not easy to separate only the hydrogenbonding effect on the 1 H chemical shift. In the solid state, chemical shifts provide information on fixed conformations and of hydrogen-bonded structures. But, in the solid state there are few studies on the high-resolution 1 H NMR of amide protons in peptides and polypeptides. One of the main reasons is the very large dipolar interaction of 1 H nuclei, which leads to a large broadening of the spectral line. The problem is how to eliminate these dipolar interactions. The most popular method is combined rotational and multi-pulse spectroscopy (CRAMPS) with magic angle spinning (MAS).[9] One of the typical homo-nuclear dipolar decoupling sequence to be used in the CRAMPS experiments is br24. Although the conventional CRAMPS experiments have been made with a relatively low spinning rate of ∼3 kHz, this MAS rate is not always enough for the removal of dipolar couplings between protons and other nuclei. This is true for the amide proton of peptides and polypeptides bonded directly to the 14 N nucleus considered under here. It is possible to eliminate this dipolar interaction between the amide 1 H bonded directly to the 14 N nucleus by high speed MAS such as 30 kHz and the observation at a high frequency such as 800 MHz. This procedure permits one to obtain, very high-resolution, 1 H spectra of peptides and polypeptides. Figure 3 shows the observed amide proton chemical shifts plotted against the hydrogen-bond length between the amide nitrogen and oxygen atoms (RN...O ). It is shown that the 1 H chemical shift values move to high frequency with a decrease in RN...O . This means that the observation of the amide proton 1 H chemical shift value leads to the determination of RN...O . From neutron diffraction and ab initio MO studies it is shown that the reduction of RN...O leads to a decrease in the hydrogen-bond length (RH...O ) between the amide proton and the carbonyl oxygen. Thus, it can be said that the 1 H chemical shift values move to high frequency with a decrease in RH...O .

Hydrogen-bonded Structure 33

Fig. 3. Plot of the observed amide 1 H chemical shifts against the N · · · O hydrogen-bond length(RN...O ).

the nucleus is quadrupolar, and the 17 O signal is broadened by nuclear quadrupolar effects in the solid. Solid-state 17 O NMR spectra of Poly(Gly) form I (PG I),Poly(Gly) form II (PG II), glycylglycine (GlyGly) and glycylglycine nitrate(GlyGly·HNO3 ) are discussed with a view to understand the relationship between the hydrogen-bonded structure and the 17 O NMR parameters. Figure 4(a) shows a plot of the observed quadrupolar coupling constant (e2 q Q/ h) values against the hydrogen bond length (RN...O ). The e2 q Q/ h values decrease linearly with a decrease of RN...O between the amide nitrogen and oxygen atoms. This change comes from a change of the q values which are the largest component of the electric gradient tensor. This experimental result shows that the decrease in the hydrogen bond length leads to a decrease in the electric field gradient. The q value seems to be very sensitive to the hydrogen-bonding length change. The results of theoretical MO calculations agree well with these experimental results. Figure 4(b) shows the plot of the observed isotropic 17 O chemical shift (δiso ) values against the hydrogen bond length (RN...O ). The isotropic 17 O chemical shift (δiso ) values in both peptides and polypeptides move to low frequencies with a decrease in the hydrogen bond length (RN...O ). The difference of the chemical shifts between peptides and polypeptides comes from the geometrical location of the amide group and the carbonyl group which forms hydrogen-bonding. From the plots of the observed principal values (δ11 , δ22 and δ33 , from higher to lower

34 Part I

Chemistry

Part I

frequency) of the 17 O chemical shifts against the hydrogen bond length (RN...O ), every principal value in both the peptides and polypeptides moves to low-frequency with a decrease in the hydrogen bond length (RN...O ). The hydrogen bond length dependence of the calculated isotropic chemical shielding (δiso ) of the Gly carbonyl oxygen in the model molecule system shows that the 17 O chemical shift moves largely to low-frequency with an increase in RN...O . This explains qualitatively the experimental trend as mentioned above.

Hydrogen-bonded Structure and 2 H Quadrupolar Coupling Constant[13] The 2 H nucleus of an amide group in which 1 H is substituted by 2 H is one of the most important nuclei involved in a hydrogen-bonded structure in peptides and polypeptides. In Figure 5, the plots of the observed e2 q Q/ h values for 2 H against the hydrogen bond length (RN...O ) are shown. The e2 q Q/ h value decreases with a decrease in RN...O . The experimental result shows that the reduction of the hydrogen-bond length leads to a linear decrease in electritic field gradient (eq). The eq value is very sensitive to change in the hydrogen bond length. This experimental finding is consistent with the experimental results

Fig. 4. (a) Plot of the observed 17 O quadrupolar coupling constant (e2 q Q/ h) values of Poly(Gly) form I (PG I),Poly(Gly) form II (PG II), glycylglycine (GlyGly) and glycylglycine nitrate(GlyGly·HNO3 ) against the hydrogen bond length(RN...O ). (b) Plot of the observed carbonyl 17 O chemical shifts of Poly(Gly) form I (PG I),Poly(Gly) form II (PG II), glycylglycine (GlyGly) and glycylglycine nitrate (GlyGly·HNO3 ) against the hydrogen bond length(RN...O ).

Fig. 5. Plots of the observed 2 H quadrupolar coupling constant e2 q Q/ h values against the hydrogen bond length (RN...O ).

13 C, 15 N, 1 H, 2 H,

Conclusion As discussed, it is concluded that solid state 13 C, 15 N, 1 H, 2 H, and 17 O NMR spectroscopy combined with theoretical MO calculations is a very useful methodology for elucidating the hydrogen-bonded structures of peptides and polypeptides in the solid state.

References 1. Ando S, Ando I, Shoji A, Ozaki T, J. Am. Chem. Soc. 1988; 110:3380. 2. Asakawa N, Kuroki S, Kurosu H, Ando I, Shoji A, Ozaki T, J. Am. Chem. Soc. 1992;114:3261.

References 35

3. Tsuchiya K, Takahashi A, Takeda N, Asakawa N, Kuroki S, Ando I, Shoji A, Ozaki T, J. Mol. Struct. 1995;350:233. 4. Kameda T, Takeda N, Kuroki S, Kurosu H, Ando I, Shoji A, Ozaki T, J. Mol. Struct. 1995;384:178. 5. Kuroki S, Ando S, Ando I, Shoji A, Ozaki T, Webb GA, J. Mol. Struct. 1990; 240: 19. 6. Kuroki S, Asakawa N, Ando S, Ando I, Shoji A, Ozaki T, J. Mol. Struct. 1991;245:69. 7. Yamauchi K, Kuroki S, Fujii K, Ando I, Chem. Phys. Lett. 2000;324:435. 8. Hori S, Yamauchi K, Kuroki S, Ando I. Inter. J. Mol. Sci. 2002;1:8. 9. Shoji A, Kimura H, Ozaki T, Sugisawa H, Deguchi K, Am. Chem. Soc. 1996;118:7604. 10. Kuroki S, Takahashi A, Ando I, Shoji A, Ozaki T, J. Mol. Struct. 1994;323:197. 11. Takahashi A, Kuroki S, Ando I, Ozaki T, Shoji A, J. Mol. Struct. 1998;422:195. 12. Yamauchi K, Kuroki S, Ando I, Ozaki T, Shoji A, Chem. Phys. Lett. 1999;302:331. 13. Yamauchi K. Kuroki S, Ando I, J. Mol. Struct. 2002;602– 603:171. 14. Ono S, Taguma T, Kuroki S, Ando I, Kimura H, Yamauchi K, J. Mol. Struct. 2002;602–603:49.

Part I

of the hydrogen-bonded amide 17 O nucleus of peptides and polypeptides and of other deuterium containing compounds. From this relationship, it is apparent that useful information about the hydrogen-bond length in peptides and polypeptides can be obtained by observation of the e2 q Q/ h value.

and 17 O NMR Chemical Shift NMR for Hydrogen Bonds

37

Isao Ando1 and Tetsuo Asakura2 1 Department

of Chemistry and Materials Science, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo 152-0033, Japan; and 2 Department of Biotechnology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184-8588, Japan

Most recently, the concept of an NMR chemical shift map has been used to characterize the conformation of synthetic polypeptides and the conformation of any specified amino acid residues of proteins. Here, we are concerned with the chemical shift map as established by a theoretical approach and experimental approach. The amino acid residues except for the proline amino acid residue have the freedom of internal rotation about the two consecutive bonds, NH–CαHR and CαHR–CO bonds, where R is the side chain. These torsional angles are defined by  and , respectively. It is very convenient to represent the chemical shift of the amino acid residue as a function of the torsional angles (, ), because the conformationdependent chemical shift can be obtained and then the conformation can be determined through the chemical shift value. This is the so-called “chemical shift contour map” or “chemical shift map.” This has similar significance to the Ramachandran map for the conformational energy of amino acid residues. First, we are concerned with the theoretical approach for establishing the concept of an NMR chemical shift map. In the crystalline state polymer chains assume a fixed conformation. In this case, the structural information obtained from the chemical shift corresponds to the fixed conformation [1]. The calculation of 13 C chemical shifts for dipeptide fragments (n-acetyl-n -methyll-alanine amide) [Ac-l-Ala-NHMe] of poly (l-alanine) and the l-alanine residue containing proteins has been attempted using the finite perturbation theory (FPT) method for chemical shift within the semi-empirical MO framework [2] in order to understand and predict the 13 C chemical shift behavior of polypeptides associated with the secondary structures such as the α-helix form, the β-sheet form, etc., and the determination of secondary structure through the observation of the 13 C chemical shift [1]. The observed 13 C chemical shifts of the Cβ carbon of the lAla residue in various peptides and polypeptides vary significantly depending on the conformation, which may be the right-handed α-helix form, β-sheet form, or another form. Such sizeable displacements of the 13 C chemical shifts can be characterized by variations in the electronic structures of the local conformation as defined by the torsion angles (, ). The chemical shift maps for the

Graham A. Webb (ed.), Modern Magnetic Resonance, 37–42.
C 2008 Springer.

Cβ and Cα carbons have been made on the basis of the calculated data. From these maps, we can estimate semiquantitatively the 13 C shielding for any specified conformation. This is a very useful representation of the chemical shift behavior resulting from changing the dihedral angles as in a Ramachandran energy map. It has been demonstrated from comparisons of the experimental data and the predicted values given by this chemical shift map that the map successfully predicts the 13 C chemical shifts of l-alanine residues in polypeptides and proteins [1–5]. More sophisticated ab initio calculations for the NMR chemical shifts have become available for medium-size molecules as a consequence of the remarkable advances in performance of workstations, personal computers, and supercomputers [3–5]. This leads to a quantitative discussion of the chemical shift behavior. From such a situation, the 13 C chemical shift map was made by ab initio MO calculations with the 4-31G basis set using the GIAO-CHF (gauge-independent atomic-orbital coupled Hartree–Fock) method on n-acetyl-n -methyl-l-alanine amide as shown in Figure 1 [3], which is the same model molecule as used in the case of the above FPT calculations. All the geometrical parameters are energy-optimized. The isotropic 13 C chemical shift map of the Cβ carbon as a function of the torsion angles was calculated as shown in Figure 1, where the positive sign indicates an increase in shielding and the calculated 13 C shielding of methane is 207.2 ppm and the observed 13 C chemical shift is −2.1 ppm relative to TMS. The overall trend of this map is similar to that obtained by the FPT method. The calculated isotropic shielding constant (σ ) for the Cβ carbon is 186.4 ppm for the torsion angles (, ), corresponding to the anti-parallel β (βA )-sheet form, 189.4 ppm for the right-handed α. (αR )-helix form, 189.6 ppm for the lefthanded α. (αL )-helix form as shown in Figure 1 (In Table 1 the calculated shieldings are converted to chemical shifts relative to TMS. Thus, the chemical shift values for the βA -sheet form, the αR -helix form, and the αL -helix form become 18.74, 15.72, and 15.4 ppm, respectively.). On the other hand, the observed isotropic chemical shifts (δ) are 21.0 ppm for the βA -sheet form, 15.5 ppm for the αR -helix form, and 15.9 ppm for the αL -helix form (Table 2). Such an experimental chemical shift behavior is well explained

Part I

NMR Chemical Shift Map

Chemistry

Part I

180

186.0

180

(a)

150

150

167 169

187.0

120

120 188.0

ϕ (deg.)

190.0189.0

186.0

60

185.0

189.0

0

183.0 182.0 181.0 180.0

189.0 186.0

187.0 188.0

188.0

181.0 182.0 183.0 184.0

180

190 188

150

30

-30

50

184 182

173 175

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

80

180

(c)

150

208 208

ϕ (deg.)

184

120

120

90

90 190 192 196

60

190

194 192

190 188

30 0

188 186

-30 -60

194

184 182

20

50

204

206

180

167 169 171173 175

165 163 161 159 157 155 153 151

177 171

20

(b)

163 165 167 169

-60

187.0 185.0 186.0

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

169 167 165 163 161 159 157 155 153

173

0

187.0

185.0

-30 -60

188.0

184.0

30

60

175

171

90

190.0

189.0

ϕ (deg.)

90

153 155 157 159 161 163 165 167 169

165

80

(d)

206

204 202

ϕ (deg.)

38 Part I

204 200

60

206 202

30 204

0

208

204 202

180

-30

178 182

-60

206

208

208 204

184

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

20

50

80

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

20

50

80

Fig. 1. The calculated 13 C chemical shift map of the Cβ and carbons of n-acetyl-n -methyl-l-alanine amide by using the GIAO-CHF method with 4-31G ab initio MO basis sets. The 4-31G optimized geometries for the peptide were employed. (a) isotropic; (b) σ 11 ; (c) σ 22 and (d) σ 33 for the Cβ carbon( in ppm), and (e) isotropic; (f) σ 11 ; (g) σ 22 and (h) σ 33 for the Cα carbon( in ppm).

by the calculated behavior. It is found that the change of the torsion angle dominates the isotropic chemical shift behavior of the l-alanine residue Cβ carbon. The principal values of the chemical shift tensor give information about the three dimensional electronic state of a molecule. However, in order to understand the behavior of the principal values, one should obtain information

about the orientation of the principal axis system of the chemical shift tensor with respect to the molecular fixed frame. The orientations of the principal axis systems of the chemical shift tensors of the l-alanine Cβ-carbons in some peptides can be theoretically determined [4], whose l-alanine moieties have different main chain torsion angles, (, ) = (−57.4◦ , −47.5◦ ) [αR -helix form],

NMR Chemical Shift Map 39

NMR Chemical Shift Map

150

180

(e)

162.8

138

142 144 146 148 150 152

120

157.3 158.4

90

159.5 160.6

60

159.5

163.9

158.4 159.5 157.3 156.2 158.4

162.8

30

ϕ (deg.)

ϕ (deg.)

90

161.7

155.1

157.3 156.2

60 30

140 138 136

136

159.6 160.6

134

136 138 140 142 144 146 148 150 152 154

-30 158.4

-60

-60

162.8 161.7 163.9 -90 -160 -130 -100 -70 -40 -10 φ (deg.)

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

146

180

20

50

161

80

180

(g)

190 188

159

150

157

120

151

150

153 155 157

120

30 0 -30

159 155

157 155

163

ϕ (deg.)

167 165

163 167 161 165 163 161 159 157 155 151 153 155 157

153

60 30 0 184

-30

-60

-60 159

161 163 -90 -160 -130 -100 -70 -40 -10 φ (deg.)

80

186 184 182 180

(h)

166 168 170 172 172 174 174 176 176 178 178

163 169

50

90

161

60

20

174 178 172 176

159

90

140 138 136

0

0

ϕ (deg.)

(f)

150

161.7

120

-30

140 138 140

182 180 176 178

180

178 176 174 172 170

174

20

50

80

-90 -160 -130 -100 -70 -40 -10 φ (deg.)

20

50

80

Fig. 1. (Continued)

(−138.8◦ , 134.7◦ ) [βA -sheet form], (−66.3◦ , −24.1◦ ) [310 R -helix form], and (−84.3◦ , 159.0◦ ) [31 -helix form]. The σ33 component nearly lies along the Cα − Cβ bond for all the peptides considered here, and also the σ11 is nearly perpendicular to the plane defined by the Cβ, the Cα, and the N atoms in the l-alanine residue; on the other hand, σ22 is parallel to the plane. These results agree with the experimentally determined direction of σ33 of the Cβ carbon in l-Ala amino acid by Naito et al. [7]. The σ11

component for the dihedral angles corresponding to the βA -sheet form is 37.06 ppm. This shows a high frequency shift of about 9 ppm with respect to that for the αR -helix form. This result means that the σ11 dominates the high frequency shift on the isotropic chemical shift of the Cβ carbon for the βA -sheet form. Since the σ11 value does not orient along a specified chemical bond, it is not easy to comprehend intuitively the chemical shift tensor behavior of the Cβ carbon. However, it is obvious that the

Part I

180

40 Part I

Chemistry

Part I

Table 1: Calculated 13 C chemical shifts (ppm) of l-alanine residue Cα- and Cβ-carbons by the 4-31G–GIAO-CHF method Cα Sample Ac-Ala-NHMe Boc-Ala-Aib-OH Boc-Ala-Pro-OH Poly (Ala)† Poly (Ala)‡

Cβ

σiso

σ11

σ22

σ33

σiso

σ11

σ22

σ33

43.62 — 45.71 — 45.52 44.73

65.79 — 64.74 — 61.93 62.02

46.46 — 55.04 — 43.69 47.53

18.91 — 26.37 — 30.93 24.64

15.94 — 15.84 — 15.72 18.74

33.80 — 32.47 — 28.16 37.06

17.97 — 19.03 — 22.14 21.70

−3.49 —* −4.00 —* −3.16 −2.53

*The chemical shifts could not be calculated because of SCF failures. † With the αR -helix conformation. ‡ With the β -helix conformation. A

through-space interaction between the Cβ methyl group and its surroundings might be important for understanding the σ11 behavior. For all the torsion angles employed in the calculations, the σ33 component of the chemical shift tensor of the lalanine Cα-carbons always lies along the Cα–C bond. R However, for the αR -helix, the 310 -helix, and the 31 -helix forms, the σ11 component lies in a slightly deviated direction from the Cα–Cβ bond: and for the βA -sheet form, the σ11 component is along this direction. The tensor component which is nearly along the Cα–Cβ bond is 47.53 ppm for the βA -sheet form, 61.93 ppm for the αR -helix form, 64.74 ppm for the 3R10 -helix form, and 65.79 ppm for the 31 -helix form. The change of the dihedral angles causes the large deviation of the chemical shift tensor component which is along the Cα–Cβ bond. Moreover, since

σ33 depends on changes from one torsion angle to another, it is obvious that there exists the explicit torsion angle dependence on σ33 . It is thought that if the carbonyl group in the l-Ala residue forms a hydrogen bond, σ33 will be probably affected [6]. Next, we are concerned about the preparation of isotropic NMR chemical shift maps for the Cα and Cβ carbons in proteins from an empirical database [8]. It seems to be important to assemble a larger database of 13 C shifts in proteins of known structure, to enable us to study the effect of protein conformation and sequence on Cα and Cβ chemical shifts experimentally. The database which contains 3,796 13 Cα and 2,794 13 Cβ chemical shifts from 40 different proteins are used for the preparation of the chemical shift maps. All the proteins have high-resolution crystal structures, and NMR studies have indicated that the

Table 2: Observed 13 C chemical shifts of l-alanine residue Cα- and Cβ-carbons for peptides including l-alanine residues in the solid state, as determined by 13 C CP-MAS NMR, and their geometrical parameters 13 C

chemical shift (ppm)

Dihedral angle (deg)

Cα

Cβ





ω

Ac-Ala-NHMe

49.3, 50.4

18.8, 21.1

−84.3

159.0

173.3

Boc-Ala-Aib-OH Boc-Ala-Pro-OH Poly (Ala)* Poly (Ala)†

52.3 49.2 53.0 48.7

17.4 17.2 15.5 21.0

−87.6 −66.3 −95.4 −57.4 −138.8

154.8 −24.1 153.6 −47.5 134.7

171.9 171.8 179.9 −179.8 −178.5

Sample

*With the αR -helix conformation. † With the β -helix conformation. A

NMR Chemical Shift Map

differences between different amino acid types in the backbone geometry dependence; the amino acids can be grouped together into five different groups with different (, ) shielding surfaces. The overall fit of individual residues to a single non-residue-specific surface, incorporating the effects of hydrogen bonding and χ 1 angle, is 0.96 ppm for both Cα and Cβ. As examples, the chemical shift maps prepared for the Cα and Cβcarbons of Ala residues are shown in Figure 2a and b, respectively, as functions of the torsion angles (, ) [9]. Here only the regions (−180◦ <  < 0◦ , −180◦ <  < 180◦ ) are shown. Data are only shown for areas with enough data points to give reliable chemical shift predictions, in which the density function is greater than 1. There is a clear conformation dependence in the chemical shifts, for example, the chemical shift in the

(a)

(b)

180

180 49

150

50

150

48.5 48

18.5 17.5

49.5

48

120

19.5

50.5

48.5

19

120

16.5

17 90

16

18

90 49

60

60 49

30

50 49.5

Y

16

15.5 16

30 51

Y

0 50.5

15.5

16.5 17.5 17

0

52

-30

15.5

-30 51.5 52.5

-60

-60

-90

-90

-120

-120

-150

-150 48.5

-180 -180

-150

-120

49

-90 f

-60

-30

0

-180 -180

19.5 19 -150

18 17.5 -120

16.5

17 -90 f

-60

-30

0

Fig. 2. Contour plots of the conformation-dependent chemical shifts (in ppm) of Cα(a) and Cβ(b) carbons of Ala residues in 40 proteins. Chemical shift values in the region (−180◦ <  < 0◦ , −180◦ <  < 180◦ ) are shown, where the density function is more than 1. Random coil chemical shifts are 50.0 ppm for Ala Cα carbon and 16.6 ppm for Ala Cβ carbon.

Part I

solution conformation of the protein is essentially identical to that in the crystal. There is no systematic effect of temperature, reference compound, or pH on the reported shifts, but there appear to be differences in the reported shifts arising from referencing differences of up to 4.2 ppm. The major factor affecting chemical shifts is the backbone geometry, which causes differences of ca. 4 ppm between typical α-helix and β-sheet geometries for Cα, and of ca. 2 ppm for Cβ. The side chain dihedral angle χ 1 has an effect of up to 0.5 ppm on the Cα shift, particularly for amino acids with branched side chains at Cβ. Hydrogen bonding to main chain atoms has an effect of up to 0.9 ppm, which depends on the main chain conformation. The sequence of the protein and ring-current shifts from aromatic rings has an insignificant effect (except for residues following proline). There are significant

NMR Chemical Shift Map 41

42 Part I

Chemistry

Part I

α-helix region is predicted at lower frequency for Cα than the chemical shift in the β-sheet region, but at higher frequency for the Cβ. These chemical shift maps in turn will help to guide efforts in protein structure refinement using 13 C chemical shifts.

References 1. Saito H, Ando I. Ann. Rep. NMR Spectrosc. 1989;21:209. 2. Ando I, Saito H, Tabeta R, Shoji A, Ozaki T. Macromolecules. 1984;17:457.

3. Asakawa N, Kurosu H, Ando I. J. Mol. Struct. 1994;323:279. 4. Asakawa N, Kurosu H, Ando I, Shoji A, Ozaki T. J. Mol. Struct. 1994;317:119. 5. Ando I, Kuroki S, Kurosu H, Yamanobe T. Prog. NMR Spectrosc. 2001;39:79. 6. Asakawa N, Kameda T, Kuroki S, Kurosu H, Ando S, Ando I, Shoji A. Ann. Rep. NMR Spectrosc. 1995;35:233. 7. Naito A, Ganapathy S, Akasaka K, McDowell CA. J. Chem. Phys. 1981;90:679. 8. Iwadate M, Asakura T, Williamson MP. J. Biomol. NMR 1999;13:199. 9. Asakura T, Iwadate M, Demura M, Williamson MP. Int. J. Biol. Macromol. 1999;24:167.

43

Hiromichi Kurosu1 and Takeshi Yamanobe2 1 School

of Natural Science and Ecological Awareness, Graduate School of Humanities and Science, Nara Women’s University, Kitauoya-Nishimachi, Nara 630-8506, Japan 2 Department of Chemistry, Gunma University, 1-5-1 Tenjin-cho, Kiryu, Gunma 376-8515, Japan

Introduction Nuclear shielding offers microscopic information about the stereochemical and crystal structures of polymers which in turn are important factors in understanding physical properties. Details of electronic structure can also be deduced from nuclear shielding data, and this is also important in controlling physical properties [1]. In order to obtain the information about the electronic structure of polymers from nuclear shielding, it is necessary to use a theoretical approach in addition to an experimental one. In general, there are two possibilities to obtain such information by theoretical methods. One is to use a fragment of a polymer such as a dimer, trimer, etc. for theoretical calculations. Such an approach is useful because the nuclear shielding is sometimes governed by the local electronic structure. However, there is doubt whether the electronic structure obtained from the model compound appropriately reproduces that of the polymer chain. Another approach is to employ directly an infinite polymer chain with periodic structure. This leads to the application of the tight-binding (TB) molecular orbital approximation, which was developed in the field of solid-state physics [2]. Its advantages are that it treats the polymer directly and that relatively long-range interactions such as hydrogen bonding in the α-helix form of the polypeptide may be included. In using the model compound approach, the electronic structure of large chains can be visualized by drawing the orbital energies associated with chains of increasing length. On the other hand, for an infinite polymer chain, the energy levels are built up as a continuous band structure. As rotation about the bond is strongly restricted in solid polymers, the periodicity of a polymer chain is retained. From the point of view of calculating nuclear shielding, use of the TB model for the calculation of the electronic structure of polymers has been successful. Herein we described the basic ideas for deducing the electronic structure and the nuclear shielding of solid polymers by the TB approximation.

Graham A. Webb (ed.), Modern Magnetic Resonance, 43–52.
C 2008 Springer.

Theoretical Aspects of Electronic State and Nuclear Shielding in Solid Polymers Polymers can be characterized by a possible periodicity in their conformational structure and a large number of electrons. In this system, the potential energy that an electron experiences is periodic. Periodic systems have the advantage of translational symmetry when compared with aperiodic systems. It is possible to exploit this symmetry in order to reduce to reasonable proportions the formidable task of computing the electronic states of an extended system. The TB molecular orbital model is employed to describe the electronic structure of linear periodic polymers within the framework of a “linear combination of atomic orbitals” approximation for electronic eigenfunctions. By means of Bloch’s theory [3], the eigenfunction n (k) for an electron at position r , belonging to the nth crystal orbital is given by n (k) = N −(1/2)

(N −1)/2

s 

ei jkb Cνn (k)φν (r − jb)

j=−(N −1)/2 ν=1

(1) where k is the wave number, ν is an orbital index for the jth cell, s is the total number of atomic orbitals in a given cell, N is the total number of cells considered, and b is the unit vector of translational symmetry. In equation (1), φ ν (r − jb) represents the ν atomic orbital in the jth cell and Cνn (k) its expansion coefficient in the linear combination of atomic orbitals. The limits of k, within a given Brillouin zone, are −π/2 and π/2. Using Equation (1), the total energy E of the polymer can be expressed as E(k) =

occ  

   n (k)  Hˆ  n (k)

(2)

n

where n (k) is the Slater determinant composed of n (k). Hˆ is the Hamiltonian, consisting of terms representing the kinetic energy, the potential energy of the electronic field

Part I

NMR Chemical Shifts Based on Band Theory

44 Part I

Chemistry

Part I

The nuclear shielding for the various nuclei in a given macromolecule can be calculated from the obtained electronic band structure, Cνn (k) and E n (k). In general, a nuclear shielding can be written as [1]

0

Energy (atomic units)

σ = σd + σp + σ

(3)

where σ d and σ p are the diamagnetic and paramagnetic contributions, respectively, σ is the contribution from neighboring atoms. For a carbon atom, σ is much smaller than 1 ppm and can be considered to be negligible. Thus, σ can be estimated by the sum of σ d and σ p . Based on the TB MO theory, σ d and σ p are obtained by a sum-overstates (SOS) method as follows [4–6]:

−1

σAd (k) = p

σA,αβ (k) = −2

A A      µ0 e2  Pνν (k) φν (r ) r −1  φν (r ) 6π m 2e ν ν

A occ unocc   −1 −µ0 h 2 e2  r −3 2p 1 E mn − 1 E 0 2 4πm e m n j

× 0

1

2

3

Wavenumber Fig. 1. Electronic band structure of polyethylene with trans zigzag conformation calculated using CNDO/2 TB MO.

of the polymer, and the electron repulsion energy. The advantage of Equation (1) is that calculation of the electronic structure of an infinite polymer can be reduced to the calculation for a unit cell (monomer unit) interacting with all other unit cells, using the periodicity of the polymer. By solving the Fock matrix, the expansion coefficients in Equation (1) can be obtained. These expansion coefficients and energies are dependent on k, which leads to a band structure for a polymer chain. Figure 1 shows the calculated band structure of polyethylene with a trans zigzag conformation. As only six valence electrons in a monomer unit cell are included in the calculation, the corresponding six valence bands can be seen in Figure 1. Calculations with model compounds give discrete energy levels. In a series of homogeneous molecules, the number of energy levels increases with the molecular size, and correspondingly the separation between these energy levels decreases. For an infinite polymer chains, the energy level is the continuous band structure as shown in Figure 1. The electronic structure of solid polymers is characterized by this band structure, with a dependence of energy level on k.

(4)

B  B  [X ( j, m, n, β, γ )X (l, n, m, γ , α) j

l

− Y ( j, m, n, β, γ )Y (l, n, m, γ , α) + X ( j, m, n, γ , α)X (l, n, m, β, γ ) − Y ( j, m, n, γ , α)Y (l, n, m, β, γ )]

(5)

Here e is the charge and m e the mass of the electron, µ0 is the permeability of free space, Pνν (k) =

occ 

∗ Cνm (k)Cν m (k)

(6)

m

where m refers to the number of occupied crystal orbitals and n to those that are unoccupied, E 0 is the electronic energy of the ground state and E mn is the energy of the state created by promoting an electron from orbital m to orbital n. j and l are atomic orbitals on centers A and B, respectively, and Iγ

Rβ

Iγ

Rβ

X ( j, m, n, β, γ ) = C jm C jn + C jn C jm Rγ

Rγ

Iβ

Iβ

− C jn C jm − C jm C jn Rγ

Rβ

Rγ

(7)

Rβ

Y ( j, m, n, β, γ ) = C jm C jn − C jm C jn Iβ

Iγ

Iγ

Iβ

+ C jm C jn − C jm C jn

(8)

where α, β, and γ refer to the x, y, and z components of the shielding tensor in cyclic order and R and I indicate the real and imaginary parts of the coefficients

Chemical Shifts Based on Band Theory

Part I

Cνn (k). As shown by Equations (4) and (5), the nuclear shielding is calculated as a function of k. In order to be able to compare the calculated and experimental values of the nuclear shielding, it is necessary to average the calculated data over k, within the first Brillouin zone, as given by π/b −(π/b)

σ (k)D(k)dk

(9)

where D(k) is the density of state, namely the number of states per unit amount of energy. Thus, the nuclear shielding can be obtained through the electronic band structure of a polymer, especially a solid polymer. The quantities calculated using Equations (4) and (5) are the nuclear shielding, σ , and so the negative sign means deshielding. On the other hand, a negative sign of the observed chemical shift, δ, means an increase in shielding. Hereinafter, calculated and observed data are expressed as the nuclear shielding and chemical shift, respectively. The absolute value of the calculated shielding should be compared directly with the observed chemical shift. Furthermore, the formalism for calculating the NMR shieldings of infinite polymer chains with a threedimensional (3D) periodicity in the crystalline state by ab initio TB MO theory has been developed [7]. For the calculations on a 3D system, the program code CRYSTALS88 [8,9] which is available for a crystal orbital calculation of 3D systems was used for calculating the electronic states and a program for calculating the shielding constant[4,5,10,11] using the SOS method was added to CRYSTAL88.

Interpretation of Nuclear Shielding by the TB Method 13

C shielding reflects the magnetic environment of the atom considered, and this depends on the conformation, configuration, and crystal structure in the case of polymers. In order to understand 13 C shielding, examples illustrating the applications of TB MO methods to some polymers will be described.

Conformation It is known from an X-ray diffraction study that a polyoxymethylene chain in the crystalline region takes an allgauche conformation with a 9/5 helix [12]. However, in the non-crystalline region the structure is not yet determined exactly, because of the complicated conformation of the chain. It has been reported that the observed 13 C

−69

σiso (ppm)

σ =

Interpretation of Nuclear Shielding by the TB method 45

−70

−71

60°

120° ψ

180°

Fig. 2. Dependence of the calculated 13 C NMR shielding of polyoxymethylene on dihedral angle ψ.

chemical shifts of polyoxymethylene in the crystalline and non-crystalline regions appear at 88.5 and 89.5 ppm, respectively [13]. Figure 2 shows the dependence of the calculated isotropic 13 C shielding on the dihedral angles (ψ) within the framework of the CNDO/2 TB MO method. The isotropic 13 C shielding increases as ψ is increased from 50◦ to 90◦ , and the shielding decreases through a minimum value as ψ is increased from 90◦ to 180◦ . The values of 60◦ and 180◦ , for ψ correspond to the all-gauche and all-trans conformations, respectively. The calculated isotropic 13 C shielding of the all-gauche conformation appears at lower frequency by about 1 ppm than that of the all-trans conformation. Dividing the difference of 13 C shielding between the all-gauche and all-trans conformations, for the diamagnetic term, the difference between the all-gauche and all-trans conformations is 0.1 ppm, and for the paramagnetic term the corresponding difference is 0.9 ppm. Thus, the contribution to the relative 13 C shielding is due mainly to a change in the paramagnetic term. The diamagnetic term is determined only by the ground state, whereas the paramagnetic term involves interactions between the ground and excited states, as seen from Equations(4) and (5). Thus, the observed chemical shift difference between the crystalline and noncrystalline regions comes from a variable interaction due to a conformational change. Therefore, it can be argued that the calculated value confirms well the experimental

46 Part I

Chemistry

Part I

finding that the 13 C shielding for the crystalline region is greater by about 1 ppm than that for the non-crystalline region. Calculated results of σ (= σ yy − σzz ; spectrum breadth) for the all-gauche and all-trans conformations are 37.7 and 1.8 ppm, respectively. σ for the all-gauche conformation is much larger than that for the all-trans conformation. On the other hand, the experimental values of δ for the crystalline and non-crystalline regions are about 35 and 7–10 ppm, respectively [13]. The calculated and experimental values agree relatively well with each other. The fact that the calculated value of σ for the all-trans conformation is rather small suggests that the electronic environment around the carbon nucleus considered here is relatively symmetric. The small value of δ may be due mainly to the high symmetry of the electronic environment in addition to the averaging of the 13 C shielding anisotropy by molecular motion. Further, we are concerned with the behavior of δ 22 , whose value is obtained from the apex of the tent-like powder pattern. In the experimental data, δ 22 for the crystalline region appears at about 5 ppm to low frequency when compared with that for the non-crystalline region. However, the calculated value of σ x x for the all-gauche conformation appears at lower frequency by about 2.5 ppm when compared with that for the all-trans conformation. Thus the calculation explains the experimental observation reasonably, despite the rough assumption of an all-trans conformation for the non-crystalline region. Table 1 shows the calculated and observed 13 C shieldings of a polyglycine chain with forms I and II [14]. The observed carbonyl 13 C signal for form I is shielded by about 4 ppm compared with that for form II. The calculated 13 C shielding for form I is larger by about 11 ppm than that for form II, which matches the experimental finding. There is no significant difference between the methylene 13 C chemical shifts for forms I and II within experimental error. At this stage, it cannot be determined whether the methylene shielding for form I or that for form II is the larger. The calculated shielding of 13 C for form I is smaller by about 3 ppm than that for II, so the

Table 1: Observed and calculated 13 C chemical shifts and shieldings of an isolated polyglycine chain δobs (ppm)∗

CH2 C=O ∗ From † The

σcalc (ppm)†

Form I

Form II

Form I

Form II

43.5 168.4

43.5 172.3

−127.2 −236.7

−124.4 −248.1

TMS. The positive sign means deshielding. negative sign means deshielding.

calculation predicts the existence of a shielding difference ( about 3 ppm) between forms I and II. It appears that the calculated shieldings are somewhat exaggerated compared with the observed values. Nevertheless, the observed trend that the 13 C chemical shift difference for the methylene carbon is very small when compared with that for the carbonyl carbon is reproduced qualitatively by the calculation.

Configuration Polyacetylene (PA) is the simplest conjugated polyene and has two configurations (cis and trans). It is reported that the 13 C shielding of cis-PA appears at a lower frequency by about 10 ppm than that of trans-PA [15]. Calculations of the 13 C shielding of PAs are carried out based on the use of CNDO/2 TB MO and INDO/S TB MO methods. The differences in the isotropic 13 C shielding between the cis- and trans-PAs, calculated by the CNDO/2 and INDO/S TB MO methods, are about 2.0 and 3.5 ppm, respectively [16], i.e. the results differ by a factor of about 2. As the observed difference is about 10 ppm, both types of calculation somewhat underestimate the real value in this case. Figure 3 shows the observed [17] and calculated components of the 13 C shielding tensors of cis- and transPAs. As is seen, the absolute values of zz and x x calculated using the INDO/S TB MO method are smaller by about 40 ppm than those obtained by the CNDO/2 TB MO method. The value of σ yy calculated by the INDO/S TB MO method is larger by about 10 ppm than that obtained using CNDO/2 TB MO procedure. Consequently, the separation between the values of σ x x and σ yy calculated using the INDO/S TB MO is much larger than that obtained by CNDO/2 TB MO’s. Further, it is shown that in going from the calculation using CNDO/2 method to that using INDO/S the values of σ zz and σ x x are changed considerably. The most remarkable difference is in the estimation of the π–π overlap integral. The values of σ zz and σ x x are affected by the distribution of π electrons. The electronic band structures of a PA consist of five valence bands within the framework of the semiempirical methods. The highest occupied band has π symmetry. Comparing the electronic band structures calculated by CNDO/2 TB MO with that of INDO/S TB MO, the energies of the three high-energy bands increase, in particular that of the highest occupied band. As can be seen from Equation (3), σ p contains a term proportional to the inverse of the excitation energy. The increase in the energies of the three high-energy bands is one of the factors leading to the deshielding of σ zz and σ x x , which implies a lower excitation energy. In fact, the band gaps (the energy difference between the highest occupied and the lowest

Chemical Shifts Based on Band Theory

δyy

δxx 80

Part I

δzz

Interpretation of Nuclear Shielding by the TB method 47

116 cis Observed 97

75

trans 100

200

σxx

σzz

σyy 30

87

cis CNDO/2 32

84

trans −250

−150 σxx

σzz

σyy 82

75

cis INDO/S 83

76

trans −250

−150 C

C

C

C σxx

σzz

C C

σxx

σyy

C σzz

σyy

Fig. 3. Observed and calculated components of 13 C NMR shielding tensors of cis- and trans-PAs. The directions of the principal axes are indicated at the bottom.

unoccupied bands) for trans-PA, calculated using the CNDO/2 and INDO/S TB MO methods, are about 8.0 and 4.2 eV, respectively. The observed band gap for trans-PA is about 1.9 eV. Thus, the description of the contributions of these high energy bands to σ x x and σ zz is remarkably improved in going from the CNDO/2 to the INDO/S TB MO procedure. It is clear from Figure 3 that the reason why the calculated difference in isotropic shielding of 13 C between the cis- and trans-PAs is underestimated is that the difference

in σ yy cannot be reasonably reproduced. In order to reproduce more reasonably the observed difference in σ yy between the cis- and trans-PAs, it may be necessary to use MOs that can properly describe band structures that have σ symmetry. Polypyrrole is one of a series of heterocyclic polymers that has attracted much attention because of its characteristic electric and electronic properties. Fundamental structural formulae have been proposed for undoped and doped polypyrroles, where the aromatic form

48 Part I

Chemistry

Part I

Table 2: Calculated 15 N shieldings and band gaps for aromatic and quinoid polypyrrole models using INDO/S TB MO Structure

Calculated shielding σiso (ppm)

Band gap (eV)

Aromatic Quinoid

−223.50 −232.21

5.12 2.86

corresponds to the undoped state and the quinoid form to the doped state. From analysis of the high-resolution solid-state 15 N NMR, it is known that the 15 N chemical shift for the quinoid form appears to high frequency of that for the aromatic form. In order to obtain information about the 15 N chemical shift, and electronic band structures of an infinite polypyrrole chain with aromatic or quinoid forms, calculations were carried out using the INDO/S TB MO’s [18]. As listed in Table 2, the calculated shielding of 15 N for the quinoid form appears to high frequency compared with that for the aromatic form. The calculated results agree with the observed ones, suggesting that the electronic band structure has been properly estimated. From the calculated electronic band structure, the band gaps for the aromatic and quinoid forms are 5.1 and 2.9 eV, respectively. This result implies that the electrical conductivity of the quinoid form is larger than that of the aromatic form. Therefore, it can be expected that if the amount of the quinoid form is increased, polypyrrole with a higher electric conductivity can be obtained.

Crystal Structure The crystal structure in the solid state, which describes how molecules condense, is an important factor when discussing physical properties. The effect of crystal structure on the nuclear shielding in principle, can be separated from the effect of conformation and configuration within the limitation of a given quantum chemical method. Experimentally, it is reported that the 13 C chemical shifts of the CH2 carbon for the n-paraffins, cyclic paraffins, and polyethylene with a trans zigzag conformation in the orthorhombic and triclinic forms are about 33 and 34 ppm, respectively [19,20]. As the conformation is the same in each case, the difference of about 1 ppm may be due to a local change in intermolecular interactions resulting from the change of the orthorhombic to the triclinic form. Paraffins can be considered as models for the crystallographic form of polyethylene. The result of X-ray diffraction studies suggests that the trans zigzag

plane of any specified chain in the orthorhombic and triclinic forms is, respectively, perpendicular and parallel to those of the neighboring chains. The 13 C chemical shift of polyethylene with a monoclinic form appears at higher frequency by 1.4 ppm compared with that of polyethylene with the orthorhombic form. The orientation of the trans zigzag plane in the monoclinic form is very close to that in the triclinic form [21]. Thus, the solid-state 13 C NMR chemical shift depends on the orientation of the C–C–C plane in trans zigzag chains. Since the chains lie periodically in the solid state, calculations of the chemical shifts should be employed that take account of the 3D structure, including interactions between chains. In order to take into account interchain interaction, a “seven-polyethylene-chains model” has been used to compute the 13 C shielding [10]. Figure 4 shows models for polyethylenes with an orthorhombic form and a monoclinic form. The setting angle ψ for the orthorhombic polyethylene is not clearly determined. Figure 5 shows the ψ dependence of difference in shielding of the 13 C between orthorhombic and monoclinic polyethylenes (σorth − σmono ) calculated using the INDO/S TB MO method. As can be seen, the calculated difference in the shielding of the 13 C between these forms is positive in the case of ψ = 25◦ –42◦ , and negative in the case of ψ = 20◦ –25◦ and ψ = 42◦ –55◦ . The largest difference occurs at ψ = 35◦ . From these results, the angle of ψ = 25◦ –42◦ is good for the setting angle of orthorhombic polyethylene, and can reasonably reproduce the observed data. This result shows that the difference in chemical shift between the orthorhombic and monoclinic forms is caused by the difference in interchain interactions through the electronic band structure. In order to elucidate the intermolecular interaction effect on the NMR chemical shift of PA crystal, the 3D ab initio TB MO calculations were carried out [22]. Before going to the 3D PA crystal, it is significant to employ a single PA chain, because the intermolecular interaction effect on the electronic structure and NMR chemical shift behavior is clear. In Table 3, the total energy per monomer unit and NMR chemical shielding for a single cis- and trans-PA chain are shown as calculated by using the ab initio TB MO method within the framework of the STO-3G minimal basis set. It is shown that, the total energy per monomer unit for trans-PA is lower by 0.0069 a.u. than that for cis-PA. This means that the trans-form is more stable than the cis-form. The tendency of the calculated results qualitatively explains the experimental results that, undoped trans-PA is thermally more stable than cis-PA. Experimentally, the 13 C chemical shift for the cis-form appears at a lower frequency by 10 ppm than that of the trans-form [17]. On the other hand, the calculation shows that the 13 C chemical shift of

Chemical Shifts Based on Band Theory

Interpretation of Nuclear Shielding by the TB method 49

Part I

(a) hydrogen

carbon aa

ψ ψ bb

(b) hydrogen

carbon

Fig. 4. The “seven-polyethylene-chains” model: (a) Orthorhombic form; (b) Monoclinic form.

Chemistry

Part I

Fig. 5. The 13 C NMR chemical shift difference between the orthorhombic and monoclinic polyethylene chains calculated using the seven-chains model, as a function of the setting angle ψ.

Chemical shift difference (ppm)

50 Part I

0.4 0.2 0.0 −0.2 −0.4

20°

the cis-form appears at a slightly lower frequency by 0.1 ppm than that of the trans-form. Comparing with the experimental results, the chemical shift difference between the cis- and trans-forms, is very small. This means that, the single chain model is insufficient to reasonably explain the experimental results. Therefore, it can be said that the interchain interactions should be taken into account by the 3D polymer crystal. The electronic structures and NMR chemical shieldings of the 3D infinite cisand trans-PA chains with orthorhombic crystallographic form were calculated using the ab initio TB MO method within the framework of the STO-3G minimal basis set. The total energies per monomer unit and chemical shieldings, along with the experimental chemical shift values are listed in Table 3. The total energy per monomer unit

30°

ψ

40°

50°

for the trans-form is lower by 0.0024 a.u. than that for the cis-form. The trans-form is thus more stable than the cis-form. This explains reasonably the experimental thermal stability of the trans-form in the PA crystal over the cis-form. As seen from Table 3, the 13 C chemical shift for the cis-form in a the PA crystal appears at a lower frequency by 1.0 ppm than that for the trans-form, calculated using the experimental lattice parameters (Case A). The calculated results approach the experimental ones more closely, compared to the single chain model. This shows that the intermolecular interaction plays an important role towards the 13 C chemical shift behavior. What is the 13 C chemical shift behavior if the intermolecular interaction is increased in the 3D PA crystal model? When the lattice

Table 3: Total energies, band gaps, and NMR chemical shieldings for a single chain of cis- and trans-polyacetylenes and for a 3D crystal of cis- and trans-polyacetylenes as calculated by ab initio TB MO method within the framework of STO-3G minimal basis set A single chain

Three-dimensional crystal 13 C

Polyacetylene form Cis-form Trans-form ∗ Relative

Total energy† (a.u.)

chemical shielding‡ (ppm)

Total energy† (a.u.)

−75.9368 −75.9437

−121.6 −121.7 ¶ = 0.1

−75.9430 −75.9454

13 C

chemical shielding‡ (ppm)

Case A§

Case B§

−117.7 −118.7 ¶ = 1.0

−114.7 −119.2 4.5

Experimental 13 C chemical shift* δ (ppm) 127.3 137.3

to TMS. The positive sign means deshielding. This is opposite to the calculation. monomer unit. ‡ The calculated 13 C chemical shielding indicates that the negative sign means deshielding. § Case A: by using experimental lattice parameters and Case B: using reduced interchain distance (reduce lattice parameter b = 5.0 A. This means that the reduced interchain distance leads to increase intermolecular interaction). ¶ Chemical shift difference between cis- and trans-forms. † Per

Chemical Shifts Based on Band Theory

−114.00

Part I

−115.00 Shielding constant / ppm

Fig. 6. The plots of the 13 C chemical shieldings for a 3D crystal of cisand trans-PAs, as calculated by the ab initio TB MO method within the framework of STO-3G minimal basis set against the length of the lattice parameter a.

References 51

−116.00

cis form trans form

−117.00 −118.00 −119.00 −120.00 4.50

5.00

5.50

6.00

6.50

7.00

7.50

8.00

8.50

interchain distance / A

parameter a is reduced to 5 A(this means an increase in the intermolecular interaction) (Case B), the 13 C chemical shift difference between the cis- and trans-forms becomes 4.5 ppm. This approaches the experimental results more closely. Therefore, in order to clarify the interchain distance effect of the 13 C chemical shift for the cis- and transforms, their 13 C chemical shieldings were calculated, with a change in the length of the lattice parameter a using the ab initio TB MO method, within the framework of the STO-3G minimal basis set. The calculated chemical shieldings are plotted against the length of the lattice parameter a as shown in Figure 6. It is seen that the 13 C chemical shift for the cis-form moves slowly to low frequency, as the length of the lattice parameter a decreases from 8 to 5 A, that is, the intermolecular interaction increases. On the other hand, for the trans-form, the 13 C chemical shift, moves slowly to low frequency, as the length of the lattice parameter a decreases from 7.7 to 5.6 A, and moves largely to high frequency as the length of the lattice parameter a decreases from 5.6 to 4.9 A. With a decrease in the length of the lattice parameter a below 5.5 A, the chemical shift difference becomes large and quantitatively approaches the experimental result. As predicted from Figure 6, the calculated chemical shift difference becomes 10 ppm (the experimental value) when the length of the lattice parameter a is about 4.7 A. The calculation with a shorter length of the lattice parameter a agrees with the experimental result. This may come from the approximations implied with the low level MO minimal basis set such as STO-3G, which is used in the calculation. If a higher level minimal basis set is used, the situation may be modified. Nevertheless, it can be said that

the present calculation explains well the experiment, and the chemical shift is very sensitive to intermolecular interactions. This means that the 13 C chemical shift is a very sensitive measure for use in investigating intermolecular interactions in a polymer crystal.

References 1. Ando I, Webb GA. Theory of NMR Parameters. Academic Press: London, 1983. 2. Imamura A. J. Chem. Phys. 1970;52:3168. 3. Bloch F. Z. Phys. 1928;52:555. 4. Yamanobe T, Ando I. J. Chem. Phys. 1985;85:3154. 5. Yamanobe T, Chujo R, Ando I. Mol. Phys. 1983;50:123. 6. Yamanobe T, Ando I, Saito H, Tabeta R, Shoji A, Ozaki T. Bull. Chem. Soc. Jpn. 1985;58:23. 7. Uchida M, Toida Y, Kurosu H, Ando I. J. Mol. Struct. 1999;508:181. 8. Dovesi R, Pisani C, Roetti C, Causa M, Saunders VR. Quantum Chemistry Program Exchange, Department of Chemistry, Indiana University, IN 47405, Program No. 577. 9. Pisani C, Dovesi R, Roetti C. In: G Berthier, MJS Dewar, H Fischer, K Fukui, GG Hall, J Hinze, HH Jaffe, J Jortner, W Kutzellnig, K Ruedenberg, J Tomasi (Eds). Lecture Notes in Chemistry, Vol. 48. Springer: Berlin, 1988. 10. Kurosu H, Ando I, Yamanobe T. J. Mol. Struct. 1989;201:239. 11. Ishii T, Kurosu H, Yamanobe T, Ando I. J. Chem. Phys. 1988;89:7315. 12. Tadokoro H, Kobayashi M, Kawaguchi Y, Kobayashi A, Murashashi S. J. Chem. Phys. 1963;38:703. 13. Kurosu H, Yamanobe T, Komoto T, Ando I. J. Chem. Phys. 1987;116:391.

52 Part I

Chemistry

Part I

14. Yamanobe T, Ando I, Saito H, Tabeta R, Shoji A, Ozaki T. Bull. Chem. Soc. Jpn. 1985;58:23. 15. Maricq MM, Waugh JS, MacDiarmid AG, Shirakawa H, Heeger AT. J. Am. Chem. Soc. 1978;100:7729. 16. Yamanobe T, Ando I, Webb GA. J. Mol. Struct. 1987;151:191. 17. Terao T, Maeda S, Yamabe T, Akagi K, Shirakawa H. Chem. Phys. Lett. 1984;103:347.

18. Kikuchi M, Kurosu H, Ando I. J. Mol. Struct. 1992;29: 193. 19. Yamanobe T, Sorita T, Komoto T, Ando I, Sato H. J. Mol. Struct. 1985;131:267. 20. VanderHart DL. J. Magn. Reson. 1981;44:117. 21. VanderHart DL, Khoury F, Polymer 1984;25:1589. 22. Fujii K, Kuroki S, Uchida M, Kurosu H, Ando I. J. Mol. Struct. 2002;602–603:3.

53

Julio C. Facelli Center for High Performance Computing, University of Utah, Salt Lake City, UT 84112-0190, USA

Abstract This article presents a discussion of the origin of the chemical shieldings, which is followed by a discussion on how they are calculated using state-of-the-art electronic structure methods. Several examples of quantum chemical calculations of chemical shieldings in common nuclei are presented to provide the reader with a general overview of the reliability of these calculations. The shortcomings of the current methods are finally discussed.

Introduction Perhaps the most important discovery after the successful detection of the NMR signal has been the observation that the nuclear spin resonance frequencies depend on the chemical or electronic environment of the nuclei [1,2], or as Norman Ramsey states in his landmark papers [3,4] of 1950: “In measurements of nuclear magnetic moments, a correction must be made for the magnetic field arising from the motions of the molecular electrons which are induced by the externally applied field.” Ramsey realized that corrections using only Lamb’s diamagnetic theory were inadequate for molecules because there are additional shielding contributions arising from the secondorder paramagnetism. To address this problem, he developed the necessary theoretical framework to explain and eventually to calculate the “chemical effect,” that would become the chemical shift commonly used in our days for structural elucidation. The calculation of chemical shieldings has been a challenge to theoreticians and computational chemists for more than 50 years. Great impetus for this theoretical and modeling work has been provided by the extraordinary sensitivity of the chemical shielding to electronic and molecular structure and environment which can only be unraveled by computational modeling. It should be noted that the chemical shieldings are tensor properties, i.e. the shift of the resonance frequency depends on the molecular orientation with respect to the external magnetic field, but the brevity of this article precludes any discussion of the tensor properties of the chemical shieldings [5]. In this chapter, as customary in the literature, we will refer to the isotropic component of the chemical shielding tensor as the chemical shielding. Graham A. Webb (ed.), Modern Magnetic Resonance, 53–62.
C 2008 Springer.

There is considerable confusion in the literature about the use of the terms “chemical shift” and “chemical shielding.” The chemical shielding is the tensor that describes the relative change in the local magnetic field at the nucleus position relative to the external magnetic field. This change in the local magnetic field, which is originated in the interaction of the electron cloud with the external magnetic field, can produce shielding or deshielding of the nucleus. In the first case, the local magnetic field is increased with respect to the external field, while in the second case the local field is decreased. In general, shielding effects are associated with diamagnetic effects from spherical charge distributions, while de-shielding effects are associated with the nonspherical charge distribution originating from p or higher angular momentum electrons. When experiments are performed at a constant magnetic field, as it is normally done in modern NMR spectrometers, a shielding effect results on a shift of the resonance to a higher frequency, while a deshielding effect will result in a lower resonance frequency. In practice, NMR experiments do not measure the chemical shielding directly, instead the common practice is to measure the chemical shifts as the change of resonance frequency of a nucleus relative to a given standard. Moreover, for historic reasons it is customary to reverse the frequency scale, i.e. nuclei more shielded than the standard are considered to have lower chemical shifts and those less shielded larger ones. The formal relation between the chemical shift and chemical shielding tensors is given by δ = 1σiso − σ.

(1)

where δ is the chemical shift tensor, σ is the chemical shielding tensor, 1 is the unit matrix and σ iso is the isotropic value or trace of the chemical shielding of the standard reference used in the NMR experiments. The determination of the value of σiso , usually known as absolute chemical shift, is quite difficult. It involves the estimation of the paramagnetic contribution to the chemical shielding in the center of mass of the molecule using its relationship with the spin rotational constant and the calculation of the corresponding diamagnetic part using quantum mechanical methods. The procedure for selecting primary and secondary reference compounds has been extensively discussed by Mason [6].

Part I

Modeling NMR Chemical Shifts

54 Part I

Chemistry

Part I

The material presented in this chapter is restricted to the chemical shifts and shielding calculations in diamagnetic molecules. When the molecular electronic ground state is not a singlet, i.e. there are unpaired electrons present in the sample, additional mechanisms contributing to the chemical shielding are present [7].

Theory of the Chemical Shieldings The theory and modeling of the chemical shifts have been described in numerous publications [8–36]. To obtain exact expressions for the calculation of chemical shieldings using the non-relativistic Born–Oppenheimer approximation [37], it is necessary to include the vector potential representing the external magnetic field and the dipolar field from the magnetic moment of the nucleus into the electronic Hamiltonian. In the gauge of Coulomb, the final expressions for the diamagnetic and paramagnetic contributions are given by    1   ψ0 rk r N k δαβ − rkα r N kβ r N−3k  ψ0 , 2 2c     −1  1 p σαβ (O) = 2 ψ0  L kα  ψn 2c n E n − E 0     × ψn  L N kβ r N−3k  ψ0 + C.C.

d σαβ (O) =

(2)

(3)

Following the derivation in reference [11], we have indicated explicitly that Eqs. [2] and [3] are valid when the origin of the vector potential is at the position O. The sum in Eq. [3] is over all the exited states of the molecule. It is important to understand the behavior of the diamagnetic, Eq. [2], and paramagnetic, Eq. [3], terms under a translation of the origin of coordinates. Both terms exhibit an explicit dependence on the origin of coordinates used in the calculations, but as demonstrated elsewhere for exact and variational wave functions in a complete space, these dependences cancel each other making the total chemical shielding independent of the origin of coordinates. Unfortunately, this is not true for a finite expansion of the wave function. Serious complications arise in chemical shielding calculations as a consequence of the imperfect cancelation of origin-dependent terms in the diamagnetic and paramagnetic components. Several methods have been proposed to mitigate the gauge problem and to make the results formally gauge invariant for incomplete basis sets and hopefully to produce better results when using moderate size basis sets. The most common approach is to use London or Gauge Invariant (or including) Atomic Orbitals or GIAOs in the atomic expansion of the molecular orbitals [17,20,21]. Other popular methods are individual gauge for localized orbitals (IGLO) [30,31,38], localized orbitals local origin (LORG) [39,40], individual gauges for atoms in molecules

(IGAIM) [41], and continuous gauge transformation (CSGT) [42]. While the methods mentioned above take different approaches to mitigate the gauge problem, all of them are exact and converge to the same chemical shielding values in the limit of very large basis [33]. Of course, the converged values are identical to those obtained with the common origin method when the same extended basis is used in the calculations. But what is more important for practical applications is that these methods produce better results when using somehow modest basis sets.

Modeling Chemical Shieldings The calculation of the diamagnetic part, Eq. [2], presents no serious complications and can be evaluated for any kind of wave function or electronic density. This calculation requires only the computation of one electron integrals of the type 1/r, 1/r3 , and xy/r3 , which are readily available for almost any approximation used to calculate the electron density in most quantum chemical codes. The more complex paramagnetic term, Eq. [3], requires, in principle, the knowledge of all the exited electronic states of the molecule, in which case direct evaluation of Eq. [3] would be immediate. Unfortunately, this is not the case and a great deal of effort is necessary to obtain reliable values of the paramagnetic contribution. It is always important, if possible, to calculate the paramagnetic contribution with the same accuracy as the diamagnetic contribution to achieve the greatest possible gauge invariance of the numerical results. Unfortunately, this may increase the computational complexity beyond practical limits, therefore making the evaluation of the paramagnetic contribution the limiting factor in the calculations of chemical shieldings. Today, there are two major approaches to the calculation of the paramagnetic component of the shielding, which are based on the two predominant types of electronic structure methods in use. Those based in the Hartree–Fock theory and its systematic improvements using perturbation methods to include the electronic correlation [37], which we will label “traditional ab initio methods” and those based on the Density Functional Theory (DFT) [43]. The reader is referred to the recent reviews by Gauss and Stanton [44] and by Wilson [45] for up-todate comprehensive reviews of these two complementary methodologies. The first approach is preferred because it provides a systematic path to improve the calculated results, but this systematic improvement arrives with a considerable increase in the computational cost of the calculations. Therefore, in practice these highly precise calculations of chemical shieldings using traditional ab initio methods are restricted to small molecules of relative minor interest to practitioner chemists. In the fourth column of Table 1, we present a compilation of the best

Modeling NMR Chemical Shifts

Modeling Chemical Shieldings 55

Stdv. Slope Interc. c.c. HF H2 O CH4 CO

19 F 17 O 13 C 13 C 17 O

N2 F2 PN

15 N 19 F 31 P 15 N

H2 S NH3 HCN

33 S 15 N 13 C 15 N

C2 H2 C2 H4 H2 CO N ON

13 C 13 C 13 C 17 O

15 N

15 N 17 O

CO2

13 C 17 O

OF2 H2 CNN

17 O 13 C 15 N 15 N

HCl SO2

35 Cl 33 S 17 O

PH3

31 P

Exp.

Ab initio

419.7 357.6 198.4 2.8 −36.7 −59.6 −192.8 53 −349 752 273.3 82.1 −20.4 117.2 64.5 −4.4 −375 99.5 11.3 200.5 58.8 243.4 −473.1 164.5 −43.4 −149 952 −126 −205 599.9

11.3 1.00 −3.5 0.9993 418.6 337.9 198.9 5.6 −52.9 −58.1 −186.5 86 −341 754.6 270.7 86.3 −13.6 121.8 71.2 4.7 −383.1 100.5 5.3 198.8 63.5 236.4 −465.5 171.9 −31.6 −142.4 962.3 −134.2 −170.4 594

HF 52.8 1.04 −39.4 0.9875 414.3 328.6 195.1 −23.7 −84.3 −110.2 −166.2 −77.3 −483.7 719.9 262.2 71.9 −48.5 115.7 59.8 −6 −436.8 63.9 −32.4 175.4 51.9 223.3 −439.5 164.7 −11.4 −299.2 950.7 −321.9 −284.9 587.9

LDA 36.4 1.09 −48.8 0.9945 416.2 334.8 193.1 −20.3 −87.5 −91.4 −284.2 −73.7 −414.9 733.9 266.3 65.3 −56.7 100.8 40.9 −40 −493.5 87.7 −2.3 179 50 209.7 −667.5 164.5 −61.5 −166.4 959.5 −242.9 −282 583.1

calculations available using traditional ab initio methods for a representative set of small molecules with shieldings spanning over 1500 ppm. The agreement between theory and experiment is quite impressive; the observed standard deviation of 11.3 ppm corresponds to a relative error of 0.7%. Also, it is remarkable that the slope of the correlation is almost exactly one, indicating that for this level of calculation there is no need for any ad hoc scaling of the calculated results. For medium size and large molecules, the computational limitations of the traditional ab initio methods make those based on DFT, with their relatively

B3LYP 30.1 1.05 −43.7 0.9959 411.1 327.7 188.7 −19 −81.1 −92.3 −250.4 −50.7 −431.3 705.2 259.9 69.1 −49.5 106.9 47.2 −24.5 −452.4 81.9 −11.4 173.1 48.9 213.5 −583.1 160 −60.1 −192.8 936.5 −262 −287.8 564.5

KT1 14.3 1.01 −6.2 0.9989 412 330.7 196.4 10.4 −56.1 −55.8 −193.6 46.6 −358.8 741.5 265.9 87.2 −18.6 120.5 64.3 −3 −383.8 106.8 14.2 184.1 65 224.5 −516.7 170.1 −37.5 −128.3 961.3 −149.5 −244.6 600.5

KT2 16.0 1.01 −10.3 0.9987 412.4 329.6 195.2 7.4 −57.1 −59.7 −211 47.1 −361.5 735.7 264.5 86 −19.4 120.4 63.2 −4.7 −379.6 102.1 12.2 177.5 63.7 221.6 −534 167.4 −41.7 −138.4 958.6 −156.8 −251.8 596

KT3 17.4 1.01 −10.2 0.9985 411.3 327.5 192.8 5.8 −55.1 −61.3 −225.4 47.2 −355.3 730.2 261.8 85.9 −17.9 121.1 63.4 −3.5 −370 101.5 13.7 175.2 63.8 220.9 −544.5 165 −42.3 −142 955.5 −134.4 −247.8 591.9

low computational cost and increasingly improved performance, highly competitive. As a result, DFT methods have become the dominant approach for modeling chemical shieldings for medium-to-large molecules. While DFT does not provide a systematic way to improve the results, recently introduced exchange-correlation functionals designed to reproduce chemical shifts (KT1, KT2, and KT3) [46–49] are able to provide results that are quite comparable to those from the best electronic structure calculations. An example of these results is given in Table 1 and Figure 1, where the KTn results are compared with the

Part I

Table 1: Comparison of the calculated chemical shieldings using the KT1, KT2, and KT3 exchange-correlation functionals with those from other electronic structure methods. The calculations were performed using the experimental geometries of the compounds. Data from references [46–49] in ppm, referenced to the bare nucleus (i.e. absolute shieldings).

56 Part I

Chemistry

Part I

1200 1000

Chemical Shieldings (ppm)

800 600 400 200 0 -600

-400

-200

0

200

400

600

-200

800

1000

1200

ab initio KT1 KT2 KT3 Linear (ab initio) Linear (KT1) Linear (KT2) Linear (KT3)

-400 -600 -800 Chemical Shifts (ppm)

Fig. 1. Comparison of the linear correlation between calculated and experimental chemical shieldings for DFT calculations using the KT1, KT2, and KT3 exchange-correlation functional with those from the best traditional ab inito electron correlated calculations. All calculations were performed using the experimental structures of the molecules. Data from references [46–49].

best traditional ab inito (electron correlated calculations), Hartree–Fock and DFT calculations using the LDA (local density approximation) [43] and the hybrid B3LYP [50] exchange-correlation functional. In this set of small molecules, the performance of the KTn functionals is quite impressive. While not achieving the level of the best ab inito, electron correlated calculation, the KTn results show relative errors of only 1.1–0.9% compared with 2% for B3LYP, 2.4% for LDA, and 3.5% for the Hartree–Fock calculations. It should be noted that the scaling of the KTn results is only a 1% correction in the slope of the correlation, while this correction is 5% for B3LYP, 9% for LDA, and 4% for Hartree–Fock. While the results presented above demonstrate that the KTn exchange-correlation functionals outperform most other DFT methods in this set of small molecules, there have not been calculations, using these functionals, reported for larger molecules. In the following, we provide a general overview of the quality of results that can be obtained routinely when using the popular B3LYP [50], MPW1PW91 [51], and OLYP [52] exchange-correlation functionals for shielding calculations in medium size organic molecules from the G2 and G3 standard sets [53,54]. The calculations were done with the popular Gaussian system for molecular modeling [55], using the GIAO [56], CSGT [42], and IGAIM [41] approaches to enforce the gauge invariance and the Dunning’s d95∗∗ basis sets [57]. In all cases, the calculations were performed using the optimized (mp2(full)/6-31g∗ ) geometries that are available for the molecules in the G2 and G3

data sets (http://chemistry.anl.gov/compmat/comptherm .htm). From the molecules selected for the shielding calculations, we have included for the analysis presented here 244 1 H chemical shieldings, 133 13 C chemical shieldings, 18 15 N chemical shieldings, and 26 17 O chemical shieldings. Figures 2–5 and Tables 2–5 depict the correlation between the calculated and their corresponding experimental chemical shifts. In the tables the slope, intercept, and standard deviation of the linear fits are given. Deviations of the values of the slopes from the ideal value of –1 (except for 15 N where it is one) provide an estimate of the systematic errors in the calculations that are usually attributed to the deficiencies in the way that the electron correlation is taken into account. In general, the values of the intercept are less informative because it is widely accepted that there are large uncertainties in determining absolute shieldings of reference compounds. Finally, for practical applications it is important to use the standard deviation to estimate the relative accuracy of the calculations, which gives an indication on how successful the method is in discriminating the resonances of nuclei with similar chemical environments. The methods considered here are able to predict the 1 H chemical shifts with relative accuracies of 2–3%. The slopes, that vary from 10% (GIAO) to 20% (IGAIM and CSGT), are independent of the exchange-correlation functional used. For 13 C the results also are quite satisfactory, providing relative accuracies of 1.4–1.9% and slopes different than −1 by less than 6%. A more

Modeling NMR Chemical Shifts

Modeling Chemical Shieldings 57

Part I

34.00 b3lyp/CSGT b3lyp/GIAO b3lyp/IGAIM

32.00

mpw1pw91/CSGT mpw1pw91/GIAO

Chemical Shieldings (ppm)

30.00

mpw1pw91/IGAIM olyp/CSGT olyp/GIAO

28.00

olyp/IGAIM Linear (b3lyp/CSGT) 26.00

Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) Linear (mpw1pw91/CSGT)

24.00

Linear (mpw1pw91/GIAO) Linear (mpw1pw91/IGAIM) 22.00

Linear (olyp/CSGT) Linear (olyp/GIAO) Linear (olyp/IGAIM)

20.00 0

1

2

3

4

5

6

7

8

9

10

Chemical Shifts (ppm)

Fig. 2. Calculated 1 H chemical shift using different exchange-correlation functionals with the GIAO, CSGT, and IGAIM approaches for selected molecules in the G2 and G3 set of molecules. (See also Plate 2 on page XVIII in the Color Plate Section.)

clear indication of deficiencies of these methods becomes apparent for 15 N and 17 O chemical shieldings, where the standard deviations reach ∼10% and ∼14%, respectively. Also significantly larger deviations in the slopes are observed for these nuclei, up to 20% for 15 N and up to 8% for 17 O. However, in these cases the agreement can be also reduced by well known medium effects on these experimental chemical shifts [6], that are not taken into account in the calculations. The results presented here to illustrate the agreement between calculated and experimental isotropic chemical shifts represent a best case scenario because it

has been recently documented that fortuitous cancelation of errors in the individual tensor components of the calculated chemical shieldings can lead to artificially high agreement in the isotropic chemical shifts [58]. In spite of the success of the chemical shielding calculations demonstrated above, there are several limitations in the current methods that should be considered. These limitations can be broadly divided as those inherent to the approximations used in the calculations and those due to the lack of knowledge of the molecular or crystalline structure.

Table 2: Parameters defining the linear correlation between calculated 1 H chemical shieldings and measured chemical shifts in selected molecules from the G2 and G3 sets. B3LYP

Stdv. Slope Intercept

MPW1PW91

OLYP

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

0.3414 −1.21 36.4

0.1950 −0.93 29.8

0.3402 −1.21 36.4

0.3423 −1.19 35.8

0.2032 −0.92 29.4

0.3411 −1.20 35.8

0.3530 −1.21 36.4

0.2017 −0.94 30.0

0.3520 −1.21 36.4

58 Part I

Chemistry

Part I

250.00 b3lyp/CSGT b3lyp/GIAO b3lyp/IGAIM 200.00 "mpw1pw91/CSGT mpw1pw91/GIAO mpw1pw91/IGAIM Chemical Shieldings (ppm)

150.00 olyp/CSGT olyp/GIAO olyp/IGAIM 100.00 Linear (b3lyp/CSGT) Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) 50.00

Linear ("mpw1pw91/CSGT) Linear (mpw1pw91/GIAO) Linear (mpw1pw91/IGAIM)

0.00 -50

0

50

100

150

200

250

Linear (olyp/CSGT) Linear (olyp/GIAO) Linear (olyp/IGAIM)

-50.00 Chemical Shifts (ppm)

Calculated 13 C chemical shift using different exchange-correlation functionals with the GIAO, CSGT, and IGAIM approaches

Fig. 3. for selected molecules in the G2 and G3 set of molecules. (See also Plate 3 on page XVIII in the Color Plate Section.)

In the first category, the greatest deficiency of the current methods is the neglect of relativistic corrections. This is of minor consequence when dealing with molecules including first or second row atoms but becomes a significant problem when the molecule includes atoms beyond the third row of the periodic table. There is a great deal of literature dealing with relativistic effects on chemical shielding calculations but there are no wellestablished methods that can be used routinely. Moreover, the most common approximations to take into account these effects have not been implemented in the most popular software used for chemical shielding calculations. The calculation of chemical shieldings in molecules

containing heavy atoms remains mostly the realm of very specialized research groups [59], a situation that may change with the recent implementation of shielding calculations using the ZORA [60,61] approach in the ADF (http://www.scm.com/) package. The second methodological challenge in the calculation of NMR chemical shielding is associated with the uncertainties in the molecular or crystalline geometry and the effects that the lattice may have on the NMR chemical shieldings. The first problem is a consequence of the great sensitivity of the chemical shieldings to the molecular geometry, a fact that has been known for some time [62,63]. This sensitivity has been instrumental in using

Table 3: Parameters defining the linear correlation between calculated 13 C chemical shieldings and measured chemical shifts in selected molecules from the G2 and G3 sets. B3LYP CSGT Stdv. 4.1056 Slope −1.02 Intercept 176.7

MPW1PW91

OLYP

GIAO

IGAIM

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

3.6650 −1.02 191.9

4.1091 −1.02 176.7

3.7800 −1.00 176.7

3.3856 −1.00 193.1

3.7829 −1.00 176.7

4.5959 −1.06 182.8

4.1226 −1.07 199.6

4.5994 −1.06 182.8

Modeling NMR Chemical Shifts

Modeling Chemical Shieldings 59

Part I

400 b3lyp CSGT b3lyp GIAO b3lyp IGAIM

300

mpw1pw91 CSGT mpw1pw91 GIAO

Chemical Shieldings (ppm)

mpw1pw91 IGAIM 200 olyp CSGT olyp GIAO olyp IGAIM 100 Linear (b3lyp CSGT) Linear (b3lyp GIAO) Linear (b3lyp IGAIM) 0 0

50

100

150

200

250

300

350

400

450

Linear (mpw1pw91 CSGT) Linear (mpw1pw91 GIAO) Linear (mpw1pw91 IGAIM)

-100 Linear (olyp CSGT) Linear (olyp GIAO) Linear (olyp IGAIM) -200 Chemical Shifts (ppm)

Fig. 4. Calculated 15 N chemical shift using different exchange-correlation functionals with the GIAO, CSGT, and IGAIM approaches for selected molecules in the G2 and G3 set of molecules. (See also Plate 4 on page XIX in the Color Plate Section.)

NMR chemical shift information to elucidate structural problems [64], but at the same time it can lead to unacceptable large errors in the calculated chemical shieldings due to the uncertainties in the position of hydrogen atoms determined by common structural methods such as X-ray. It has become a common practice to optimize the position of the hydrogen atoms, determined by X-ray structures before performing shielding calculations. Unfortunately, the practice of using optimized or partially optimized structures to calculate NMR chemical shieldings leads to

significant questions about possible cancelation of errors between the method used to optimize the geometry and the one used to calculate the NMR chemical shieldings. It is conceivable that good agreement could be achieved due to the use of a optimized geometry that underestimates the interatomic bond distances, in conjunction with a method to calculate NMR chemical shieldings that also underestimates the shieldings or vice versa. Note that almost always the derivative of the chemical shielding with respect to the interatomic bond distances is negative, δσ ≤ 0 [65]. δr

Table 4: Parameters defining the linear correlation between calculated 15 N chemical shieldings and measured chemical shifts in selected molecules from the G2 and G3 sets. B3LYP

Stdv. Slope Intercept

MPW1PW91

OLYP

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

38.4057 0.83 141.4

40.1664 0.81 134.1

38.4026 0.83 141.4

40.3231 0.81 140.5

41.8324 0.80 132.6

40.3200 0.81 140.5

36.6648 0.89 130.5

38.4835 0.87 122.4

36.6598 0.89 130.5

60 Part I

Chemistry

Part I

400.00 b3lyp/CSGT b3lyp/GIAO 300.00 b3lyp/IGAIM mpw1pw91/CSGT 200.00 mpw1pw91/GIAO mpw1pw91/IGAIM Chamical Shielding (ppm)

100.00 olyp/CSGT olyp/GIAO 0.00 0

100

200

300

400

500

600

700

olyp/IGAIM Linear (b3lyp/CSGT)

-100.00 Linear (b3lyp/GIAO) Linear (b3lyp/IGAIM) -200.00 Linear (mpw1pw91/CSGT) Linear (mpw1pw91/GIAO) -300.00 Linear (mpw1pw91/IGAIM) Linear (olyp/CSGT) -400.00 Linear (olyp/GIAO) Linear (olyp/IGAIM) -500.00 Chemical Shifts (ppm)

Fig. 5. Calculated 17 O chemical shift using different exchange-correlation functionals with the GIAO, CSGT, and IGAIM approaches for selected molecules in the G2 and G3 set of molecules. (See also Plate 5 on page XIX in the Color Plate Section.)

This situation has been recently discussed in the case of the calculations of the 15 N chemical shifts of 15 NO2 in aminonitropyridines, aminonitropyrimidines, and their N -oxides [66]. Comprehensive studies, including intermolecular effects (see below) and vibrational corrections are needed to fully understand the interplay between geometry optimization and chemical shielding calculations. The inclusion of intermolecular effects in the calculation of chemical shieldings has attracted a great deal of

attention over the years [8,16,34,67–69] but no “off the shelf” methods are available to take into account these effects in solid or in liquid phase. Unfortunately, in many cases the intermolecular interactions cannot be neglected without losing the quantitative agreement between experimental and calculated results. In these situations, it is necessary to exercise care and complement the standard methods available to calculate chemical shieldings with appropriate ways to take into account the intermolecular effects.

Table 5: Parameters defining the linear correlation between calculated 15 N chemical shieldings and measured chemical shifts in selected molecules from the G2 and G3 sets. B3LYP

Stdv. Slope Intercept

MPW1PW91

OLYP

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

CSGT

GIAO

IGAIM

91.7201 −0.96 219.7

90.3712 −0.94 237.7

91.7200 −0.96 219.8

92.2539 −0.94 221.9

91.4189 −0.92 239.7

92.2538 −0.94 221.9

89.5895 −1.02 227.7

88.0658 −0.99 246.5

89.5893 −1.02 227.7

Modeling NMR Chemical Shifts

Support from the National Science Foundation, The National Institutes of Health, and the Office of Science of the Department of Energy, which over the years have provided funding for the NMR program at Utah, is gratefully acknowledged. The calculations presented here were performed at the CHPC Arches Metacluster, which was partially funded by grant 1 S10 RR17214-01 from the NIH National Center for Research Resources.

References 1. 2. 3. 4. 5. 6. 7. 8.

9.

10. 11. 12. 13.

14. 15. 16. 17. 18.

19.

Purcell EM. Phys. Rev. 1949;76:1262. Anderson HL. Phys. Rev. 1949;76:1460. Ramsey NF. Phys. Rev. 1950;77:567. Ramsey NF. Phys. Rev. 1950;78:699. Grant DM. Chemical shift tensors. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. Wiley & Sons: London, 1996, p 1298. Mason J. Multinuclear NMR. Plenum Press: New York, 1987. Berini I, Luchiant C, Parigu G. Solution NMR of Paramagnetic Molecules. Elsevier: Amsterdam, 2001. Jameson CJ, de Dios AC. Theoretical and physical aspects of nuclear shielding. In: GA Webb (Ed). Specialist Periodical Reports on Nuclear Magnetic Resonance. Royal Society: London, 2004, p 47. Webb GA. Shielding: overview of theoretical methods. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4307. Facelli JC. Shielding calculations. In: DM Grant and RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 2002, p 323. Facelli JC. Concepts Magn. Reson. 2004;20A:42. Facelli JC, De Dios AC. Modeling NMR chemical shifts: gaining insights into structure and environment. ACS Symp. Ser. 1999;732. Facelli JC. Shielding calculations: perturbation methods. In: DM Grant and RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4299. Facelli JC. Shielding tensor calculations. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4327. de Dios AC, Jameson CJ. Annu. Rep. NMR Spectrosc. 1994;29:1. de Dios AC, Oldfield E. Solid State Nucl. Magn. Reson. 1996;6:101. Pulay P, Hinton JF. Shielding theory: GIAO method. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4334. Lazzeretti P, Malagoli M, Zanasi R. Shielding in small molecules. In: DM Grant and RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4318. Buhl M, Kaupp M, Malkina OL, Malkin VG. J. Comput. Chem. 1999;20:91.

20. Cheeseman JR, Trucks GW, Keith TA, Frisch MJ. J. Chem. Phys. 1996;104:5497. 21. Ditchfield R. Mol. Phys. 1974;27:789. 22. Fukui H. Prog. Nucl. Magn. Reson. Spectrosc. 1997;31:317. 23. Geertsen J. Chem. Phys. Lett. 1991;179:479. 24. Hansen AE, Bilde M. Shielding calculations: LORG and SOLO approaches. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4292. 25. Malkin VG, Malkina OL, Salahub DR. Chem. Phys. Lett. 1994;221:91. 26. Malkin VG, Malkina OL, Eriksson LA, Salahub DR. The calculation of NMR and ESR spectroscopy parameters using density functional theory. In: JM Seminario, P. Politzer (Eds). Modern Density Functional Theory. Elsevier Science: Amsterdam, 1995, p 273. 27. Chesnut DB. Annu. Rep. NMR Spectrosc. 1989;21:51. 28. Chesnut DB. Annu. Rep. NMR Spectrosc. 1994;29:71. 29. Chesnut DB. The ab initio computation of nuclear magnetic resonance chemical shielding. In: KB Lipkowitz, DB Boyd (Eds). Reviews in Computational Chemistry. VCH Publishers: New York, 1996, p 245. 30. Kutzelnigg W, Fleischer U, Schindler M. NMR Basic Principles and Progress 1990;23:165. 31. Kutzelnigg W, Fleischer U, van W¨ullen C. Shielding calculations: IGLO Method. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 4284, p 4284. 32. Fleischer U, van W¨ullen C, Kutzelnigg W. NMR chemical shift computation: ab initio. In: P von Ragu´e Schleyer (Ed). Encyclopedia of Computational Chemistry. John Wiley & Sons: London, 1998, p 1827. 33. Schreckenbach G. Theor. Chim. Acta 2002;108:246. 34. Helgaker T, Jaszunski M, Ruud K. Chem. Rev. 1999;99: 293. 35. Mauri F, Pfrommer BG, Louie SG. Phys. Rev. Lett. 1996;77:5300. 36. Sebastiani D, Goward G, Schnell I, Parrinello M. Comput. Phys. Communications 2002;147:707. 37. Simons J, Nichols J. Quantum Mechanics in Chemistry. Oxford University Press: New York, 1997. 38. Kutzelnigg W. Isr. J. Chem. 1980;19:193. 39. Hansen AE, Bilde M. Shielding calculations: LORG and SOLO approaches. In: DM Grant, RK Harris (Eds**). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: London, 1996, p 4292. 40. Facelli JC, Grant DM, Bouman TD, Hansen AE. J. Comput. Chem. 1990;11:32. 41. Keith TA, Bader RFW. Chem. Phys. Lett. 1992;194:1. 42. Keith TA, Bader RFW. Chem. Phys. Lett. 1993;210:223. 43. Parr RG, Yang W. Density-Functional Theory of Atoms and Molecules. Oxford University Press: Oxford, 1989. 44. Gauss J, Stanton JF. Electron-correlated approaches for calculation of chemical shifts. In: I Prigogine, SA Rice (Eds). Advances in Chemical Physics. John Wiley & Sons, Inc., Somerset, NJ08875, 2002, p 355. 45. Wilson PJ. Annu. Rep. NMR Spectrosc. 2003;49:118. 46. Keal TW, Tozer DJ. J. Chem. Phys. 2003;119:3015. 47. Keal TW, Tozer DJ. J. Chem. Phys. 2004;121:5654. 48. Keal TW, Tozer DJ, Helgaker T. Chem. Phys. Lett. 2004;391:374.

Part I

Acknowledgments

References 61

62 Part I

Chemistry

Part I

49. Allen MJ, Keal TW, Tozer DJ. Chem. Phys. Lett. 2003;380:70. 50. Becke AD. J. Chem. Phys. 1993;98:5648. 51. Adamo C, Barone V. Chem. Phys. Lett. 1997;274:242. 52. Handy NC, Cohen AJ. Mol. Phys. 2001;99:403. 53. Curtiss LA, Raghavachari K, Redfern PC, Rassolov V, Pople JA. J. Chem. Phys. 1998;109:7746. 54. Curtiss LA, Raghavachari K, Trucks GW, Pople JA. J. Chem. Phys. 1991;94:7221. 55. Frisch MJ, Trucks GW, Schlegel HB, et al. Gaussian, Inc.: Pittsburgh, PA, 2003. Gaussian 03 Software System for Molecular Modeling, http://www.gaussian.com. 56. Wolinski K, Hinton JF, Pulay P. J. Am. Chem. Soc. 1990;112:8251. 57. Dunning TH, Jr. J. Chem. Phys. 1989;90:1007. 58. Sefzik T, Turco D, Iuliucci RJ, Facelli JC. J. Chem. Phys. 2005;109:1180. 59. Melo JI, Ruiz de Azua MC, Giribet CG, Aucar GA, Romero RH. J. Chem. Phys. 2003;118:471. 60. Autschbach J, Ziegler T. Relativistic computation of NMR shieldings and spin–spin coupling constants. In: DM Grant and RK Harris (Eds). Encyclopedia of Nuclear

61.

62. 63. 64. 65. 66. 67. 68. 69.

Magnetic Resonance, Supplementary Volume. John Wiley & Sons: London, 2002, p 306. Author: Please replace “Supplementary Volume” with correct volume number, if available for reference number 60. Autschbach J. The calculation of NMR parameters in transition metal complexes. In: N Kaltsoyannis, JE McGrady (Eds). Principles and Applications of Density Functional Theory in Inorganic Chemistry. Springer-Verlag GmbH: Berlin, 2004, p 1. Facelli JC, Grant DM. Nature. 1993;365:325. Grant DM, Liu F, Iuliucci RJ, Phung CG, Facelli JC, Alderman DW. Acta Crystallogr. B1995;51:540. Harper JK, Facelli JC, Barich DH, McGeorge G, Mulgrew AE, Grant DM. J. Am. Chem. Soc. 2002;124:10589. Jameson CJ, Osten HJ. Annu. Rep. NMR Spectrosc. 1986;17:1. Anderson KL, Merwin LH, Wilson WS, Facelli JC. Int. J. Mol. Sci. 2002;3:858. Jameson CJ, Jameson AK, Parker H, Cohen SM, Lee C-L. J. Chem. Phys. 1978;68:2861. Besley NA, Hirst JD. J. Am. Chem. Soc. 1999;121:8559. Chalmet S, Ruiz-Lopez MF. J. Chem. Phys. 1999;111:1117.

63

Peter B. Karadakov Department of Chemistry, University of York, Heslington, York YO10 5DD, UK

Introduction The ab initio calculation of NMR shielding constants is one of the “hot” areas in contemporary theoretical chemistry. The reasons for this follow from the extremely high popularity of NMR as an experimental approach and from the fact that, while in many other areas the existing quantum chemical methodology and codes promise much but frequently deliver less than what is required by the practicing chemist, the ab initio theoretical NMR results even at this stage reproduce many experimental measurements to a high degree of accuracy. The aim of the present text is to provide a concise account of the quantum chemical methods for calculating NMR shielding constants, placing the emphasis on the approaches that already are, or are very likely to become available to the research community through implementations in widely available ab initio program packages. Although this choice will inevitably lead to certain omissions, it should be emphasized that there is no shortage of detailed review articles covering a wide range of topics within the area. Central place amongst these is taken by the very comprehensive account of the ab initio methods for the calculation of NMR shielding and indirect spin–spin coupling constants presented by Helgaker et al.[1] Various aspects of the theory and its applications have been discussed by Gauss,[2,3] Jameson,[4] de Dios,[5] Fukui,[6] B¨uhl et al.[7] The Encyclopedia of NMR[8] contains a wealth of information on pre-1996 work. In the next section, we have given a brief overview of the theoretical background. This is followed, in the third section, by an overview of the ab initio program packages capable of evaluating NMR shielding tensors. The concluding section discusses the methods suitable for the calculation of NMR shielding tensors in large molecules.

Overview of the Theoretical Background The electronic Hamiltonian which describes a molecule in the presence of a uniform magnetic field B and the field of fixed nuclear magnetic moments {m J } at positions R J Graham A. Webb (ed.), Modern Magnetic Resonance, 63–70.
C 2008 Springer.

can be written as[9] (in atomic units) Hˆ (B, {m J }) =

 j

1  −1 hˆ j (B, {m J }) + r 2 j =k jk

= Hˆ 1 (B, {m J }) + Hˆ 2 ,

(1)

where all differences from the standard non-relativistic time-independent many electron Hamiltonian are confined to the modified one-electron operator 2 1

−i∇ j + c−1 A(r j − R) hˆ j (B, {m J }) = 2  − Z J r −1 jJ .

(2)

J

In these expressions c is the velocity of light, Z J stands for the charge of nucleus J , and r j J and r jk are the distances between electron j and nucleus J , and between electrons j and k, respectively. The vector potential at electron j A(r j − R) =

 1 (m J × R j J ) r −3 B × (r j − R) + jJ , 2 J (3)

depends on the gauge origin R which can be chosen arbitrarily. Physical intuition suggests that the energy spectrum of Hˆ (B, {m J }), as well as all other electronic properties of the molecule should be independent of the choice of R. A rigorous proof of the gauge invariance of properties calculated with the exact eigenfunctions of Hˆ (B, {m J }) has been provided by Hameka.[10] As shown by Epstein,[11] variational methods employing approximate wavefunctions made of orbitals, such as the Hartree–Fock (HF) approach and multiconfigurational self-consistent field (MCSCF) theory, also produce gauge-invariant results, if the orbitals are expanded in a complete basis. By implication, the same is assumed for results obtained using complete-basis-set non-variational approximate wavefunctions, for example, second-order and fourth-order Møller–Plesset (MP2 and MP4) constructions and coupled–cluster (CC) theory.

Part I

Ab Initio Calculation of NMR Shielding Constants

64 Part I

Chemistry

Part I

Calculations with finite basis sets based on the Hamiltonian (1) produce results which, as a rule, are gauge-dependent. One way to minimize the associated errors is to use larger basis sets which is computationally inefficient, especially for larger molecules. Another possibility is to employ techniques that introduce local gauge origins to define the vector potential of the external magnetic field. Two approaches of this type have become particularly popular: the first of these uses London’s gaugeincluding atomic orbitals (GIAOs), while the second one, individual gauge for localized orbitals (IGLOs) associates an individual gauge origin with each of the localized molecular orbitals (MOs) in a molecular system. Both approaches exist in HF, as well as in post-HF implementations, based on multi-configuration self-consistent field (MCSCF) wavefunctions (MC-IGLO[12] and MCSCFGIAO[13] ), second- and higher-order MP perturbation theory expansions (MP2-GIAO,[14,15] MP3-GIAO[16] and MP4-GIAO[17] ), CC constructions (CCSD-GIAO[18] and CCSD(T)-GIAO[19] ). Density functional theory (DFT) incarnations of the IGLO and GIAO ideas are also available (DFT-IGLO[20,21] and DFT-GIAO[22] ). In the GIAO scheme, which was first applied to NMR shielding calculations by Ditchfield,[23] the MOs for a molecule in a magnetic field are constructed from basis functions that depend on the field explicitly

i χ p (B) = exp − [B × (R p − R)] · r χ p (0), 2c

(4)

where χ p (0) is the usual field-independent AO associated with the atomic centre at R p . Each GIAO has its own local gauge origin placed at its centre. The GIAO approach became very popular after Wolinski et al. developed a highly efficient HF-level implementation[24] incorporating computational techniques similar to those used in the calculation of analytical gradients. In a GIAO calculation, in addition to the two-electron integrals over the original basis functions χ p (0), one needs to evaluate two-electron integrals over an extended basis involving the original basis functions χ p (0) and their products with x, y and z. Integrals of this type are calculated by the analytic gradients routines present in most standard ab initio packages. This has much facilitated the provision of HF and postHF GIAO-based approaches within large ab initio codes: the HF-GIAO implementation in TEXAS90[24] (based on Pulay’s TEXAS[25] ) was soon followed by similar developments in GAUSSIAN94 (the current version of the GAUSSIAN suite is GAUSSIAN03, see Ref. 26), ACES II[27] and TURBOMOLE.[28] The IGLO approach[29] takes an alternative route and assumes that the local gauge origins are associated

with localized MOs, rather than AOs. The wavefunction in the presence of a magnetic field is constructed in terms of field-dependent MOs defined similar to Equation (4),

i φ j (B) = exp − [B × (R j − R)] · r φ j (0), 2c

(5)

where φ j (0) is a localized occupied MO in the zero-field wavefunction and R j is usually chosen as the corresponding orbital centroid, R j = φ j (0)|r|φ j (0) .

(6)

From a computational viewpoint, the main difference between the GIAO and IGLO methods is that IGLO avoids calculating the additional two-electron integrals required by GIAO by means of a completeness insertion. This approximation is well-justified when using MOs, but would not work for AOs (that is, in the GIAO case). As a result, an IGLO calculation can be much cheaper computationally than a corresponding GIAO calculation, especially for large molecules. Despite this advantage, the IGLO scheme is currently less popular than its GIAO counterpart which is related to the lower availability of IGLO-based codes and the absence of MPn-IGLO (especially MP2-IGLO) approaches. The elements of the NMR chemical shielding tensor of a nucleus J can be expressed as the second partial derivative of the molecular energy in the presence of an external magnetic field B with respect to the components of the nuclear magnetic moment m J and B (in the following expression α, β ∈ {x, y, z}): σ J,αβ =

 ∂ 2 E  . ∂m J,α ∂ Bβ B=0,∀m J =0

(7)

σ J is a second-rank tensor, which can be written as the sum of three tensors of ranks zero, one and two, respectively,[30] ⎛

⎞ ⎛ 0 1 0 0 A σ J = σ J, iso ⎝ 0 1 0 ⎠ + ⎝ −σ J,x y A 0 0 1 −σ J,x z ⎛ ⎞ d J,x x d J,x y d J,x z + ⎝ d J,x y d J,yy d J,yz ⎠ d J,x z d J,yz d J,zz

A σ J,x y 0 A σ J,yz

⎞ A σ J,x z A ⎠ σ J,yz 0 (8)

where σ J,iso is the isotropic shielding σ J,iso = 13 (σ J,x x + σ J,yy + σ J,zz ),

(9)

Ab Initio Calculation of NMR Shielding Constants

A = 12 (σ J,αβ − σ J,βα ), σ J,αβ

(10)

and the quantities d J,αβ are given by d J,αβ = 12 (σ J,αβ + σ J,βα − 2σ J,iso ).

(11)

As a rule, shielding tensors are quoted in the principal axis system (PAS), in which the second-rank tensor with elements d J,αβ in Eq. (8) is diagonal (and so is the symmetrized shielding tensor σ SJ = 12 (σ J + σ TJ ), where σ TJ is the transpose of σ J ). It is usual to assume that the PAS shielding tensor is diagonal, which amounts to discarding the first-rank tensor involving the antisymmetry parameters in Eq. (8), ⎛

σ PAS J

1 = σ J,iso ⎝ 0 0

0 1 0

⎞ ⎛ PAS d J,x x 0 0⎠ + ⎝ 0 1 0

0 PAS d J,yy

0

⎞ 0 0 ⎠.

PAS d J,zz

(12) The shielding anisotropy, σ J , and asymmetry, η J , are defined as PAS PAS PAS − 12 (σ J,22 + σ J,33 ) σ J = σ J,11

(13)

The energy expectation value of the Hamiltonian (1) for an arbitrary wavefunction constructed from MOs expanded in a GIAO basis can be represented in terms of the elements of the one- and two-electron density matrices in the GIAO basis, Ppq (B) and  pq,r s (B), in analogy to a well-known expression for the B = 0 case,[32] as E(B, {m J }) =

+

≥ ≥ tities that can be used to characterize the shielding tensor are the span,  J , and the skew, κ j ,[31]

PAS where it is assumed that σ J,33

J =

PAS σ J,33

PAS σ J,22

−

PAS σ J,11 . Other quan-

PAS σ J,11

(15)

and    PAS PAS PAS − σ J,11 σ J,33 . κ J = 3 σ J,iso − σ J,22

(16)

Theoretical values for NMR chemical shifts can be obtained through the expression δ J = (σ J,iso,ref − σ J,iso )/(1 − σ J,iso,ref ) ≈ σ J,iso,ref − σ J,iso when |σ J,iso,ref |  1,

(17)

where σ J,iso,ref is the (absolute) isotropic shielding of nucleus J in a reference molecule.

1   pq,r s (B)χr (B)χs (B) 2 p,q,r,s

−1 × |r12 |χ p (B)χq (B) .

(18)

In view of the fact that the terms dependent on the nuclear magnetic moments are just those involving the matrix elˆ ements of h(B, {m J }), it is convenient to start the derivation of the expression for the elements of the shielding tensor (7) based on Equation (18) with the derivative (see Ref. 17)  ∂E ∂h qp (B, {m J }) = Ppq (B) ∂m J,α ∂m J,α p,q

(19)

ˆ m1 , m2 , . . .)|χ p (B) . where h qp (B, {m J }) = χq (B)|h(B, A second derivation differentiation, with respect to the elements of B, yields σ J,αβ =

(14)

ˆ Ppq (B)χq (B)|h(B, {m J })|χ p (B)

p,q

and    PAS PAS PAS PAS σ J,33 − σ J,11 − σ J,11 η J = σ J,22



 p,q

 ∂ 2 h qp (B, {m J })  Ppq (0)  ∂m J,α ∂ Bβ 

  ∂ Ppq (B)  +  ∂ Bβ  p,q

B=0

B=0,∀m J =0

 ∂h qp (B, {m J })    ∂m J,α

.

B=0,∀m J =0

(20) This equation provides a convenient starting point for the calculation of NMR shielding tensors by means of GIAO approaches utilizing different wavefunction constructions (see e.g., Refs. 17, 33). Detailed expressions for the derivatives of the matrix elements of the one-electron ˆ operator h(B, {m J }) have been reported by Gauss.[33] The derivatives of the one-electron density matrix with respect to the magnetic field are evaluated using linear response theory which, in the case of a HF wavefunction, is well-known as the coupled-perturbed HF (CPHF) approach. An alternative expression for the elements of the NMR shielding tensor can be obtained through direct differentiation of the expectation value of the Hamiltonian Hˆ (B, {m J }) [see Equation (2)] with respect

Part I

A σ J,αβ stands for the antisymmetry parameter

Overview of the Theoretical Background 65

66 Part I

Chemistry

Part I

to a field-dependent wavefunction (B) as follows:

one of the reasons why the DFT-level NMR results are not systematically better than their HF-level counterparts.    ∂  However, it should be pointed out that the results of (B)  Hˆ (B, {m J }) (B) Lee et al.[34] indicate that the use of a local exchange∂m J,α correlation functional, which depends on both the elec     ∂ Hˆ (B, {m })  tron density and the paramagnetic current density in the J   1 = (B)   (B) , DFT-GIAO framework leads only to a slight deshield  ∂m J,α ing effect. There is computational evidence (see Ref. 35) (21)     showing that extension of the basis set in DFT-GIAO  ∂ 2 Hˆ (B, {m })  1 J   calculations performed using GAUSSIAN can lead to (B)  σ J,αβ =  (B)  ∂m J,α ∂ Bβ  poorer-quality NMR shielding constants. This suggests     that although the DFT-GIAO approach is one attractive    ∂(B)  ∂ Hˆ 1 (B, {m J })   way of including correlation effects in shielding calcula. + 2 Re   (B)  tions on large molecules, the results should be treated with   ∂ Bβ  ∂m J,α B=0,∀m J =0 care. The HF-GIAO and DFT-GIAO codes incorporated When dealing with complicated wavefunction construcinto the GAUSSIAN suite are reasonably fast and have tions, it may prove simpler and more straightforward to relatively modest memory and disk space requirements, evaluate the first derivatives of the wavefunction with reespecially when use is made of the direct routines which spect to the elements of the magnetic field, rather than recompute all integrals as required instead of storing them the corresponding derivatives of the elements of the oneafter the initial evaluation. The GAUSSIAN MP2-GIAO electron density matrix required by Equation (20). module is much slower, uses the conventional procedure of storing integrals to disk and, as a result, needs large amounts of temporary disk space. It can treat systems inAb Initio Program Packages Capable of volving more than 255 basis functions (up to 361 on a Calculating NMR Chemical Shielding Tensors 32-bit computer system), although the disk space requireThere are several general-purpose ab initio program pack- ments can become prohibitive: a calculation with a large ages which incorporate modules for calculating NMR locally dense basis set on hexafluorobenzene (90 elecchemical shielding tensors. Apart from these packages, trons, 270 basis functions) produces a read–write file of most of which are commercial, there are other codes with over 52 Gb. This is well beyond the capabilities of 32-bit similar or even better functionality, developed in differ- computer systems, which cannot handle files larger than ent research groups, which are less readily available to 16 Gb. The ACES II[27] package (see http://www.qtp.ufl.edu/ the scientific community. The focus in this section will be on the first group of packages, as their features and Aces2/) can calculate NMR chemical shielding tensors at the HF-GIAO and MP2-GIAO levels of theory. The functionality are much better documented. The GAUSSIAN suite (see http://www.gaussian.com) HF-GIAO and MP2-GIAO codes in this package were is probably the most widely used general-purpose ab initio contributed by J. Gauss and realize the theory presented package. It has been capable of computing NMR chemical in Ref. 15. The memory requirements for an MP2-GIAO shielding tensors since the 1994 version, GAUSSIAN94, calculation in ACES II are about 2n 2 N 2 / h double (eightwhich introduced implementations of the GIAO, contin- byte) words, where n and N denote the numbers of ocuous set of gauge transformations (CSGT), individual cupied and virtual orbitals, respectively, and h stands for gauges for atoms in molecules (IGAIM) and single origin the order of the molecular point group. The MP2-GIAO approaches at the HF and DFT levels of theory. GAUS- code in ACES II uses the conventional procedure of storSIAN94 and later versions of the package can also cal- ing integrals to disk, and the size of each of the three culate HF and DFT-level magnetic susceptibilities with largest temporary files, associated with the perturbations the CSGT, IGAIM and single origin approaches. GAUS- corresponding to the elements of B, is about 1.5M 4 /(4h) SIAN98 extended the NMR-related capabilities of the double words, where M stands for the number of basis package through the addition of an MP2-GIAO module. functions. The program can treat the perturbations caused In addition to this, the current version, GAUSSIAN03,[26] by the three magnetic field components separately, which allows the calculation of indirect spin–spin coupling con- can lead to a substantial reduction in the disk space restants using HF and DFT wavefunctions, as well as quirements at the expense of some loss of efficiency. The the calculation of NMR properties in the presence of a maximum number of basis functions in a MP2-GIAO calculation performed within ACES II is 255. solvent. Turbomole[28] (see http://www.ipc.uni-karlsruhe.de/ The DFT functionals in the GAUSSIAN suite do not include magnetic field dependent terms, which may be tch/tch1/index.de.html) is another ab initio package that

Ab Initio Calculation of NMR Shielding Constants

Ab Initio Calculation of NMR Chemical Shielding Tensors for Large Molecules 67

The RPAC molecular properties package developed mainly by Bouman and Hansen[44] (e-mail address for enquiries concerning RPAC: [email protected]) is a post-SCF code that can work with the US version of GAMESS[45] or with GAUSSIAN. It calculates electronic excitation and response properties employing first-order (random-phase approximation/coupled HF) and secondorder (SOPPA/second-order LORG (SOLO)) linear response theory. In addition to a number of other electronic ground state response properties, RPAC can calculate NMR shielding tensors with the inclusion of electron correlation effects in the case of the SOLO approach.

Ab Initio Calculation of NMR Chemical Shielding Tensors for Large Molecules In recent years, the calculation of NMR chemical shielding constants of large molecules, using a variety of theoretical methods, has become largely a routine task. However, the computational cost scales with at least the fourth power of the number of basis functions and the calculations can quickly become prohibitively expensive in terms of the computational resources required, particularly when electron correlation is included. The demand on resources can be reduced by taking into account the fact that nuclear shielding is predominantly governed by local effects. This idea is behind the locally dense basis set (LDBS) technique suggested by Chesnut and Moore[46] in which the atoms of interest are described using larger basis sets, while a smaller basis set is employed for the rest of the molecule. All ab initio packages discussed in the previous section allow straightforward specification of different basis set for different atoms, which has made the use of LDBS constructions the most popular way of reducing the computational effort associated with NMR shielding tensor calculations. A now classic example of the advantage of LDBS involves the study of a cluster of 17 H2 O molecules,[47] in which the results obtained using a 6-311G(d,p) basis on the central H2 O molecule and its two-nearest hydrogen-bonded partners were found to be virtually identical to those obtained using the larger basis throughout, but were produced in just one-sixth of the time required for the larger calculation. The treatment of very large molecular systems may require more drastic approximations. In one approach, advanced by de Dios et al, distant parts of the molecular system, solvent molecules, etc., are modeled by partial point charges, creating a classical electrostatic field. In spite of its simplicity, this idea has been demonstrated to work very well for the 13 C shielding tensors in crystals of l-tyrosine and l-threonine.[48] This idea was taken further by Cui and Karplus,[49] who suggested a general approach to chemical shielding calculations within the quantum mechanics/molecular mechanics (QM/MM) framework,

Part I

can computes NMR chemical shieldings within the GIAO ansatz at the HF and MP2 levels of theory. The MPSHIFT module of Turbomole implements the direct MP2-GIAO approach developed by Kollwitz and Gauss[36] , which allows calculations on large molecules using machines with limited amounts of memory and disk space. For example, the largest MP2-GIAO calculation reported by Kollwitz and Gauss,[36] on the anthracenium cation (94 electrons, 288 basis functions), was carried out on a workstation with just 128 Mb RAM and 2 Gb of scratch disk space. The most recent version of this code makes full use of molecular point group symmetry, including non-Abelian point groups,[37] and has been used in calculations on highly symmetric molecules involving more than 600 basis functions.[37] DALTON[38] (see http://www.kjemi.uio.no/software/ dalton/dalton.html) is an ab initio package that can calculate a very wide range of molecular properties at different levels of theory. Gauge origin independent nuclear shieldings and magnetizabilities can be obtained through the use of GIAOs, or through the continuous transformation of the origin of the current density (CTOCD) approach,[39] in its CTOCD-DZ form.[40] Dalton allows the use of GIAOs with HF and MCSCF wavefunctions, while the CTOCDDZ technique can be combined with the second order polarization propagator approximation (SOPPA), and with SOPPA with CCSD amplitudes. Dalton can also calculate indirect spin–spin coupling constants using the triplet linear response function. deMon (Density of Montr´eal)[41] (see http://www. demon-software.com) is a DFT package that incorporates the deMon-NMR code written by Malkin et al. deMon-NMR can calculate chemical shifts, coupling constants and electron pair resonance quantities. It implements the sum-over-states-density functional perturbation theory (SOS-DFPT) approach[21,42] in combination with the IGLO choice of gauge origins and in many cases produce more accurate results than the standard DFT-GIAO and DFT-IGLO methods, and the more common HF-GIAO and HF-IGLO approaches. The reasonable accuracy and lower computational costs of the SOSDPFT-IGLO method makes it an attractive alternative to MP2-GIAO in shielding calculations on larger molecular systems (for example, biosystems or transition metal complexes), which require inclusion of correlation effects. DGauss (see http://www.cachesoftware.com/cache/ dgauss/index.shtml) is another DFT code for calculating NMR shielding constants, which is also largely based on theory developed by Malkin et al.[20,43] DGauss uses DFT in combination with the IGLO and LORG (localized orbital/local origin) techniques. The Cambridge analytic derivatives package (CADPAC) (see http://www-theor.ch.cam.ac.uk/software/ cadpac.html) incorporates HF-LORG and DFT-LORG modules for calculating NMR shielding constants.

68 Part I

Chemistry

Part I

in which the most important region of a molecule is described by a QM wavefunction, while its surroundings are described using MM. These authors showed that the MM atoms, which polarize the wavefunction of the QM region as point charges, make a two-fold contribution to the elements of the NMR shielding tensor in Equation (20): firstly, the charges on the MM atoms modify the one-electron density matrix Ppq (0) and, secondly, these atoms influence the derivatives of the density matrix, ∂ Ppq (B)/∂ Bβ |B=0 . As a consequence, the correct treatment of the effects due to the charges on the MM atoms requires a modified variant of linear response theory. The results of Cui and Karplus show that their QM/MM approach, with an appropriate QM/MM partitioning, allows a good description of the environmental effects on the chemical shifts. The typical errors relative to full QM calculations within the same basis set were found to be about 1–2 ppm for distances between QM and MM atoms ˚ At shorter distances, such as those corgreater than 2.5 A. responding to hydrogen bonds, the deviations from the full QM results become more significant as the QM/MM model does not account for the Pauli repulsion and the magnetic susceptibility of the environment. One way to correct for this is to extend the QM region so that it includes all atoms that interact directly with the atom of interest. The fact that it is possible to perform higher level calculations on the nuclei of interest in a large molecule, as long as their local chemical environment is adequately

described, without having to use the same level of theory for the whole system, is fully exploited in the ONIOMNMR approach,[50] which is quickly gaining popularity (for example, see Refs. 51, 52). The ONIOM approach involves splitting the system into two or more layers that can be described using different levels of theory and/or basis sets (for an in-depth discussion of the theory behind the ONIOM, our own n-layer integrated molecular orbital and molecular mechanics, approach see, e.g., Refs. 53, 54). A two-layer ONIOM-NMR construction requires the performance of three NMR calculations in order to obtain the shielding of each of the nuclei of interest. First, the shieldings for the whole system must be calculated at the selected lower level of theory, and then those for the molecular fragment surrounding the nucleus of interest have to be evaluated at both the selected higher and lower levels of theory. The expression for the elements of the NMR chemical shielding tensor for nucleus J in the two-layer ONIOM-NMR approach is given by σ J,αβ [ONIOM2(H-GIAO : L-GIAO)] = σ J,αβ (H-GIAO, model) + σ J,αβ (L-GIAO, real) − σ J,αβ (L-GIAO, model), (22) where ‘H’ and ‘L’ represent the higher and lower levels of theory, respectively, both of which would normally

Fig. 1. Application of the ONIOM2(MP2-GIAO:HF-GIAO) approach to the water dimer (MP2//6-31G∗∗ geometry of Cs symmetry). Abbreviations in the shielding descriptions: HE = HF-GIAO/6-31G∗∗ , MP2 = MP2-GIAO/6-31G∗∗ , MP2:HF = ONIOM2(MP2-GIAO/6-31G∗∗ :HF-GIAO/6-31G∗∗ ), Ml = model system 1 (left water molecule), M2 = model system 2 (right water molecule), R = real system (the whole dimer). All shieldings in ppm.

Ab Initio Calculation of NMR Shielding Constants

References 1. Helgaker T, Jaszunski ´ M, Ruud K, Chem. Rev. 1999;99:293. 2. Gauss J, Ber. Bunsenges. Phys. Chem. 1995;99:1001. 3. Gauss J In: Grotendorst J (Ed). Modern Methods and Algorithms of Quantum Chemistry, Proceedings, Second Edition, NIC Series, Vol. 3, John von Neumann Institute for Computing, J¨ulich, 2000, p. 541. 4. Jameson CJ, Ann. Rev. Phys. Chem. 1996;47:135. 5. de Dios AC, Progr. Nucl. Magn. Res. Spectr. 1996;26:229. 6. Fukui H, Progr: Nucl. Magn. Res. Spectr. 1997;31:317. 7. B¨uhl M, Kaupp M, Malkina OL, Malkin VG, J. Comput. Chem. 1999;20:21. 8. Grant DM, Harris RK (Eds). Encyclopedia of NMR, Wiley: NY, 1996. 9. Ditchfield R, J. Chem. Phys. 1972;56:5688. 10. Hameka HF, Rev. Mod. Phys. 1962;34:87. 11. Epstein ST, J. Chem. Phys. 1964;42:2897. 12. van W¨ullen C, Kutzelnigg W, Chem. Phys. Lett. 1993; 205:563. 13. Ruud K, Helgaker T, Kobayashi R, Jørgensen P, Bak K, Jensen H, J. Chem. Phys. 1994;100:8178. 14. Vauthier EC, Comenau M, Odiot S, Eliszar S, Can. J. Chem. 1988;66:1781.

15. Gauss J, Chem. Phys. Lett. 1992;191:614. 16. Fukui H, Baba T, Matsuda H, Miura K, J. Chem. Phys. 1994;100:6608. 17. Gauss J, Chem. Phys. Lett. 1994;229:198. 18. Gauss J, Stanton J, J. Chem. Phys. 1995;103:3561. 19. Gauss J. Stanton J, J. Chem. Phys. 1996;104:2574. 20. Malkin VG, Malkina OL, Salahub DR, Chem. Phys. Lett. 1993;204:87. 21. Malkin VG, Malkina OL, Casida ME, Salahub DR, J. Am. Chem. Soc. 1994;116:5898. 22. Cheeseman J, Trucks GW, Keith TA, Frisch M, J. Chem. Phys. 1996;104:5497. 23. Ditchfield R, Mol. Phys. 1974;27:789. 24. Wolinski K, Hinton JF, Pulay P, J. Am. Chem. Soc. 1990;112:8251. 25. Pulay P, Theor. Chim. Acta. 1979;50:299. 26. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Montgomery JA, Jr, Vreven T, Kudin KN, Burant JC, Millam JM, Iyengar SS, Tomasi J, Barone V, Mennucci B, Cossi M, Scalmani G, Rega N, Petersson GA, Nakatsuji H, Hada M, Ehara M, Toyota K, Fukuda R, Hasegawa J, Ishida M, Nakajima T, Honda Y, Kitao O, Nakai H, Klene M, Li X, Knox JE, Hratchian HP, Cross JB, Bakken V, Adamo C, Jaramillo J, Gomperts R, Stratmann RE, Yazyev O, Austin AJ, Cammi R, Pomelli C, Ochterski JW, Ayala PY, Morokuma K, Voth GA, Salvador E, Dannenberg JJ, Zakrzewski VG, Dapprich S, Daniels AD, Strain MC, Farkas O, Malick DK, Rabuck AD, Raghavachari K, Foresman JB, Ortiz JV, Cui Q, Baboul AG, Clifford S, Cioslowski J, Stefanov BB, Liu G, Liashenko A, Piskorz P, Komaromi I, Martin Rl, Fox DJ, Keith T, Al-Laham MA, Peng CY, Nanayakkara A, Challacombe M, Gill PMW, Johnson B, Chen W, Wong MW, Gonzalez C, Pople JA, Gaussian 03, Revision C.02, Gaussian, Inc., Wallingford, CT, 2004. 27. Stanton JF, Gauss J, Watts JD, Nooijen M, Oliphant N, Perera SA, Szalay PG, Lauderdale WJ, Kucharski SA, Gwaltney SR, Beck S, Balkov´a A, Bernholdt DE, Baeck KK, Rozyczko P, Sekino H, Hober C, Bartlett RJ, ACES II, An Ab Initio Program System, Release 2.0, Quantum Theory Project, Departments of Chemistry and Physics, University of Florida, Gainesville, 1997. 28. Ahlrichs R, B¨ar M, Baron H-P, Bauernschmitt R, B¨ocker S, Ehrig M, Eichkorn K, Elliott S, Furche F, Haase F, H¨aser M, Horn H, Huber C, Huniar U, Kollwitz CKM, Ochsenfeld C, ¨ Ohm H, Sch¨afer A, Schneider U, Treutler O, von Arnim M, Weigend F, Weis P, Weiss H, TURBOMOLE, Program Package for ab initio Electronic Structure Calculations, Version 5.1, Quantum Chemistry Group, University of Karlsruhe, Karlsruhe, 1999. 29. Kutzelnigg W, Isr. J. Chem. 1980;19:193. 30. Smith SA, Palke WE, Gerig JT, Cont. Magn. Reson. 1992;4:107. 31. Mason J, Solid State Nut. Magn. Reson. 1993;2:285. 32. McWeeny R, Methods of Molecular Quantum Mechanics, London: Academic Press, 1992. 33. Gauss J, J. Chem. Phys. 1993;99:3629. 34. Lee AM, Handy AC, Colwell SM, J. Chem. Phys. 1995;103:10095. 35. Karadakov PB, Webb GA, England JE, ACS Symp. Series, 1999;732:115. 36. Kollwitz M, Gauss J, Chem. Phys. Lett. 1996;260:639.

Part I

be GIAO-based approaches. The ‘model’ system corresponds to the inner layer formed by nucleus J and its local environment, and the ‘real’ system represents the entire molecule. One important advantage of the ONIOM-NMR approach is that the expression (21) can be evaluated using any ab initio package that implements the ‘H-GIAO’ and ‘L-GIAO’ methods, without any need for additional programing. The choice of suitable molecular fragments is a very important aspect of the use of the ONIOM-NMR approach. In most cases, when calculating the shielding of a particular nucleus, inclusion of its nearest neighbours in the “model” system is sufficient to achieve an adequate description of its local environment. Usually, the definition of the “model” system requires breaking of chemical bonds, and within the ONIOM approach the resulting “free valencies” are saturated through the addition of terminal hydrogen atoms. The ONIOM-NMR approach works particularly well for hydrogen-bonded systems, as illustrated by Figure 1, which shows the results of its application to the water dimer.[50] Due to the local nature of the NMR shielding tensor, local-correlation treatments (see e.g. ref. 55) should be a suitable way of reducing the computational effort associated with post-HF shielding calculations on large molecules. Gauss and Werner have developed a LMP2GIAO scheme,[56] which has been shown to be comparable in accuracy to the standard MP2-GIAO approach. However, the limited data available at the moment does not allow an estimate of the potential computational savings associated with the use of this local-correlation treatment.

References 69

70 Part I

Chemistry

Part I

37. Kollwitz M, H¨aser M, Gauss J, J. Chem. Phys. 1998; 108:8295. 38. Angeli C, Bak KL, Bakken V, Christiansen O, Cimiraglia R, Coriani S, Dahle P, Dalskov EK, Enevoldsen T, Fernandez B, H¨attig C, Hald K, Halkier A, Heiberg H, Helgaker T, Hettema H, Jensen HJA, Jonsson D, Jørgensen P, Kirpekar S, Klopper W, Kobayashi R, Koch H, Ligabue A, Lutnæs OB, Mikkelsen KV, Norman P, Olsen J, Packer MJ, Pedersen TB, Rinkevicius Z, Rudberg E, Ruden TA, Ruud K, Satek P, de Meras AS, Saue T, Sauer SPA, Schimmelpfennig B, Sylvester-Hvid KO, ˚ Taylor PR, Vahtras O, Wilson DJ, Agren H, DALTON Release 2 Program Manual, 2005. 39. Lazzeretti P, Malagoli M, Zanasi R, Chem. Phys. Lett. 1994;220:299. 40. Ligabue A, Sauer SPA, Lazzeretti P, J. Chem. Phys. 2003;118:683O. 41. K¨oster AM, Calaminici P, Escalante S, Flores-Moreno R, Goursot A, Patchkovskii S, Reveles JU, Salahub DR, Vela A, Heine T, The deMon User’s Guide, Version 1.0.3, 2003– 2004, deMon Software, 2004. 42. Malkin VG, Malkina OL, Eriksson LA, Salahub DR, In: Seminario J, Politzer P (Eds), Theoretical and Computational Chemistry, vol. 2, Modern Density Functional Theory: A Tool For Chemistry, Amsterdam: Elsevier, 1995, p. 273.

43. Malkin VG, Zhidomirov GM, Zh. Strukt. Khim. 1988;29:32. 44. Bouman TD, Hansen AE, Chem. Phys. Lett. 1990;175:292. 45. Schmidt MW, Baldridge KK, Boatz JA, Elbert ST, Gordon MS, Jensen JH, Koseki S, Matsunaga N, Nguyen KA, Su SJ, Windus Tl, Dupuis M, Montgomery JA, J. Comput. Chem. 1993;14:1347. 46. Chesnut DB, Moore KD, J. Comput. Chem. 1989;10:648. 47. Hinton JE, Guthrie P, Pulay P, Wolinski k, J. Am. Chem. Soc. 1992;114:1604. 48. de Dios AC, Laws DD, Oldfield E, J. Am. Chem. Soc. 1994;116:7784. 49. Cui Q, Karplus M, J. Phys. Chem. B. 2000;1O4:3721. 50. Karadakov PB, Morokuma K, Chem. Phys. Lett. 2000;317:589. 51. Molina PA, Jensen JH, J. Phys. Chem. B. 2003;1O7:6226. 52. Markwick PRL, Sattler M, J. Am. Chem. Soc. 2004;126:11424. 53. Svensson M, Humbel S, Froese RDJ, Matsubara T, Sieber S, Morokuma K, J. Phys. Chem. 1996;100:19357. 54. Humbel S, Sieber S, Morokuma K, J. Chem. Phys. 1996;105:1959. 55. Saebø S, Pulay P, Ann. Rev. Phys. Chem. 1993;44:213. 56. Gauss J, Werner H-J, Phys. Chem. Chem. Phys. 2000;2:2083.

71

Ulrich Sternberg1 , Raiker Witter1 , and Anne S. Ulrich1,2 1 Institute

of Biological Interfaces, Forschungszentrum Karlsruhe, POB 3640 D-76021 Karlsruhe, Germany; and 2 Institute of Organic Chemistry, University of Karlsruhe, Fritz-Haber-Weg 6, D-76131 Karlsruhe, Germany

Introduction The NMR chemical shift is available from practically every conventional NMR experiment. In contrast to X-ray diffraction it is mainly caused by the density distribution of the valence electrons, hence it contains genuine information about the valence structure of the molecular system. High-resolution solid-state investigations on crystalline systems revealed a considerable dependence of the chemical shift on the 3D arrangement of the atoms and on their packing within the unit cell [1]. In many cases, an asymmetric content of the unit cell could be deduced from NMR line splittings. The point group symmetry of the molecule under study is frequently reflected within the NMR spectra and especially within the chemical shift tensors [2]. It was demonstrated by Taulelle [3] that even the complete space group could be deduced from NMR results. The success of diffraction methods originates from the direct dependence of the measured intensities on the coordinate of the scattering particles. NMR structure analysis is more complex, but the better we understand how chemical shifts change with the 3D arrangement of atoms, the more reliably we can construct molecular models from our experiments. In the case of chemical shifts, this connection is far from simple and requires quantum chemical computations. In some cases, the molecular packing could be directly deduced by calculating chemical shifts induced by intermolecular interactions [4]. These effects become dominant with increasing polarity of the lattice. In ab initio chemical shift calculations molecules are embedded into point charge lattices to simulate the crystallographic surrounding [5]. These computations are highly demanding and for many solid state problems prohibitive. The quest for simple empirical and semi-empirical approaches to structure analysis using chemical shifts has been recently reviewed by Sternberg et al. [6]. The ultimate goal will be to perform an accurate chemical shift calculation coupled to a 3D structure refinement. In many cases, ab initio CS calculations are used to supplement NMR investigations to extract aspects of the spatial arrangement, but for complete structure refinement a number of severe problems still have to be solved. Chemical shift calculations often require extended basis sets or correlation Graham A. Webb (ed.), Modern Magnetic Resonance, 71–78.
C 2008 Springer.

corrections that may easily lead to multiples of the original computational time. If refinements are to be performed in addition to the chemical shift calculation, their derivatives with respect to the molecular coordinates need to be evaluated. This is again much more demanding than the chemical shift calculations alone. We believe that the success story of liquid state NMR in protein structure elucidation is going to continue in the solid state once chemical shifts can be successfully exploited. In many cases, the chemical shift will be only one parameter amongst many others that can yield constraints for structure calculation. In solution, we can measure Nuclear Overhauser Enhancements (NOEs), J-couplings, and residual dipolar couplings, and for these parameters there exist well-established relationships to aspects of the molecular structure. In solid-state investigations, we face the situation that up to now there is no general protocol for structure calculations from NMR data. Direct dipolar couplings are often used as distance constraints, provided that the interaction between two spins can be singled out. Likewise, anisotropic chemical shift information is highly useful to determine orientational constraints in macroscopically oriented samples such as membranes or fibers. For solid-state NMR protein structure analysis there exist a number of excellent reviews covering most aspects of chemical shift approaches [7–10]. To solve crystal structures by NMR or to at least augment diffraction studies, the following computational requirements have to be fulfilled: (i) the energy of a unit cell including the influence of the lattice has to be calculated, (ii) extremely fast methods are required for calculating chemical shifts, and (iii) the chemical shift derivatives with respect to the atomic coordinates need to be evaluated. The following paragraphs will give a short introduction into the appropriate methods, before several applications will be presented.

Computational Methods Bond Polarization Theory for Chemical Shifts In principle, chemical shielding calculations are performed by investigating the perturbation of a molecular wave function due to the presence of an external field B Z

Part I

Crystal Structure Refinement Using Chemical Shifts

72 Part I

Chemistry

Part I

and magnetic moments caused by the nuclei. Within the context of the Hartree-Fock (HF) theory we have to solve perturbed HF equations. Even if more powerful computers will be available in the future and ab initio nuclear shielding calculations are sped up further [11], the calculations will still be very computationally demanding. In many cases, the HF level is not sufficient for reliable results. Correlated methods, such as coupled clusters, have to be applied. However, for MD calculations of crystal lattices these methods are far too time consuming. Therefore, we have to apply a highly efficient method that concentrates on the key points of quantum mechanical chemical shift calculations. The first key point is that chemical shift is a local quantity, which depends mainly on the first sphere of chemical bonds around the nucleus under consideration. Therefore, bond orbitals are best suited for our task. Bond orbitals can be constructed from the geometry of the system, provided that their occupation numbers and polarity parameters are known. Secondly, we account for polarization by invoking anti-bonds as excited configurations. Instead of parametrizing the Hamiltonian matrix like other semi-empirical methods the expectation value for the chemical shift is parametrized directly. The complete wave function is not needed, because we are only interested in the change of a local quantity and rather than its absolute value. Finally, we arrive at an expression were two (in the tensorial case 6) linear empirical parameters per bond type have to be determined. Within the bond polarization theory BPT approach [12, 13] the chemical shift tensor is expressed by i∈A  





   αβ αβ i i 0 |δˆαβ |0 = Dαα n i δi + n i2 Ai

D

ββ i



αβ

         . × χAi Vˆ  χAi − χBi Vˆ  χBi

(1)

The matrix elements Dαα’ describe the coordinate transformation from the bond orbital frame to the reference frame. The first sum runs over all bond contributions of atom A. The bond polarization matrix elements are given (in atomic units) by 

     charges   ˆ χλ V χλ = h 2k φkλ (r ) x

k

Qx φ λ (r )dr 3 , × |Rx − r | k

(2)

with the charges Q x at position Rx , the Slater type orbitals [14] φkλ (r ), and the bond coefficients h k . The first sum runs over all polarizing atomic charges. The bond tensor

αβ

αβ

increments δi and polarization tensor parameters Ai (in ppm/Hartree) are obtained by calibration procedures, [15]. A collection of crystal structures and single crystal chemical shift measurements has been used to establish a set of linear equations for the parameters [16][17], and in some cases we also included ab initio results. The correlation coefficient obtained from the parameter calibration is R = 0.994 with a standard deviation αβ of SD = 7.2 ppm. Once the bond increments δi and αβ polarization parametersAi are determined, only the matrix elements χλi |Vˆ |χλi and the occupation numbers n i (from valences [18]), have to be calculated. Introducing point charges in the expression for the potential Vˆ leads to compact analytic expressions for the integrals, hence calculations within the BPT approach are highly efficient. In Equation (1) there are two sums, which run over all bond contributions of the atom under consideration and over all polarizing charges of the potential Vˆ . If the charge distribution is known, the computational cost for a chemical shift calculation is thus proportional to the number of atoms N.

BPT Calculation of Atomic Charges As can be seen from Equation (2), accurate atomic charges, Q x , are also prerequisite of BPT chemical shift calculations. The chemical shifts in this theory are proportional to bond polarization integrals that account for the change of the chemical shift caused by surrounding charge distributions. Since semi-empirical polarization parameters are introduced into the chemical shift calculations, the absolute values of the charges are not of concern. αβ The polarization parameters Ai , on the other hand, will depend on the type of ab initio calculation and on the procedure for the population analysis. The atomic charges can be calculated within the BPT approach in a manner analogous to Equation (1) [19]:     i∈A
  Qx  i q χ QA = n i qi + n i2 Ai χAi  |Rx − r |  A i      Qx  i χ − χBi  . (3) |Rx − r |  B Overlap contributions are omitted when calculating the bond polarization integrals. By investigating the charge equations, it is obvious that the charge on atom A, Q A , has to be estimated from all other charges Q x . By taking the Q x as factors out of the integrals, we end up with a system of linear equations for the Q x and Q A with the sum over n i qi as inhomogeneities. The parameters qi and q Ai have been calibrated against atomic charges of a set of 175 structures consisting of H, C, N, O, F, Si, P, S,

Crystal Structure Refinement Using Chemical Shifts

Molecular Force Fields and Chemical Shift Pseudo-Forces For a complete solid-state NMR structure determination the most desirable approach would be to calculate the energy, the chemical shifts and their derivatives using ab initio methods. Even on fast computers this does not seem to be feasible for systems with much more than 10 heavy atoms. With the advent of DFT calculations the situation improved, but a breakthrough in chemical shift calculations was not achieved. One of the most promising developments is the combination of quantum mechanical ab initio methods with molecular mechanics calculations (QM/MM). Cui and Karplus [21] combined the empirical CHARMM force field [22] with HF and DFT calculations. In this framework, the chemical shielding calculations can be performed on the GIAO-DFT, -HF or -MP2 level in the QM part of the system under the influence of a larger MM surrounding. The electrostatic perturbations of all relevant matrix elements are treated by a point charge distribution from the MM part of the system, and their influence on the chemical shielding can be studied. Even with the limitations on the size of the QM part this method will be of great value, especially in the treatment of reaction centers in large molecules. Traditional methods for the treatment of large molecular systems using NMR constraints are molecular mechanics force fields like DYANA [23]. It was demonstrated that such empirical force field reproduce the 3D structures rather well and can compete in this aspect with elaborate ab initio calculations. The limitations of molecular mechanics in system size are due to the calculation of the intermolecular energy terms, which scale with the second power of the number of atoms. Hundreds or even thousands of atoms are no real problem for molecular mechanics force field calculations. The most popular method for the search of global minima in NMR force field calculations is to run MD simulations at elevated temperatures, to surmount most conformational barriers and populate extended areas of the configurational space. A larger number of coordinate snapshots is then cooled down in the so-called simulated annealing procedure, or the energy minimum is determined by geometry optimization. The latter method has the advantage that we can weight every structure by its minimum energy. A combi-

nation of simulated annealing and geometry optimization is preferred in some investigations to avoid side minima. The problem of all previous molecular mechanics force field methods is that electronic properties cannot be calculated without wave functions or electron distributions. Even atomic charges are mostly fixed parameters of the force field, and polarizations of the electron distribution are excluded. One possibility to overcome the limitations of the traditional molecular mechanics is the combination of the BPT with a force field. Within the COSMOS force field [24, 25] two-center bond orbitals are constructed for every bond defined in the force field. If the hybridization coefficients and the bond occupation numbers n i are calculated from the geometry, only the bond polarities are left as free parameters. Within the framework of the BPT, bond polarities or atomic charges can be readily calculated. Therefore, this force field works with charges that depend in the same way on the 3D structure of the molecular systems as the ab initio charge values that were used in the parametrization. For σ -bonds the occupation numbers n i are set to two, and for conjugated πbonds the value is estimated from the bond distance[25]. Using the COSMOS force field it is possible to divide the molecular system into an MM part and a BPT-QM part, which considers only the polarizations caused by the point charges of the MM part. BPT calculations are very fast and scale with the number of atoms multiplied by the number of bonds. Nevertheless, the BPT atomic charge calculation is the most time consuming step in the force field cycle, hence cutoffs or smaller QM parts help to run efficient simulations. Since all polarizations can be included into the Coulomb energy, the COSMOS force field can be used to study interactions of highly charged systems as for instance ions and peptides [25]. To apply molecular mechanics calculations to crystal structures, the force field has to represent the influence of the crystal lattice in an adequate way. This is achieved by surrounding a central unit cell by one or two shells of translationally created images of itself. The number of shells depends on the electrostatic cutoff radius in relation to the size of the unit cell. In most molecular crystal structures some of the bonds span the borders of the unit cell, therefore, one also has to account for these periodic intramolecular contributions besides the intermolecular energy. To perform realistic crystal structure simulations, the force field has to maintain strict lattice periodicity throughout the calculations:  r ) = F(  r + i a + j b + k c), {i, j, k} = 0, ±1, ±2. F( (4) For every part of a molecule that is not within the unit cell, a code is stored to update the positions of the atoms, forces, and charges from the central unit cell (analogous to

Part I

Cl, and Zn atoms [20]. The calculations were performed using the 6-31G(d, p) basis set, and the atomic charges are obtained by a natural population analysis (NPA) of the ab initio charge distributions. BPT and ab initio charges of small molecules correlate regularly with R = 0.996. For details of the parametrization and the formalism see Witter [20].

Computational Methods 73

74 Part I

Chemistry

Part I

Equation 4). Therefore, all energies and forces are onlycalculated only for the central unit cell, but under the influence of one or two shells of neighboring cells. Space group symmetry was not enforced, but the chemical shift constraints conserve the symmetry relations within the unit cell if two or more sites display the same chemical shift value. Given the COSMOS force filed, NMR parameters are now introduced as constraints in the energy calculations. When searching for the most probable structure of a polyatomic molecule like a peptide, one has to find an energy minimum on a hypersurface possessing a vast multitude of minima. Every experimental constraint will limit the free configurational space and drive the system toward the genuine structure. Even with a large number of constraints we have to search for a global minimum on a multidimensional energy hypersurface. It is important to realize that the minimum structure will not be the most probable structure, because this will depend on Gibbs free energy, G, containing not only the enthalpy but also the entropy. Broad shallow minima may thus be preferred because of the entropic term. Our NMR constraints, on the other hand, will contain an average over the most probable structures in solution or the solid state. Therefore, in most cases the experimental constraints will drive our molecular system energetically uphill.

BPT Pseudo-Forces In order to obtain energetic corrections, the contribution of the bonds around nucleus A to the polarization energy has to be calculated [20]. The total energy can be approximated by E =

i∈A 

A



 2n i E 0i + n i2

i

       χAi Vˆ  χAi − χBi Vˆ  χBi . E0 − Ei ∗ (5)

E 0i are the energies of un-polarized bond contributions, E 0 is the total ground state energy of the molecular system, E i ∗ is the excited state energy for the polarized bond contribution i, and n i is the occupation number. In a force field approach, we are only interested in relative energies and disregard the constant contributions, hence we introduce the molecular polarization energy E P as well as the atomic polarization energy E AP E = P

i∈A  A

i

 n i2

        χAi Vˆ  χAi − χBi Vˆ  χBi = E AP E 0 − E i∗ A (6)

The chemical shift can be expressed in terms of the atomic polarization energy [20]. By expanding it with respect to theo the chemical shift tensor δαβ around the experimental exp value δαβ and evaluating the gradient, the BPT pseudoforce can be deduced (for isotropic chemical shifts) as: F j = k CS (δ theo − δ exp )

∂δ theo . ∂xj

(7)

The chemical shift derivatives can be calculated within the BTP approach mainly from the derivatives of the polarization energy integrals (see Equation 2). In this case, the force constant becomes k CS =

i∈A  A

i

q

2Ai . n i2 (Aiδ )2

(8)

The computational cost depends, to a first degree, on the charge calculation, which is proportional to the cube of the number of atoms, N3 . Calculations on systems of about 104 atoms are feasible within a day on current standard GHz machines.

Applications in Crystal Structure Refinement Refinement of Proton Positions The first application of the COSMOS–NMR force field to a crystallographic problem was the refinement of proton positions from 13 C chemical shifts [26]. Accurate proton positions are not so readily determined by X-ray diffraction, especially for large molecules, because protons have no core electrons. Even if there are good X-ray data available, the refinement of the proton sites using NMR chemical shifts will lead to better-defined structures, and can provide valuable insights into the formation of hydrogen bonds. In our first example of β-d-mannitol, both the highresolution X-ray structure and the solid-state 13 C chemical shifts were known. The BPT calculations of the chemical shifts from the X-ray atomic positions gave a mean deviation from the experimental NMR data of 1.7 ppm, with a maximum difference of 2.7 ppm. A force field optimization of the protons, while keeping the positions of the heavy atoms unchanged, lead to a structure with an even larger mean deviation of 2.5 ppm for the calculated 13 C chemical shifts from their experimental values. Next, 13 C chemical shift pseudo-forces were switched on, to act only on the proton positions. Even though 13 C chemical shifts are used as target parameters, this does not mean that the pseudo-forces act only on the carbons. All atoms that contribute to the polarization of a carbon bond acquire pseudo-forces and can thus be influenced by the geometry optimization. The pseudo-forces were scaled in a range starting from 10−3 up to 103 . Significant changes started

Crystal Structure Refinement Using Chemical Shifts

Applications in Crystal Structure Refinement 75

Fig. 1. Superposition of the X-ray structure of d-mannitol with its refinement from 13 C NMR chemical shifts. The 50% probability spheres of the protons from the X-ray investigation are shown for comparison.

to show up around 10−1 , and at for a scaling constant of 102 a lower limit of 0.02 ppm for the chemical shift difference is reached. Notably, the refined proton positions do not violate the limits of the X-ray diffraction, even if the pseudo-forces exceed all other force field energies. The average proton displacement parameter derived ˚ The standard from the temperature factor is about 0.2 A. deviation of the NMR-refined structure with respect to the ˚ [26]. X-ray structure is only about 0.13 A In Figure 1, the superposition of the X-ray and the porton-refined structure (scaling factor 1000) is shown. The spheres at the proton positions are the isotropic 50% probability ellipsoids. It is thus possible to refine crystal structures using chemical shifts as target functions, and thereby resolve structural features such as proton positions that are not well represented in diffraction investigations.

The lack of crystalline order in polymers such as cellulose or silk fibers significantly reduces the number of interference spots in diffraction investigations, hence the patterns often cannot be analyzed unambiguously. In these cases, a crystal structure refinement using NMR chemical shifts will be of great value, because it does not require longrange crystalline order. For cellulose three major polymorphs are known, namely natural occurring cellulose Iα and Iβ , as well as regenerated cellulose II. When starting our investigations, good crystal structures had been published for cellulose II, while for the other polymorphs only preliminary models based on electron diffraction were available. Optimizations with 13 C chemical shift target functions succeeded to produce structures that fulfill the requirements of both the NMR and diffraction data [27]. It was thus demonstrated that some inner-chain hydrogen bonds induce geometry changes of the glycosidic linkage, which cause the C4 resonance to shift from an amorphous value of typically 75 ppm to the observed crystalline value of about 88 ppm. Since this chemical shift value is observed in all cellulose I and II polymorphs, it was concluded that their hydrogen bond patterns have to be similar. The chemical shift of the C4 carbon site can thus be taken as indicator of crystallinity for all three cellulose polymorphs [27]. Recently, native cellulose structures was reinvestigated by Nishiyama et al. [28], using X-ray and neutron diffraction experiments. For the first time, data concerning the hydrogen bond network could be extracted, hence it was of interest to compare the diffraction results with newly refined NMR structures. 2D NMR investigations on 13 C-enriched bacterial cellulose made it possible to assign all six resonances of the glucose units, and moreover to obtain chemical shift anisotropy data [29]. Nishiyama et al. [28] had discussed two possible schemes of how the hydroxyl groups could form two alternative hydrogen bond networks. These two networks were proposed to coexist within the cellulose crystallites, and the authors presented occupation numbers for the alternative deuteron positions. We used these two sets of data to derive two alternative structural models (A and B) with the COSMOS–NMR force field. There are furthermore two possibilities of assigning the two simultaneously observed sets of chemical shifts to the two glucose rings in the unit cell (designated with I and II), hence four sets of MD calculations had to be performed (see Table 1). To assess the stability of the hydrogen bond schemes and to clarify the glucose assignments, we ran 100 ps crystal dynamics simulations on each model, in which the coordinates and atomic charges were recalculated every 0.5 femtoseconds. The simulations

Part I

Structure of Cellulose Polymorphs from 13 C Chemical Shifts

76 Part I

Chemistry

Part I

Table 1: Energy contributions and chemical shift differences of the original and chemical shift refined cellulose Iα structures

Structure and method

NMR Van der Electrostatic Hydrogen bond assignment of Waals energy energy scheme glucose units (kJ/mol) (kJ/mol)

Neutron diffraction

A

CS optimized

A

Neutron diffraction

B

CS optimized

B

I II I II I II I II

40.4

−622.6

67.0 93.8 71.1

−1369.1 −1557.6 65.2

139.6 77.9

−1389.7 −1624.8

were performed at 293 K to create structures near all minima that could be populated at room temperature. A total of 200 coordinate snapshots were sampled for geometry optimizations with 13 C isotropic chemical shifts as target functions. First, all structures were geometry optimized, and in a second step the chemical shift pseudoforces were switched on. The results of the optimization procedures are given in Table 1. The first remarkable observation is that for both hydrogen bond schemes, A and B, deep minima for the electrostatic and total energies are obtained, which makes a spontaneous conversion of the two forms at room temperature very unlikely. Since in most cases the chemical shift pseudo-forces drive the structures energetically uphill, we selected the most preferable structure according to the sum of the total and pseudo energies. The most favorable structural model thus corresponds to hydrogen bond scheme A and assignment I (A-I). This structure is moreover in 6th best position of all optimized MD conformations with respect to the total energy, and it has the lowest RMS deviation between the calculated and observed chemical shifts (see Table 1). The drop in total energy in all cases where chemical shift pseudo-energies are applied, is a clear indication that the calculated chemical shifts are of high quality. After optimization we reach minima with small pseudoenergies and with chemical shift values that lie within the error range of the NMR experiment. The pseudo-energy of 71 kJ/mol is only about 5% of the electrostatic energy in the case of the A-I structure. To test the reliability of the chemical shift refined structured, a least-squares superposition of structure A-I onto the original neutron diffraction coordinates was performed. The RMSD of the two cel˚ for all atoms, and this lulose chain fragments is 0.51A ˚ if only heavy atoms are condifference drops to 0.37 A sidered (see Figure 2). Most atomic positions fall clearly within the error bounds of the fiber diffraction analysis.

RMSD of Pseudo-energy Total energy CS values (kJ/mol) (kJ/mol) (ppm) 2970.4 2991.0 71.0 323.0 3677.3 3705.1 389.0 702.8

161.8 −1111.6 −1249.7 876.4 −1061.7 −1294.6

5.4 5.4 0.83 1.77 6.0 6.0 1.93 2.61

a c

b

Fig. 2. Least-squares superposition of the cellulose Iα coordinates from neutron diffraction by Nishiyama et al. [28] (transparent model) together with the NMR-refined structure A-I (hydrogen bond scheme A and chemical shift assignment I, see Table 1) that was optimized according to chemical shift restraints. ˚ and 0.37 A ˚ for the carbon The RMSD for all atoms is 0.51 A, and oxygen atoms. (See also Plate 6 on page XX in the Color Plate Section.)

Crystal Structure Refinement Using Chemical Shifts

Figure 3 shows an excellent correlation between the observed and calculated principal CS tensor components of all 12 carbons in the unit cell (filled circles). Notably, the chemical shift components derived from the original diffraction coordinates gave no correlation at all (open circles). This must be regarded as strong evidence for the validity of the new NMR-refined structure (see Figure 2). It has moreover been possible for the first time to derive the orientations of the individual chemical shift tensors in the molecular framework of the glucose rings, as illustrated in Figure 4. Based on these chemical shift calculations, some characteristic rules for CS tensors in carbohydrates can be tested. As discussed by Koch et al. [30], for carbons carrying a hydroxyl group the principal component δ 33 should be oriented along the C–O bond. As seen from Figure 4, the δ 33 directions deviate only by a few degrees from the respective C–O bond directions. In the case of C1, which is bound to two oxygen atoms, the δ 33 component lies within the O1–C1–O5 plane, and the δ 22 direction is aligned with the bisector of the angle formed by the three atoms.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. Fig. 4. Orientation of the 13 C chemical shift tensors in a glucose ring, calculated from a structure obtained by a geometry optimization with chemical shift boundary conditions. The δ 33 principal axis components of the tensors (calculated by BPT) are nearly parallel with the C–O bond directions. In the case of carbon C1 with two neighboring oxygen atoms, the δ 22 component lies on the bisector of the O–C–O angle. (See also Plate 7 on page XX in the Color Plate Section.)

16. 17. 18. 19. 20.

Brown SP, Spiess HW. Chem. Rev. 2001;101:4125. Klaus E, Sebald A. Magn. Reson. Chem. 1994;32:679. Taulelle F. Solid State Sci. 2004;6:1053. Ochsenfeld C, Brown SP, Schnell I, Gauss J, Spiess HW. J. Am. Chem. Soc. 2001;123:2597. de Dios AC, Laws DD, Oldfield E. J. Am. Chem. Soc. 1994;116: 7784. Sternberg U, Witter R, Ulrich AS. Annu. Rep. NMR Spectrosc. 2004;52:53. Case DA, Dyson HJ, Wright PE. Methods Enzymol. 1994; 239:392. Szilagyi L. Prog. Nucl. Magn. Reson. Spectrosc. 1995;27: 325. Williamson MP, Asakura T. Methods Mol. Biol. 1997;60: 53. Wishart DS, Sykes BD. Methods Enzymol. 1994;239:363. Ochsenfeld C, Kussmann J, Koziol F. Angew. Chem. Int. Ed. 2004;43:4485. Sternberg U. J. Mol. Phys. 1988;63:249. Sternberg U. Priess W. J. Magn. Reson. 1997;125:8. Slater JC. Phys. Rev. 1930;36:57. Priess W, Sternberg U. J. Mol. Struct. (Theochem) 2001; 544:181. Veeman WS. Prog. NMR Spectrosc. 1984;20:193. Sherwood MH, Alderman DW, Grant MG. J. Magn. Reson. 1989;84:466. O’Keefe M, Brese NE. J. Am. Chem. Soc. 1991;113:3226. Koch F-T, M¨ollhoff M, Sternberg U. J. Comp. Chem. 1994; 15:524. Witter R. Three Dimensional Structure Elucidation with the COSMOS-NMR Force Field, thesis, 2003, www.dissertation. de.

Part I

Fig. 3. Calculated principal 13 C chemical shift tensor components (δ11 , δ22 , δ33 ) of cellulose Iα , plotted against the experimental values. The values after optimization with chemical shift-pseudo-forces are displayed as filled circles, while the result evaluated from the original (non-optimized) diffraction structure are open circles.

References 77

78 Part I

Chemistry

Part I

21. 22. 23. 24. 25.

Cui Q, Karplus M. J.Phys. Chem. B. 2000;104:3721. Cui Q, Karplus M. J. Chem. Phys. 2000;112:1133. G¨untert P. Q. Rev. Biophys. 1998;31:145. M¨ollhoff M, Sternberg U. J. Mol. Model. 2001;7:90. Sternberg U, Koch F-Th, Br¨auer M, Kunert M, Anders E. J. Mol. Model. 2001;7:54. 26. Witter R, Prieß W, Sternberg U. J. Comp. Chem. 2002;23: 289.

27. Sternberg U, Koch F-Th, Prieß W, Witter R. Cellulose. 2003;10:189. 28. Nishiyama Y, Sugiyamam J, Chanzy H, Langan P. J. Am. Chem. Soc. 2003;125:14300. 29. Witter R, Sternberg U, Hesse S, Kondo T, Koch F-Th, Ulrich AS. Macromolecules 2006: in press. 30. Koch F-Th, Prieß W, Witter R, Sternberg U. Macromol. Chem. Phys. 2000;201:1930.

79

Hiroyuki Fukui Kitami Institute of Technology, Kitami, Hokkaido, Japan

Introduction One of the reasons for difficulties in explaining indirect nuclear spin–spin coupling constants is that this phenomenon has no analogs in classical physics. The main driving force for inducing nuclear spin–spin couplings in molecules is not electromagnetic interactions but the Pauli’s exclusion principle, operating between electrons with the same spin. It was demonstrated that Fermi correlation, due to the Pauli’s exclusion principle, can be considered to be the mechanism whereby distant atoms communicate with each other [1]. The indirect nuclear spin–spin coupling is described by the form of JMN IM · IN in which IM and IN are the nondimensional nuclear spin vectors, and JMN is called an isotropic nuclear spin–spin coupling constant [2,3]. JMN has the units of hertz (2π rad/s) Unlike the direct interaction of magnetic dipoles, an energy of this sort of nuclear spin–spin coupling does not average out to zero when the molecules are rotating, so its effect still remains in the spectra of liquids. This fact indicates that the indirect nuclear spin–spin coupling comes from an indirect coupling mechanism via the electrons in the molecule. The indirect coupling mechanism between nuclear spins will be considered in the next section.

Origin of the Indirect Nuclear Spin–Spin Coupling Interaction A full and successful theory of the indirect nuclear spin–spin coupling based on the complete Hamiltonian for electron–nuclear interactions was first outlined by Ramsey and Purcell [4] and developed in more detail in a later paper by Ramsey [5]. We will describe the origin of the indirect nuclear spin–spin coupling interaction below. The electron–nuclear magnetic interactions come from the interactions between nuclear spins and electronic motions or electronic spins. The magnetic interactions in an electronic system are well described with the use of the relativistic Dirac equation [6,7]. Let us consider an electronic system consisting of one electron and two hypothetical nuclear spins which have the nuclear magnetogyric ratios, γ M and γ N , respectively, but do not possess nuclear Graham A. Webb (ed.), Modern Magnetic Resonance, 79–83.
C 2008 Springer.

charges. The time-independent Dirac equation for the electron is then given by    0 cσ −3   + (µ0 /4π )e¯h γ M r M · −i¯h ∇ IM cσ 0 × rM + (µ0 /4π )e¯h γ N r N−3 IN × rN   0 0 ψ = Eψ, + m e c2 0 −2

(1)

where σ is the 2 × 2 Pauli spin matrix vector. The three components of the σ vector are given by       0 1 0 −i 1 0 σx = , σy = , σz = . (2) 1 0 i 0 0 −1 σ is the double of the electronic spin vector s, i.e. σ = 2s · µ0 is the permeability of the vacuum, c is the velocity of light, and m e and −e are the rest mass and electronic charge of the electron, respectively. rM and rN are defined by rM = r − R M and rN = r − R N , respectively, with the electronic position r and the nuclear positions, R M and R N . The wave function ψ has four components, i.e. large two-component spinor φL (the first and second components of ψ) and small two-component spinor φS (the third and fourth components of ψ). Equation (1) is decomposed into the two component equations: cσ · πφ  S = EφL

(3)

cσ · πφ  L − 2m e c2 φS = EφS ,

(4)

and where −3   + (µ0 /4π)e¯h γ M r M I M × rM π = −i¯h ∇

+ (µ0 /4π)e¯h γ N r N−3 IN × rN .

(5)

π is called the mechanical momentum of the electron. In order to eliminate the small component φS , we solve Equation (4) for φS and obtain φS = (2m e c2 + E)−1 cσ · πφ  L.

(6)

Part I

The Theory of Nuclear Spin–Spin Couplings

80 Part I

Chemistry

Part I

We substitute Equation (6) into Equation.(3) and get  −1 2  2 2m e c2 + E c σ · π − E φL = 0. (7) Equation (7) gives us the energy of the system 1/2

σ · π)  2 . E ± = −m e c2 ± m 2e c4 + c2 (

(8)

We are usually interested in the positive energy of the system. So we discard the negative energy E − and keep the positive energy E + alone. We expand the positive energy E + in terms of a power series of the reciprocal velocity of light c−1 . E + = ( σ · π)  2 /2m e − ( σ · π )4 /8m 3e c2 + O(c−4 ) + · · · . (9) The first term is the nonrelativistic energy of the system. The second term proportional to c−2 is the lowest order of relativistic correction to the energy. In this chapter, we consider only the nonrelativistic term and ignore the relativistic corrections to the energy. Using the identity [6,7] ( σ · π)  2 = π 2 + i σ · π × π, 

(10)

we obtain E + = ( σ · π )2 /2m e = −(¯h 2 /2m e ) PSO + (e2 /2m e )( A2M + A2N ) + h DSO MN + h M FC SD SD + h PSO + h FC N M + hN + hM + hN ,

(11)

where Aa = (µ0 /4π)¯h γa ra−3 Ia × ra ,

a = M or N , (12)

and µB = e h¯ /2m e .

(17)

h DSO and h PSO are the diamagnetic spin orbital (DSO) and paramagnetic spin orbital (PSO) interactions, respectively. h FC and h SD are the Fermi contact (FC) and spin dipole (SD) interactions, respectively. µB is the Bohr magneton. In the calculation of π × π in Equation (10), we used the identity [7] r )δuv + δuv /r 3 − 3ru rv /r 5 , ∇u (ru /r 3 ) = (4π/3)δ( (u, v ∈ {x, y, z}).

(18)

The field-independent splitting of NMR lines, usually described in hertz, is due to the isotropic part JMN of the indirect nuclear spin–spin coupling tensor JˆMN . The ! nuclear spin–spin coupling energy is written as uv JMN,uv I Mu I N v (u, v ∈ x, y, z). The (u,v) component JMN,uv (u, v ∈ {x, y, z}) of the tensor JˆMN has five different contributions [5], all of which result from electroncoupled interactions between the two nuclear spin components, I Mu and I N v . Four of these contributions are due to second-order perturbations and depend on the first-order wave function, whereas one contribution is due to a firstorder perturbation type and can be expressed using only the zeroth-order wave function. Four of these contributions can be expressed as a sum-over-states (SOS) formula [8]      B   A   0  HM,u m m  HN ,v  0 AB −1 JM N ,uv = h E0 − Em m>0   B    A   0  HM,u  m m  HN ,v  0 +(1 − δAB ) + C.C., (19) E0 − Em

where |0 and |m represent the many-electron ground and excited states of the unperturbed system, respectively. C.C. means the complex conjugate of the former term. A and B represent the type of interaction Hamiltonians, that is, the PSO, FC, and SD interactions whose one-electron (13) interaction operators are given by Equations (14)–(16). The isotropic part JMN is equal to the diagonal average of  h aPSO = 2(µ0 /4π )µB γa ra−3 Ia · la , la = −i¯h ra × ∇, the tensor, i.e. (JMN,x x + JMN,yy + JMN,zz )/3. The FC and SD interactions have matrix elements between the sina = Mor N , (14) glet ground state and the triplet excited states, whereas the PSO term has matrix elements between the singlet ra ) σ · Ia , h aFC = (8π/3)(µ0 /4π)µBh¯ γa δ( ground and excited states. So we have the four possible a = M or N , (15) combinations of FC–FC, SD–SD, FC–SD, and PSO–PSO contributing to JMN,uv . The FC–FC contribution is fully  isotropic, namely, all the diagonal elements of the FC– SD −3 h a = (µ0 /4π)µB h¯ γa −ra σ · Ia FC contribution are equal to each other, and all the off  diagonal elements of it are zero. On the other hand, the   + 3ra−5 σ · ra Ia · ra , a = M or N (16) FC–SD contribution is fully anisotropic and makes no

 2 2  −3 −3 2 ¯ /m e γ M γ N r M rN h DSO MN = (µ0 /4π ) e h       × IM · IN rM · rN − IM · rN IN · rM ,

The Theory of Nuclear Spin–Spin Couplings

ge = 2 + (α/π) − 0.65696(α/π)2 + 1.49(α/π)3 +O((α/π)4 ) + . . . = 2.02319,

(20)

where α = e /4π ε0 h¯ c = 1/137.0359895. 2

(21)

The nondimensional constant α is called the fine structure constant. The four kinds of one-electron interaction operators can be easily written using Equations (13)–(16). The nonrelativistic unperturbed Hamiltonian H0 is written as    H0 = h (0) (k) + e2 /4πε0 rkl , (22) k

k s and n > 0,

 

   

  (P1 )vn = 0  P, qv+  0 0  P, K n+  0 , (34a) (P2 )vn = [0 |[P, qv ]| 0 0 |[P, K n ]| 0 ] , (34b)   0|[qv , Q]|0 , (Q 1 )vn = 0|[K n , Q]|0   0|[qv+ , Q]|0 , (35) (Q 2 )vn = 0|[K n+ , Q]|0  % $   %⎤ ⎡ $     0 qv , qv+ 0 0 qv , K n+ 0  % $   %⎦, (S)vn,v n = ⎣ $     0 K n , qv+ 0 0 K n , K n 0 (36)   0|[qv , qv ]|0 0|[qv , K n ]|0 , (37) ()vn,v n = 0|[K n , qv ]|0 0|[K n , K n ]|0

(32) It is well known that ia (FC) yields a triplet excitation energy which is too small, especially for molecules having multiple bonds such as C2 H4 , C2 H2 , etc. Sometimes ia (FC) becomes negative and the CHF calculation gives us meaningless results [13]. This phenomenon is called the “triplet instability.”

(A)vn,v n

⎡ $    0 qv , H0 , qv+

⎢ = ⎣ $    0 K n , H0 , qv+

 %  0  %  0

$   0 qv , H0 , K n+

  $   0 K n , H0 , K n+

 %⎤  0 ⎥  % ⎦,  0 (38)

The Theory of Nuclear Spin–Spin Couplings

References 83

(B)vn,v n

  0|[qv , [H0 , qv ]]|0 0|[[qv , H0 ], K n ]|0 = . (39) 0|[K n , [H0 , qv ]]|0 0|[K n , [H0 , K n ]]|0 The CC method is an attempt to introduce interactions among electrons within a cluster and to permit the wave function to contain all possible “disjoint couplings” among the clusters. The second-order correction to the wave function due to quadruply excited configurations arises as products of doubly excited configurations. This is an example of disjoint couplings among the two-electron clusters. We write the exact ground state wave function |0 of the system Hamiltonian H as |0 = e T |φ0 ,

(40)

where T is called a cluster operator and |φ0 is the normalized HF wave function. It is now assumed that the wave function |0 satisfies the intermediate normalization condition, φ0 |0 = 1. For a system containing an even number of electrons 2n, the cluster operator T generates one-, two-, . . . , 2n electron clusters: T = T1 + T2 + · · · + T2n ,

(41)

where Tk is the k-electron cluster. The CC energy E of the ground state is determined by the Schr¨odinger equation H e T |φ0 = Ee T |φ0 ,

(42)

from which the system energy is given by     E = 0 e−T H e T  φ0 ,

(43)

    E = φ0  H e T  φ0 .

(44)

or

An analytical differentiation of the CC energy E is obtained by using the equation-of-motion coupled-cluster (EOM-CC) method [22] or the coupled-cluster polarization propagator (CCPPA) method [23]. As the start of the EMO-CC approach, Perera et al. [22] assumed that 0| = φ0 | (e T )† = φ0 | (1 + ),

(45)

E = φ0 | (1 + )e−T H e T |φ0

(46)

is variational under the condition that φ0 |(1 + )| φ0 = 1. The operator  is expanded as  = 1 + 2 + · · · + 2n . The operator  is determined variationally using the stationary condition of E. The CCPPA uses the linear response function in the framework of perturbation theory at the level of CC approximation. The detail is shown elsewhere [24].

References 1. Bader RFW, Streitwieser A, Neuhaus A, Laidig KE, Speers P. J. Am. Chem. Soc. 1996;118:4959. 2. Gutowsky HS, McCall DW, Slichter CP. Phys. Rev. 1951;84: 589. 3. Hahn EL, Maxwell DE. Phys. Rev. 1951;84 :1286. 4. Ramsey NF, Purcell EM. Phys. Rev. 1952;85:143. 5. Ramsey NF. Phys. Rev. 1953;91:303. 6. Schiff LI. Quantum Mechanics (Chapter 13), 3rd ed. McGraw Hill: New York, 1968. 7. Moss RE. Advanced Molecular Quantum Mechanics. Chapman and Hall: London, 1973. 8. Fukui H, Miura K, Matsuda H, Baba T. J. Chem. Phys. 1992; 97:2299. 9. Berestetskii VB, Lifshitz EM, Pitaevskii LP. Quantum Electrodynamics, 2nd ed. Pergamon: New York, 1982. 10. Matsuoka O, Aoyama T. J. Chem. Phys. 1980;73:5718. 11. Pople JA, Schneider WG, Bernstein HJ. High-Resolution Nuclear Magnetic Resonance. McGraw Hill: New York, 1959. 12. Pople JA, Beveridge DL. Approximate Molecular Orbital Theory. McGraw Hill: New York, 1970. 13. Guest MF, Saunders VR, Overill RE. Mol. Phys. 1978;35:427. 14. Dalgaad E. J. Chem. Phys. 1980;72:816. 15. Coester F. Nucl. Phys. 1958;7:421. 16. Cizek J, Paldus J. Int. J. Quant. Chem. 1971;5:359. 17. Harris FE. Int. J. Quant. Chem. 1977;S11:403. 18. Harris FE, Phariseau P, Scheive L (Eds). Electrons in Finite and Infinite Structures. Plenum Press: New York, 1977. 19. Bartlett RJ, Purvis GP. Int. J. Quant. Chem. 1978;16:561. 20. Laaksonen A, Kowalewski J, Saunders VR. Chem. Phys. 1983;80:221. 21. Jørgensen P, Simons J. Second Quantitization-Based Methods in Quantum Chemistry. Academic Press: New York, 1981. 22. Perera SA, Nooijen M, Bartlett RJ. J. Chem. Phys. 1996;104: 3290. 23. Geertsen J, Oddershede J. J. Chem. Phys. 1986;85:2112. 24. Fukui H. Prog. Nucl. Magn. Reson. Spectrosc. 1999;35:267.

Part I

where  is the de-excitation operator. It is assumed that the energy functional

and

Part I

Fibrous Proteins

87

G¨oran Zernia and Daniel Huster Junior Research Group “Solid-State NMR Studies of Membrane-Associated Proteins”, Biotechnological Biomedical Centre, Institute of Medical Physics and Biophysics, University of Leipzig, D-04107 Leipzig, Germany

Introduction Collagen is the most abundant protein on the earth. It is found in many different tissues of animals and humans. The major property of collagen is to provide tensile strength to tissues such as tendons, ligaments, skin, cartilage, blood vessels, and bone [1]. The remarkable tensile strength of collagen can be understood from its unique secondary structure. The polypeptide chain of collagen forms a slightly twisted lefthanded helix with three amino acids per turn. Three collagen chains are coiled together to form the three-stranded collagen triple helix. In the amino acid sequence of each polypeptide chain, every third residue is glycine (Gly). In the triple helix, the Gly residues of two chains come in close proximity to form a hydrogen bond. This structural arrangement is too dense to allow a larger side chain explaining the high Gly content of collagen. Further, collagen consists of approximately 21% proline (Pro) and hydroxyproline (HyPro). These amino acids impart rigidity and stability to the structure, especially by hydrogen bonds between the hydroxyl groups of HyPro. Together with 9% alanine (Ala) and 5% glutamic acid (Glu), these five amino acids account for about 70% of all the residues in collagen. Collagen forms fibrils, which are superstructures of collagen triple helices. The individual structure of this arrangement determines the tensile strength of collagen. The collagen triple helices are linked by covalent lysine–hydroxylysine bridges [1,2]. There are about 30 structural variants of collagen depending on the function of the respective tissue. The most relevant species are collagen type I (as found in bone, tendon, or ligament) and type II (as found in cartilage). Each collagen type has a slightly different amino acid sequence [1]. A characteristic feature of many biological tissues are the viscoelastic properties. Tendons, ligaments, or cartilage must respond quickly, robustly, and reversibly to deformations caused by mechanical load or dynamic stresses [3]. These elastic properties of many biological tissues are a consequence of the structural arrangement of fibrillar collagen and proteoglycans. The versatile molecular Graham A. Webb (ed.), Modern Magnetic Resonance, 87–92.
C 2008 Springer.

dynamics of these different macromolecules, the osmotic pressure, and the flow of aqueous tissue fluids represent the physical basis of the unique viscoelasticity of these tissues. Therefore, to cope with the various compressive stresses, acting on the tissue, these molecules undergo dynamic reorientations of very different geometries within a broad time window [3]. A sketch of the molecular organization in articular cartilage is given in Figure 1. NMR techniques have been successfully applied to investigate the macromolecular species in tissues and to describe their dynamic properties. This short review focuses on the application of solid-state NMR methods to investigate the dynamical properties of isolated and tissue collagen. Solid-state NMR methods have to be applied since collagen is largely rigid due to its fibrillar structure and large molecular weight. Therefore, the anisotropic NMR interactions such as the chemical shift anisotropy and dipolar couplings are not averaged out by motions and the NMR spectra show the characteristic orientation dependence of the NMR frequency that is observed for solid materials. Solid-state NMR spectroscopy has unique capabilities for studying the molecular dynamics with correlation times from picoseconds to seconds by relaxation time measurements, lineshape analysis, or exchange methods [4,5]. All molecular motions are studied in the absence of an overall tumbling of the molecules that is present in solution NMR and might interfere with the motional analysis [6]. In this chapter, we will discuss static and magic angle spinning (MAS) solid-state NMR methods that have been applied to investigate the dynamics of tissue collagen. Further, recent data of our ongoing research on the dynamics of cartilage collagen will be discussed.

Investigation of Collagen Dynamics by Static Solid-State NMR First NMR studies on collagen have been published by Torchia and co-workers [7,8]. Because of the rigidity of collagen fibrils, solution NMR fails to resolve their signals and solid-state NMR methods are most appropriate to

Part I

Investigation of Collagen Dynamics by Solid-State NMR Spectroscopy

88 Part I

Chemistry

Part I

B)

A)

C)

Fig. 1. (A) Schematic structure of articular cartilage with collagen molecules (green) and proteoglycans. (B) Proteoglycans consist of a strand of hyaluronic acid (red), to which a core protein (black) is attached. On the core protein, glycosaminoglycans (blue) such as chondroitin sulfate and keratan sulfate are covalently bound in a bottle-brush fashion. (C) The typical triple helix structure of collagen molecules and the chemical structure of the major amino acids in collagen are depicted. (See also Plate 8 on page XX in the Color Plate Section.)

study these molecules. Typically, solid-state NMR spectra of large molecules consist of a superposition of many broad anisotropic lineshapes that cannot be resolved in one dimension. Therefore, selective isotopic labeling is a prerequisite for any further analysis of the NMR spectra. To this end, collagen fibrils were labeled in animal tissue cultures. 2 H-, 13 C-, 15 N-, or 19 F-labeled amino acids were fed or injected to rats, chicken, or rabbits [9]. Thus, isotopically labeled collagen was produced and isolated from tendon, calvaria, tail, or sternum of those animals. Besides structural data, the static solid-state NMR lineshapes contain information about the dynamics of a respective site. Since molecular motions decrease the strength of anisotropic interactions, they partially or fully abolish the orientation dependence of the NMR frequency. Therefore, lineshape measurements are sensitive both to the amplitude and to the correlation time of the molecular motions. Consequently, the presence of fast motions leads to a reduction of the width of the anisotropic spectrum. Amplitudes and correlation times of motions can be determined from 2 H NMR spectra. The deuterium electric field gradient tensor is axially symmetric along the C–D bond, which simplifies the analysis. For the analysis, experimental 2 H NMR spectra are compared to simulated spectra that are calculated by applying a specific motional model. In particular, side chain motions have

been studied by 2 H NMR spectroscopy. Using collagen molecules with 2 H-labeled Ala, leucine (Leu), Pro, or methionine (Met), these spectra showed typical features of motionally averaged lineshapes at ambient temperature [10–13]. Only at low temperature, the characteristic Pake spectra with the full quadrupolar splitting were detected. Application of a two-site jump hop for the Ala side chains in chick calvaria collagen revealed fast reorientations of the Cα–Cβ bond vector over an angle of ∼30◦ with a correlation time ∼10−7 s [11,13]. Fast two-site exchange with an amplitude of ∼55◦ and a correlation time of 8 × 10−7 s were also found for Leu side chains [10]. For Pro and HyPro puckering motions have been identified from the 2 H NMR spectra with root mean square amplitudes in the 11◦ –30◦ range [12]. A typical example of static 2 H NMR lineshapes for [2 H10 ] Leu-labeled collagen as a function of temperature is given in Figure 2. The backbone motions of collagen molecules have been investigated by static 13 C solid-state NMR methods using 13 CO Gly-labeled collagen. Similar to 2 H NMR lineshapes, anisotropic 13 C NMR spectra contain information about motional amplitudes and provide at least an upper limit for the correlation times. Root mean square amplitudes of 41◦ , 33◦ , and 14◦ were calculated for the backbone motions in uncross-linked, cross-linked, and mineralized collagen, respectively [14]. These findings

Collagen Dynamics by Solid-State NMR

Application of CP MAS Methods to Study the Molecular Properties of Collagen 89

Part I

Fig. 2. Static 2 H NMR lineshapes of collagen to determine side chain mobility in [2 H10 ] Leu-labeled collagen. The left column shows 38.5 MHz 2 H NMR spectra of [2 H10 ] Leu-labeled collagen at various temperatures (a, −85 ◦ C; b, −43 ◦ C; c, −18 ◦ C; d, −6 ◦ C; e, +1 ◦ C; f, +15 ◦ C; g, +30◦ C). In the right column, lineshape simulations of the experimental spectra are displayed. These simulations assume a two-site hop model in which the Cγ–Cδ bond axes are assumed to hop between two sites separated by 108◦ –112◦ and κ defines the hopping rate (h, κ ≤ 6 × 103 rad/s; i, κ = 1.9 × 104 rad/s; j, κ = 3.1 × 104 rad/s; k, κ = 3.7 × 105 rad/s; l, κ = 6.3 × 105 rad/s; m, κ = 8.7 × 105 rad/s; n, κ = 1.2 × 106 rad/s). Reproduced with permission from Ref. [10].

were derived from the anisotropic carbonyl chemical shift tensor measurements indicating that the upper limit for the correlation times of these motions is 10−4 s. In addition to these somewhat slower motions, relaxation studies on collagen labeled with 13 Cα Gly revealed fast motions with correlation times in the 1–5 ns range exhibiting small amplitudes of 10◦ , 9◦ , and 5.5◦ uncross-linked, cross-linked, and mineralized collagen, respectively [15]. Because of their very fast correlation times, these motions must be segmental. While there is an obvious dependence of the backbone motion on the degree of cross-linking and mineralization, the side chain motions are only slightly affected by mineralization of collagen [12]. Recently, the static NMR data from the Torchia group have been reanalyzed and interpreted in terms of a librational rod model [16]. This analysis revealed that the 2 H NMR data are also consistent with small-angle librations about internal bond directions.

Application of CP MAS Methods to Study the Molecular Properties of Collagen While static solid-state NMR spectra are broad and typically signals of only one or very few sites can be resolved in one dimension, MAS methods allow to resolve many relatively sharp signals at once. The first applications of cross-polarization (CP) MAS solid-state NMR methods have been demonstrated by the groups of Schaefer [17] and Saitˆo [18,19]. By application of MAS, the spectral intensity of the broad anisotropic solid-state NMR spectra is collected into a sharp central band of a line width on the order of one or a few ppm and a series of spinning sidebands. The number and intensity of spinning sidebands depends on the MAS frequency and the Larmor frequency of the NMR spectrometer. Thus, relatively well-resolved 13 C NMR spectra from isolated collagen samples have been obtained at natural abundance. The NMR signals of

90 Part I

Chemistry

Part I

Fig. 3. Proton decoupled 188.6 MHz 13 C CP MAS spectra of fully hydrated porcine articular cartilage (A), dried porcine articular cartilage (B), and dry collagen type II (C) at a MAS frequency of 7 kHz and a temperature of 37 ◦ C. The amino acid assignment is given. Spectra were externally calibrated with respect to TMS.

13

C CP MAS spectra could be assigned to the most abundant amino acids in collagen (Gly, Ala, Pro, HyPro, and Glu) [18]. Further, in comparison to dry collagen spectra, sharper lines have been detected in hydrated collagen indicating the presence of fast motions [19]. Thus, besides cross-linking and the degree of mineralization, the level of hydration seems to be a determining factor for the molecular mobility of collagen in tissue. This is also consistent with the static NMR spectra that indicated that the collagen dynamics is dependent on the state of the surrounding water [16]. Therefore, a systematic investigation of collagen mobility as a function of hydration has been carried out by 13 C CP MAS techniques [20]. While anisotropic NMR lineshapes contain information about the molecular dynamics, the anisotropic contributions to the NMR spectra need to be reintroduced by slow spinning [21] or recoupling methods [22–24] to exploit these quantities to obtain information about molecular dynamics. If carried out as a separated local field experiment [25], these techniques provide comprehensive dynamical information about all resolved signals in one experiment. For instance, in a separated local field experiment, a motional amplitude for each site that provides a resolved signal in the MAS spectrum can be determined. For the study of collagen mobility as a function of hydration [20], the 1 H–13 C dipolar couplings have been measured in a DIPSHIFT experiment [26–28]. Thus, the dipolar coupling of each resolved collagen signal could be

determined. Fast motions average out dipolar couplings, the amplitude of these motions can be described by a molecular order parameter, which is calculated as the ratio of motionally averaged and full dipolar coupling. Generally, motional amplitudes were found to be larger in the side chains compared to the backbone of collagen underlying the importance of hydration for the molecular dynamics of collagen. Further, with increasing hydration level, a decrease in the order parameters has been observed. The upper limit for the correlation times of motion calculated from motionally averaged 1 H–13 C dipolar couplings is on the order of 4 × 10−5 s. With the recent introduction of high field magnets, it is now possible to study collagen in native tissues such as cartilage [29]. This is particularly remarkable since cartilage consists of ∼82 wt% water, ∼6 wt% proteoglycans, and only ∼12 wt% collagen [30–32]. Due to this high water and ion content, the sample volumes have to be restricted to avoid sample heating by the application of high power decoupling fields. This means that only milligram quantities of collagen can be investigated, which calls for extremely sensitive instrumentations. Figure 3A shows a 188.6 MHz 13 C CP MAS spectrum of porcine articular cartilage obtained on approximately 15 mg cartilage tissue at natural abundance. While still relatively noisy, the signals of the main amino acid of collagen could be identified and assigned. For comparison, the 13 C CP MAS spectra of dried porcine articular cartilage and isolated collagen type II are shown in Figure 3B and C, respectively.

Collagen Dynamics by Solid-State NMR

What Has Been Learned from Solid-State NMR Studies of Collagen? 91

This indicates that almost exclusively collagen signals are detected in 13 C CP MAS spectra of articular cartilage. In addition, signals from rigid proteoglycans of cartilage (mostly hyaluronan) can be detected in the 13 C CP MAS spectrum of cartilage. These signals are assigned to the ring carbons of the proteoglycans with typical chemical shifts between 65 and 80 ppm [33–35]. In fully hydrated cartilage, these signals are strongly attenuated in 13 C CP MAS spectra because of their high mobility, but contribute significantly to the NMR spectrum of dried cartilage. Figure 4 shows typical order parameters of collagen in dried and native porcine articular cartilage. In the dry sample, the backbone signals exhibit order parameters between 0.9 and 0.94 in agreement with the rigid molecular structure. For the side chains, order parameters between 0.64 and 0.87 indicate motions with root mean square amplitudes between 42.5◦ and 24.4◦ . In contrast, much smaller order parameters have been observed in fully hydrated cartilage. For the backbone, order parameters between 0.73 and 0.78 are consistent with motions of amplitudes in the range of 36.1◦ –32.3◦ . Even larger amplitudes of 48.6◦ –41.1◦ are observed in the side chains of collagen in fully hydrated cartilage with order parameters of 0.55– 0.66. For Ala Cβ, order parameters < 0.33 are obtained, characteristic for the fast rotation of methyl groups about the Cα–Cβ bond.

What Has Been Learned from Solid-State NMR Studies of Collagen? Solid-state NMR techniques have strongly contributed to our understanding of the molecular dynamics in isolated and tissue collagen. The first interesting observation was that even dry collagen is not entirely rigid. The amplitude of collagen motions is not greatly influenced by crosslinking, however, mineralization reduces collagen flexi-

bility in bone. The amino acids in collagen undergo fast segmental reorientations that can be described by root mean square amplitude fluctuations. Very small amplitudes are observed for the collagen backbone, while the motional amplitudes increase into the side chains of the amino acids. For Pro and HyPro, puckering motions of the entire ring structure have been identified. The methyl groups of amino acid side chains undergo fast rotations about the C–C bond axis. In hydrated collagen, a more versatile molecular dynamics was found. While the segmental motions of dry collagen occur on a fast timescale of the order of a few nanoseconds, in hydrated collagen also slower motions with correlation times of the order of 10−4 s have been observed. The motional amplitudes of hydrated collagen are significantly increased in comparison to dry collagen. Tissue collagen of fully hydrated articular cartilage shows the largest motional amplitudes. This is mostly due to the high water content. Different types of collagen do not show any dynamical diversity as concluded from comparison of collagens I and II. Although most amino acids in collagen are uncharged, there are several polar groups in the backbone and the side chains that represent water binding sites. In particular, the hydroxyl groups of HyPro have been identified as water binding sites according to X-ray studies since they have both hydrogen bond donor and acceptor properties [36]. It has been suggested that the collagen triple helices acquire extra hydrogen bonding capacity by prolyl hydroxylation [37]. The possible functional significance of the collagen mobility has already been suggested [15]. Due to their high tensile strength, collagen fibers provide mechanical stability to connective tissues. When tension is applied, collagen molecules are able to make rapid conformational changes. Thus, stress is distributed uniformly and the mechanical energy can be dissipated and adsorbed by the segmentally flexible molecules. The motions that have been

Part I

Fig. 4. 1 H–13 C order parameters of collagen in native (fully hydrated, open bars) and dried porcine articular cartilage (filled bars) at a temperature of 37 ◦ C. Order parameters were determined from DIPSHIFT experiments carried out at a MAS rotational frequency of 7 kHz.

92 Part I

Chemistry

Part I

identified so far occur on a sub-microsecond timescale. However, the stress that is exerted on connective tissues by our daily tasks such as walking, climbing the stairs, or exercising stresses the connective tissue on a slower timescale of tens to hundreds of milliseconds. Therefore, motions with these correlation times may also be relevant for collagen. At the moment, only preliminary data for collagen motions in that time window are available [20]. However, several newly developed solid-state NMR methods will allow to investigate such motions in collagen as well [38]. Besides understanding of the basic properties of collagen in regard to the viscoelasticity of biological tissue, recent tissue engineering developments have led to an increasing interest in the quantitative characterization of artificial tissues. For various applications in regenerative medicine (stem) cells are seeded into organic or inorganic scaffolds to produce extracellular matrix in vitro. The monitoring and quality control of the engineered tissues represents a major challenge to produce high quality replacements. NMR spectroscopy is very well suited to contribute to this field. Artificially grown tissues need to exhibit very similar properties as the natural tissue in order to be built into cartilage, bone, or other defects. The methods described in this chapter may be used to characterize artificial tissue and compare its properties with those of natural specimen. Thus, the optimal procedures for tissue engineering may be determined aided by a comprehensive monitoring of the dynamical properties of the tissue macromolecules as a prerequisite for a successful implantation.

6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28.

Acknowledgments This research has been funded by the European Funds for Regional Development (EFRE, Project #: 4212/03-12). D.H. would like to thank Jan Keller for help in preparing the figures.

29. 30. 31. 32.

References 33. 1. Nelson D, Cox M. Lehninger Biochemie. Springer-Verlag: New York, 2001. 2. Creighton TE. Proteins: Structures and Molecular properties. W. H. Freeman and Company: New York, 1993. 3. Scott JE. J. Physiol. 2003;553:335. 4. Palmer AG III, Williams J, McDermott A. J. Phys. Chem. 1996;100:13293. 5. Tycko R. In: R Tycko (Ed). Nuclear Magnetic Resonance Probes of Molecular Dynamics. Kluwer Academic Publishers: Dodrecht, 1994, p 1.

34. 35. 36. 37. 38.

Opella SJ. Methods Enzymol. 1986;131:327. Torchia DA. Methods Enzymol. 1982;82:174. Torchia DA. Annu. Rev. Biophys. Bioeng. 1984;13:125. Jelinski LW, Torchia DA. J. Mol. Biol. 1979;133:45. Batchelder LS, Sullivan CE, Jelinski LW, Torchia DA. Proc. Natl. Acad. Sci. U.S.A. 1982;79:386. Jelinski LW, Sullivan CE, Torchia DA. Nature. 1980;284:531. Sarkar SK, Hiyama Y, Niu CH, Young PE, Gerig JT, Torchia DA. Biochemistry. 1987;26:6793. Jelinski LW, Sullivan CE, Batchelder LS, Torchia DA. Biophys. J. 1980;32:515. Sarkar SK, Sullivan CE, Torchia DA. J. Biol. Chem. 1983;258:9762. Sarkar SK, Sullivan CE, Torchia DA. Biochemistry. 1985;24:2348. Aliev AE, Chem. Phys. Lett. 2004;398:522. Stejskal EO, Schaefer J. J. Am. Chem. Soc. 1976;98:1031. Saitˆo H, Tabeta R, Shoji A, Ozaki T, Ando I, Miyata T. Biopolymers. 1984;23:2279. Saitˆo H, Yokoi M. J. Biochem. (Tokyo). 1992;111:376. Reichert D, Pascui O, deAzevedo ER, Bonagamba TJ, Arnold K, Huster D. Magn. Reson. Chem. 2004;42:276. Antzutkin ON. Prog. Nucl. Magn. Reson. Spectrosc. 1999;35:203. Griffin RG. Nat. Struct. Biol. 1998;5:508. Bennett AE, Griffin RG, Vega S. In: NMR Basic Principles and Progress. Springer Verlag: Berlin Heidelberg, 1994, p 3. Dusold S, Sebald A. Annu. Rep. NMR Spectrosc. 2000;41:185. Waugh JS. Proc. Natl. Acad. Sci. U.S.A. 1976;73:1394. Munowitz MG, Griffin RG, Bodenhausen G, Huang TH. J. Am. Chem. Soc. 1981;103:2529. Schaefer J, Stejskal EO, McKay RA, Dixon WT. J. Magn Reson. 1983;52:123. Hong M, Gross JD, Griffin RG. J. Phys. Chem. 1997;101:5869. Huster D, Schiller J, Arnold K. Magn. Reson. Med. 2002;48:624. Maroudas A. In: MAR Freeman (Ed). Adult Articular Cartilage. Pitman Medical: Tunbridge Wells, 1973, p 131. Maroudas A. In: A Maroudas, K. Kuetter (Eds). Methods in Cartilage Research. Academic Press: London, 1990, p 211. Ratcliffe A, Mow VC. In: WD Comper (Ed). Extracellular Matrix, Volume 1, Tissue Function. Harwood Academic Publishers GmbH: Amsterdam, 1996, p 234. Brewer CF, Keiser H. Proc. Natl. Acad. Sci. U.S.A. 1975;72:3421. Torchia DA, Hasson MA, Hascall VC. J. Biol. Chem. 1977;252:3617. Naji L, Kaufmann J, Huster D, Schiller J, Arnold K. Carbohydr. Res. 2000;327:439. Bella J, Eaton M, Brodsky B, Berman HM. Science. 1994;266:75. Bella J, Brodsky B, Berman HM. Structure. 1995;3:893. Luz Z, Tekely P, Reichert D. Prog. Nucl. Magn. Reson. Spectrosc. 2002;41:83.

93

Kristin K. Kumashiro Department of C hemistry, University of aHw aii at Manoa, oHnolulu, H aawii 6 892 , USA

Introduction The skin and blood vessels of a vertebrate have a uniquely resilient and elastic quality to them. These properties are traced to elastin, the principal protein component of the fibers that comprise a large portion of these elastic tissues. Numerous reviews have been written on the biology and chemistry of this unique protein [1–3]. Briefly, the elastin gene encodes tropoelastin, which is crosslinked in post-translational modification to form insoluble elastin or, simply, elastin. The molecular weight of tropoelastin is large, with molecular weights ranging from 70 to 80 kDa, and its high-resolution structure is not yet solved. Generally, tropoelastin and insoluble elastin are considered to have two “domains,” namely, hydrophobic and crosslinking. Much attention has been focused on the hydrophobic regions that are dominated with the small nonpolar amino acids, glycine, alanine, proline, and valine. A number of repeating polypeptide sequences are found in this domain. Among them are (VPGVG)n and (PGVGVA)n . The crosslinking domain is rich with alanines, with a typical repeat sequence of (KAAK)n or (KAAAK)n . In portions of tropoelastin, the hydrophobic and crosslinking domains alternate. Figure 1 illustrates several domains of rat tropoelastin, as reported by Pierce et al. [4]. To date, there is limited information on the threedimensional structure of insoluble elastin. If one were to consider the size and nature of tropoelastin, then it would be easy to see why this problem is so difficult. That is, the predominance of the small hydrophobic residues and the presence of the crosslinks are the root causes for the insolubility of amorphous, or “mature,” elastin in all but the harshest conditions. Hence, solution nuclear magnetic resonance (NMR) and X-ray crystallography are virtually useless for high-resolution structure determination. Indeed, elastin has more in common with synthetic organic polymers than with many proteins characterized thus far with NMR spectroscopy. In the past, numerous models have emerged to explain the elasticity of elastin [5–8]. They range from the most disordered and globular to ones with significant degrees of order. As examples of the former, Hoeve and Flory used thermodynamic measurements to suggest that elastin was much like rubber, with long hydrophobic Graham A. Webb (ed.), Modern Magnetic Resonance, 93–99.
C 2008 Springer.

chains interspersed randomly with crosslinks [5]. In contrast, the “oiled coils” from predictive methods [6] and the “β-spiral” from structural studies of elastin-based peptides [7–10] suggest that this polymer has a much greater degree of order. More recent computational studies on the elastin peptides have provided some new insights [11,12]. It is generally accepted, though, that the Alarich crosslinking domains are mostly α-helical, whereas the hydrophobic domain’s organization is much less clear. Again, the lack of site-, residue-, and sequence-specific data, such as those obtained by solution and solid-state NMR spectroscopy, has greatly hampered the understanding of the native protein’s structure–function relationships. Two basic approaches have emerged as viable ways to characterize this intriguing protein by solid-state NMR spectroscopy. One focuses on the native (or nativelike) elastin, while the other uses smaller model peptides. Studies of the native protein would be most physiologically relevant, when drawing conclusions regarding structure–function relationships. The preparation of elastin from connective tissue is straightforward [13–15], and large quantities are easily obtained. With purified elastin samples, various groups [13,16–22] have characterized the natural-abundance 13 C populations present in the native protein, complete, in most cases, with the waters of hydration. To complement this approach, methods for isotopic enrichment of a given residue type have also emerged [23–25]. These labeling schemes are essential for NMR studies targeting key amino acid types in elastin, and the power of solid-state NMR as a high-resolution structural tool is becoming more evident as these findings are reported. Alternatively, now-classic approaches in elastin biochemistry have focused on mimetics, most notably, the repeating polypenta and hexapeptides, as studied by Urry [7–10,26] and Tamburro [27–29]. The use of these smaller peptides circumvents the problems associated with the polymeric nature and insolubility of elastin. Typically, the rationale for using these peptides is based on the fact that the hydrophobic regions of elastin have an abundance of these somewhat unusual repeating motifs, and elasticity has been assumed to originate from this domain. In addition, the repeating polypeptides possess properties similar to the native tropoelastin, such as coacervation, and can

Part I

Solid-State NMR Studies of Elastin and Elastin Peptides

94 Part I

Chemistry

Part I Fig. 1. Amino acid sequence of rat tropoelastin as encoded by exons 23–30, as reported by Pierce et al. [4]. Exons 23, 25, 27, and 29 are the Ala-rich crosslinking domains. Exons 24, 26, 28, and 30 are the hydrophobic domains, dominated by small hydrophobic residues, such as glycines.

also be crosslinked by reaction with oxidizing reagents or by exposure to γ-irradiation. Modern approaches, such as those based on recombinant methods, has facilitated the synthesis of biopolymers with elastinlike subunits for solid-state NMR characterization. In comparison to the task of labeling native elastin, the isotopic enrichment strategies for the elastin-based peptides or mimetics, whether by synthesis or bacterial expression, tend to be more straightforward. However, by virtue of the inherent simplicity of these systems, one may wonder about the relevance of the repeating polypeptides to the questions surrounding elastin’s elasticity. One major and valid concern focuses on the fact that many of the model peptides reported thus far do not contain the Ala-rich crosslinking domains. In this chapter, we focus on recently reported and current work using solid-state NMR spectroscopy to characterize elastin and elastin peptides. After a short review with a description of important results obtained over two decades ago, recent work by this author’s lab and others will be described. Many of these studies are based on techniques in “high-resolution solid-state NMR,” utilizing cross-polarization magic-angle-spinning (CPMAS) as a cornerstone, although some projects incorporate nonspinning methods. It will also be shown that the unusual nature of elastin requires new and creative approaches for

the continued use of NMR spectroscopy as a powerful tool for characterizing the structure and dynamics of one of nature’s most novel and unique elastomers.

Studies of Native Elastin Focus Mainly on the Natural-Abundance 13 C Populations In the 1970s and early 1980s, Torchia and co-workers reported a series of studies using NMR to characterize elastin [16,17,23,24]. Their earliest work [16] used 13 C NMR to examine calf ligamentum nuchae, which is rich in elastin. Their results indicated that elastin, unlike collagen, was comprised of “highly mobile chains.” A subsequent study [17] provided a more quantitative picture of this unusual mobility in elastin with relaxation data, including T1 and NOE determinations, for elastin swollen in various solvent environments. With this data, they again concluded that there was significant and unusually high mobility in this amorphous protein, although molecular motion was slightly more restricted in the Ala-rich regions. Other groups made additional contributions to this early picture of elastin. Urry and Mitchell reported a 13 C NMR study of α-elastin and fibrous elastin [30], with most of their focus on qualitative comparisons of the various

NMR of Elastin

Studies of Native Elastin Focus Mainly on the Natural-Abundance 13 C Populations 95

Part I

protein samples with each other and with the elastin polypentapeptide. Ellis and Packer also contributed to this early picture with their 2 H relaxation measurements of hydrated elastin [18]. Their work identified the existence of three types of water in these samples. To close out this era, Kricheldorf and Muller reported the 13 C chemical shifts for a sample of commercially available elastin [19]. To assist with their interpretation of the elastin spectra, they also looked at the 13 C chemical shifts of several Pro-rich polypeptides. Their analysis focused particularly on the backbone carbonyl population. It was concluded that ∼25% of the protein was α-helical, another population was “helical segments of unknown pitch,” and a third was simply described as an “amorphous phase.” Furthermore, they emphasized that elastin’s structure possessed no long-range order, although local segments were structured. In the years since Torchia, Kricheldorf, and others first reported their compelling findings on elastin, the area of biological solid-state NMR has greatly evolved. From the development of progressively higher-field instrumentation and other hardware to the boom in new pulse sequences for defining, e.g. progressively more refined structural data, solid-state NMR spectroscopy has seen a tremendous growth. Furthermore, the tools of molecular biology for protein expression are much more accessible. As a result, many more questions surrounding protein structure and function may be addressed in this day and age. To begin our review of more recent results, the effects of temperature and hydration on the structure and dynamics of elastin are first discussed [13]. In this study, 13 C solid-state NMR experiments were applied to samples of bovine nuchal ligament elastin that were purified using the cyanogen bromide (CNBr) method [13–15]. Samples of elastin at various hydration levels at four different temperatures were first characterized. 13 C CPMAS data showed that elastin with little or no water (0–30% hydration) had similar profiles; i.e. in the lyophilized and drier samples, chemical shifts were identified for the center-of-mass of the backbone carbonyl, the aromatic, and 8–9 resolved aliphatic peaks. In contrast, spectra of wetter samples (40–100% hydration) at physiological and room temperatures were observed with markedly lower signal-to-noise than the drier samples. Only the 53 ppm peak was clearly resolved with relatively high signal-tonoise in the Cα region, and several peaks were noted in the upfield region of ca. 10–30 ppm. The differences between the profiles of the wetter and drier samples were negligible as the sample temperature is lowered to −20 ◦ C. To further examine the nature of elastin, a number of additional experiments were conducted. For instance, the CPMAS spectrum of the hydrated sample was compared to DPMAS data, as shown in Figure 2. With DPMAS, all sites are observed. In contrast, CPMAS data reflect

Fig. 2. CPMAS (top) and DPMAS (bottom) spectra of hydrated elastin at 37 ◦ C, as originally reported by Perry et al. [13]. Major differences are identified for the backbone carbonyl and in selected regions of the aliphatic carbons, such as the Cα-Gly and methyls.

marked differences for the backbone carbonyl carbons, as well as selected carbons in the aliphatic region. With regards to the latter, the lack of much 13 Cα-Gly signal is striking. This simple experiment highlighted the heterogeneous nature of elastin; i.e. some segments of elastin, particularly those that are Gly-rich, are so mobile that CP efficiency is greatly compromised. In addition, static experiments showed the presence of unusually narrow lines in the spectrum of the wet sample, clearly unusual for a typical solid. These qualitative measurements were complemented by T1 experiments, which again showed portions of the hydrated sample at 37 ◦ C to have almost liquidlike mobility —a term first used by Torchia and coworkers in their examination of elastin in the 1970s [24]. While the drier samples tended to behave more like typical peptides in the solid state, the hydrated samples tended to exhibit more heterogeneity, particularly in terms of their dynamics. Another recent solid-state NMR study of unenriched elastin samples purified from tissue was also reported by Pometun et al. [20]. In their study, solid-state NMR methods were employed to characterize “elastin fibers” obtained from a commercial source. A number of techniques were employed to characterize these samples. First, 2 H and 17 O NMR were used to identify the dynamics of the exchangeable backbone amides, as well as the waters of hydration. The data showed that there was no evidence for the tightly bound waters of hydration, unlike the work of Ellis and Packer [18]. The 2 H spectra were also used to conclude that the entire protein was mobile, whereby typical powder patterns found for most solid peptides were not observed in these conditions. Two-dimensional spectra

96 Part I

Chemistry

Part I

of dry and hydrated elastin yielded small 13 C shielding anisotropies of the resolved carbons in hydrated elastin. The order parameter S (10 kDa). Also, α-elastin retains properties of the native system, such as coacervation. As with the above-described study of the normal versus atherosclerotic elastin [21], there were no discernible differences in 13 C chemical shift for the normal versus undercrosslinked [22]. (Indeed, it has been shown that the

chemical shifts of the key amino acid types have nearly identical chemical shifts over a broad range of samples.) However, it was noted that the differences in peak intensities of the two samples did not correspond to the simple change in composition, i.e. fewer desmosine and isodesmosine crosslinks and more underivatized lysines. Few differences were observed for the two types of samples based on 13 C T1 and 1 H T1ρ measurements. However, 13 C T1ρ values indicated that motion on the kHz scale was slowed in the undercrosslinked sample, leading to the arguably counterintuitive conclusion that some mobility is lost with fewer crosslinks. As with the study on the elastin from atherosclerotic tissue, impaired function of the elastic fiber is correlated with a change in dynamics, again lending support to the idea that mobility of the protein is a salient feature of its elasticity.

A New Approach for Production of Isotopically Labeled Elastin Utilizes a Mammalian Cell Culture This section begins with a description of the earliest successful attempts of isotopic enrichment of elastin, as established by Torchia and co-workers [23,24]. In these early experiments, chick aorta was cultured with media containing the isotopic label. First, 14 C-labeled glycine or alanine was used to show that isotopic scrambling was relatively low (>10%) and also to provide a basis for estimating incorporation (10–20%) [23]. Subsequently, results were reported for chick aortic elastin that was labeled at the carbonyl carbons of Ala, Val, and Lys. Incorporation of the 13 C label was modest, with 6.4, 10.5, and 19.5% enrichment for [1-13 C]Ala, [1-13 C]Val, and [1-13 C]Lys, respectively. However, these levels were enough to observe these key amino acids. As with their natural-abundance results on the nuchal ligament, Torchia and co-workers obtained T1 , linewidth, and NOE values for the various residue types. In addition, the CP efficiency and the effect of dipolar decoupling were also observed to obtain a model that was slightly more refined than the ones obtained previously. Specifically, these experiments indicated that all of the Val, ∼75% of the Ala, and ∼40% of the Lys residues (and its derivatives) were found in very mobile regions. The remaining Ala and Lys residues were attributed to the crosslinking domains that were “motionally restricted.” Linewidth and NOE data were also recorded as a function of temperature. More recently, this author’s lab has shown that a cell line well-known in cardiovascular biology could be successfully exploited for isotopic labeling and, hence, NMR spectroscopy [25]. The neonatal rat smooth muscle cell (NRSMC) line had been well-documented as a viable means of studying elastin synthesis [33,34]. These primary cultures are grown in a mixture of standard growth

NMR of Elastin

Information on the Hydrophobic Domain of Elastin is Gleaned from Repeating Polypeptides 97

Fig. 3. 2 H spectrum of hydrated [2,2-2 H2 ]Gly NRSMC elastin at 37 ◦ C, swollen in 2 H-depleted water.

protocols for labeling elastin well in hand, a number of experiments targeting the key amino acids of elastin are underway.

Information on the Hydrophobic Domain of Elastin is Gleaned from Repeating Polypeptides As noted above, the hydrophobic domains of elastin are rich with repeating polypeptides. Most well-known are the polypentapeptides (VPGVG)n [9,10,36]. Much attention has been drawn to Urry’s model [7,8], in which each (VPGVG) subunit has the structure of a type II, β-turn and the repeating polypentapeptide, hence, forms a “βspiral.” Recent solid-state NMR results, however, show little support for such a regular and highly ordered structure. Rather, an inhomogeneous structure seems more likely, even for a simplified repeat like (VPGVG)n . A collaborative study between the Asakura and Kumashiro groups [37], for example, used solid-state spectral editing techniques [38] in combination with 2D spin diffusion under off-magic-angle-spinning conditions to provide another structural picture of the (VPGVG) subunit, as it is found in novel, recombinant silk-elastinlike peptides. These results showed that distorted β structures are predominant in the protein, with significant structural disorder about the central glycine residue in the (VPGVG) subunits. Hong and co-workers also reported a number of studies on elastin mimetics that incorporated the (VPGVG) subunit [39–42]. Their earliest report focused on the central glycine residue found in each of the pentapeptidyl subunits of the 81-kDa elastin-mimetic peptide [(VPGVG)4 (VPGKG)]39 [39]. This paper utilized a new technique for selective detection of a residue pair to identify the chemical shift anisotropy of a site that is poorly resolved in the 1D spectra. At the end of this article, they concluded that the Gly3 CSA is consistent with type I and II β-turn structures. A subsequent study [40] found that the type II β-turn was predominant at the Pro–Gly pair of each subunit, based on Pro 15 N and 13 C chemical shift measurements. The most recent work on this same pair in (VPGVG)3 , however, identified a “bimodal structure distribution” of an “extended and distorted β-strand” and a turn, either as a β-turn or a “previously unidentified turn” as its major and minor forms, respectively [41]. The relative populations of the two general types of conformers are 65 and 35%, which is roughly 2:1, consistent with an earlier report for the 1D 13 C CPMAS of (LGGVG)n [43] (described below). It appears that the structural studies by Hong and co-workers have evolved to the same general conclusions as those obtained with the silk-elastinlike peptides [37], namely, that the β-spiral model of regular, repeating β-turns is not supported by solid-state NMR studies of peptides incorporating the (VPGVG) motif.

Part I

medium components that have been supplemented with the appropriate isotopically enriched amino acid. The label is introduced upon seeding of the cells. Normally, after 6–8 weeks of growth, the cells and the elastin-rich matrix are harvested, and insoluble elastin is purified from the mixture using the CNBr method. To assay incorporation, we followed the example of Meier and co-workers in their study of isotopically enriched spider silk [35]; acid hydrolysis followed by solution NMR spectroscopy was used to determine the amount of stable isotope that had been incorporated into the protein. For samples of [1-13 C]Gly, [2-13 C]Gly, and [15 N]Gly NRSMC elastin, the product of the acid hydrolysis was dissolved in D2 O and then observed using 13 C NMR spectroscopy. Typically, the analysis focused on Cα-Gly peak and the doublets that would result for the enriched sites. Using this approach, 30–40% incorporation of labeled Gly into elastin was confirmed. Solution NMR data were also used to show that isotopic scrambling is minimal. Analysis of other labeled elastin samples is done by an analogous method (unpublished data). Early NMR results of the elastin samples enriched at the glycines are promising [25]. The 13 C CPMAS data tend to be significantly lower in signal intensity than typically seen with samples of this mass and incorporation level, and DPMAS data yield extremely narrow lines for a polymer with the complexity of elastin. Short T1 values are also consistent with our earlier observations of the natural-abundance 13 C populations of hydrated elastin [13]. In addition to earlier work [25], Figure 3 illustrates the 2 H spectrum of NRSMC elastin with enrichment at the glycines. The detailed analysis of this sample is forthcoming. However, again we note the remarkably narrow lines observed for this amorphous protein. Finally, we note that this approach has also proven to be feasible for enrichment at the alanines and valines (unpublished data). With the

98 Part I

Chemistry

Part I

As a final, interesting note, the work from Hong and co-workers [42] also demonstrated that the dynamics of the [(VPGVG)4 (VPGKG)]39 peptide were similar to those of the native elastin [13], underscoring the fact that these simplified systems yield results, in terms of structure, relaxation, and mobility, that are reasonably consistent with those of the more complicated biopolymer found in nature. Other elastin repeats are good candidates for characterization with solid-state NMR spectroscopy. In collaboration with Martino and Tamburro, this author’s lab used a series of 13 C CPMAS NMR experiments to characterize poly(LGGVG), a repeating motif found in elastin [43]. Martino et al. had earlier reported that the solution structure of this peptide is best described as a “conformational ensemble” that includes both type I and type II βturns, in addition to some unstructured regions [28]. Solidstate spectral editing techniques were employed to assist in making tentative peak assignments [38]; 1D CPMAS spectra showed clearly that two peaks are observed for the Cβ-Val carbon. The features at 32.7 and 29.2 ppm are present in the ratio of 2:1, implying that the conformation(s) corresponding to the downfield peak is dominant. Clearly, this simple result eliminates from consideration the possibility that each subunit may fold into only one type of structure. Furthermore, the nature of the backbone carbonyl lineshape supported this picture; i.e. simple deconvolution subroutines found that the backbone carbonyl peak did not yield results that would support a model such as the β-spiral. Finally, it was noted that, although the T1 ’s obtained were similar to other lyophilized peptides, we found that they tended to be on the shorter end of the range. That is, even though the peptides are lyophilized, simplified mimetics of the native protein, they tend to mirror the characteristics that set elastin apart from other proteins in the solid-state. Namely, they tend towards structurally heterogeneous samples with dynamics that reflect unusually fast motion.

Concluding Remarks Studies by Torchia and co-workers, among several others, provided the first glimpse of the unusual nature of the structure and dynamics of elastin, using solid-state NMR spectroscopy. Since then, the field has evolved to utilize a wealth of newer approaches, combining techniques like cell culture and recombinant methods with sophisticated ways to manipulate samples and spins. Overall, it appears that the diverse range of philosophies and approaches are leading to a convergence of themes, to some extent. From the relatively straightforward measurement of T1 ’s and other relaxation parameters to the multidimensional methods for measuring residual anisotropies, it is clear that the molecular dynamics of elastin is unlike most other

proteins in the solid-state. Furthermore, the solid-state NMR studies to-date have all but eliminated any structural model with long-range order. Instead, it appears that even the simplest repeating polypeptides based on the amino acid sequence of elastin have structural parameters that call for a “structure distribution” [41] or “conformational ensemble” [1,28] in proposing a three-dimensional model for this protein. In future years, we can expect that the discrepancies that exist amongst the various results will be reconciled with additional NMR experiments and with the growing number of elastin and elastinlike peptides.

Acknowledgments Various portions of the work presented in this chapter were supported by the NSF, NIH, and the Hawaii Community Foundation. Current and former members of this author’s research laboratory and Dr. W.P. Niemczura (NMR facility, University of Hawaii) are also acknowledged for their participation in this ongoing project.

References 1. Debelle L, Tamburro AM. Int. J. Biochem. Cell Biol. 1999;31:261. 2. Rosenbloom J, Abrams WR, Mecham R. FASEB J. 1993;7:1208. 3. Sandberg LB. Int. Rev. Connect. Tissue Res. 1976;7:160. 4. Pierce RA, Deak SB, Stolle CA, Boyd CD. Biochemistry. 1990;29:9677. 5. Hoeve CAJ, Flory PJ. Biopolymers. 1974;13:677. 6. Gray WR, Sandberg LB, Foster JA. Nature. 1973;246: 461. 7. Urry DW. Adv. Exp. Med. Biol. 1974;43:211. 8. Urry DW, Long MM. Adv. Exp. Med. Biol. 1977;79:685. 9. Venkatachalam CM, Urry DW. Macromolecules. 1981;14:1225. 10. Chang DK, Venkatachalam CM, Prasad KU, Urry DW. J. Biomol. Struct. Dyn. 1989;6:851. 11. Li B, Alonso DOV, Bennion BJ, Daggett V. J. Am. Chem. Soc. 2001;123:11991. 12. Li B, Alonso DOV, Daggett V. J. Mol. Biol. 2001;305:581. 13. Perry A, Stypa MP, Tenn BK, Kumashiro KK. Biophys. J. 2002;82:1086. 14. Starcher BC, Galione MJ. Anal. Biochem. 1976;74:441. 15. Rasmussen BL, Bruenger E, Sandberg LB. Anal. Biochem. 1975;64:255. 16. Torchia DA, Piez KA. J. Mol. Biol. 1973;76:419. 17. Lyerla JR, Torchia DA. Biochemistry. 1975;14:5175. 18. Ellis GE, Packer KJ. Biopolymers. 1976;15:813. 19. Kricheldorf HR, Muller D. Int. J. Biol. Macromol. 1984;6:145. 20. Pometun MS, Chekmenev EY, Wittebort RJ. J. Biol. Chem. 2004;279:7982. 21. Tarnawski R, Tarnawski R, Grobelny J. Atherosclerosis. 1995;115:27.

NMR of Elastin

34. Barone LM, Faris B, Chipman SD, Toselli P, Oakes BW, Franzblau C. Biochim. Biophys. Acta. 1985;840: 245. 35. Kummerlen J, vanBeek JD, Vollrath F, Meier BH. Macromolecules. 1996;29:2920. 36. Urry DW, Trapane TL, Sugano H, Prasad KU. J. Am. Chem. Soc. 1981;103:2080. 37. Ohgo K, Kurano TL, Kumashiro KK, Asakura T. Biomacromolecules. 2004;5:744. 38. Kumashiro KK, Niemczura WP, Kim MS, Sandberg LB. J. Biomol. NMR. 2000;18:139. 39. Hong M, McMillan RA, Conticello VP. J. Biomol. NMR. 2002;22:175. 40. Hong M, Isailovic D, McMillan RA, Conticello VP. Biopolymers. 2003;70:158. 41. Yao XL, Hong M. J. Am. Chem. Soc. 2004;126:4199. 42. Yao XL, Conticello VP, Hong M. Magn. Reson. Chem. 2004;42:267. 43. Kumashiro KK, Kurano TL, Niemczura WP, Martino M, Tamburro AM. Biopolymers. 2003;70:221.

Part I

22. Kumashiro KK, Kim MS, Kaczmarek SE, Sandberg LB, Boyd CD. Biopolymers. 2001;59:266. 23. Torchia DA, Sullivan CE. Adv. Exp. Med. Biol. 1977;79:655. 24. Fleming WW, Sullivan CE, Torchia DA. Biopolymers. 1980;19:597. 25. Perry A, Stypa MP, Foster JA, Kumashiro KK. J. Am. Chem. Soc. 2002;124:6832. 26. Luan C-H, Krishna NR, Urry DW. Int. J. Quantum Chem.: Quantum Biol. Symp. 1990;17:145. 27. Tamburro AM, Guantieri V, Gordini DD. J. Biomol. Struct. Dyn. 1992;10:441. 28. Martino M, Coviello A, Tamburro AM. Int. J. Biol. Macromol. 2000;27:59. 29. Martino M, Tamburro AM. Biopolymers. 2001;59:29. 30. Urry DW, Mitchell LW. Biochem. Biophys. Res. Commun. 1976;68:1153. 31. Partridge SM, Davis HF, Adair GS. Biochem. J. 1955;61:11. 32. Partridge SM, Davis HF. Biochem. J. 1955;61:21. 33. Jones PA, Scott-Burden T, Gevers W. Proc. Natl. Acad. Sci. U.S.A. 1979;76:353.

References 99

101

Tetsuo Asakura and Yasumoto Nakazawa Department of Biotechnology, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184-8488, Japan

Introduction Recently, much attention has been paid to silks from textile engineers to polymer chemists and biomedical scientists. The silk fibers produced by silkworms or spiders are the nature’s most highly engineered structural materials with combinations of strength and toughness not found in today’s man-made materials [1]. In addition, there are many kinds of silks from silkworms and spiders with different structures and properties. The silk fibroin from the domesticated silkworm, Bombyx mori, is a well-known fibrous protein whose amino acid composition (in mol%) is 42.9 Gly, 30.0 Ala, 12.2 Ser, 4.8 Tyr, and 2.5 Val. The fibroin consists mostly of the sequence (Ala-Gly-Ser-Gly-Ala-Gly)n and comes in silk I (the structure before spinning) and silk II (the structure after spinning) structural forms. Despite a long history of studying silk I, the structure remains poorly understood because any attempt to induce a macroscopic orientation of the sample for X-ray diffraction, electron diffraction, or solid-state NMR, readily causes a conversion of the silk I form to the silk II form. Employing several highresolution solid-state NMR techniques and analyzing 13 C CP/MAS NMR chemical shifts quantitatively, in conjunction with molecular simulations, we proposed a repeated β-turn type II structure stabilized by intra-molecular hydrogen bond for the silk I form. On the other hand, the structure of silk II has been proposed as a regular array of anti-parallel β-sheet firstly by Marsh et al. about half century ago, based on a fiber diffraction study of native B. mori silk fibroin fiber [1]. Later, Fraser et al., Lotz et al., and Takahashi et al. pointed out some intrinsic structural disorder in the silk II structure although they essentially supported the general features of this anti-parallel β-sheet model [2,3]. The solid-state NMR techniques that have been successfully used for the structure of silk I were also used for the detailed structural determination of silk II. The primary structure of Samia cynthia ricini silk fibroin is considerably different from that of B. mori silk fibroin [4]. The basic repeat sequence is made of alternating (Ala)12–13 regions and the Gly-rich regions which is similar to the sequence of spider dragline silk (major ampullate) although the length of polyalanine is shorter (Ala)5–6 in the latter case. The use of appropriate stable Graham A. Webb (ed.), Modern Magnetic Resonance, 101–106.
C 2008 Springer.

isotope-labeled model peptides for the repeated sequences of S. c. ricini silk fibroin and spider dragline silks coupled with the use of solid-state NMR methods have applied to determination of the precise local structure. In this chapter, we overview our recent studies on the structural determination of these silks with solid-state NMR.

Structure of B. mori Silk Fibroin Before Spinning (Silk I) The structural features of B. mori silk fibroin are conveniently studied using synthetic peptide (AG)15 , as a model for crystalline region because the lack of Ser in the model peptide (AG)15 does not make any difference in the 13 C CP/MAS NMR chemical shifts of the Ala and Gly residues in the repeated sequence (AGSGAG)n of native silk fibroin [5–7]. By combining several solid-state NMR techniques, we have determined the conformation of the model peptide (AG)15 in the silk I form: The torsion angles of Ala and Gly residues were (−60◦ ± 5◦ , 130◦ ± 5◦ ) and (70◦ ± 5◦ , 30◦ ± 5◦ ), respectively. 2D spin-diffusion NMR was used to determine the torsion angles. Figure 1A and B show the observed 2D spin-diffusion NMR spectrum (only the carbonyl region was expanded) of (AG)6 A-[113 C]G14 [1-13 C]A15 G(AG)7 and the spectrum calculated by assuming the torsion angles, (φ, ψ) = (−60◦ , 130◦ ), for Ala residue, respectively. Similarly, Figure 1C shows the experimental 2D spin-diffusion NMR spectrum of (AG)6 [1-13 C]A13 [1-13 C]G14 (AG)8 together with the spectrum D calculated with the torsion angles, (φ, ψ) = (70◦ , 30◦ ), for Gly residue. In both cases, the observed spectra could be reproduced well with the calculated spectra. With these torsion angles of Ala and Gly residues determined here, the structural model of an (AG)15 chain with silk I form was prepared and shown in Figure 2. This can be called as a repeated β-turn type II structure. In order to confirm this model, REDOR experiments were performed. Namely, the atomic distance between the 13 C = O carbon atom of the 14th Gly residue and the 15 N nitrogen atom of the 17th Ala residue of (AG)15 was determined precisely as shown in Figure 3. The distance was ˚ independent of the dilution determined to be 4.0 ± 0.1 A with unlabeled (AG)15 peptide which agrees very well

Part I

Structural Analysis of Silk Fibroins using NMR

102 Part I

Chemistry

Part I Fig. 1. The (A) experimental and (B) simulated 2D spin-diffusion NMR spectra of (AG)6 A[1-13 C]G14 [1-13 C] A15 G(AG)7 and the (C) experimental and (D) simulated spectra of (AG)6 [1-13 C]A13 [1-13 C]G14 (AG)8 . The torsion angles of Ala15 residue used for the simulation of former spectrum were (φ, ψ) = (−60◦ , 130◦ ), while the torsion angles of Gly14 residue were (φ, ψ) = (70◦ , 30◦ ).

Fig. 3. Observed plots of S/S0 (= 1 − S/S0 ) values against the corresponding NcTr values for REDOR experiments of (AG)6 A[1-13 C]GAG[15 N]AG(AG)6 for the determination of distance between the 13 C = O carbon of the 14th Gly residue and the 15 N nitrogen of the 17th Ala residue. Solid and dotted lines show the theoretical dephasing curves corresponding to the designated distances. The data marked by
are observed for the isotope-labeled compound without dilution of natural abundance (AG)15 and those by , for a mixture of equivalent amount of the isotope-labeled compound and natural abundance (AG)15 . By comparing the REDOR data and the theoretical dephasing curve, the 13 C–15 N inter-atomic distance was determined to be 4.0 ± ˚ which agrees with the 4.0 A ˚ calculated for intra-molecular 0.1 A, hydrogen bond for the repeated β-turn type II-like structure.

˚ calculated for the corresponding with the distance, 4.0 A, atomic distance of the intra-molecular hydrogen bonding site in the repeated β-turn type II-like structure. This supports the structural model proposed here (Figure 2). By adding X-ray diffraction data of poly(Ala-Gly) in the silk I form to the solid-state NMR data, a more precise model with intra- and inter-molecular hydrogen bond formations alternatively was proposed for the structure in the solid state [8].

Structure of B. mori Silk Fibroin After Spinning (Silk II) Fig. 2. The conformation of a repeated β-turn type II-like molecule as a model for silk I. There are intra-molecular hydrogen bonds between the carbonyl oxygen atom of the ith Gly residue and the amide hydrogen atom of the (i + 3)th Ala residue.

As mentioned in the section “Introduction”, although Lotz et al. [3] and Fraser et al.[2] generally supported the anti-parallel β-sheet model proposed by Marsh et al. [1], the former researchers also pointed out the presence of an irregular structure in the silk fibers. More recently,

NMR of Silks

Structure of B. mori Silk Fibroin After Spinning (Silk II) 103

Part I

Fig. 4. Expanded Ala Cβ peak of (AG)15 in silk II form, model peptide of the crystalline fraction of B. mori silk fibroin fibers. Shown as dotted lines underneath are the spectral deconvolutions with Gaussian peaks.

Takahashi et al. [9] proposed that each crystal site of B. mori silk fiber is statistically occupied by two anti-parallel β-sheet chains with different relative orientations. Actually, we recently found out that the Ala Cβ peak in the 13 C CP/MAS NMR spectrum of B. mori silk fiber in the silk II form is broad and asymmetric, reflecting the heterogeneous structure of the silk fiber [6,7]. The Ala Cβ peak of the model peptide (AG)15 in silk II form was also asymmetric which consists of three peaks with isotropic chemical shifts of 22.2 (27%), 19.6 (46%), and 16.7 (27%) ppm, respectively (Figure 4). The broad peak at the highest field has essentially the same chemical shift as the sharp Ala Cβ peak at 16.7 ppm of silk I [7]. Therefore, the broad component at 16.7 ppm in Figure 4 was assigned to distorted β-turn where the averaged φ, ψ angles are the same as those of β-turn type II-like structure, but the distribution of the φ, ψ angles is larger [10]. The other two components with the chemical shifts of 19.6 and 22.2 ppm can be assigned to anti-parallel β-sheet conformation [7,10,11]. Actually, the 2D spin-diffusion NMR study indicates that the conformation of (AG)15 in silk II form is mainly an anti-parallel β-sheet with the torsion angles (φ, ψ) = (−150◦ , 150◦ ) of the Ala residue. Since the Ala Cβ methyl groups are located outside of the protein backbone, the occurrence of two peaks suggests that there is a difference in the mode of side chain packing: The

19.6 ppm peak is assigned to the Ala Cβ carbons which point in the same direction, while 22.2 ppm peak to the Ala Cβ carbons which alternately point in opposite directions as shown in Figure 4. The relative peak intensities at 22.2 and 19.6 ppm are approximately 1:2, which is in good agreement with the ratio of different packing modes suggested from X-ray diffraction analysis of B. mori silk fiber [9].

Structure of Silk Fibroin from S. c. ricini Before Spinning The 13 C CP/MAS NMR chemical shifts of Ala residue clearly indicate that the silk fibroin prepared from the silk gland of S. c. ricini and then dried mildly, take a typical α-helix structure and that the structure changed to β-sheet after spinning. This was also supported using the 2D DOQSY (double-quantum single-quantum correlation experiment) NMR measurements [12]. In order to obtain more precise structural information for the sequence of the poly-Ala region, several stable isotope-labeled peptides with the sequence, GGAGGGYGGDGG(A)12 GGAGDGYGAG, were synthesized. The torsion angles of the central Ala residue, Ala19 , in the peptide, GGAGGGYGGDGG (A)5 -[1-13 C]

104 Part I

Chemistry

Part I

A18 [1-13 C]A19 (A)5 GGAGDGYGAG were determined using 2D spin-diffusion NMR after TFA treatment. The angles were determined to be (φ, ψ) = (−59◦ , −48◦ ) which are typical angles of α-helical structures [13,14]. The torsion angles of the N- and C-terminal Gly residues adjacent to poly-Ala region were also determined using the 2D spin-diffusion NMR method for two model peptides, GGAGGGYGGD[1-13 C]G11 [113 C]G12 (A)12 GGAGDGYGAG and GGAGGGYGGDGG(A)11 [1-13 C]A24 [1-13 C]G25 GAGDGYGAG. From the error analysis of the observed and calculated spindiffusion NMR spectra [14], the torsion angles of the Gly12 and Gly25 residues were determined to be (φ, ψ) = (−70◦ , −30◦ ) and (φ, ψ) = (−66◦ , −22◦ ), respectively. In order to obtain further structural information on the C-terminal region, REDOR experiments were performed for GGAGGGYGGDGG (A)8 [1-13 C]A21 AAA[15 N]G25 GAGDGYGAG. The 13 C–15 N inter-atomic distance of the stable isotope-labeled site was deter˚ When the torsion anmined to be 4.8 ± 0.1 A. gles of Ala22 residue are (φ, ψ) = (−59◦ , −48◦ ) and those of the other two Ala residues, Ala23 and Ala24 , are (φ, ψ) = (−66◦ , −22◦ ), the atomic distance between the [1-13 C]A21 and [15 N]G25 atoms was cal˚ The similar REDOR method culated to be 4.8 A. was applied to determination of the local structure at the N-terminal region. With the torsion angles determined here, the structure of the model peptide, GGAGGGYGGDGG(A)12 GGAGDGYGAG, was proposed in Figure 5. As shown in the left side, the local structure of N- and C-terminal residues besides the α-helical poly-Ala chain is more strongly wound than those found in a typical α-helix. Namely, at the terminals of the helical region, five residues, Gly12 , Ala21 , Ala22 , Ala23 , and Ala24 contribute to the formation of i → i + 3 hydrogen bonding (see right side in Figure 5), suggesting that there are mechanisms to stabilize the α-helix structure of poly-Ala region of the silk fibroin in S. c. ricini silkworm.

Structure of Nephila clavipes Dragline Silk (MaSp1) The dragline filaments produced by orb weaving spiders have been the focus of numerous recent studies because they are the toughest protein fibers known [15–17]. The dragline silk of the golden orb web spider N. clavipes contains two structural proteins, designated spidroin 1 (MaSp1) and spidroin 2 (MaSp2) [18,19]. The dominant MaSpl protein can be described as a block copolymer consisting of poly-Ala and Gly-rich regions, which is similar to the primary structure of S. c. ricini silk fibroin. Several kinds of solid-state NMR [20–30] and X-ray diffraction methods [31,32] have been applied to clarify the structure and dynamics of native spider silk fibers.

Fig. 5. The structure of poly (L-alanine) region of the model peptide, GGAGGGYGGDGG(A)12 GGAGDGYGAG of polyalanine region of S. c. ricini silk fibroin before spinning. Both structures are the same, but different presentation. The left side presentation shows that α-helix structure of polyalanine region tend to be winded strongly at the both terminal ends. The right side presentation shows the corresponding intra-molecular hydrogen bonding pattern by broken lines.

It has been shown that silk fibroins undergo substantial structural change from gland silk to native dragline silk. In particular, it has been shown using conformationdependent 13 C chemical shifts of Ala residues that the poly-Ala region in dragline silk fiber adopts a β-sheet structure [21–25]. In contrast, the Gly-rich region in the final silk has been described as: rubber-like [33] or amorphous [31], or recently as mainly 31 -helical as shown by 2D spin diffusion [22] and by DOQSY and DECODER (direction exchange with correlation for orientation-distribution evaluation and reconstruction) solid-state NMR techniques [27]. However, it is still difficult to judge their precise local structure by solidstate NMR, because heterogeneity in the repeated sequences and resulting large variations in structural distributions must also be taken into account. To avoid the large variations in structural distributions resulting from the heterogeneity in the primary structure, we prepared both a non-labeled peptide with a sequence containing

NMR of Silks

Structure of Nephila clavipes Dragline Silk (MaSp1) 105

Part I

Fig. 6. 13 C CP/MAS NMR spectra of I (A), Ia (B), and II (C) after dissolving these peptides in 9 M LiBr and then dialyzing against water (ssb means spinning side band).

106 Part I

Chemistry

Part I

both the polyalanine and the repeated GGA regions, QGAG(A)6 GGAGA(GGA)3 GAGRGGLGG (I), and the 13 C-labeled peptides, QGAGAAA[1-13 C]A8 AAGG[213 C]A13 GAGGAG[2-13 C]G20 [3-13 C]A21 GGAGAGRGGLGG (Ia) and QGAGAAAAAAGGAGAGGAG[113 C]G20 [1-13 C]A21 GGAGAGRG-GLGG (II), as a local structural model of MaSpl protein. Solvent treatments prior to the NMR measurements induce structural change of these model peptides and provide a model to reproduce the structure of the silk fiber. Conformationdependent 13 C NMR chemical shifts were mainly used to determine the local structure, including the evaluation of the fraction of several conformations. As shown in Figure 6, the characteristic structure; 65% β-sheet for Ala8 residue in poly-Ala region, and 70% 31 -helix for Ala21 residue and mainly 31 -helix for Gly20 residue in the GG20 A21 sequence was observed after dissolving the peptides (Ia) and (II) in 9 M LiBr followed by dialysis against water. The 2D spin-diffusion 13 C solid-state NMR spectrum of the Ala21 residue of the peptide (II) after this treatment was also reproduced by 70% 31 -helix (φ, ϕ = −90◦ , 120◦ ) and 30% β-sheet structure (φ, ϕ = −150◦ , 150◦ ). However, the Ala Cβ peak assigned to 31 -helix in the spectrum of (Ia) is broad, implying that the torsion angles of Ala21 residue are distributed, but with an average that corresponds approximately to the torsion angles of the 31 -helix. Increase in the fraction of β-sheet in both poly-Ala and GG20 A21 regions was observed for (Ia) after dissolving it in formic acid and then drying in air. Moreover, after dissolving (Ia) in formic acid and then precipitating it in methanol, the spectrum showed a tightly packed β-sheet structure with further increase in the fraction of β-sheet although 15% 31 -helix still remained in the GG20 A21 region. The β-sheet structure of poly-Ala region, and both 31 -helix and β-sheet structures in the repeated GGA sequence is in agreement with the structural model for the native spider dragline silk fiber from N. clavipes from a previous NMR study. On the other hand, α-helical conformation was found to be dominant for the peptide treated with trifluoroacetic acid together with a significant contribution from other structures. The fraction of the other structures was 20–40% depending on the position of 13 C-labeled Ala residue.

References 1. Marsh RE, Corey RB, Pauling L. Biochem. Biophys. Acta. 1955;16:1.

2. Fraser RD, MacRae TP, Stewart FH. J. Mol. Biol. 1966;19: 580. 3. Lotz B, Brack A, Spach G. J. Mol. Biol. 1974;87:193. 4. Asakura T, and Nakazawa Y. Macromol. Biosci. 2004;4:175. 5. Asakura T, Ashida J, Yamane T, Kameda T, Nakazawa Y, Ohgo K, Komatsu K. J. Mol. Biol. 2001;306:291. 6. Asakura T, Yao J, Yamane T, Umemura K, Ulrich AS. J. Am. Chem. Soc. 2002;124:8794. 7. Asakura T, Yao J. Protein Sci. 2002;11:2706. 8. Asakura T., Ohgo K., Komatsu K., Kanenari M., Okuyama K. Macromolecules 2005;38:7397. 9. Takahashi Y, Gehoh M, Yuzuriha K. Int. J. Biol. Macromol. 1999;24:127. 10. Asakura T, Kuzuhara A, Tabeta R, Saito H. Macromolecules. 1985;18:1841. 11. Ishida M, Asakura T, Yokoi M, Saito H. Macromolecules. 1990;23:88. 12. van Beek JD, Beaulieu L, Schafer H, Demura M, Asakura T, Meier BH. Nature. 2000;405:1077. 13. Nakazawa Y, Bamba M, Nishio S, Asakura T. Protein Sci. 2003;12:666. 14. Nakazawa Y, Asakura T. J. Am. Chem. Soc. 2003;125: 7230. 15. O’Brien J, Fahnestock S, Termonia Y, Gardner K. Adv. Mater. 1998;10:1185. 16. Gosline JM, Guerette PA, Ortlepp CS, Savage KN. J. Exp. Biol. 1999;202:3295. 17. Vollrath F, Knight DP. Nature. 2001;410:541. 18. Xu M, Lewis RV. Proc. Natl. Acad. Sci. U.S.A. 1990;87:7120. 19. Hinman MB, Lewis RV. J. Biol. Chem. 1992;267:19320. 20. Simmons A.H., Ray E.D., Jelinski L.W. Macromolecules 1994;27:5235. 21. Simmons A.H., Michal, C.A., Jelinski, L.W. Science 1996;271:84 22. K¨ummerlen J., Van Beek J.D., Vollrath F., Meier B.H. Macromolecules 1996;29:2920. 23. Michal C.A., Jelinski L.W. J. Biol. NMR. 1998; 12:231. 24. van Beek J.D., K¨ummerlen J., Vollrath F., Meier B.H. Int. J. Biol. Macromol. 1999;24:173 25. Seidel A., Liivak O., Calve S., Adaska J., Ji G., Yang Z., Grubb D., Zax D.B., Jelinski L.W. Macromolecules 2000;33:775. 26. Yang Z., Liivak O., Seidel A., Laverda G., Zax D.B., Jelinski L.W. J. Am. Chem. Soc. 2000;122:9019 27. van Beek J.D., Hess S., Vollrath F., Meier B.H., Proc. Natl. Acad. Sci. USA 2002;99:10266. 28. Eles P.T., Michal C.A. Biomacromolecules 2004;5:661. 29. Eles P.T., Michal C.A. Macromolecules 2004;37:1342. 30. Holland G.P., Lewis R.V., Yarger J.L. J. Am. Chem. Soc. 2004;126:5867. 31. Grubb D.T. and Jelinski L.W. Macromolecules 1997;30:2860. 32. Riekel C., Br¨anden C., Craig C., Ferrero C., Heidelbach F., M¨uller M. Int. J. Biol. Macromol. 1999;24:179. 33. Gosline J.M., Denny M.W., DeMont M.E. Nature 1984;309:551.

Part I

Field Gradient NMR

109

William S. Price Nanoscale Organization and Dynamics Group, College of Science, Technology and Environment, University of Western Sydney, Penrith South, NSW 1797, Australia

Diffusion as a Probe Translational diffusion is inherently important in the chemical and biological world since it constitutes the most basic form of transport. But its study is also important by virtue of the translational motion of a species being affected by molecular interactions (e.g. binding) or restricted diffusion by physical barriers (e.g. within a pore). Thus, the diffusion of a species provides a rich source of information regarding the interactions of a species with other molecules and its environment. NMR provides a conceptually simple but direct and extremely powerful method for measuring diffusion down to about 10−14 m2 /s. In contrast to traditional methods, and of particular significance since the species of interest is likely to be a small molecule or ion, the NMR method (generally) does not require labeling and is effectively non-invasive. This chapter provides a brief introduction to the Pulsed Gradient Spin-Echo (PGSE) NMR method (also commonly referred to as Affinity NMR, DOSY, or PFG NMR) for measuring diffusion and its application to solution dynamics and probing porous media. Already a very large literature exists on NMR diffusometry and applications, and thus the literature cited here is only as an example and is in no way comprehensive. A number of reviews have already appeared including of a general nature [1–8] and specializing in supramolecular and combinatorial chemistry [9], polymer gels [10], proteins [11], transport and binding [12–14], and surfactants [15,16].

Gradient-Based Diffusion Measurements Although technically difficult the concept underlying the PGSE technique is breathtakingly simple. All PGSE sequences are, as the name suggests, based on some form of spin-echo sequence. We will illustrate the operation of the PGSE sequence with the simplest case, that of a Hahn spin-echo based sequence. From the earliest days of NMR it was realized that the refocusing of the echo in the Hahn sequence could be compromised by the effect of magnetic field gradients. Since should a spin move to a region with a different magnetic field during the sequence, the phase change acquired during the first τ period would not be Graham A. Webb (ed.), Modern Magnetic Resonance, 109–115.
C 2008 Springer.

counteracted by that experienced in the second τ period (recall that the effect of the π pulse is to reverse the sign of the phase change that has accumulated prior to its application). Theoretical modeling of this attenuation of the echo due to spins experiencing different magnetic fields is facilitated if the applied magnetic gradient is constant (often mistakenly referred to as “linear”). The imposition of a magnetic field gradient during the rf pulses and acquisition is deleterious: much stronger rf pulses are required to overcome the gradient induced spreading of the spectrum and chemical shift information is lost during acquisition. Further, it would also result in the timescale of the diffusion measurement being tied to τ . A much better, albeit technically more demanding solution, is to apply the magnetic gradient in the form of two equal pulses of length δ and magnitude and direction g as depicted in Figure 1. The area of such gradient pulses (i.e. δg) leads to the definition of the reciprocal space vector   q = (2π)−1 γ gδ m−1 .

(1)

It is easily imagined that an infinitely short gradient pulse (SGP) (i.e. δ → 0 and |g| → ∞ while δg remains finite) with g directed along the long axis of a cylindrical sample would wind (i.e. spatially encode) the transverse magnetization into a helix with pitch q−1 (m). If instead the pulse had finite duration, the effect of translational diffusion during the pulse would corrupt the helix formation. Assuming the SGP condition so that motion during the gradient pulse can be neglected, but accounting for motion during the period  between the first gradient pulse and the second (i.e. spatially decoding) gradient pulse leads to the SGP relation for the spin-echo attenuation [17] E=

ρ(r0 ) P(r0 , r1 , ) ei2πq(r1 −r0 ) dr0 dr1

(2)

where ρ (r0 ) is the equilibrium spin-density and P (r0 , r1 , ) is the diffusion propagator [18] (or Green function [19]) derived using appropriate boundary conditions and a delta function initial condition. The integral

Part I

NMR Diffusometry

110 Part I

Chemistry

Part I

A

B τ

τ

π/2

π

t1 δ

q -1

t2

S

g ∆ Fig. 1. (A) A PGSE sequence based on a Hahn-spin echo where two equal gradient pulses of duration δ and magnitude g are inserted into each τ period. Typically δ is in the range of 1–10 ms, whilst the separation  between the leading edges of the gradient pulses is normally in the range of 10 ms to 1 s. The second half of the echo is used as the NMR signal. Normally the echo attenuation, defined by E(g) = S(g)/S(g = 0), is used to determine the diffusion coefficient as it allows the effects of spin relaxation to be normalized out. A gradient pre-pulse is included before the π/2 rf pulse to reduce eddy current and gradient mismatch effects. (B) An example of a magnetization helix (the arrows represent nuclear spins and the spiral curve is a guide for the eye), with pitch q−1 that would be formed by applying a gradient pulse along the long axis of a cylindrical sample. Any imperfection in gradient constancy or motion during the pulse results in a distorted helix.

is taken over all starting (r0 ) and finishing (r1 ) positions. Equation (2) states that the echo attenuation is given by the Fourier transform of the diffusion propagator with respect to q. In the case of free diffusion, Equation (2) leads to   E = exp −4π 2 q 2 D .

(3)

Importantly, diffusion is measured along the direction of the gradient. Equation (3) states that the echo attenuation for the simple case of free diffusion is given by a single exponential. Despite the relative simplicity of the SGP approximation, apart from free isotropic diffusion, solutions to Equation (2) are only available for simple symmetrical geometries, such as between reflecting planes separated by a distance a, viz [17]. 2[1 − cos(2πqa)] E = + 4(2πqa)2 (2πqa)2 
2 2 ∞  n π D 1 − (−1)n cos(2πqa) × exp −

2 a2 (2πqa)2 − (nπ)2 n=1 (4) At long , the second term in Equation (4) disappears leaving the long-time diffractive behavior with diffractive minima, which arise from the first term, appearing at q = n/a (n = 1, 2, 3, . . .). In cases where P (r0 , r1 , ) is unknown, expansion of Equation (2) reveals that for small q with respect to the characteristic distance of the restricting geometry, R, the

echo attenuation is given by,    (2πq)2 z 2 () E q  R , ≈ 1 − (5) 2   where z 2 () is the mean squared-displacement along the direction of the gradient (i.e. taken to be along the z-axis) and in such cases the PGSE data can be analyzed on the basis of an effective diffusivity [20,21] 

−1

 Deff () =

 z 2 () . 2

(6)

Experimental Complications Here we consider some experimental complications peculiar to PGSE NMR measurements. In almost all cases experimental imperfections lead to faster “apparent” diffusion coefficients.

Finite Gradient Pulses Experimentally, the SGP approximation is never completely justified and including its effect in the theoretical analysis is difficult. For example, the analytical solution obtained from solving the Bloch equations is [22,23]   E = exp −γ 2 g 2 D( − δ/3) . (7) Comparison with Equation (3) reveals that the δ/3 term is a correction for the finite length of the gradient pulses. In

NMR Diffusometry

Background Gradients Due to magnetic susceptibility differences resulting from sample heterogeneity and sample interfaces, the presence of background magnetic gradients is unavoidable. And their effects are insidious on the PGSE measurement [26]. Assuming a simple case of a constant background gradient through the sample of direction and magnitude g0 , the analytical solution starting from the Bloch equations is [22,27] ⎡ ⎛

measurements of strong NMR resonances and can cause effects similar to those caused by background gradients except that the non-exponential behavior is insensitive to the polarity of the applied gradient [26]. Apart from using a very small sample, the only three practicable and generally applicable means by which accurate PGSE experiments can be conducted in conditions that radiation damping will occur are: (1) by keeping all transverse magnetization spatially encoded during as much of the sequence as possible, (2) allowing part of the magnetization to (reproducibly) decay before starting the diffusion part of the sequence or (3) to use Q-switching [31].

Convection

⎢ ⎜ ⎢ ⎜ E (g, g0 ) = exp ⎜−γ 2 ⎢g 2 Dδ 2 ( − δ/3) ./ 0 ⎣⎝ g term

⎤⎞   ⎥⎟ 2 ⎥⎟ + g · g0 Dδ t12 + t22 + δ (t1 + t2 ) + δ 2 − 2τ 2 ⎥⎟ ⎦⎠ 3 ./ 0 g·g0 cross terms

(8) where the delays t1 and t2 are defined in Figure 1A. The difference between Equations (7) and (8) is that the inclusion of the g·g0 cross terms results in the attenuation being no longer described by a single exponential. The presence of background gradients can be detected by reversing the sign of the applied gradient and thus the cross term. Consequently, sequences incorporating bipolar pulses have been devised to ameliorate PGSE measurements in the presence of background gradients [28–30].

In PGSE measurements of low viscosity samples away from ambient temperature, convective motion can have particularly deleterious effects [32,33]. Importantly, whereas the diffusion coefficient depends on molecular size, the (convective) flow velocity is common to all of the species in the sample. Whereas a net flow of spins along the direction of the gradient is clearly indicated by the resulting net phase change in the PGSE spectra, convection currents do not produce a phase change since the flow of the spins along the direction of the gradient is exactly matched by the flow in the anti-parallel direction [34,35]. Convection causes a cosine modulation of the PGSE signal attenuation (for a single diffusing species). But due to the similarity between the cosine and Gaussian functions, the PGSE data appears to be well described by an exponential [e.g. Equation (3)] but with an apparent diffusion coefficient that increases with . Apart from improving the temperature regulation to decrease any temperature gradients modifying the sample holder to limit flow, convection can also be minimized by using specialized pulse sequences (e.g. Ref. [36]).

Radiation Damping Gradient Constancy In samples with a high concentration of spins (e.g. a sample of water), a feedback loop can arise in which the precessing spin magnetization generates an oscillating current in the receiver coil, which in turn generates an oscillating magnetic field, which rotates the magnetization back to its equilibrium position—generally much more rapidly than that would occur due to longitudinal relaxation. The effect increases with the strength of the static magnetic field. In the PGSE experiment, radiation damping is active in the periods in which the magnetization is not spatially encoded (i.e. during the periods t1 and t2 in the sequence as depicted in Figure 1A). Radiation damping complicates the performing of diffusion

Small deviations from constancy of the applied gradient throughout the sample volume do not generally cause serious errors [37]. Nevertheless, as gradient coils only produce a constant gradient over a small volume, to ensure reasonable constancy, the NMR active volume of the sample must be restricted. Often the effective sample volume will be sufficiently restricted by virtue of the size of the rf coils. More generally it is necessary to physically limit the sample volume (although care must be taken not to introduce background gradients) or by including a slice selective element in the PGSE sequence (e.g. Ref. [38]).

Part I

general, analytical solutions are impossible and numerical approaches are indicated [24,25].

Experimental Complications 111

112 Part I

Chemistry

case of a sphere under stick boundary conditions the friction coefficient, f is given by (Stokes law)

Eddy currents in the surrounding conducting surfaces around the gradient coils (e.g. probe housing, etc.) arise from the rapid rise and fall of the gradient pulses and their severity increases with the speed of the rise time and the strength of the gradient pulses. The advent of shielded gradient coils has greatly decreased their effects, but they can still be significant when using large rapidly rising and falling gradient pulses. The decay time of the eddy currents (and their associated magnetic fields) determine the minimum delay required between the end of the gradient pulse and the start of spectral acquisition. Eddy currents can result in: (i) gradient induced broadening of the observed spectrum, (ii) phase changes and anomalous changes in the attenuation, and (iii) time-dependent but spatially invariant B0 shift effects (which appears as “ringing” in the spectrum). Gradient pulse mismatch can produce similar artifacts to eddy currents [39]. Even extremely small mismatches can cause a severe loss in echo signal intensity due to the resulting phase twist. If the mismatch increases as a function of gradient strength it has the potential to produce artifactual “diffraction” peaks. The presence of such phase-based artifacts is verified, for example, by performing measurements on a freely diffusing sample with a very small diffusion coefficient (e.g. large polymer). Apart from the obvious solution of better gradient generation, the easiest means for removing eddy current and gradient mismatch effects is to use shaped gradient pulses [40] or prefixing a number of -spaced gradient prepulses (see Figure 1A) [41]. In the case of gradient mismatch it is also possible, although considerably less convenient, to empirically match the gradient pulse pairs, or, at the expense of chemical shift information, use the imagingbased MASSEY sequence approach [42].

f = 6πη R

(10)

where R is the effective hydrodynamic (or Stokes) radius, η is the solvent viscosity. Since f is determined by the overall dimensions of the diffusing species (which may include the effects of solvation and rugosity), few species are well described by a simple geometry. Consequently, f must normally be determined numerically (e.g. Ref. [44])—indeed exact solutions are only known for some simple geometries (e.g. see Table 1 in Ref. [45]). When NMR diffusion measurements are used to separate mixtures on the basis of diffusion differences, it is often referred to as DOSY NMR [7] with the resulting data presented in a two-dimensional format with the diffusion coefficient on one axis and the chemical shift on the other. Some examples of the applications of diffusion measurements are given in the following subsections.

Solution Dynamics and Surfactants NMR diffusometry finds particular application in studying solution dynamics—especially since it is capable of determining the diffusion behavior of many of the species in a solution simultaneously. For example, the diffusion coefficient of all species in an ethanol–water solution are given in Figure 2. Such detailed data allows inferences to

2.0

2 -1

In the absence of restriction, the diffusion coefficient of a species reports directly on its size, geometry, and the medium in which it is diffusing. This connection is conveniently formulated using the Stokes–Einstein equation, which is derived assuming that the solute sees the solvent as a continuum (e.g. see Ref. [43]),

-9

1.5

Diffusion in Complex Systems

D0 =

kT f

D × 10 m s

Part I

Eddy Currents and Gradient Mismatch

1.0

0.5

0.0

(9)

where D 0 is the diffusion coefficient of the solute at infinite dilution (hence the superscript 0), k is the Boltzmann constant, and T is temperature. For the particularly simple

0.0

0.2

0.4

0.6

0.8

1.0

X2 Fig. 2. Diffusion coefficients of the alkyl (
), hydroxyl (∗), water (•), and water-hydroxyl (䊎) groups at 285 K at various ethanol mole fractions, X A , in the ethanol–water system.

NMR Diffusometry

Kd

PL  P + L). The coupled differential equations describing the echo signal intensities at the free and bound sites are (e.g. see [21,39–41]), dSf Sf Sb + = −γ 2 g 2 Df δ 2 Sf − dt τf τb

4

D (× 10

-10

2 -1

ms )

6

(12)

Sb Sf dSb + = −γ 2 g 2 Db δ 2 Sb − dt τb τf

2

0.01

0.1 c (wt%)

1

Fig. 3. Determination of the cmc of SDS in D2 O from NMR diffusion measurements as a function of surfactant concentration. The break in the data at Ct = 0.2 wt% represents the cmc (modified from Ref. [48]).

where τ f and τ b are the lifetimes in the free and bound sites, respectively. The initial conditions are given by Sf |t=0 = Pf = (1 − Pb ) and Sb |t=0 = Pb where Pf and Pb are the populations in the free and bound sites, respectively. At t =  and in the case of fast exchange, this reduces to the particularly simple single exponential form,   E = Sb + Sf = exp −γ 2 g 2 Dobs δ 2  (13) where Dobs = (1 − Pb ) Df + Pb Db

be drawn regarding the complicated solution chemistry of this system [46]. Since the diffusion coefficient directly reflects molecular size, diffusion measurements have been used to great effect in determining the critical micellar concentration (cmc) of surfactants. Typically the associating surfactant solution is modeled using a two-site exchange model, in which the observed diffusion coefficient is expressed as a population weighted average between “free” and “bound” (i.e. surfactants in micelles) surfactant [47]:


D = Df

Cf Cf + Db 1 − Ct Ct

 (11)

where Df,b are the diffusion coefficients of free and micellized surfactants, respectively. Cf,t is the concentration of free and total concentration of surfactant (NB C f /Ct is the free population), respectively. An example of determining the cmc from diffusion data is given in Figure 3.

Ligand Binding and Aggregation Since diffusion is an excellent probe of molecular size and mobility, NMR diffusometry is becoming an increasingly important tool in drug development where it is sometimes referred to as “affinity NMR” [49]. As an illustration, consider the simple two site system where a drug (i.e. ligand, L) exchanges between being in free solution to any one of n equivalent binding sites on the protein (P) with a dissociation constant K d (i.e.

(14)

is the population-weighted average diffusion coefficient. In the case of this simple two-site model, the bound population is given by Pb = α −

2

α2 − β

(15)

and α=

(CL + nCP + K d ) 2CL

and β =

nCP CL

(16)

where CL and CP are the total concentrations of drug and protein, respectively. Df can be determined by measuring the diffusion of the drug in protein-free solution, and Db can normally be taken as equal to the protein diffusion coefficient since the binding of the drug should have negligible effect on the diffusion coefficient of the (much larger) protein molecule. An example of an NMR diffusometry study of drug binding is given in Figure 4.

Restricted Diffusion As noted above, when diffusion occurs within a restricting geometry, the geometrical restrictions result in characteristic echo attenuation curves. Thus, diffusion measurements provide a powerful means of probing porous materials. An example of probing a simple pore in which water is diffusing between two parallel planes is given in Figure 5.

Part I

10 8

Diffusion in Complex Systems 113

114 Part I

Chemistry

Part I Fig. 4. An example of an NMR diffusion measurement for studying drug binding: A 500 MHz 1 H PGSE-WATERGATE spectra of 80 mM salicylate and 0.5 mM bovine serum albumin in water at 298 K. The (residual) water resonance gives rise to the peak at 4.7 ppm and the three peaks to the left originate from salicylate (from left to right: H-6, H-4, H-3/H-5; also see inset) (modified from Price et al. [50]).

Acknowledgment 1 The NSW State Government is acknowledged for support through a BioFirst award. -1

E(q)

10

References

-2

10

-3

10

-4

10

0.0

0.5

1.0

1.5 5

2.0

2.5

3.0

-1

q (10 × m ) Fig. 5. 1 H PGSE NMR attenuation profile for water diffusing between planes separated by distance a = 128 µm at 318 K. The gradient is directed perpendicular to the planes. The experimental parameters were  = 2 s and δ = 2 ms. The solid black line denotes the result of fitting the data with the SGP formula [Equation (4)]. The diffractive minima appear at q = n/a(n = 1, 2, 3, . . .) (modified from Ref. [51]).

1. Callaghan PT. Aust. J. Phys. 1984;37:359. 2. Stilbs P. Prog. NMR Spectrosc. 1987;19:1. 3. K¨arger J, Pfeifer H, Heink W. Adv. Magn. Reson. 1988; 12:1. 4. Callaghan PT. Principles of Nuclear Magnetic Resonance Microscopy. Clarendon Press: Oxford, 1991. 5. Price WS, In: GA Webb (Ed). Annual Reports on NMR Spectroscopy. Academic Press: London, 1996, p. 51. 6. Kimmich R. NMR: Tomography, Diffusometry, Relaxometry. Springer Verlag: Berlin, 1997. 7. Johnson CS Jr. Prog. NMR Spectrosc. 1999;34:203. 8. Stilbs P. In: JC Lindon, GE Tranter, JL Holmes (Eds). Encyclopedia of Spectroscopy and Spectrometry. London, 2000, p. 369. 9. Cohen Y, Avram L, Frish L. Angew. Chem. Int. Ed. 2005;44:520. 10. Matsukawa S, Yasunaga H, Zhao C, Kuroki S, Kurosu H, Ando I. Prog. Polym. Sci. 1999;24:995. 11. Price WS. In: GA Webb (Ed). Annual Reports on the Progress in Chemistry Section C. Royal Society of Chemistry: London, 2000, p. 3. 12. Waldeck AR, Kuchel PW, Lennon AJ, Chapman BE. Prog. NMR Spectrosc. 1997;30:39.

NMR Diffusometry

33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.

Hedin N, Yu TY, Fur´o I. Langmuir. 2000;16:7548. Jerschow A. J. Magn. Reson. 2000;145:125. Mohoric A, Stepiˇsnik J. Phys. Rev. E. 2000;62:6628. Jerschow A, M¨uller N. J. Magn. Reson. 1997;125:372. H˚akansson B, J¨onsson B, Linse P, S¨oderman O. J. Magn. Reson. 1997;124:343. Xia Y. Concepts Magn. Reson. 1996;8:205. Price WS, Hayamizu K, Ide H, Arata Y. J. Magn. Reson. 1999;139:205. Price WS, Kuchel PW. J. Magn. Reson. 1991;94:133. von Meerwall E, Kamat M. J. Magn. Reson. 1989;83:309. Callaghan PT. J. Magn. Reson. 1990;88:493. Tyrrell HJV, Harris KR. Diffusion in Liquids: A Theoretical and Experimental Study. Butterworths: London, 1984. Garc´ıa de la Torre J, Huertas ML, Carrasco B. Biophys. J. 2000;78:719. Price WS. In: Atta-Ur-Rahman (Ed). New Advances in Analytical Chemistry. Harwood Academic Publishers: Amsterdam, 2000, p. 31. Price WS, Ide H, Arata Y. J. Phys. Chem. A. 2003;107: 4784. S¨oderman O, Stilbs P. Prog. Nucl. Magn. Reson. Spectrosc. 1994;26:445. Pettersson E, Topgaard D, Stilbs P, S¨oderman O. Langmuir 2004;20:1138. Lin M, Shapiro MJ, Wareing JR. J. Org. Chem. 1997;62:8930. Price WS, Elwinger F, Vigouroux C, Stilbs P. Magn. Reson. Chem. 2002;40:391. Price WS, Stilbs P, S¨oderman O. J. Magn. Reson. 2003;160:139.

Part I

13. Fielding L. Tetrahedron. 2000;56:6151. 14. Price WS. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. Wiley: New York, 2002, p. 364. 15. S¨oderman O, Stilbs P. Prog. NMR Spectrosc. 1994;26:445. 16. Fur´o I. J. Mol. Liquids. 2005;117:117. 17. Tanner JE, Stejskal EO. J. Chem. Phys. 1968;49:1768. 18. K¨arger J, Heink W. J. Magn. Reson. 1983;51:1. 19. Duffy DG. Green’s Functions with Applications. CRC: Boca Raton, 2001. 20. K¨arger J, Fleischer G, Roland U. In: J K¨arger, P Heitjans, R Haberlandt (Eds). Diffusion in Condensed Matter. Vieweg: Braunschweig, 1998, p. 144. 21. Ben-Avraham D, Havlin S. Diffusion and Reactions in Fractals and Disordered Systems. Cambridge University Press: Cambridge, 2000. 22. Stejskal EO, Tanner JE. J. Chem. Phys. 1965;42:288. 23. Price WS. Concepts Magn. Reson. 1997;9:299. 24. Callaghan PT. J. Magn. Reson. 1997;129:74. 25. Price WS, S¨oderman O. Isr. J. Chem. 2003;43:25. 26. Price WS, Stilbs P, J¨onsson B, S¨oderman O. J. Magn. Reson. 2001;150:49. 27. Price WS. Concepts Magn. Reson. 1998;10:197. 28. Cotts RM, Hoch MJR, Sun T, Markert JT. J. Magn. Reson. 1989;83:252. 29. Wu D, Chen A, Johnson CS Jr. J. Magn. Reson. A. 1995;115:260. 30. Seland JG, Sørland GH, Zick K, Hafskjold B. J. Magn. Reson. 2000;146:14. 31. Price WS, W¨alchli M. Magn. Reson. Chem. 2002;40;S128. 32. Hedin N, Fur´o I. J. Magn. Reson. 1998;131:126.

References 115

117

Istv´an Fur´o1 and Sergey V. Dvinskikh2 1 Department

of Chemistry, Division of hPysical Chemistry, Royal Institute of Technology;and 2P hysical Chemistry, Stockholm nUiversity, Stockholm, Sw eden

Abstract Field-gradient NMR applications in liquid crystals (LCs) are dominantly experiments that detect the translational self-diffusion of the various structural units of the liquid crystalline phases. The anisotropy of LCs often leads to line broadening effects that must typically be suppressed in order to accommodate sufficient gradient dephasing of the nuclear spins. Having dealt with this problem, diffusion studies provide important insights into both lyotropic and thermotropic liquid crystal systems.

Introduction Liquid crystals (LCs), although anisotropic, consist of highly mobile molecules just like “usual” isotropic liquids. Hence, all NMR methods that are used for extracting molecular information in isotropic liquids could, in principle, furnish the same type of information in LCs, with one important difference: anisotropy of LCs results in tensor instead of scalar properties. Thus, for example, translational self-diffusion, characterized by a scalar diffusion coefficient D in the isotropic case, becomes instead dependent in LCs on a diffusion tensor D [1–5]. By diagonalizing D, one can in general extract three pieces of information, represented by the principal components Dαα , Dββ , and Dγ γ of D [4,6]. Of these three components, two are equal for LCs with a symmetry axis of higher than threefold symmetry. Here, we intend to provide a brief survey, with representative examples and directions to relevant reviews, of magnetic-field-gradient-based NMR investigations of LCs. Most of the involved studies aimed at measuring self-diffusion in those materials and therefore diffusion studies dominate here, with a few other gradient applications mentioned at the end. Our survey is not chronological and, with over 103 relevant publications, cannot be comprehensive. Neither can we elaborate upon related issues, such as the large and still emerging area of diffusion tensor imaging [4,7–10]. With very few exceptions (such as lyotropic cubic phases), LCs [11] are formed by anisomeric objects, often called mesogenic units. Liquid crystalline ordering stems from the anisotropic pair potential among those objects Graham A. Webb (ed.), Modern Magnetic Resonance, 117–122.
C 2008 Springer.

[12]. In thermotropics, one of the two broad classes of LCs, the mesogenic units are typically elongated or flat molecules with no other principal system component. A typical example is 4-pentyl-4 -cyanobiphenyl (5CB) where a rigid central biphenyl group lends the molecule an elongated shape. The other broad LC class contains the lyotropic systems where the common element is a solvent (typically water) that embeds the mesogenic units that can be multi-molecular aggregates but also single molecules. Typical example for the former are lyotropic LCs formed by elongated (e.g. in hexagonal phases) or flat (e.g. in lamellar phases) aggregates of amphiphilic molecules, either surfactants or lipids. Lipid-based LCs, akin to lipid bilayer [13] structures of cells, have a clear biological relevance.

NMR Methods and Diffusion in LCs Since LCs are anisotropic, their NMR properties depend on the orientation with respect to the applied magnetic field [1,2,4,5,14]. Moreover, anisotropic spin couplings, such as the dipole–dipole or quadrupole interactions, are not averaged by molecular motions to zero but to a residual value. Often, the relation between the instantaneous and residual couplings is simply defined by a scalar order parameter S of the LC phase while in some phases, such as in biaxial ones, this relation may take instead a more complex form [2]. In some LCs, with cubic phases as an example, the residual coupling may vanish by symmetry and the spectra containing narrow lines become similar to those recorded in isotropic liquids. The manifestation of non-vanishing residual couplings depends on the macroscopic orientational order within the sample. For simplicity, we exemplify this with a uniaxial LC phase where the average orientation of molecules in a given spatial region is defined by a unique direction, the LC director d. The NMR signal given by those molecules depends on the orientation of d with respect to the applied static magnetic field B0 . Some LCs can be prepared, either through mechanic [15–18] or electromagnetic [19–25] interactions, in a homogeneous state with the same d all over the sample volume. Such macroscopically oriented LCs are contrasted to “powder” samples with d that varies randomly from domain-to-domain: if the domain

Part I

Field Gradient NMR of Liquid Crystals

118 Part I

Chemistry

Part I

size is small, on the experimental time scale translational diffusion may average the residual coupling to zero. The appearance of the spectra also depends on the involved spin couplings. The pairwise dipole–dipole interaction typically acts in LCs among a manifold of nuclei. Hence, static spectral splitting typically renders the spectra of dipole–dipole coupled 1 H nuclei wide and featureless, even in macroscopically oriented samples, because it consists of many overlapping lines. The only exception is having the director of a macroscopically oriented sample at the “magic angle” to B0 , where all static splitting vanishes. In contrast to the dipole–dipole case, singlespin quadrupole interactions for spin I > 1/2 nuclei in a macroscopically oriented LC may result in spectra split into (2I + 1) sharp lines. Powder samples exhibit large static broadening, both for the dipole–dipole and quadrupole cases but often with a discernible “powder pattern” line shape [26] for the latter one. Spectral broadening translates into a quick decay of spin coherences in the time domain that has a direct bearing on field-gradient NMR experiments for diffusion [27–31]. Irrespective of the specific experiment (pulsedfield-gradient spin or stimulated echo and variations, the former ones often abbreviated as PGSE and PGSTE), NMR detects displacement via gradient-assisted de- and re-phasing of spin coherences. In the absence of motion, all de-phased magnetization can be recovered. Random diffusive displacement of individual molecules introduces random re-phasing errors that manifest themselves in an increasing signal loss upon increasing de-phasing (∼γ gδ, where g and δ are the strength and the length of the applied gradient pulses, respectively and γ is the gyromagnetic ratio), diffusion time , and diffusion coefficient D. In isotropic liquids, the result is described by the wellknown Stejskal-Tanner [32–34] expression I (g, δ, ) ∼ exp[−Dγ 2 g 2 δ 2 ( − δ/3)],

(1)

whose Gaussian appearance is a direct consequence of the Gaussian spatial propagator for translational self-diffusion. In anisotropic systems this expression is straightforwardly modified to accommodate for the anisotropy of the system as [1,2,4,33] I (g, δ, ) ∝ exp[−(γ δ)2 ( − δ/3)gDg];

(2)

note that 2 the relative orientation of the gradient vector g (g = g2 ) and the principal axes of the diffusion tensor D affects the experimental outcome. Clearly, diffusion can be measured only if the spins can be sufficiently dephased which is strongly limited in LCs with quick decays of spin coherences. For the same reason, most diffusion experiments were performed in nuclei with large γ (such as 1 H and 19 F).

The presented solutions suppress the spectral broadening effects by residual spin couplings and involve either the mechanical manipulation of the sample orientation or the radiofrequency (rf) manipulation of the involved spins or both. The simplest and earliest [35–37] method in the former class involves preparation of a homogeneously oriented sample which is then placed by its director at the magic angle with respect to B0 which often results in a sufficient reduction of the static line broadening [1,38,39]. The disadvantage with the technique is its demand on homogeneous orientation that must be both settable and sustainable at the magic angle. An offspring of this technique is a diffusion experiment performed under MAS conditions [40–43] with the gradient field set along the spinning axis. The other broad option involves various decoupling or echo-based refocusing techniques applied under the de-phasing and re-phasing periods of a diffusion experiment [44–52]. The difficulty with those experiments is to maintain the performance of the selected rf pulse sequences under the far-offset conditions set by the simultaneous application of the field gradient. Slice selection, although at the cost of signal-to-noise ratio, is a straightforward option [45–52]. Spatial slice selection and decoupling can also be replaced by spectral slice selection by long selective pulses [53]. Finally, we note that instead of suppressing line broadening one may try to use stronger field gradients; one way of obtaining such is to use static instead of pulsed ones [54,55]. The disadvantage of that technique is the additional line broadening and connected reduction in signal-to-noise ratio caused by the static gradient. In whichever case, a full characterization of the diffusion tensor D requires experiments performed at several relative gradient orientations. This can be achieved on two principally different ways. First, in a homogeneously oriented sample several experiments can be performed with different gradient directions [56–58] with respect to the sample. Conventionally, D|| and D⊥ denote diffusion along and perpendicular to d. Our examples, shown in Figure 1, are taken from such type of studies. The other option, applicable in unoriented powder samples, exploits the spectral broadening itself by anisotropic spin interactions. In favorable cases, those interactions provide correspondence between the spectral frequency and domain orientation. Hence, differential decay of powder spectra (either by quadrupole interaction [59] or by chemical shift anisotropy [48,60]) for just one gradient direction reveals the complete diffusion tensor D. If diffusion within individual domains of unoriented powder samples cannot be orientationally assigned (as is typical for 1 H nuclei), the diffusional decay becomes the composite of decays for different gradient orientations [4,62–66]. If the orientational distribution is completely random and other effects, such as restricted diffusion, do not complicate the evaluation, the diffusion tensor can also

Field Gradient NMR of Liquid Crystals

0.9

310

300

T (K) 290

280

0.6

D

0.4

Diso

0.7

D//

0.6

L

N

I

D

=

D(10-10m2/s)

D/D0

0.8 0.2

0.1 0.08

D

0.06

-12

-10

-8

-6

-4

-2

0

T--TNl

A

2 0.04

Isotropic Nem Smectic A 3.1

3.2

3.3

3.4

3.5

3.6

1000/T (1/K)

B Fig. 1. (A) Temperature dependence of anisotropic diffusion across several lyotropic, [61] and (B) thermotropic [51] phases, measured by pulsed-field-gradient spin-echo-type experiments. In (A), 2 H pulsed-field-gradient quadrupole-echo experiment is applied to heavy water in its mixture with cesium perfluorooctanoate (CsPFO). This fluorinated surfactant forms in water flat aggregates, which exhibit nematic order (with their short axis along the field direction) upon cooling below the isotropic–nematic transition temperature TNI . Upon further cooling, the system enters into a lamellar phase (L) consisting of defective CsPFO-bilayers. In (B), the diffusion of the mesogenic unit 4-octyl-4 -cyanobiphenyl (8CB) is followed by 2 H pulsed-field-gradient stimulated-echo experiments performed under simultaneous decoupling. Reproduced with permission.
C American Physical Society, 1996, 2002.

be extracted either from the composite decay [4,62–66] or from more advanced diffusion–diffusion correlation (or exchange) experiments [67–69].

Lyotropic Applications Although some molecular lyotropic phases have been investigated by NMR, here we restrict ourselves to systems where the mesogenic units are aggregates of simple surfactants and/or lipids (and thereby also exclude discussion of, e.g. block-copolymer-based lyotropic materials [64,65,70]). The corresponding isotropic phase, typically termed as micellar, has a liquid-like orientational order of the aggregates; diffusion NMR in micellar or related systems has been extensively studied and reviewed [71–74]. Among the LC phases, there exist many different symmetries with an underlying variation of aggregate geometry [24,75–81]. Irrespective of that, the NMR properties of the two main system components, water (solvent) and amphiphile differ: if any, the residual coupling and thereby the static broadening/splitting of water is typically small.

Hence, water diffusion is accessible by conventional diffusion experiments, modified if necessary to ascertain refocusing in presence of static dipole or quadrupole splitting [61]. The same is also frequently the case for hydrophobic solubilizates within the amphiphile phase [82,83]. On the other hand, the broadening by residual coupling for the amphiphiles is typically large and must be suppressed in a diffusion experiment. The only exception is formed by cubic phases that, although crystalline, exhibit no static broadening. Hence, most amphiphile diffusion data are from cubic systems [1,3]; those ones from bicontinuous phases where curved amphiphile bilayers separate waterand oil-rich regions are relevant for and representative of bilayer diffusion in lamellar phases, too. Diffusion experiments were also carried out on some surfactant counterion species. As concerning anisotropic lyotropic LCs, we discuss below nematic phases that consist of orientationally ordered anisomeric micelles, hexagonal phases where elongated aggregates (or water channels in the inverse versions) arrange themselves in a 2D hexagonal lattice, and lamellar phases where flat amphiphilic bilayers exhibit

Part I

320

Lyotropic Applications 119

120 Part I

Chemistry

Part I

a 1D translational order. In all those systems, there are two issues that have decisive influence on the obtained diffusion data. First, D varies with the molecular environment and molecules exchange quickly among those; therefore, the observed diffusion coefficient is typically a population average. For water, the two environments are the bulk (fast diffusion) and the amphiphilic headgroup region (slow diffusion), while for the amphiphile there exists a small population of quickly diffusing monomers and a large population of slowly diffusing aggregates. The second consideration is topological [84–86]. Pathways for diffusion of water, for example, are obstructed from the hydrophobic interior the aggregates and therefore the spatial average of the diffusion coefficient is lower than the bulk value. If the obstruction is topologically enclosing (or directionally limiting) a region, the diffusion there becomes restricted that may result in very low average diffusion coefficient and/or to a composite diffusional decay. In nematic lyotropics that consist of closed (finite) aggregates, diffusion in the hydrophobic domain [66], although informative, is less representative of the phase and aggregate structure (except in the region of phase transition into continuous aggregate topologies). Instead, water diffusion can be used to report on these issues but to draw quantitative conclusions require very high accuracy and precision (see data example in Figure 1A) [56]. In the CsPFO/water mixture, large obstruction by flattened aggregates for diffusion parallel to their axis has been used to calculate average aggregate size and the orientational order of the aggregates as function of temperature [61]. The same system forms, upon cooling, a lamellar phase where the bilayer structural unit is pierced by water-filled defects; similar structures appear in lipid-based systems, where amphiphile lateral diffusion becomes indicative of the defects [66,87]. Water diffusion along the defective lamellar phase director (and therefore across the bilayers) is much higher [61,66,86,88–92] than in defect-free lamellar phases [58,93]. In the latter systems, lateral diffusion of the amphiphilic molecules along the bilayers has been addressed by several different methods [1,3,4,36– 38,60,94–96] yielding liquid-like mobilities and activation energies where the latter is often dominated by the strong headgroup–headgroup interactions. In contrast to lamellar phases, defects in hexagonal lyotropic LCs break the continuity of the aggregate [82,97]. Since the aggregate shape in hexagonal phases is elongated, obstruction to water diffusion [84,85,98] is weaker than in systems consisting of flattened aggregates.

Thermotropic Applications Diffusion in thermotropic LCs [11] has been addressed by a broad range of NMR techniques [4,99,100].

These include combining pulse-field-gradients with (i) “nematic” echo [101–103], (ii) multiple pulse decoupling [45–52,104–106], magic-angle sample orientation [25,35,57,102,107–112], (iii) deuterium stimulated (alignment) echo [50], (iv) soft-pulse excitation [53], and (v) multiple quantum NMR [51,113,114]. Besides, static field gradient techniques have also been used [115,116]. Most experiments have been performed in nematic phases that lack translational but possess orientational order. Nematic phases were also frequent choices for development and testing of new techniques for selfdiffusion LCs: the nematic phase of 5CB is a benchmark compound. Recently, diffusion data in 5CB obtained by several methods, including also non-NMR techniques, have been compared [49]. While earlier results strongly disagree, obviously due to methodological problems, more recent studies [49,53,115] by advanced NMR techniques demonstrate a better agreement: the diffusion coefficient in 5CB ranges from 10−9 to 10−10 m2 /s, depending on temperature and diffusion directions. The diffusion anisotropy, D|| /D⊥ decreases from 2.7 to 1.5 toward the nematic–isotropic transition and, hence, it reflects the decrease of molecular orientational order; the elongated mesogenic units diffuse easier along the director than across it. In the isotropic–nematic transition region, the average diffusion coefficient matches the diffusion coefficient in the isotropic phase, with similar (∼30 kJ/mol) activation energies. Clearly, diffusional transport in the nematic phase is liquid-like and broadly consistent with some available diffusion models [49,51]. Conventional PGSE NMR measurements performed on the isotropic side of the isotropic–nematic phase transition indicate the formation of locally ordered nematic clusters [117,118]. The temperature dependencies of the principal components of diffusion tensors were also reported for homologous series of alkoxy–azoxy benzenes and nOCB [111,112]. Characteristic alternation of diffusion coefficients and activation energies as functions of the number of chain segments, i.e. the familiar odd–even effect, has been observed. In a cholesteric–nematic phase [119], PGSTE NMR has detected diffusion anisotropy (D|| /D⊥ ≈ 1.7) similar to that in 5CB but D was strongly dependent on the diffusion time. Since smectic phases have layered structures with 2D liquid-like order within the layers, their diffusion anisotropy typically becomes D|| /D⊥ < 1 [99], opposite to that in nematics. Exceptions are smectics that exhibit a significant temperature range of a nematic phase: for them D|| /D⊥ > 1 may occasionally be found in the vicinity of the nematic region [51,108] and upon cooling deeper into the smectic phase the diffusion anisotropy changes sense [102]. As concerning activation energies, the relation E || ≥ E ⊥ is always fulfilled in smectics. At the nematic–smectic phase transition D⊥ changes nearly

Field Gradient NMR of Liquid Crystals

Other Applications of Field Gradients Magnetic field gradients are useful in high-resolution NMR for selecting and filtering coherence transfer pathways and may advantageously replace or be combined with phase cycling [131]. Hence, magnetic-fieldgradient pulses have been used for various multidimensional [42,132,133] and multiple-quantum experiments [134,135] in LCs and for selective suppression of, e.g. water signals in 1 H HR-MAS NMR in lipid membrane samples [136]. As another tool for improving spectral quality, field-gradient pulses have been applied in combination with frequency-selective rf pulses to limit the sensitive volume in liquid-crystalline sample in multiple-pulse decoupling PGSE experiments [45,46,48–52,60]. There are also examples of NMR imaging experiments applied to LCs. Velocity imaging of liquid crystalline polymers flowing through an abrupt contraction

was performed by pulsed-field-gradient NMR techniques [137]. Magnetic-field-gradient pulses were also incorporated in rheo-NMR experiments on various LC samples as reviewed by Callaghan [138].

References 1. Lindblom G, Or¨add G. Progr. Nucl. Magn. Reson. Spectrosc. 1994;26:483. 2. Dong RY, Nuclear Magnetic Resonance of Liquid Crystals. Springer: New York, 1994. 3. Lindblom G, Or¨add G. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance, Vol. 4. Wiley: Chichester, 1996, p. 2760. 4. Fur´o I, Dvinskikh SV. Magn. Reson. Chem. 2002;40:S3. 5. Burnell EE, de Lange CA (Eds). NMR of Ordered Liquids, Kluwer: Dordrecht, 2003. 6. Glicksman ME. Diffusion in Solids. Wiley: New York, 2000. 7. Basser PJ et al. Biophys. J. 1994;66:259. 8. Basser PJ. NMR Biomed. 1995;8:333. 9. Basser PJ, Pierpaoli CJ. Magn. Reson. B 1996;111:209. 10. Gabrieli J et al. (Eds). White Matter in the Cognitive Neurosciences: Advances in Diffusion Tensor Imaging and Its Applications. New York Academy of Sciences: New York, 2005. 11. Demus D et al. (Eds). Handbook of Liquid Crystals, Vol. 1–4. Wiley-VCH: Weinheim, 1998. 12. de Gennes PG, Prost J. The Physics of Liquid Crystals, Clarendon: Oxford, 1993. 13. Katsaras J, Gutberlet T (Eds). Lipid Bilayers. Springer: Berlin, 2000. 14. Halle B, Fur´o I. In: P Tol´edano, AM Figueiredo (Eds). Phase Transitions in Complex Fluids, World Scientific: Singapore, 1998, p. 81. 15. Safinya CR et al. Science 1993;261:588. 16. Imperor-Clerc M et al. Macromolecules 2001;34:3503. 17. Lukaschek M et al. Langmuir 1995;11:3590. 18. Lukaschek M et al. Colloid Polym. Sci. 1996;274:1. 19. Diehl P, Khetrapal CL. In: P Diehl et al. (Eds). NMR Basic Principles and Progress, Vol. 1. Springer: Berlin, 1969, p. 1. 20. Forrest BJ, Reeves LW. Chem. Rev. 1981;81:1. 21. Amaral L. Mol. Cryst. Liq. Cryst. 1983;100:85. 22. Fur´o I et al. J. Phys. Chem. 1990;94:2600. 23. Quist PO et al. J. Chem. Phys. 1991;95:6945. 24. Stegemeyer H (Ed.) Lyotrope Fl¨ussigkristalle, Steinkopff: Darmstadt, 1999. 25. Holstein P et al. J. Magn. Reson. 2000;143:427. 26. Spiess HW In: P Diehl et al. (Eds). Dynamic NMR Spectroscopy, Vol. 15. Springer: Berlin, 1979, p 55. 27. Stilbs P. Prog. Nucl. Magn. Reson. Spectrosc. 1987;19:1. 28. K¨arger J et al. Adv. Magn. Reson. 1988;12:1. 29. Callaghan PT, Principles of Nuclear Magnetic Resonance Microscopy. Clarendon Press: Oxford, 1991. 30. Price WS, Concepts Magn. Reson. 1997;9:299. 31. Price WS, Concepts Magn. Reson. 1998;10:197. 32. Stejskal EO, Tanner JE. J. Chem. Phys. 1965;42:288. 33. Stejskal EO. J. Chem. Phys. 1965;43:3597. 34. Tanner JE. J. Chem. Phys. 1970;52:2523. 35. Kr¨uger GJ et al. Phys. Lett. A. 1975;51:295.

Part I

continuously, while D|| and/or its activation energy may jump, in accordance to the layer-like smectic structure with liquid-like in-layer diffusion and solid-like jumps between adjacent layers [51,99,116]. In contrast to conventional thermotropics built by elongated mesogenic units, discotic materials are formed by flat molecules. In a columnar (smectic) phase of those, 2 H PGSE NMR detected very slow (∼10−14 m2 /s) diffusion with a large activation energy (115 kJ/mol) that suggest solid-like or, perhaps, collective diffusion mechanisms in discotics [52]. Thermotropic LC behavior can also be found for polymeric molecules. Hence, anisotropic diffusion with D|| > D⊥ relative to α-helical chain axis has been observed in LC phase formed by rod-like polypeptides [120– 123], with activation energies recorded as function of the main-chain length [122]. In the polymeric LC formed by the less rigid poly(diethysiloxane), the diffusion was found faster than that in the isotropic phase: this interesting effect was attributed to more entanglements between the polymer chains in the isotropic phase [124]. Due to much lower orientational order, small organic solute molecules in LCs exhibit long decays of spin coherences. Hence, conventional PSGE experiments are typically sufficient to access the diffusion coefficients of solutes [99,125]. In nematic phases, the solute diffusion is fast and the diffusion anisotropy is small [99,100]. This contrasts the strong diffusion anisotropy D|| /D⊥  1 observed in smectic phases [99]. A particularly interesting and simple solute is the noble gas 129 Xe, whose diffusional behavior was studied in detail [126–130]. While no diffusion anisotropy (D|| /D⊥ ∼1) was detected in a nematic phase [126], weak anisotropy with D|| /D⊥ > 1 has been observed in a related mixture [127]. This contrasts the D|| /D⊥  1 found in smectic phases of ferroelectric LCs [128,129].

References 121

122 Part I

Chemistry

Part I

36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86.

Roeder SBW et al. J. Chem. Phys. 1976;64:1848. Lindblom G, Wennerstr¨om H. Biophys. Chem. 1977;6:167. Ukleja P et al. Liquid Cryst. 1991;9:359. Or¨add G, Lindblom G. In: EE Burnell, CA de Lange (Eds). NMR of Ordered Liquids, Kluwer: Dordrecht, 2003, p 399. Maas WE et al. J. Magn. Reson. 1999;141:29. Pampel A et al., Chem. Phys. Lett. 2002;357:131. Gaede HC, Gawrisch K. Magn. Reson. Chem. 2004;42:115. Polozov IV, Gawrisch K. Biophys. J. 2004;87:1741. Zhang W, Cory DG. Phys. Rev. Lett. 1998;80:1324. Dvinskikh SV et al. J. Magn. Reson. 2000;142:102. Dvinskikh SV, Fur´o I. J. Magn. Reson. 2000;144:142. Dvinskikh SV, Fur´o I J. Magn. Reson. 2000;146:283. Dvinskikh SV, Fur´o I. J. Magn. Reson. 2001;148:73. Dvinskikh SV, Fur´o I. J. Chem. Phys. 2001;115:1946. Dvinskikh SV et al. J. Magn. Reson. 2001;153:83. Dvinskikh SV et al. Phys. Rev. E. 2002;65:061701. Dvinskikh SV et al. Phys. Rev. E. 2002;65:050702. Kim MJ et al. J. Chem. Phys. 2004;120:11327. Kimmich R et al. J. Magn. Reson. 1991;91:136. Karakatsanis P, Bayerl TM. Phys. Rev. E 1996;54:1785. Fur´o I, J´ohannesson H. J. Magn. Reson. A. 1996;119:15. Oishi O, Miyajima S. J. Magn. Reson. A. 1996;123:64. W¨asterby P et al. J. Magn. Reson. 2002;157:156. Callaghan PT et al. J. Chem. Phys. 1983;79:6372. Kadi M et al. Langmuir 2002;18:5015. J´ohannesson H et al. Phys. Rev. E 1996;53:4904. Callaghan PT, S¨oderman O. J. Phys. Chem. 1983;87:1737. Blum FD et al. Langmuir 1985;1:127. Fleischer G et al. J. Chem. Phys. 1999;111:2789. F. Rittig et al., Macromolecules 1999;32:5872. Gaemers S, Bax A. J. Am. Chem. Soc. 2001;123:12343. Callaghan PT, Komlosh ME. Magn. Reson. Chem. 2002; 40:S15. Callaghan PT et al. Magn. Reson. Imag. 2003;21:243. Callaghan PT, Fur´o I. J. Chem. Phys. 2004;120:4032. Rittig F et al. Macromolecules 2001;34:868. S¨oderman O, Stilbs P. Prog. Nucl. Magn. Reson. Spectrosc. 1994;26:445. S¨oderman O, Olsson U. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance, Vol. 4. Wiley: Chichester, 1996, p. 3046. S¨oderman O, Olsson U. Curr. Opinion Colloid Interf. Sci. 1997;2:131. Fur´o I. J. Mol. Liq. 2005;117:117. Sonin AS. Sov. Phys. Usp. 1987;30:875. Larsson K. J. Phys. Chem. 1989;93:7304. Seddon JM, Templer RH. Phil. Trans. Roy. Soc. London A. 1993;344:377. Laughlin RG. The Aqueous Phase Behavior of Surfactants. Academic Press: London, 1994. Gelbart WM et al. (Eds). Micelles, Membranes, Microemulsions, and Monolayers. Springer: New York, 1994. Seddon JM. Ber. Bunsenges. Phys. Chem. 1996;100:380. Tol´edano P, Figueiredo Neto AM (Eds). Phase Transitions in Complex Fluids. World Scientific: Singapore, 1998. Joabsson F et al. J. Phys. Chem. B 1997;101:9710. Jeong SW et al. Langmuir 2002;18:1073. J¨onsson B et al. Colloid Polym. Sci. 1986;264:77. J´ohannesson H, Halle B. J. Chem. Phys. 1996;104:6807. Celebre G et al. Gazz. Chim. Italiana 1996;126:489.

87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138.

Soong R, Macdonald PM. Biophys. J. 2005;88:255. Chidichimo G et al. Chem. Phys. Lett. 1985;117:514. Chidichimo G et al. Mol. Cryst. Liq. Cryst. 1986;135:223. Chidichimo G et al. Mol. Cryst. Liq. Cryst. 1987;150b: 221. Chidichimo G et al., Chem. Phys. Lett. 1988;145:85. Holmes MC et al. Langmuir 1995;11:356. Wassall SR. Biophys. J. 1996;71:2724. Ukleja P, Doane JW. In: SK Sinha (Ed). Ordering in Two Dimensions. North-Holland: Amsterdam, 1980, p. 427. G. Lindblom et al. Biochemistry 1981;20:2204. Filippov et al., Biophys. J. 2004;86:891. Coppola L et al. Langmuir 2000;16:4180. Coppola L et al. Langmuir 2003;19:1990. Kr¨uger GJ. Phys. Rep. 1982;82:229. Noack F. In: D Demus et al. (Eds). Handbook of Liquid Crystals. Wiley-VCH: Weinheim, 1998, p. 582. Kr¨uger GJ, Spiesecke H. Z. Naturforsch. 1973;28a:964. Kr¨uger GJ et al. Mol. Cryst. Liq. Cryst. 1977;40:103. Kr¨uger GJ, Weiss R. J. Phys. (Paris) 1977;38:353. Blinc R et al. Phys. Rev. Lett. 1973;30:546. Blinc R et al. Phys. Rev. Lett. 1974;33:1192. Zupancic I et al. Solid State Comm. 1974;15:227. Kr¨uger GJ et al. J. Phys. Colloque (Paris) 1976;37:123. Miyajima S et al. Chem. Phys. Lett. 1993;212:277. Oishi O, Miyajima S. J. Phys. Soc. Jpn. 2002;71:2373. Oishi O, Miyajima S. J. Magn. Reson. 2003;160:74. Noack F et al. Annu. Rep. NMR Spectrosc. 1997;33:1. Noack F. Mol. Cryst. Liq. Cryst. 1984;113:247. Martin JF et al. J. Chem. Phys. 1982;76:2632. Zax D, Pines A. J. Chem. Phys. 1983;78:6333. Vilfan M et al. Magn.Reson.Imag. 2001;19:433. Cifelli M et al. Phys. Chem. Chem. Phys. 2004;6:4701. Kashirin NV et al. Colloid J. 2000;62:68. Gasilova ER et al. Liq. Cryst. 2000;27:573. Blinc R et al. Phys. Rev. Lett. 1985;54:438. Ando I et al. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance, Vol. 9. Wiley: Chichester, 2002, p. 770. Yin YG et al. J. Chem. Phys. 2000;113:7635. Yin Y et al. Macromolecules 2002;35:2335. Yin YG et al. Macromolecules 2002;35:5910. Kanesaka S et al. Macromolecules 2004;37:453. Moseley ME, Loewenstein A. Mol. Cryst. Liq. Cryst. 1982;90:117. Long HW et al. J. Phys. Chem. 1995;99:11989. Ruohonen J, Jokisaari J. Phys. Chem. Chem. Phys. 2001;3:3208. Ruohonen J et al. Mol. Phys. 2001;99:711. Cifelli M et al. J. Phys. Chem. A 2004;108:3973. Jokisaari J. In EE Burnell, CA de Lange (Eds). NMR of Ordered Fluids, Kluwer: Dordrecht, 2003, p 109. Zhu J-M, Smith CP, Concepts Magn.Reson. 1995;7:281. Lafon O et al. J. Magn. Reson. 2004;171:135. Soubias O et al. J. Magn. Reson. 2003;165:303. Field LD. In: EE Burnell, CA de Lange (Eds). NMR of Ordered Fluids, Kluwer: Dordrecht, 2003, p. 67. Ramadan SA et al. Molec. Phys. 2003;101:1813. Chen JH et al. J. Magn. Reson. 2004;171:143. Gentzler M et al. Rheol. Acta 2000;39:1. Callaghan PT. Rep. Prog. Phys. 1999;62:599.

123

Yuji Yamane and Sunmi Kim Department of Chemistry and Materials Science, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8552, Japan

Introduction Pulse field gradient (PFG) NMR method has become a useful technique for studying self-diffusion of probe molecules in polymer systems. Recently, high fieldgradient NMR system with a maximum strength of more than 1,000 G/cm and new pulse sequences are possible to measure the diffusion coefficient (D) with the order of 1–10−11 cm2 /s in polymer systems. It is expected that the use of this system leads to sophisticated knowledge on nature of polymer systems as well as diffusional behavior of probe molecules in polymer systems. A number of papers, reviews, and monographs in this field have appeared [1–7]. In this section, some of most recent topics especially characterization of the polymer systems such as polymer gels and polymer media with controlled cavity through the diffusion experiments rather than principle of PFG NMR system and diffusional behavior of polymer chains have been described.

Diffusion in Polymer Gel Systems Probe Diffusion in Polymer Gel Systems Polymer gel systems consist of network polymer chains, solvents, and probe molecules, and also apply to many industrial fields. These functionalities are closely associated with diffusional behavior of solvents and probe molecules, and intermolecular interactions between networked polymer chains and probe molecules. Matsukawa, et al. have studied the diffusional behavior of water and poly(ethylene glycol) (PEG) in chemically cross-linked hydrogels. The D values of HDO (D(HDO) ) in gels are well explained by modified free volume theory in a wide range of the degree of swelling (Q = Mswollen /Mdry ) [1], 0 and the DPEG values in gels are followed by D/DPEG = exp(−κ R), the dynamical screening length κ −1 is proportional to c−0.71 . The value of −0.71 is close to that proposed theoretically by de Gennes [8,9]. Masaro, et al. have studied the diffusional behavior of PEG in aqueous poly(vinyl alcohol) solution and gels as functions of the polymer concentration and molecular size of the diffusant. Several theoretical models based on different physical concepts have been used to fit the exGraham A. Webb (ed.), Modern Magnetic Resonance, 123–127.
C 2008 Springer.

perimental data [10,11], and intermolecular interactions of probe molecules with networked polymer chains have been elucidated from the diffusion data [12–15]. Further, they have studied the effect of shapes of the diffusant on the diffusional behavior in gels [16–18]. Most recently, some groups have measured the D values of 7 Li and 19 F ions in polyelectrolyte gels to understand deeply the mechanism of ionic conduction [19–23].

Characterization of the Inhomogeneities in Polymer Gels Polymer gels have generally inhomogeneity of the network size, and then properties of polymer gels depend on their spatial inhomogeneity. The existence of spatial inhomogeneity has been studied by light scattering as speckles [24,25]. One of the clearest manifestations of the inhomogeneity is an appearance of speckle pattern. As for chemically-crosslinked polymer gels, the relationship between speckles and spatial inhomogeneity has been elucidated [26]. Nevertheless, some problems on intermolecular interactions between the network and probe molecules associated with inhomogeneity of the network size in polymer gels remain. The measurement of D by PFG NMR has emerged as a powerful method for detecting inhomogeneities in gels. The method is based on the interpretation of the dependence of D on the time  (“diffusing time” or “observation time”) between the two gradient pulses in the pulse sequences. The time dependens and the distribution of D are observed in gels [27] and other heterogeneous systems [28,29]. For convenience, we consider a probe molecule in a gel with homogeneous network size distribution (homogeneous gel) and with inhomogeneous network size distribution (inhomogeneous gel). As for probe molecules diffusing within a network cell in both of the gels the D value depends on . If a probe molecule is diffusing through sufficiently many network cells in the homogeneous gel, the diffusion component is a single and the D value is independent of , but in inhomogeneous gel the diffusion components are two or more and then the D value is dependent of . When observed over longer times, the diffusion component becomes a single because the D values that comes from the distribution of the network size

Part I

Field Gradient NMR for Polymer Systems with Cavities

124 Part I

Chemistry

Fraction of the slow diffusion component

Part I

are averaged out, and are independent of . Therefore, if appropriate  in the PFG experiments is selected, useful information on the inhomogeneous network size distribution of gels can be obtained. The inhomogeneity of polymer gels such as polystyrene (PS) gel and cross-linked ethoxylate acrylate (CLEAR) gel with deuterium dimethylformamide (DMF-d7 ) as solvent has been characterized by using time-dependent diffusion NMR [30]. From the experimental results on the D values of probe amino acid, tert-butyloxylcarbonyl-l-phenylalanine (Boc-Phe), in the gels, it is cleared that in the short diffusion time range the amino acid in the gels has two components in diffusion as influenced by the distribution of network size, but in the long diffusion time range has a single component in diffusion. Here, we focus on the diffusing time  that the diffusion changes from the two components to the single component in diffusion. This specified  value is named as the “specific” diffusion time (Stime ). Then, the √ diffusion distance is d = 2D and the “specific” √ diffusion distance (Sdistance ) is defined as Sdistance = 2DStime . Figure 1 shows the  dependence of the fraction of the slow diffusion component ( f slow ) value for Boc-Phe in PS(2) gels, PS(1) gels and CLEAR gels with DMF-d7 as solvent at 30 ◦ C, where PS(2) and PS(1) gels are crosslinked by 1 and 2% divinylbenzene (DVB), respectively, and the Boc-Phe concentration is 10 wt%. As seen from

1 0.8 0.6 0.4 0.2 0 0

20

40

60

80

Gradient pulse interval ∆ / ms Fig. 1. Dependence of the fraction of the slow diffusion component of Boc-Phe in PS(2) gels (
), in PS(1) gels (3) and CLEAR gels (2) with DMF-d7 as solvent at 30 ◦ C on the gradient pulse interval .

this figure, the f slow value increases with an increase in  and changes from the two components to the single component at larger . The Stime as estimated from these plots for Boc-Phe in PS(2) gels, PS(1) gels, and CLEAR gels are 20, 40, and 30 ms, respectively. It is found that the Stime depends on the kinds of gels. As for Boc-Phe in PS(2) gels, PS1(1) gels and CLEAR gels at 30 ◦ C, the Sdistance are 1.1, 1.7, and 2.7 µm, respectively. It can be said that the Sdistance depends on the kind of gels, and that the Sdistance of Boc-Phe in CLEAR gels is much larger than that of Boc-Phe in PS gels. The cross-linker of PS gels is DVB, and the cross-linker of CLEAR gels is acrylate. One of the goals of this study is to detect inhomogeneities in polymer gels differing in their degree of crosslinking [31]. Then, the networks are prepared at 70 ◦ C by simultaneous polymerization and cross-linking of a mixture of acrylic acid (AA), sodium carbonate, cross-linker (1,4-butanedioldiacrylate), and the redox couple sodium persulfate/sodium isoascorbate as the initiator. Two types of networks are prepared, using the same monomer and sodium carbonate concentrations, but different amounts of the cross-linker, 1.1 and 0.5 wt%, respectively, in the monomer mixture. The corresponding notations are PAA(1.1) and PAA(0.5) , respectively. Detection of inhomogeneities is based on measuring the D of the probe molecule PEG by time-dependent diffusion NMR. Diffusion measurements are performed as a function of the degree of swelling, Q = Mswollen /Mdry , with Q in the range 2.8–10.0. The different diffusional behavior of the two gel systems emerged as their degree of swelling is varied. For PAA(1.1) gel with Q = 10.0 and 5.2, and for PAA(0.5) gel with Q = 10.0, 5.1, and 4.5, only single diffusion component is detected, independent on the  in the range 30–500 ms. For less swollen gels (Q in the range 2.9– 4.5 for PAA(1.1) gel and 2.8–3.9 for PAA(0.5) gel), two diffusion components (Dfast and Dslow ) are detected as influenced by the distribution of network size, and both of the Dfast and Dslow values depend on . Here, for all gels, The dfast and dslow , f fast , and f slow values are calculated as a function of . A useful parameter in the interpretation of results is the “specific” degree of swelling (SQ ) above which the diffusion of the probe in the two gel systems changed from single to two components. A larger value of SQ in PAA(1.1) gel is taken as an indicator of a more inhomogeneous gel. Analysis of the effect of  on the D, d, and f of the slow and fast diffusion components indicates that both of the gels form a highly cross-linked region in a narrow Q range. In this Q range, the polymer chains interact and form a highly restricted diffusion region. The density and distribution of the cross-links form different restricted diffusion regions in PAA(1.1) and PAA(0.5) gel systems, and the heterogeneity in terms of the network size distribution and corresponding

Field Gradient NMR for Polymer Systems with Cavities

Characterization of Smart Gels with Regular Structure Recently, it is necessary to prepare and characterize the smart gel with regular structure, and PFG NMR is more powerful tool for characterizing the soft materials such as gels. It has been reported that highly-oriented poly(γbenzyl l-gultamate) (PBLG) gel is prepared by crosslinking reaction of highly-oriented PBLG chains in 1,4dioxane with cross-linker in the magnetic field of an NMR magnet, by monitoring the orientation of the PBLG chains with solid-state static 13 C NMR by using the relationship between the observed 13 C chemical shift of the main-chain carbonyl carbon in PBLG [32], and then solvents and rod-like molecules in the PBLG gel are anisotropically diffusing in the direction parallel and perpendicular to the α-helix axis by PFG NMR [32,33]. The degree of orientation is 0.81, and the D of 1,4-dioxane molecule in the direction parallel to the α-helical PBLG axis(D|| ) to be 5.4 × 10−6 cm2 /s is larger than that perpendicular to the α-helical PBLG axis(D⊥ ) to be 4.5 × 10−6 cm2 /s and the D|| /D⊥ value is 1.20. This shows that there exist ˚ (the inlong channels with small diameter of 15–20 A terchain distances between the nearest-neighboring two PBLG chains determined by the wide angle X-ray diffraction pattern) in the highly-oriented PBLG gel, and that the diffusional behavior of solvents and probe molecules is significantly influenced by the microstructure of network polypeptide chains in gels. Further, highly-oriented PBLG gels having channel cavities with µm-scale diameters due to phase separation in cross-linking reaction process have been prepared and characterized the structure of the polypeptide gels by PFG NMR and three-dimensional (3D) NMR imaging [34,35]. PBLG gel with channel cavity has two regions consisting of long channel cavity (with µm-scale diameters) region and the remaining gel matrix region without long channel cavity. In the PFG 1 H NMR experiments, spin echo signals coming from solvent molecules in the corresponding two regions are observed. In the PBLG gel matrix region, the D|| /D⊥ value is 1.29. This shows that the PBLG gel matrix region has anisotropic channel cavity in a nm-scale, and that nm-scale structure of the PBLG gel matrix region in PBLG gel with channel cavity is similar to nm-scale structure of highly-oriented PBLG gel with no µm-scale channels. While, in µm-scale channel cavities of PBLG gel, the 1,4-dioxane molecules may be trapped in channel cavities or may permeate partially and slowly from the channel cavity region to the gel matrix region and from the gel matrix region to the channel cavity region through network of the wall of the channel cavity

because the network density near the wall of the channel cavity as formed by cross-linking reaction with phase separation may be higher than that in the PBLG gel matrix region. Thus, if 1,4-dioxan in channel cavity is trapped and mainly diffuse in channel, this reflecting model on diffusion in the space between two infinitely large perfectly reflecting parallel planes (separation 2R) [36–40] may be approximately used to analyze the diffusion behavior of 1,4-dioxane in the channel cavity of the PBLG gel. The plots of PFG spin echo attenuation E(q,  = 5 ms) for diffusion of 1,4-dioxane in PBLG gels with channel cavities against “tunable” parameter q(= (2π )−1 γ gδ) show the two diffraction minima (not shown) [33]. Here, γ is the gyromagnetic ratio of proton, g is the strength of the field gradient pulse and length δ. When the probe molecules are trapped in restricted space, the diffraction minima corresponding to the size scale of restricted space are often observed as seen from Equation(1). E (q, ∆ = ∞) =

2 [1 − cos (4πq R)] (4πq R)2

(1)

The tendencies for the simulated curves (2R = 50 and 60 µm) and the experimental plots are very similar to each other. The slight difference between the experimental and simulated curves may come from the fact that solvent molecules may permeate partially. From this, it can be said that the simulated results do not conflict with the experimental results and thus the mode diameter of the long channel cavities may be estimated to be about 50– 60 µm. This is very close to the result by 3D NMR imaging (about 70 µm). Further, it can be said that most of the 1,4dioxane molecules is reflected at the surface of the wall of channel cavities.

Characterization of the Polymer Media with Nano Cavities Polymer media with nano cavities has potential as smart membrane, and we need to understand deeply the diffusional behavior of probe molecules and property of cavities. K¨arger, et al. have shown that PFG experiments gives useful information on zeolites, the molecular dynamics simulations have reasonably explained with the experimental results on self-diffusivity for a binary mixture adsorbed inside zeolite [41–44] and for zeolite and porous media [45,46]. In this section, characterization of the media with controlled nano cavities has been briefly described. The channels in poly( p-biphenylene terephthalate) with long n-dodecyl side chains(PBpT-O12) have been characterized in order to elucidate the nature of the inside of the cylindrical channel cavities as studied by PFG NMR [47]. PBpT-O12 forms the hexagonal columnar phase, and honeycombed network is formed and then

Part I

populations is higher in the gel with the higher degree of cross-linking.

Diffusion in Polymer Gel Systems 125

126 Part I

Chemistry

Part I

has cylindrical cavity channels with diameter of about 3 nm. They aim to prepare PBpT-O12 charged methane and ethane molecules into the cylindrical channel cavities in the hexagonal columnar phase, to measure the D values of the gas in the directions perpendicular (D⊥ ) and parallel (D|| ) to the channel cavity axis. Methane and ethane molecules in the cylindrical channel cavities have a single diffusion component in the  used here. The D||(ethane) value in the cylindrical channel cavities at  = 6 ms is determined to be 2.9 × 10−7 cm2 /s. This D value is extremely much smaller than that of ethane gas (0.22 cm2 /s). This means that the diffusion of the ethane molecules is strongly restricted by the nature of the inside of cylindrical channel cavity by intermolecular interactions between the ethane molecules and, especially, long n-dodecyl side chains of the polyester. While, D⊥(ethane) value in the cylindrical channel cavities at  = 6 ms is determined to be 9.5 × 10−9 cm2 /s. This D value is extremely much smaller than that of the D||(ethane) value, and also shows that the ethane molecules are not trapped, but diffuse through the wall of the cylindrical channel cavity to the neighboring cylindrical channel cavities. By using D⊥ = 9.5 × 10−9 cm2 /s,  = 6 ms, we can estimate the d to be 107 nm. Therefore, it can be said that the wall of cylindrical channel cavity has some defects and the ethane molecules are possible to pass through these defects, and that the furthermore ethane molecules in the cylindrical channel cavities are clearly moving through the wall of the cylindrical channel cavity to neighboring cylindrical channel cavities in the direction perpendicular to the cylindrical channel cavity axis in a rod piece of oriented PBpT-O12 polyester media with a diameter of 0.6 mm (=6.0 × 105 nm). The ratio of D||(ethane) /D⊥(ethane) is 31, thus, it can be said that the inside of cylindrical cannel cavity is anisotropic field for gas diffusion. As compared with the D(methane) values in the channel cavities, the D||(ethane) and D⊥(ethane) values are much smaller than those of methane(D||(methane) = 4.2 × 10−7 cm2 /s and D⊥(methane) = 6.5 × 10−8 cm2 /s), and then the ratio of D||(ethane) /D⊥(ethane) for ethane is much larger than that for methane(D||(methane) /D⊥(methane) = 6.5). This means that the cylindrical cannel cavities of PBpT-O12 have high anisotropy in diffusion and the ability for the recognition of the molecular size.

Diffusional Behavior of Linear Molecules in Channels In general, the inclusion compound is defined as a chemical substance consisting of a lattice of one type of molecule (host) trapping and containing a second type of molecule (guest). The host molecules form a cavity such as a crystal lattice containing spaces with long tunnels or

channels in a crystal. According to many kinds of pairs between host and guest molecules, many kinds of inclusion compounds can be formed. Urea usually forms tetragonal structure in the crystalline state. However, in the case of urea adducts, the structure of urea changes to the hexagonal form that is a parallel channel with diam˚ by strong hydrogen-bonds between eter of about 5.5 A urea molecules and n-paraffin molecules are included in the channel. When the host molecule is urea, the inclusion compounds are often called “urea adducts” instead of “urea inclusion compounds.” As a consequence of the requirement for the size and shape compatibilities between the guest molecules and the host channels, typical guest molecules for the urea channel structure are linear molecules such as higher n-alcohols and n-paraffins (with six or more carbon atoms), some n-olefines, ncarboxylic acids, ketones, and esters. The structure and dynamics of urea adduct which has n-paraffin molecules as a guest molecule has been widely studied by various methods such as X-ray diffraction, neutron scattering, Raman, IR, solid-state NMR, molecular dynamics calculation, etc. However, there is little work for elucidating whether guest molecule diffuses in urea channels or not. It is very interesting to think that if guest molecules diffuse in urea channels, how is diffusional behavior? Most recently, the phenomenon that n-paraffins diffuse in urea channels has been successfully detected. It is found that n-paraffin molecules are diffusing in long urea channels and has the two diffusion components such as the fast diffusion component (D = 10−6 cm2 /s) and the slow diffusion component (D = 10−7 cm2 /s) by using PFG NMR [48]. According to the single-file diffusion model applied in case that molecules cannot pass each other, it can be explained that the two diffusion components such as Dfast and Dslow are correspond to n-paraffin molecules in the two external regions near the ends of the urea channel and n-paraffin molecules in the central inner region of the urea channel. Furthermore, D of the fast and slow diffusion components of n-paraffin molecules in urea channel are greatly decreased as the carbon number is increased from 8 to 21, but the diffusion coefficients D are slowly decreased as the carbon number is increased from 21 to 32, and from the activation energy of self-diffusion, they say intermolecular interactions between n-paraffin chains and urea channels wall are very smaller than those between n-paraffin chains in the rotator phase [49]. From the diffusing-time () dependence of the diffusion coefficients, it is cleared that n-paraffin molecules are diffusing colliding with molecules on sides within urea channel. Consequently, the diffusion process of n-paraffins in urea channels is cooperative diffusion such as single file diffusion from the time-dependent diffusion NMR by compared the results of the simulation [50] of the relationships between single file diffusion and diffusion experiments.

Field Gradient NMR for Polymer Systems with Cavities

It is demonstrated that PFG experiments gives useful information on the characterization of the polymer systems as well as probe diffusion, and also will have a great potentiality for applications to characterization of smart media with controlled cavities [51,52], aggregation process, lattice-forming process, phase separation systems, and heterogeneous systems [53] as well as gels. Especially, it is important to consider the diffusing time, if so, information about the structure of polymer system and nature of diffusant can be derived from PFG experiments.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

Hahn EL. Phys. Rev. 1950;80:580. Stejskal EO, Tanner JE. J. Chem. Phys. 1965;42:288. Nose T. Ann. Rep. NMR Spectrosc. 1993;27:217. Price WS. Ann. Rep. NMR Spectrosc. 1996;32:51. Callaghan PT, Principles of Nuclear Magnetic Resonance Microscopy. Claredon Press: Oxford, England, 1991. Kimmich R, NMR: Tomography, Diffusiometry, Relaxometry. Springer: Berlin, 1997. Ando I, Kobayashi M, Zhao C, Yin Y, Kuroki S. Encyclopedia of NMR, Vol. 19. Interscience: New York, 2002, p. 770. Matsukawa S, Ando I. Macromolecules. 1996;29:7136. de Gennes PG. Macromolecules. 1976;9:594. Masaro L, Zhu XX, Macdonald PM. Macromolecules. 1998;31:3880. Masaro L, Zhu XX, Macdonald PM. J. Polym. Sci. Polym. Phys. Ed. 1999;37:2396. Matsukawa S, Ando I. Macromolecules. 1997;30:8310. Matsukawa S, Ando I. Macromolecules. 1999;32:1865. Masaro L, Zhu XX. Langmuir 1999;15:8356. Kwak S, Lafleur M. Macromolecules 2003;36:3189. Masaro L, Zhu XX. Macromolecules. 1999;32:5383. Baille WE, Malveau C, Zhu XX, Kim YH, Ford WT. Macromolecules. 2003;36:839. Baille WE, Zhu XX, Fomine S. Macromolecules. 2004;37:8569. Saito Y, Kataoka H, Stephan AM. Macromolecules. 2001;34:6955. Hayamizu K, Aihara Y, Arai S, Price WS. Electrochem. Acta. 2000;45:1313. Ward IM, Williamson MJ, Hubbard HVSt.A, Southall JP, Davies GR. J. Power Sources. 1999;81–82:700. Forsyth M, Sun J, Zhou F, MacFarlane DR. Electrochem. Acta. 2003;48:2129.

23. Darwish MIM, van der Maarel JRC, Zitha PLJ. Macromolecules. 2004;37:2307. 24. Pusey NP, van Megen W. Physica A. 1989;157:705. 25. Ikkai F, Shibayama M. Phys. Rev. E. 1997;56:R51. 26. Mallam S, Horkay F, Hecht AM, and Geissler E. Macromolecules. 1989;22:3356. 27. Ros´en O, Bostr¨om M, Nyd´en M, Piculell L. J. Phys. Chem. B. 2003;107:4074. 28. Cicerone MT, Wagner PA, Ediger MD. J. Phys. Chem. B. 1997;101:8727. 29. Lin G, Zhang J, Cao H, Jones AA. J. Phys. Chem. B 2003;107:6179. 30. Yamane Y, Matsui M, Kimura H, Kuroki S, Ando I. Macromolecules. 2003;36:5655. 31. Yamane Y, Ando I, Buchholz FL, Reinhardt AR, Schlick S. Macromolecules. 2004;37:9841. 32. Zhao C, Zhang H, Yamanobe T, Kuorki S, Ando I. Macromolecules. 1999;32:3389. 33. Zhao C, Kuorki S, Ando I. Macromolecules. 2000;33:4486. 34. Yamane Y, Kanekiyo M, Koizumi S, Zhao C, Kuroki S, Ando I. J. Appl. Polym. Sci. 2004;92:1053. 35. Yamane Y, Koizumi S, Kuroki S, Ando I. J. Mol. Struct. 2005;739:137. 36. Price WS. Ann. Rep. NMR Spectrosc. 1996;32:51. 37. Mitra PP, Sen PN. Phys. Rev. B. 1992;45:143. 38. Snaar JEM, Van As H. J. Magn. Reson. A. 1993;102:318. 39. Mitra PP, Sen PN, Schwartz LM. Phys. Rev. B. 1993;47:8565. 40. Callachan PT. J. Magn. Reson. A. 1995;113:53. 41. McDaniel PL, Coe CG, K¨arger J, Moyer JD. J. Phys. Chem. 1996;100:16263. 42. Snurr RQ, K¨arger J. J. Phys. Chem. B. 1997;101:6469. 43. Heink W, K¨arger J, Naylor T, Winkler U. Chem. Commun. 1999;57–58:57. 44. Geier O, Snurr RQ, Stallmach F, K¨arger J. J. Chem. Phys. 2004;120:367. 45. Rittig F, Coe CG, and Zielinski JM. J. Am. Chem. Soc. 2002;124:5264. 46. Mair RW, H¨urlimann MD, Sen PN, Schwartz LM, Patz S, Walsworth RL. Magn. Reson. Imaging. 2002;19:345. 47. Matsui M, Yamane Y, Kuroki S, Ando I, Fu K, Watanabe J. J. Mol. Struct. 2005;739:131. 48. Kim S, Kimura H, Kuroki S, Ando I. Chem. Phys. Lett. 2003;367:581. 49. Yamakawa H, Matsukawa S, Kuroki S, Kurosu H, Ando I. J. Chem. Phys. 1999;111:5129. 50. Aslangul C. Europhys. Lett. 1998;44:284. 51. Appel M, Fleischer G, K¨arger J, Dieng AC, G. Riess. Macromolecules 1995;28:2345. 52. Challa V, Kuta K, Lopina S, Cheung HM, von Meerwall E. Langmuir 2003;19:4154. 53. Seland JG, Ottaviani M, Hafskjold B. J. Colloid Interface Sci. 2001;239:168.

Part I

Conclusion Remarks

References 127

129

Shingo Matsukawa Department of Food Science and Technology, Tokyo University of Marine Science and Technology, Minato-ku, Tokyo 108-8477, Japan

Introduction An application of field gradient attaches a spatial information in NMR signal, therefore, it can produce a spatial distribution of nuclei, that is, NMR imaging [1,2]. When two field gradients for the diphase and rephrase applied, the NMR signal decays due to the displacement of nucleus during the time between two field gradients [3,4]. This gives the diffusion coefficient for Fickian diffusion in free space and the space size for a spatially restricted diffusion [5]. Recently, the field gradient is used for the selection of desired coherence pathway by rephasing the desired coherence and dephasing the undesired coherence [6]. In this chapter, these three important uses of field gradient are described.

Diffusion Coefficient Measurements The Larmor precession frequency depends on the magnetic field experienced by the nucleus, therefore, it has a spatial dependence under the field gradient. The spatial dependent Larmor frequency ω(r) at the position r under a spatially linear field gradient g is expressed as follows ω(r) = γ (H0 + gr) = ω0 + γ gr

(1)

where H0 is the externally applied magnetic field and gr = 0 at the position of r = 0. When the duration time of the gradient is δ, the difference of the phase angle at r from that at r = 0 is φ(r) = γ grδ

(2)

The distance in the direction of g where φ(r) = 2π is q −1 = 2π/γ gδ

(3)

q −1 is the scale of length with the gradient. For example, q −1 becomes 235 µm for g = 10 G/cm with δ = 1 ms. When the sample size, or the size of detection area, is several times larger than q −1 , the total signal intensity vanishes because of the dephasing. For the diffusion coefficient measurements, a second gradient is applied Graham A. Webb (ed.), Modern Magnetic Resonance, 129–134.
C 2008 Springer.

in order to rephase the dephased magnetization. Figure 1 shows a typical pulse sequence with two pulsed field gradients (PFG) with rectangular shape along the z-axis, and the dephasing and rephasing behavior of the magnetization when the individual nucleus did not change their positions in the interval  between the two PFG. In Figure 1 (a) The magnetizations are aligned along the y-axis by an rf π/2 pulse. (b) Under the first PFG, the magnetizations precess at the angular velocity of γ gr corresponding to the z coordinate. (c) At the end of the first PFG, the magnetizations are spirally twisted at a pitch of q−1 . (d) The application of an rf π pulse along the y-axis rotates the individual magnetizations along the y-axis through 180◦ , which makes the mirror-symmetrical arrangement of the magnetizations with respect to the y–z plane. (e) Under the second PFG, the individual magnetizations precess at the same angular velocity with that under the first PFG. (f) At the end of the second PFG, the magnetizations are aligned along the y-axis. When the nucleus has a displacement of z in the z direction during , it has a phase angular shift φ(z) = 2π

z = γ gdz q −1

(4)

The echo signal intensity I (2τ , gδ) at 2τ is proportional to the vector sum of magnetizations in the sample, therefore, expressed as follows I (2τ ,gδ) = I (2τ , 0) cos (φ (z)) ×ρ (r) p (r, z) dr dz

(5)

where ρ(r) is the density of the nucleus and is constant for homogeneous sample, p(r,z) is the probability of the displacement during  for the nucleus at r and I (2τ ,0) is the total signal intensity without PFG and expressed as follows I (2τ , 0) = I (0, 0) exp(−2τ/T2 )

(6)

where I (0,0) is the initial signal intensity just after the rf π/2 pulse. For the free diffusion in an isotropic medium,

Part I

NMR Measurements Using Field Gradients and Spatial Information

130 Part I

Chemistry

Part I

δ

π 2x

g

a) z

b) z

y

x

g

c) z

γ gr

x

δ

πy

y

d) z

γ gr

q-1

x

y

f) z

e) z

x

y

x

y

x

y

Fig. 1. A typical pulse sequence with two PFG of rectangular shape and the dephasing and rephasing behavior of the magnetization. (a) The magnetizations are aligned along the y-axis by an rf π/2 pulse. (b) Under the first PFG, the magnetizations precess at the angular velocity of γgr corresponding to the z coordinate. (c) At the end of the first PFG, the magnetizations are spirally twisted at a pitch of q −1 . (d) An rf π pulse along the y-axis rotates the individual magnetizations along the y-axis through 180 degree. (e) Under the second PFG, the individual magnetizations precess at the same angular velocity with that under the first PFG. (f) At the end of the second PFG, the magnetizations are aligned along the y-axis.

p(r,z) becomes the Gaussian distribution 
z 2 −1/2 exp − p (r, z) = (4π D) 4D

(7)

where D is the diffusion coefficient. Taking the diffusion during δ into account, Equation (5) is rewritten as follows

2τ δ I (2τ ,gδ) = I (0, 0) exp − − (γ gδ)2 D  − T2 3 (8) In common measurements of the free diffusion, gδ is varied under constant . For the diffusion in restricted space, the D value obtained by applying Equation (8) is an apparent diffusion constant  2 z () Dapp () = (9) 2 where z 2 () is the mean square of z 2 during .

z 2 () is proportional to  for the free diffusion, however, becomes smaller than the proportional value due to the spatial restriction, which gives the information of the space size of the restriction. The signal decay by the displacement under the field gradient can be used to remove peaks for small molecules from the mixture spectrum of small and large molecules. Conversely, the spectrum of peaks for small molecules can be obtained by the subtraction of the decayed spectrum for large molecules from the mixture spectrum. The mixture spectrum is composed of the peaks contained in different molecules that exponentially decay with the square of γ gδ according to individual diffusivity. By changing Equation (8) into a sum of each component, the intensity at each frequency in the mixture spectrum is expressed as follows, I (2τ, gδ) = I (0, 0)

 2τ δ × f i exp − − (γ gδ)2 Di  − T2,i 3 i (10)

NMR Measurements Using Field Gradients and Spatial Information

Fi (2τ ) = f i exp(−2τ/T2,i )

(11)

Equation (10) is rewritten as follows, I (2τ, gδ) = I (0, 0) 

 δ × Fi exp − (γ gδ)2 Di  − 3 i

(12)

When F is expressed as a continuos function of D, Equation (12) is rewritten as follows, I (2τ, gδ) = I (0, 0) F(Di ) exp(−KD)dD (13) where K = (γ gδ)2 (-δ/3). F(D) is a T2 enhanced distribution of D. Equation (13) shows that I (2τ ,gδ) is the Laplace transformation of F(D), therefore, an inverse Laplace transformation of I(2τ ,gδ) will give F(D). The 2D spectrum with dimensions of frequency and D, diffusion-ordered NMR spectroscopy (DOSY), gives separated spectrum for each species in the mixture on the diffusivity and the distribution of the diffusivity for each species [7].

NMR Imaging An application of field gradient along one direction gives one-dimensional profile of spin density. The use of gradients along three direction of x, y and z gives a three dimensional NMR imaging. Figure 2 shows a typical pulse sequence with field gradients of gx , g y and gz along x, y

and z, respectively. The gz applied during rf pulse selects the layer where the Larmor frequency ω(r) expressed by Equation (1) is equal to the rf frequency. The rf pulse is shaped into sine form (Sin(t/d)/(t/d)) which has a rectangular shape in the frequency domain, which is a Fourier transform of the rf pulse in time domain, and slices the selected layer sharply (Figure 3). The thickness of the layer d is inverse to the rf duration, therefore, a weak rf pulse with long duration is applied for a thin slice of the layer. The gz causes the phase shift corresponding to the excess magnetic field at r, therefore, a reversed gradient is set after the π/2 rf pulse in order to rephase the phase shift. The phase shift during the first half of gz at the π pulse is rephased during the latter half of the period. The pairs of the dephasing and rephasing followed by the slice selection gradient are represented as shadowed portions with upper right and left in Figure 2. The gx is applied when the magnetizations are aligned along the y-axis. At the end of the gradient, the magnetization has a phase shift depending on the position along the x-axis, which is given by the rewriting Equation (2) as follows, φ(x) = γ gx xtx = 2πk x x

(14)

where k x = γ gx tx /2π . Then the signal intensity is I (k) = I (0) ρ (x) exp (−2πikx)dx (15) where, I (0) is the signal intensity without the gradient of gx and ρ(x) is the projection of the spin density along x-axis in the slice. Equation (15) indicates that I (k) is the Fourier transform of ρ(x), therefore, an inverse Fourier transformation on a series of measured data varying k value (usually varying gx ) gives ρ(x). Figure 4 illustrates the phase shift of magnetization under various gx . The

πy π/2x r.f. gx gy gz Fig. 2. A typical pulse sequence for three dimensional NMR imaging.

acquisition

Part I

where f i and T2,i are fraction and T2 for the component with the diffusion coefficient Di . By using the T2 weighted fraction of i-th component

NMR Imaging 131

132 Part I

Chemistry

Part I

d 2/d Fourier transformation

t

ω0

ω (r)

ω

Fig. 3. The rf pulse with the shape of sinc form (Sin(t/d)/(t/d)) and its Fourier transform which has a rectangular shape in the frequency domain.

magnetization without the gradient has an intensity profile corresponding to the projection of the spin density along x-axis in the slice. The application of gx induces the excessive magnetic field with the strength of gx x at the x coordinate, which is indicated by gray arrow in Figure 4. The excessive magnetic field rotates each magnetization at the individual rate of γ gx x and the magnetization has the phase shift corresponding to the x coordinate. The pitch of the phase shift k −1 is decreased with increasing gx , and the total magnetization under each gx is a Fourier component of ρ(x) at k. The application of g y during the acquisition period gives each magnetization the difference of precession rate corresponding to the y coordinate, which is reflected in the frequency in spectrum. Therefore, the spectrum obtained by the Fourier transformation of the echo signal is the projection of the spin density along y-axis ρ(y) in the slice. The echo signal S(k, t) is a Fourier component of ρ(x) at k, therefore, the two-dimensional Fourier transform of S(k, t) gives the spin density ρ(x, y) in the slice.

When there are nuclei in different environments, the effect of the excessive magnetic field induced by the field gradient coexist with the shielding effect of surrounding electron clouds, which is the origin of the chemical shift in spectra without field gradients. By using a chemical shift selective pulse, it is possible to obtain NMR imaging of desired nuclei at the chemical shift. It is also possible to remove the undesired signal by a presaturation of the nuclei.

Selection of Coherence The field gradient is used for the selection of coherence pathways. The coherence selection also achieved by phase cycling, which causes a subtraction of undesired peaks. On the other hand, the field gradient method is based on a spin echo method, which rephases desired peaks and dephases undesired peaks, needs only one scan for a spectrum or an increment in multidimensional measurements.

Fig. 4. The phase shift of magnetization under various gx . The magnetization without the gradient has an intensity profile corresponding to the projection of the spin density along x-axis in the slice. The application of gx induces the excessive magnetic field with the strength of gx x (gray arrows), which rotates each magnetization at the individual rate of γgx x and induces the phase shift corresponding to the x coordinate. The pitch of the phase shift k−1 is decreased with increasing gx .

NMR Measurements Using Field Gradients and Spatial Information

π/2x acquisition

r.f. g1

g2

Fig. 5. A pulse sequence for Gradient Selected COSY in magnitude mode. The solid line and dotted line indicate coherence pathways that have the coherence order of –1 at the acquisition period. The former is rephased and the latter remains a phase shift when the gradients are set as g1 δ1 = γ2 δ2 .

g δ1

δ2

+1 p=0 -1

Further, three is no problem of the residue for undesired peaks caused by imperfection of the subtraction in the phase cycling method. Because of this, the application of the gradient gives remarkable improvement for many measurements developed on the basis of the phase cycling method. The gradient method takes advantage of the fact that the space dependent phase shift caused by the gradient depends on the coherence order p. The phase shift is expressed by Equation (2) when p = 1. For general orders of p, the phase shift is expressed as follows, φ(r) = pγ grδ

π/2x

(16)

In a homonuclear spin system, the final phase shift after the gradient pulses becomes φ(r) = γ

n 

pi gi δi r

(17)

i

Therefore, the desired coherence pathway corresponding a set of coherence orders p1 , p2 , . . . pi , . . . pn is rephased by using the set of gradients g1 δ 1 , g2 δ 2 , . . . gi δ i , . . . gn δ n , which satisfies φ(r) = 0. Other pathways remain the dephase expressed by Equation (17). Figure 5 shows a pulse sequence for gs (Gradient Selected)-COSY in magnitude

πx acq.

r.f.(1H) π/2x

π/2x

1/2JXH

1/2JXH

t1

r.f.(X) g1 g

g3

g2 δ1

δ2

δ3

+1 1H

p=0 -1

+1 X p=0 -1

Fig. 6. A pulse sequence for Gradient Selected HMQC. The coherence pathway of solid line is selected when (γH + γX )γ1 δ1 + (−γH + γX )γ2 δ2 − γH γ3 δ3 = 0.

Part I

π/2x

Selection of Coherence 133

134 Part I

Chemistry

Part I

mode. The interval of gradients should be short in order to reduce the decay of the signal for the desired pathway by the molecular diffusion. The sinusoidal shaped gradients are used in order to reduce the effect of the eddy current induced by the gradient on the rf pulses. The solid line and dotted line in Figure 5 indicate two coherence pathways that have the coherence order of –1 at the acquisition period. When the gradients are set as g1 δ1 = g2 δ2 , φ(r) = 0 for the coherence pathway of the solid line, on the other hand, φ(r) = 2γg1 δ1 r for that of the dotted line. Consequently, the former pathway is selected. In a homonuclear spin system, the final phase shift after the gradient pulses becomes φ(r) =

n   i

γ j pi, j gi δi r

(18)

j

Figure 6 shows a pulse sequence for gs-HMQC. For the selection of coherence pathway indicated the solid line, the gradients is set as satisfies the equation (γ H + γ X )g1 δ1 + (−γ H + γ X )g2 δ2 − γ H g3 δ3 = 0

(19)

When the nuclear X is 13 C, γ13C /γH ⱌ 0.25. By using same duration time for each gradient, Equation (19)

becomes 1.25g1 − 0.75g2 − g3 = 0

(20)

For example, Equation (20) is satisfied when g1 :g2 : g3 = 2 : 2 : 1, 5:3:4 or 3:5:0. The 1 H signals of other pathways such as 1 H attached to 12 C, which gives a main peak in usual measurements, remains the phase shift and vanishes.

References 1. Callaghan PT, Principles of Nuclear Magnetic Resonance Microscopy, Clarendon Press Oxford, 1991, p. 93. 2. Yasunaga H, Kobayashi M, Matuskawa S, Kurosu H and Ando I, In: GA Webb and I Ando (Eds), Annual Reports on NMR Spectroscopy, Vol. 34, Academic Press: London, 1997, p. 39. 3. Stejskal EO, and Tanner JE. J. Chem. Phys. 1965;42:288. 4. Karger J, Pfeifer H, Heink W. Adv. Magn. Reson. 1988;12:1. 5. (a) Matsukawa S, Ando I. Macromolecules 1996;29:7136; (b) 1997;30:8310; (c) 1999;32:1865; (d) Matsukawa S, Yasunaga H, Zhao C, Kuroki S, Kurosu H, Ando I. Prog. Polym. Sci. 1999;24:995. 6. Claridge TDW. High-Resolution NMR Techniques in Organic Chemistry, Elsevier Science Ltd: Oxford, 1999, Chapter 5. 7. Johnson Jr. CS. Progress in Nuclear Magnetic Resonance Spectroscopy 1999;34:203.

135

Torsten Brand1 , Eurico J. Cabrita2 , and Stefan Berger1 1 Institut

f¨ur Analytische Chemie, Universit¨at Leipzig, Johannisallee 29, 04103 Leipzig, Germany; and 2 REQUIMTE/CQFB, Department de Qu´ımica, Faculdade de Ciˆ encias e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal

Theoretical Aspects Concepts of diffusion Self-diffusion is the random translational motion of molecules driven by their internal kinetic energy [1]. Translational diffusion and rotational diffusion can be distinguished. Diffusion is related to molecular size, as becomes apparent from the Stokes–Einstein equation: D = kB T / f

(1)

where D is the diffusion coefficient, kB is the Boltzmann constant, T is the temperature, and f is the friction coefficient. If the solute is considered to be a spherical particle with an effective hydrodynamic radius (i.e. Stokes radius) rS in a solution of viscosity η, then the friction coefficient is given by: f = 6πηrS

(2)

Pulse Sequences for PFG NMR Diffusion Measurements Using the pulsed-field gradient (PFG) method, motion is measured by evaluating the attenuation of a spin echo signal [2]. The attenuation is achieved by the dephasing of nuclear spins due to the combination of the translational motion and the imposition of gradient pulses. In contrast to relaxation methods, no assumptions concerning the relaxation mechanism(s) are necessary. The PFG NMR sequence (Figure 1) is the simplest for measuring diffusion [2]. During application of the gradient, which is along the direction of the static, spectrometer field, B0 , the effective magnetic field for each spin is dependent on its position. Therefore, the precession frequency is also position dependent which leads to the development of position dependent phase angles. The 180◦ pulse changes the direction of the precession. Hence, the second gradient of equal magnitude will cancel the effects of the first and refocus all spins, provided that no change of position, with Graham A. Webb (ed.), Modern Magnetic Resonance, 135–143.
C 2008 Springer.

respect to the direction of the gradient, has occurred. If there is a change of position, the refocusing will not be complete. This results in a remaining dephasing which is proportional to the displacement during the period  between the two gradients. Since diffusion is a random motion, there is a distribution of gradient-induced phase angles. These random phase shifts are averaged over the ensemble of spins contributing to the observed NMR signal. Hence, this signal is not phase shifted but attenuated, with the degree of attenuation depending on the displacement. In Ref. [1], this phenomenon is explained in more detail. It is shown in Ref. [1] that the signal intensity S(2τ ) after the total echo time 2τ is given by:
 

2τ δ S(2τ ) = S(0) exp − exp −γ 2 g 2 Dδ 2  − T2 3

 δ (3) = S(2τ )g=0 exp −γ 2 g 2 Dδ 2  − 3 where S(0) is the signal intensity immediately after the 90◦ pulse, T2 is the spin–spin relaxation time of the species, γ is the gyromagnetic ratio of the observed nucleus, g is the strength of the applied gradient, and δ and  are the length of the rectangular gradient pulses and the separation between them, respectively. Typically, δ is in the range of 0–10 ms, the diffusion time  is in the range of milliseconds to seconds, and g is up to 20 T/m [1]. To determine diffusion coefficients, a series of experiments is performed in which either g, δ, or  is varied while keeping τ constant to achieve identical attenuation due to relaxation. Normally, the gradient strength g is incremented in subsequent experiments. Non-linear regression of the experimental data can be used for the determination of D. Nowadays, the BPPLED pulse sequence (see Figure 2) is most often used for measuring diffusion since it allows eddy currents to decay and uses bipolar gradients which enables double effective strength as well as compensation for imperfections. This sequence is not affected by spin– spin coupling since it is based on the stimulated echo sequence.

Part I

Theory and Application of NMR Diffusion Studies

136 Part I

Chemistry

Part I

τ

Fig. 1. The Stejskal and Tanner pulsedfield gradient NMR sequence. Narrow and wide filled bars correspond to 90◦ and 180◦ pulses, respectively. Open bars with horizontal stripes correspond to pulsed-field gradients whose strength is varied during the experiment. The pulse phases are φ1 = x and φ2 = y. Phase cycling can be included to remove spectrometer artifacts.

τ

φ1

φ2

g G δ

t1

The signal intensity of the BPPLED sequence is given by:

τg δ S = Sg=0 exp −γ 2 g 2 Dδ 2  − − 3 2

(4)

∆−δ

δ

t2

the data obtained in PFG NMR measurements where the chemical shift is plotted in one (or two) dimension and the diffusion coefficient in the other dimension. This presentation allows the identification of signals belonging to one component (or at least to components showing the same diffusion coefficient). Because of this separation, DOSY can be considered as “non-invasive chromatography” [4].

Processing of Diffusion Data In the chemical shift dimension(s), Fourier transformation (FT) is applied as usual. For each frequency ν, the signal can (in general) have contributions from several components (1, . . . n) which individually decay with their respective diffusion coefficient: S(g,v) =

n  i=1



δ Si (0,v) exp −γ 2 g 2 Di δ 2  − 3

(5)

The individual diffusion coefficients Di and the signal intensities Si (0, ν) have to be extracted in order to construct the diffusion spectra. The name DOSY (diffusion ordered spectroscopy) refers to the presentation of Fig. 2. The LED pulse sequence using bipolar gradients [3]. Narrow and wide filled bars correspond to 90◦ and 180◦ pulses, respectively. Phase cycling: φ1 = φ2 = φ5 = x; φ3 = 2(x), 2(−x), φ4 = φ7 = 4(x), 4(−x), 4(y), 4(−y); φ6 = x, −x, x, 2(−x), x, −x, x, y, −y, y, 2(−y), y, (−y), y.

φ1

φ2 τ1

Applications of Diffusion NMR In PFG spin echo NMR experiments, the interesting observable variable is the diffusion coefficient, therefore, in principle, any phenomena that affects the diffusion coefficient can be studied with this technique. The concept behind the application to the study of molecular interactions is very simple and is based on the fact that the diffusion coefficient of a molecule is altered upon addition of another molecule if there is an interaction between them. Diffusion NMR has been applied to the study of intermolecular interactions both qualitatively, to identify compounds that bind to a specific receptor in NMR screening

φ3

φ4

τ1

τ2

φ5 τ1

φ6

φ7

τ1

g

g G −g

δ/2

−g

τg

te ∆

Theory and Application of NMR Diffusion Studies

Size and Shape Determination by Diffusion Measurements Since diffusion NMR allows spectral resolution by size or shape, and this resolution being especially visible in DOSY experiments, it is not surprising that the qualitative or semi-quantitative application of DOSY to the distinction of compounds according to their size is one of the most popular [21]. Examples of this type of application of DOSY can be found in many diverse areas such as in the characterization of polymer additives [22], hydrocarbon mixtures [23], in food chemistry [24,25], or carbohydrate mixture analysis [26,27], just to name a few examples. If some cautions are taken, the experimental diffusion coefficients can be used to obtain quantitative information about the size and shape of a molecule or a particular assembly. As was already mentioned, the connection between the diffusion coefficient (D) and structural properties arises because diffusion coefficients depend on friction factors ( f ) which are associated with the molecular size and the viscosity of the solution. The Stokes–Einstein equation [Equation (1)] relates the translational self-diffusion coefficient at infinite dilution of a spherical particle to its hydrodynamic radius rS , and in spite of the difficulty to justify this equation at a molecular level [28], its simplicity and success in relating experimental diffusion coefficients to molecular radii is the basis for its extensive use in the literature. Examples of the application of experimental diffusion coefficients and the Stokes–Einstein equation for size determination can be found in fields ranging from organometallic chemistry to biochemistry. This relation is usually also the starting point for the development of

other models that connect the diffusion coefficient with shapes different from spherical and to expressions related to the molecular weight of the diffusing species. A simple but very elucidative example is the characterization of THF solvated n-butyllithium aggregates by DOSY [29]. Diffusion ordered spectroscopy was used to distinguish the tetrasolvated dimeric and tetrasolvated tetrameric aggregates (see Figure 3) in THF solution. Theoretical diffusion values for the dimer and tetramer, calculated from the Stokes–Einstein equation, predict measurable differences in diffusion coefficients. For the calculation of the theoretical diffusion the viscosity of neat THF was used, and the hydrodynamic radii were determined from molecular volumes based on crystal structures and gas-phase minimized structures. A good agreement between experimental diffusion coefficients and theoretical values was obtained [29]. As was shown by Waldeck et al. [30], by considering the relation between the Stokes radius of a molecule (rS ), its experimentally determinable partial specific volume, V , and its molecular weight, M, a useful expression relating diffusion to molecular weight can be derived: 3 D1 = D2

3

M2 M1

(6)

This general relationship shows that for “ideal” spherical models there is a reciprocal cube dependence of the diffusion coefficient on molecular weight and this allows the calculation of a set of theoretical diffusion coefficients using a reference diffusion (experimental) value. It is therefore worth considering when accounting for the effect of molecular association on the apparent diffusion coefficient, expected to be measured in an experiment. Another impressive example of the applicability of the Stokes–Einstein equation is found in a study dealing

Fig. 3. PM3 optimized structures of [n-BuLi]4 ·THF4 and [n-BuLi]2 ·THF4 . Reprinted with permission from Ref. [29]. Copyright (2000) American Chemical Society.

Part I

or in studies related to host–guest chemistry [5–8], and quantitatively, in the determination of association constants [9–12] and complex or aggregate sizes [13–18]. For binding and screening studies it is usually sufficient to identify compounds that bind to a certain receptor from a mixture of non-binding compounds, or to establish a relative binding affinity, but the determination of association constants or size requires quantitative determination of the diffusion coefficients with precision and accuracy. A very comprehensive work about the factors that affect data quality in PFG spin echo NMR methods for chemical mixture analysis was published recently by Antalek [19], both data acquisition (including a discussion about experimental conditions and available pulse programs) and data analysis were considered in detail by the author. This chapter complements well the previous work of Price on the experimental aspects of PFG spin echo NMR [2]. After completing this article, an outstanding and comprehensive review on NMR diffusion experiments by Cohen et al. has been published [20].

Applications of Diffusion NMR 137

138 Part I

Chemistry

Part I

with ubiquitin [32]. The hydrodynamic radius of this protein was calculated from its diffusion coefficient determined by DOSY-HSQC experiments using an accurately calibrated temperature and the viscosity calculated for this temperature. Using Equations (1) and (2) yielded ˚ Furthermore, the NMR structure of ubiqrS = 15.8 A. uitin [33] was used for the calculation of its size, which ˚ which is in reasonably was then converted to rS = 16.3 A, good agreement with the value found in DOSY experiments. Thus, the numerical factor 6 given in Equation (2) also holds true for complex situations such as a protein in aqueous solution. This demonstrates that the assumption of a spherical solute moving in a continuous solvent is fulfilled fairly well in this case, which can be verified by inspecting Figure 4. The field of organometallic chemistry provides several examples of the application of diffusion measurements for size determination, since this is one of the fields where the use of diffusion NMR is becoming more and more popular. Pregosin is among the leaders in the application of PFG diffusion methods in organometallic chemistry and his contributions and perspectives about the technique as well as the most important applications in this field have been the subject of several publications [34]. 13 C detected DOSY was used by Schl¨orer et al. to study the unstable intermediate (2) in the reaction of CO2 with [Cp2 Zr(Cl)H] (1) (see Scheme 1) which was impossible to characterize by other means [35]. 13 CO2 was used for the reaction which was observed in situ by 13 C NMR.

Fig. 4. Schematic representation of the three dimensional structure of ubiquitin. The structure presented here is taken from the data set “1D3Z” [33] in the pdb data bank [31]. (See also Plate 9 on page XXI in the Color Plate Section.)

Cl Z r H

Cl O Z r O C H

1

H H Cl C Cl Z r O O 2

1

- H2CO

H2CO

Z r

Z r

Cl H H O C H

Z r

Cl O rZ Cl

3

Scheme 1. The reaction of [Cp2 Zr(Cl)H] (1) with CO2 [35].

Following the formation of sufficient amounts of 2, the mixture has been cooled to −78◦ C, and 13 C INEPT DOSY spectra were recorded. The intermediate 2 was shown to have a smaller diffusion coefficient than the mononuclear complex 3 (see Figure 5) and was therefore proven to be binuclear. Furthermore, its hydrodynamic radius calculated from the experimental results was found to be in good agreement with an estimation based on a minimized gas-phase structure. Still in the field of ionic interactions, a very recent paper from the group of Pregosin explored the application of PGSE NMR studies within the context of chiral cation/anion recognition [36]. According to the authors, this is the first reported example that shows that the diffusion data are sensitive enough to recognize a subtle diastereomeric structural effect on ion translation. The work investigated the dependence of the diffusion value on the diastereomeric structure of the ion pair for chiral organic salts (see Scheme 2). Investigated were the pairs of diastereomers formed between two novel chiral hexacoordinate phosphate anions, known to induce efficient NMR chiral-shifts, and chiral quaternary ammonium cations. Diffusion constants were determined for the salts [6][-4], [6][-5], [6][PF6 ], [7][-4], [7][-5], and [7][PF6 ] at different concentrations and in chloroform, dichloromethane, acetone, and methanol. To facilitate the comparison of results, hydrodynamic radii derived from the Stokes–Einstein equation, using the viscosity of the non-deuterated solvents, were calculated. The methanol data were employed to estimate the size of solvated and independently moving anions and cations. For the cations in methanol, the rS values were found to be independent ˚ for 6 and 5.0 A ˚ for 7). The values of the anion (5.2–5.3 A ˚ both in [6][-5] and [7][-5], for the anion -5 are 7.0 A − ˚ ˚ whereas for PF− 6 the values are 2.7 A in [6][PF6 ] and 2.6 A − in [7][PF6 ] in agreement with previous results for other salts of PF− 6 from the same group [37,38].

Theory and Application of NMR Diffusion Studies

2

2

Part I

3

Applications of Diffusion NMR 139

3

slow −10.2

lgD / m2s−1 −10.0

−9.8

fast 116

115

114

103

102

101

64

63

13

d ( C) Fig. 5. 100 MHz 13 C INEPT DOSY spectrum obtained during the reaction of 1 with 13 CO2 at –78 ◦ C in [D8 ] THF. The sections show the signals of 2 (δC = 114.6 ppm (Cp) and 101.7 ppm (CH2 )) and 3 (δC = 114.9 ppm (Cp) and 63.5 ppm (OCH3 )). See Scheme 1 for the chemical structure of 1, 2, and 3. Figure taken from Ref. [35]. Reproduced with permission of John Wiley & Sons Limited.

Cl Cl

Cl

Cl

Cl

O

Cl

O

Cl

Cl Cl

Cl

O

O

O

O

P Cl

P

O Cl

TRISPHAT ∆-4

O

O O

Cl

Cl

Cl Cl

Pr N

N

O O

BINPHAT ∆-5

Cl

Cl Cl

Pr N

O O

6

Cl

7

Scheme 2. Chiral anions and cations investigated by Pregosin and coworkers [36].

140 Part I

Chemistry

Part I

Internal Standards for Diffusion Measurements The direct determination of hydrodynamic radii, and thus size, through the Stokes–Einstein equation requires a knowledge of the solution viscosity at the measurement temperature. Additionally, in order to have accurate information about size in studies related to molecular interactions where the comparison of diffusion coefficients obtained in different conditions is usual, it is crucial to be able to separate contributions due to changes in viscosity and effective changes in hydrodynamic radii. In PFG NMR, two major approaches are frequently used to avoid additional experimental work to measure the viscosity of the solution. The simplest approach is to consider that the viscosity of the solution is approximately the same as the viscosity of the pure non-deuterated solvent. This approximation has been shown to be legitimate in a number of cases, especially when considering pure solvents and diluted solutions and some examples have already been mentioned above for the determination of hydrodynamic radii of, n-butyllithium aggregates [29] and solvated anions and cations [36] but many more can be found in the literature. In complex solutions it may be more difficult to obtain a value for the viscosity of the exact solvent mixture, and in these cases the interpretation of size or molecular mass derived from diffusion data has to take into account the validity of the approximations made and the possibility of under/over valuating the viscosity. The other solution to the problem is the back calculation of the solution viscosity, through the Stokes– Einstein equation, by using the diffusion measured for a non-interacting reference compound of known hydrodynamic radius. This internal probe should be of similar size with respect to the molecules of interest so that it experiences a similar microscopic environment and can act as an internal viscosity standard. The use of such a standard allows the estimation of size even in complex solution mixtures and the comparison of diffusion coefficients in series of experiments where the composition of the solution is altered, a situation that commonly arises in studies related to molecular interactions. The use of a diffusion standard allows one also to separate the contributions due to changes in viscosity and effective changes in hydrodynamic radii even if the hydrodynamic radius of the standard is not known. In fact, the ratio of the diffusion of a particular solute and the reference compound will be independent of the viscosity (D/Dref = rSref /rS ) and relative information about changes in hydrodynamic radius can be obtained when comparing ratios measured in different conditions. This procedure is well exemplified in a study by Cabrita et al. where tetramethylsilane (TMS) was used as a standard for the diffusion measurements to account for viscosity changes, and was proposed as a reference for the study of intermolecular interactions involving hydrogen bonds in organic solutions [14]. Kapur et al. [39] have shown that DOSY can be a useful technique for the quali-

tative study of the relative strengths of hydrogen bonds in solution. Since the formation of an intermolecular H-bond leads to a decrease of the diffusion coefficient of a certain molecule, the relative decrease in the diffusion coefficient of a particular molecule in a mixture of molecules, interacting by H-bond with a common H-bond acceptor or donor, was interpreted in terms of the tendency for the molecules in the mixture to be involved in association by H-bonds with the H-bond donor or acceptor. As an example, it was shown that when dimethylsulfoxide (DMSO), a strong H-bond acceptor, is added to a solution containing phenol (8) and cyclohexanol (9), two molecules with a similar shape, a higher relative decrease in the diffusion coefficient of phenol was observed. This different behavior was attributed to the greater tendency of phenol to be involved in H-bonding with DMSO, since phenol is more acidic than cyclohexanol [39]. OH

8

OH

9

Binding, Screening, and Determination of Association Constants In the previous section, we have shown examples of applications that explore the relation between size and diffusion coefficient primary as a source of information on molecular size. However, this relation can be explored in a different way in order to get information about the strength of intermolecular interactions. The majority of the reports on the application of diffusion NMR to the study of intermolecular interactions deal with the alteration of the diffusion coefficient due to binding phenomena in solution. In fact, when a small molecule binds to a large receptor, its diffusion coefficient may decrease more than one order of magnitude. This means that at least for some time the small molecule will have the diffusion coefficient of the large receptor, and if we consider the fast exchange limit, its observed diffusion coefficient (Dobs ) is described by: Dobs = f free Dfree + f bound Dbound

(7)

where f and D denote the molecular fractions and diffusion coefficients of the free and bound molecule. If the difference in size is large enough, it can be assumed that the diffusion coefficient of the receptor or host (DH ) is not greatly modified and that Dbound is the same as the DH alone. This relation is the starting point for the majority of the diffusion NMR-related binding studies.

Theory and Application of NMR Diffusion Studies

M eO

H N

M eO

O

f HG =

M eO O OM e

10 For this reason, a model peptide containing the 12 Cterminal residues of α-tubulin (VEGEGEEEGEEY) was investigated with respect to the pH dependence of the binding to 10 [40]. Although binding studies on this system have only been computational using docking programs, it was shown in the diffusion studies that the model peptide adopts different conformations depending on the pH, this being reflected by different observed diffusion constants. In this work, NOE data have also been used, but only for determining the conformation of the single peptide. The determination of association constants (K a ) from NMR data has been recently reviewed by Fielding [9] with a section dedicated to diffusion experiments. The starting point for the determination of the association constant is Equation (7) and the mathematical treatment to get K a from Dobs is exactly the same as for any other NMR observable, such as δobs . As Fielding points out in his review, the advantage of measuring Dobs instead of δobs is that the diffusion coefficient of the host–guest complex can be assumed to be the same as that of the non-complexed host molecule, thus reducing one unknown in Equation (7). In principle, this allows one to determine K a within a single experiment and without the need of titrations, as exemplified below. The formation of a host–guest complex of stoichiometry 1:1 is described by: [HG] [H] [G]

Dobs = f G DG + f HG DHG

(8) (9)

where [HG], [H], and [G] are the equilibrium concentrations of the host–guest complex, host, and guest, respectively and f G and f HG are the molar fractions of

DG − Dobs DG − DHG

(10)

If, as it was mentioned earlier, DHG is assumed to be the same as the measurable diffusion coefficient of the host (DH ), then f HG can easily be determined. Accounting for mass balance and combining Equations (8) with (10), we arrive to the expression for the association constant: Ka =

H

Ka =

non-complexed guest and complex, respectively. From Equation (9) it follows that:

f HG (1 − f HG )([H]0 − f HG [G]0 )

(11)

where [H]0 and [G]0 represent the total concentrations of host and guest, respectively. The procedure before is straightforward and examples of its application can be found in recent literature related to host–guest chemistry studies [12,41]. Rather than exemplifying the examples in detail here, we prefer to take a closer look at the limitations of the approximation that DHG = DH . The assumption that DHG = DH is valid for the majority of studies involving small molecules binding to macromolecules (typically biological), but may not necessarily be true for smaller host molecules usually employed in host–guest chemistry studies. To test the assumption that DHG = DH for a typical medium-sized host molecule, Cameron et al. [10] have studied the β-cyclodextrin (11) complexes of cyclohexylacetic acid (12) and cholic acid (13). They have shown that caution should be taken when determining the association constant by the single experiment method, and have employed a data treatment which takes into account the diffusion of all species. With this treatment, the 1 H NMR chemical shift titration method and the diffusion coefficient method give the same results for K a . Simova and Berger presented a comparison of DOSY experiments and chemical shift titrations with respect to the determination of association constants [42]. The authors investigated camphor and cyclodextrins (CD) in D2 O. They showed that precise association constants are more easily determined by chemical shift titration. Diffusion measurements using HR-DOSY allow easy determination of the complex composition at different concentration ratios and an estimation of the binding energy if a viscosity reference, in this case tetramethylammonium bromide, is present. Linear dependence of the diffusion coefficients on the molecular mass of free and associated CD has been observed (see Figure 6). The solution structures of α- and β-CD complexes of camphor in D2 O were deduced from intermolecular cross-relaxation data obtained by using the ROESY sequence. Different preferential orientation in the 2:1 α-CD and 1:1 β−CD species have been derived in contrast to the weak 1:1 complex

Part I

In this field two main lines of application can be identified, one more qualitative, related to the screening of complex mixtures or individual molecules, usually with the aim of identifying potential new drug compounds, and another, more quantitative, concerned with the determination of association constants. A recent example of the first type of application mentioned above is the binding of cholchicine 10 to α/β tubulines, which is of large interest in cancer-related studies.

Applications of Diffusion NMR 141

142 Part I

Chemistry

Part I

OH

O OH

O

O

O

H O

O OH H O

H O

O

OH

OH

H O O H O

H O

O OH

OH O OH

OH

O

H O OH O

OH O OH

H O

O

O OH

11

OH O OH

OH O

H O

OH

H

12

13

2,8

α-CD β-CD

2,6

D (10

-10

2 -1 m .s )

3,0

γ-CD

2,4 2,2

(α-CD)2-camphor

2,0 0 9

130 0

170 0

210 0

M (g.mol-1) Fig. 6. Dependence of the diffusion coefficients of cyclodextrins and the α-CD complex of camphor on the molecular mass [42].

Theory and Application of NMR Diffusion Studies

References 1. Price WS. Concepts Magn. Reson. 1997;9:299. 2. Price WS. Concepts Magn. Reson. 1998;10:197. 3. Wu D, Chen A, Johnson CS Jr. J. Magn. Reson. A. 1995;115:260. 4. Huo R, Wehrens R, van Duynhoven J, Buydens LMC. Anal. Chim. Acta. 2003;490:231. 5. Meyer B, Peters T. Angew. Chem. Int. Ed. 2003;42:864. 6. Shapiro MJ, Wareing JR. Curr. Opin. Drug Discov. Devel. 1999;2:396. 7. Avram L, Cohen Y. Org. Lett. 2002;4:4365. 8. Avram L, Cohen Y. J. Am. Chem. Soc. 2002;124:15148. 9. Fielding L. Tetrahedron. 2000;56:6151. 10. Cameron KS, Fielding L. J. Org. Chem. 2001;66:6891. 11. Wimmer R, Aachmann FL, Larsen KL, Petersen SB. Carbohydr. Res. 2002;337:841. 12. Avram L, Cohen Y. J. Org. Chem. 2002;67:2639. 13. Price WS, Tsuchiya F, Arata Y. J. Am. Chem. Soc. 1999;121:11503; and references therein as an example of the application to the study of protein aggregation. 14. Cabrita EJ, Berger S. Magn. Reson. Chem. 2001;39: S142. 15. Cameron KS, Fielding L. J. Org. Chem. 2001;66:6891. 16. Valentini M, Pregosin PS, R¨uegger H. Organometallics. 2000;19:2551. 17. Zuccaccia C, Bellachioma G, Cardaci G, Macchioni A. Organometallics. 2000;19:4663. 18. Timmerman P, Weidmann J-L, Jolliffe KA, Prins LJ, Reinhoudt DN, Shinkai S, Frish L, Cohen Y. J. Chem. Soc. Perkin Trans. 2000;2:2077. 19. Antalek B. Concepts Magn. Reson. 2002;14:225. 20. Cohen Y, Avram L, Frish L. Angew. Chem. 2005;117: 524. 21. Johnson CS Jr. Prog. NMR Spectrosc. 1999;34:203.

22. Jayawickrama DA, Larive CK, McCord EF, Roe DC. Magn. Reson. Chem. 1998;36:755. 23. Kapur GS, Findeisen M, Berger S. Fuel. 2000;79:1347. 24. Gil AM, Duarte I, Cabrita E, Goodfellow BJ, Spraul M, Kerssebaum R. Anal. Chim. Acta. 2004;506:215. 25. Nilsson M, Duarte IF, Almeida C, Delgadillo I, Goodfellow BJ, Gil AM, Morris GA. J. Agric. Food Chem. 2004;52:3736. 26. Schraml J, Blechta V, Soukupov´a L, Petr´akov´a E. J. Carbohydr. Chem. 2001;20:87. 27. Diaz MD, Berger S. Carbohydr. Res. 2000;329:1. 28. Walser R, Mark AE, van Gunsteren WF. Chem. Phys. Lett. 1999;303:583. 29. Keresztes I, Williard PG. J. Am. Chem. Soc. 2000;122: 10228. 30. Waldeck AR, Kuchel PW, Lennon AJ, Capman BE. Prog. NMR Spectrosc. 1997;30:39. 31. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE. Nucleic Acids Res. 2000;28:235. 32. Brand T, Cabrita EJ, Morris GA, Berger S. (in preparation). 33. Cornilescu G, Marquardt JL, Ottiger M, Bax A. J. Am. Chem. Soc. 1998;120:6836. 34. Valentini M, R¨uegger H, Pregosin PS. Helv. Chim. Acta. 2001;84:2833; and references therein. 35. Schl¨orer NE, Cabrita EJ, Berger S. Angew. Chem. Int. Ed. 2002;41:107. 36. Mart´ınez-Viviente E, Pregosin PS, Vial L, Herse C, Lancour J. Chem. Eur. J. 2004;10:2912. 37. Kumar PGA, Pregosin PS, Goicoechea JM, Whittlesey MK. Organometallics. 2003;22:2956. 38. Mart´ınez-Viviente E, Pregosin PS. Inorg. Chem. 2003;42: 2209. 39. Kapur GS, Cabrita EJ, Berger S. Tetrahedron Lett. 2000;41: 7181. 40. Pal D, Mahapatra P, Manna T, Chakrabarti P, Bhattacharyya B, Banerjee A, Basu G, Roy S. Biochemistry. 2001;40:512. 41. Frish L, Sansone F, Casnati A, Ungaro R, Cohen Y. J. Org. Chem. 2000;65:5026. 42. Simova S, Berger S. Journal of Inclusion Phenomena and Macrocyclic Chemistry. J. Incl. Phenom. (in press).

Part I

with γ -CD. Proton NMR chemical shift values proved to be much more sensitive to diastereomeric complex formation than are diffusion coefficients.

References 143

Part I

Host–Guest Chemistry

147

John A. Ripmeester and Christopher I. Ratcliffe Steacie Institute for Molecular Sciences, National Research Council Canada, Ottawa, ON, Canada K1A 0R6

Introduction This contribution will focus on the use of NMR spectroscopy to study host–guest chemistry in the solid state. NMR spectroscopy already is a well-established approach for the study of complex formation in solution, for instance, for measurement of equilibrium constants and stoichiometry. In the solid state, guest–host chemistry is a rather more complex issue, as the materials in question range from molecular receptor–guest complexes that have crystallized, to extended framework materials that have cavities, channels, and interlamellar void space which may or may not be easily accessible to guest species. Examples of the first instance are crown ether and cyclodextrin complexes. For the latter, many diverse organic and metal-organic materials have been constructed using the principles of supramolecular chemistry and crystal engineering to assemble frameworks out of building blocks. On the inorganic side, there are zeolites, clathrate hydrates, clathrasils and zeosils, aluminum phosphates (ALPOs), metal cyanides and oxides, clays, carbons (graphite, nanotubes), and mesoporous materials such as the siliceous MCMs, and related materials. Details on all these different kinds of host–guest materials can be found in the series Comprehensive Supramolecular Chemistry [1]. NMR spectroscopy can contribute a great deal to understanding the structure and properties of such host–guest materials, as also reviewed in a chapter in Vol. 8 of the aforementioned series and references therein [2]. The topic of host–guest materials is a very broad area with an extensive literature, and for this reason and the limited length of this chapter we have chosen to illustrate with examples largely from our own work and to refer the interested reader to more extensive reviews. In the study of solid-state host–guest chemistry, the problems that stand out are the accurate determination of the host structure, the location and orientation of the guest, and the understanding of specific interactions that lead to molecular recognition and selective adsorption or inclusion. This also includes the understanding of the electronic structure as obtainable from the chemical shift. Applications include gas separation and storage, shape-selective catalysis, molecular sensing, drug delivery, and a variety of systems have been developed that are meant to Graham A. Webb (ed.), Modern Magnetic Resonance, 147–154.
C 2008 Springer.

mimic processes in biology in their entirety or in part, for instance, biocatalysis involving enzymes, ion channels, etc. [1,2].

The Solid-State Spectrum The main interactions that dominate the NMR lineshape in the solid state include the nuclear dipolar interaction, giving information on the distance between magnetic nuclei (up to ∼0.8 nm), chemical shielding, giving information on the electronic structure, and the quadrupolar interaction, which is determined by the interaction of the nuclear electric quadrupole with the electric field gradient at the nucleus [3]. Occasionally J-coupling also has an influence. In many if not most cases, several or all of these interactions are present simultaneously. Since all have a spatial dependence (interactions described by second rank tensors), the solid-state NMR lineshape in powdered materials is broad and complex. Most materials have a number of inequivalent atoms, so the selective elimination of some of the interactions in order to obtain high-resolution spectra was a major focus in the NMR spectroscopy of solids for many years. The most commonly used techniques include high power decoupling (HPDEC), generally to eliminate dipolar couplings to abundant spins such as protons, magic angle spinning (MAS), which can reduce dipolar couplings as well as the anisotropic chemical shift [4], and combinations of pulse schemes and spinning to separate chemical shifts and quadrupolar couplings (MQMAS) [5]. Recently, much effort has been expended into re-introducing dipolar couplings into high-resolution spectra to make use of these to measure a variety of internuclear distances [6]. Unfortunately, the spectroscopy of the ubiquitous 1 H is rather difficult, as its chemical shift scale is small and homonuclear dipolar couplings dominate the resonance lines which are often very broad. A number of multiple pulse schemes have been developed to give a measure of high resolution [7], and the more recent developments in fast MAS have proved to be particularly powerful [8]. Especially in soft materials such as polymers and biosolids the use of fast MAS has allowed applications of many of the techniques that are routinely used in solution NMR to derive information on structure in the solid state [9].

Part I

Solid-State NMR in Host–Guest Chemistry

148 Part I

Chemistry

Part I

As the design of organic and metal-organic guest–host systems is a heavily researched topic today, many applications deal with the use of 13 C NMR spectroscopy in the solid state, which is easily accessible to most researchers, and the interpretation of the spectrum is relatively straightforward. The approaches used are similar to those applied to a number of other spin 1/2 nuclei (15 N, 31 P, 29 Si), as in all of these cases dipolar couplings are mainly heteronuclear, often to 1 H, and can be removed by HPDEC. The use of 2D techniques such as HETCOR and WISE are able to provide information on H atoms that are attached to the above spin 1/2 nuclei with a measure of resolution that can be considerably greater than the 1 H spectrum itself.

General Characterization One of the first things to realize is that some very simple NMR experiments can provide a great deal of information in the characterization of a host–guest material. For example 13 C cross-polarization (CP) and MAS are frequently used to determine which guests are taken up by the host and HPDEC/MAS can be used to quantify the relative amounts of host and guest by comparing intensities of distinct lines (This can also be done with CP provided the CP behavior is calibrated for different resonances). Quite often the question with new materials is simply whether the guest is indeed included, and this can be confirmed by CP/MAS since it detects only material in the solid state, even in the presence of excess gaseous or liquid guest, which is sometimes necessarily present to keep the complex stable. The presence of dynamics can also be established using the dipolar dephasing CP/MAS experiment, since, while only quaternary carbons should appear in this spectrum, non-quaternaries will appear with reduced intensity if their C–H dipolar interactions are considerably reduced by dynamics. 13 C chemical shifts have been used to determine conformations, e.g. in crown ethers the methylene shifts relate to torsional angles [10]. Chemical shifts can also be used to identify products in “ship-in-abottle” type synthesis, e.g. 31 P NMR of P4 Sn (n = 3 − 7) synthesized inside zeolites [11] and 13 C of organics in zeolites, or to show the existence of unusual or unstable species which are stabilized by inclusion, e.g. a novel P4 S4 isomer in zeolite [11], Na− and diamagnetic metal clusters in zeolites [12]. The NMR of metal nuclei such as 113 Cd and 63 Cu has been used to determine local environments and connectivities in metal cyanide host lattices [13,14], e.g. 113 Cd shifts reflect the numbers of C and N atoms attached to tetrahedrally coordinated Cd, with CdC4 to low field ranging to CdN4 at high field; quadrupolar effects on 63 Cu lineshapes indicate whether its environment is strictly tetrahedral (isotropic line), or slightly distorted CuC4 (2nd order lineshape) or very distorted CuC3 N (too

broad to detect); J-couplings can show connectivities, e.g. Cu–13 C4 or 67 Zn–15 N4 . Chemical shift and quadrupolar lineshapes (and corresponding asymmetry parameters) also reflect the local crystal symmetry, providing another link to structure, e.g. 77 Se NMR of H2 Se clathrate hydrate distinguishes between H2 Se in the small spherical cage (isotropic line) and the oblate larger cage (axially anisotropic line), Figure 1 [15]. Chemical shifts can also be sensitive to the presence of other guests in the same or in neighboring cavities, e.g. 129 Xe in Dianin’s compound [16]. Similarly, quadrupolar nuclei can be particularly sensitive to loss of local crystal symmetry when guests are removed from neighboring cages, e.g. 23 Na in the cages of silicon clathrates [17]. 63

Structural Information from Spin 1/2 Nuclei Organic and Metal-Organic Hosts NMR spectroscopy is primarily a local order technique, which makes it a particularly strong ally to X-ray crystallography, which is a technique that depends on the presence of long-range order. However, the 13 C spectrum is sensitive not only to the presence of chemical inequivalence, but also to crystallographic inequivalence. For most materials, the 13 C NMR spectrum should confirm the asymmetric unit as determined by diffraction. Hence, in many cases 13 C NMR spectroscopy provides a rapid way of detecting structural similarities or differences in guest–host materials with a common host lattice: the spectrum gives a quick assessment of the content of the asymmetric unit, a very useful piece of information before attempting a complete structural determination [18]. This is also particularly useful when dealing with issues of polymorphism or pseudo-polymorphism [19]. In the presence of disorder, which may be dynamic, there may be disagreements between diffraction and NMR, usually with more line splittings in the NMR spectrum than one would expect based on the determined asymmetric unit. This means that either locally, or even on a larger scale, the symmetry of the lattice is lower than expected. For instance, in Dianin’s compound with p-xylene as guest there are many more splittings than one would expect from the X-ray symmetry, from which the cavity where the p-xylene resides has three-fold symmetry and an inversion center [20]. However, the p-xylene is statically disordered so that each cage has lost its threefold axis and inversion center. Since the disorder is not correlated throughout the lattice, it is spatially averaged, which gives a high symmetry in X-ray diffraction, and NMR sees the local symmetry, which is far lower. A different situation exists for the compound of calixarene

Host–Guest Chemistry

Small cage

Large cage Large cage

180K

230K

270K

240K

100 ppm

with toluene [21]. NMR spectroscopy shows a symmetry lowering transition to occur at ∼250 K, which is not seen by X-ray diffraction (Cu radiation), Figure 2. However, when shorter wavelength Mo radiation is used, the lower symmetry phase is indeed observed. This can be understood in terms of the volume over which the ordered lattice is coherent. Cu radiation requires a larger volume than Mo, so when using the longer wavelength the spatial averaging inherent in diffraction again shows the lattice to be of higher symmetry than it actually is [22]. As in solution, in the solid state it is possible to use complexation-induced shifts. For instance, guest nuclei that deeply penetrate into the cavity of calix[4]arene are deshielded by ring current effects from four aromatic rings, which can give 13 C methyl resonances a high field shift of up to ∼6 ppm [21]. Of course, these may be modified by dynamic processes that involve exchange of methyl groups over different sites. The fact that the toluene methyl has a much larger shift than pentane methyls (with the methyls equivalent) suggests that the pentane molecule can invert itself in the cavity. The X-ray structure shows that there are two positions for the methyl groups of the pentane that are quite different (one in–one

Fig. 1. Symmetry and dynamics in hydrogen selenide structure I clathrate hydrate: 77 Se NMR of H2 Se/7H2 O (left) and 2 H NMR of D2 Se/7D2 O (right). The spherical small cage gives rise to isotropic lineshapes, whereas the oblate large cage gives axially symmetric 77 Se chemical shift anisotropy and 2 H quadrupolar lineshapes. The intensities of the two components give the relative cage occupancies. The lines are considerably narrower than the static lineshapes due to rapid reorientation of the guest molecule. The broadening and reduced sharpness of the features, especially of the 2 H lineshape, as temperature decreases is due to freezing out of the host–water reorientational motions.

5 kHz

out), so that the two methyl groups indeed must exchange between the two positions. This approach has allowed the determination of the order of preference of a variety of moieties for occupancy of the deep cavity in calix[4]arene; generally methyls, methylenes, methines, and hydroxyls are preferred over halogens [23].

Inorganic Hosts Some of the best-known inorganic host materials are the zeolites. In this case, 29 Si NMR spectroscopy has allowed the measurement of the distribution of Si and Al in the lattice, as Si with 0–4 Al nearest neighbors are easily distinguished [24]. Since Al is a quadrupolar nucleus, the Si lines are rather broad, so that further distinctions of inequivalence are not possible. However, on going to an all Si lattice it becomes possible to obtain extremely high resolution spectra, where it is possible to measure through-bond connectivity (COSY), as well as all interSi distances less than ∼0.7 nm [25]. A recent example, using a symmetry-based dipolar recoupling scheme has shown that complete three-dimensional structures can be traced out [26].

Part I

Small cage

Structural Information from Spin 1/2 Nuclei 149

150 Part I

Chemistry

Part I

observed nucleus that arises when it is observed with or without dipolar coupling to one or more nearby heteronuclei. The difference, known as the REDOR fraction, can be modeled in terms of the internuclear distances, e.g. as for the guest–host compound of p-t-butylcalix[4]arene with Cs [27]. However, an added complication in many guest– host systems would be the presence of molecular motion, giving a distance that is averaged [28]. REAPDOR [29], TRAPDOR, and QUEDOR [30] all are versions related to the SEDOR/REDOR family of experiments but that involve quadrupolar nuclei. K

Spin Counting

337

179

129

200

100

0

−100

−200

kHz Fig. 2. 2 H NMR lineshapes of toluene-d5 in t-butylcalix[4]arene. At 129 K the toluene is static, at 179 K it is undergoing rapid 180◦ rotations about the molecular axis, and at 337 K it has rapid four-fold reorientation. The weak broader line visible in the 337 K spectrum is from the para-deuteron, which lies along the axis and is largely unaffected by the motion. The switch from two-fold to four-fold motion of the guest is associated with a phase transition at 248 K. In the low temperature phase, the calix host is locked into a two-fold symmetry, but in the higher temperature phase it alternates dynamically between two two-fold structures at 90◦ to each other.

Distance Measurements Other methods for measuring distances that give significant information on conformations of the host, or guest–host distances are techniques such as spin-echo double resonance (SEDOR), which is a low-resolution method, and the spinning version, REDOR, which gives high-resolution information. This approach uses the time evolution of the difference in the magnetization for the

Spin-counting, a technique that measures the order of the multiple quantum coherence that can be generated, can be used to measure the number of nuclei coupled by dipolar interactions. This can be used to probe the size of clusters of atoms or molecules, which may or may not be bound together, or that fill a cavity in a host material. With increased time, longer distances can be probed, so that intercluster distances can be obtained as well. 1 H spin counting in solids was first demonstrated by Munowitz and Pines [31]. One application to host–guest materials was a study of the numbers of protons present in “Si clusters,” and their precursors, formed inside the large cage of Y zeolite [32]. The 1 H spin count for the precursor, consisting of Si2 H6 molecules and Si2 H5 groups attached to the framework, reaches a plateau at 38 spins corresponding to the number in the cluster held within one cavity, and the final Si-cluster was found to contain 5 H atoms. This evidence helped to show that the clusters must be rather smaller than had been suggested previously.

Probing Pore Spaces The last two decades have seen the growth of 129 Xe NMR [33], and more recently of hyperpolarized (HP) Xe [34,35], as a valuable tool for probing the pore space of host–guest materials. Early work on clathrate hydrates and a number of organic clathrates [36] established a relationship between the size and shape of a closed cavity and the Xe chemical shift and its anisotropy: smaller pores give larger shifts, and anisotropy reflects whether the cavity is spherical (zero anisotropy), oblate (positive anisotropy or skew), prolate (negative anisotropy or skew), or nonaxial. More recently a broad correlation between 129 Xe shifts and pore sizes in mesoporous silicas has been determined [37]. The observed Xe lineshape is a dynamic average over all the space accessible to the Xe over the timeframe of the experiment, and in open pore systems this can lead to loading dependent shifts and lineshapes [38]. In some materials incomplete filling leads to signals corresponding to cages with different numbers of Xe [39].

Host–Guest Chemistry

Dynamics 151

Part I

2D EXSY experiments can then be used to follow the exchange of Xe between cages with different occupancies, e.g. in NaA and AgA zeolites [39,40]. Thus Xe can be used to determine the interconnectivity of different pores and the exchange barriers. A recent example is a study of organic aerogels where the hierarchy of exchange between micro and mesopore spaces and the gas phase could be determined [41]. HP Xe is particularly useful due to its enormously enhanced sensitivity. This allows the use of very low Xe concentrations and thus removes the effects caused by Xe–Xe interactions, probing only the Xe– host interaction [42]. Another example is the void space access test, as illustrated in Figure 3 for a low density phase of p-t-butylcalix[4]arene [42]. HP Xe also makes it possible to follow real time processes, such as the formation/decomposition of gas hydrates [43]. Another important use has been in the study of structural transformations and competitive exchange between Xe and another guest in a metal-organic framework [44]. In recent work on dipeptides, Xe NMR spectroscopy has demonstrated both a new type of material (biozeolites) [45] and experimental and modeling calculations illustrating an application to biomaterials [46]. 131 Xe studies have revealed some interesting effects where the quadrupolar interaction probes longer range order/disorder compared to the chemical shift [47].

MRI While MRI has very broad application in the medical arena, it is finding increasing uses in studies of materials, particularly of porous host–guest systems [48]. This is particularly true of porous host–guest systems where the long T2 of the guest (as gas or liquid) makes imaging feasible. There has been some use of HP 129 Xe as the probe nucleus, but the easiest nucleus is 1 H and this is present in many guests, e.g. organic gases or liquids. MRI can be used to study diffusion of a guest into a host, both by direct imaging as a function of time and by MRI spectroscopy [49], and to probe whether the distribution of pore space in a material is homogeneous [41]. HP Xe chemical shift imaging can be used to probe composites of different porous materials [50]. By monitoring the disappearance of the signal from the water as it solidifies into the host framework,1 H MRI was used to establish that growth of gas hydrate from small droplets was not continuous, but had a random component [51].

Dynamics NMR is the technique par excellence for the detailed study of dynamics in solids (c.f. neutron scattering, dielectric relaxation, and torsional/librational spectroscopy), and

Fig. 3. Void space access test with hyperpolarized 129 Xe NMR. The sample, a low density phase of p-t-butylcalix[4]arene without obvious channels, was exposed to a flow of HP Xe gas at various temperatures in the NMR probe. Static and MAS spectra are shown at each temperature. X marks spinning side bands. At room temperature, the HP Xe signals correspond to adsorbed, or interparticle gas, and included gas. At high temperature much more gas is included, and it has the shape typical of an approximately axial tensor. One can see that the MAS spectrum in intermediate temperature regions consists of two signals. This means that there are two phases, one transforming to the other with increased temperature. The transition is not sharp, as the transition temperature depends on the degree of loading.

dynamics plays a large role in the behavior of host–guest materials [2]: The generally weak interaction between host and guest means that in many instances the guest rotates even at very low temperatures, and is often free to roam about the pore spaces (rotational and translational freedom). This creates problems for structure determination due to dynamic disorder. In many instances the symmetry of the guest does not mesh with the symmetry of the cavity, and the result is an accommodation

152 Part I

Chemistry

Part I

with the guest fitting into several possible positions to create a pseudo-symmetry. From the diffraction point of view this disorder can be either static, resulting from spatial averaging over many unit cells, or dynamic, as the molecule hops between the various positions, and NMR can be used to distinguish the two [20,21]. A number of NMR techniques are used to probe dynamics: lineshapes, linewidths, relaxation times, 2D-exchange, dipolar fadeout, etc. The earliest works on host–guest materials used the traditional low field methods of 1 H lineshapes, second moments and T1 vs. temperature to obtain motional models, correlation times and activation energies (E a ), mainly of guests but sometimes of host, as in the case of clathrate hydrates. 2 H NMR, however, has perhaps been the greatest tool available for studying reorientational dynamics [52,53] (though it does have the disadvantage of not probing translational motion), and in the process it also sometimes yields geometrical information. As temperature is increased the motion initially distorts the lineshapes which go through a series of changes resulting in a narrowed lineshape in the fast motion limit. The static or fast motion limit lineshapes are readily analyzed to give the three components of the effective quadrupolar coupling tensor, and from these it is usually possible to determine a unique dynamic model. Once a model is determined the lineshapes at different jump rates can be calculated and matched with experiment. A plot of log(rate) vs. temperature then yields an E a , e.g. 18-crown-6 in its complexes [54]. NMR has shown how surprisingly easy it can sometimes be for rather large molecules, such as 18-crown-6, to reorient in the solid. A guest molecule in cages with different symmetry can have quite different dynamics, resulting in different 2 H and chemical shift lineshapes, e.g. H2 Se clathrate hydrate described above [15], and cyclohexane-d12 in different metal cyanide frameworks [2]. This sensitivity to symmetry also frequently shows as sudden changes in the dynamic lineshapes of guests at phase transitions [21]. 2 H NMR has also been used to study the diffusion of organic molecules as they jump between adsorption sites inside zeolites, e.g. benzene-d6 in H-SAPO-37 [55]. The analysis of dynamics can lead to a separate determination of the orientation of the guest with respect to crystal axes, and this can help in developing structural models for X-ray refinement, e.g. pyridine-d5 in t-butyl-calix[4]arene [56]. More recently a few cases of dynamics involving noninteger quadrupolar nuclei have come to light, e.g. 131 Xe in xenon clathrate hydrate is sensitive to the motion of the framework water molecules [47], and 17 O in the water of THF hydrate [57]. NMR and diffraction constitute the two primary tools for the study of structure and dynamics in host–guest materials, and, as will be evident from this brief review, their roles are very much complementary. On the other hand, NMR truly comes into its own when used to provide

insight into host–guest materials for which suitable single crystals are not available, or which by their very nature are not crystalline. NMR has a long and bright future in this very diverse and expanding field of chemistry.

References 1. Atwood JL, Davies JED, MacNicol DD, Vogtle F, Lehn JM (Eds). Comprehensive Supramolecular Chemistry, Vols. 1– 11. Elsevier: Oxford, 1996. 2. Ripmeester JA, Ratcliffe CI. Solid-State NMR Spectroscopy: Applications to Supramolecular Chemistry. In: JL Atwood, JED Davies, DD MacNicol, F Vogtle, JM Lehn (Eds). Comprehensive Supramolecular Chemistry, Physical Methods, Vol. 8. Elsevier: Oxford, 1996, p 323. 3. Fyfe CA. Solid State NMR for Chemists. CFC Press: Guelph, 1983. 4. Bryce DL, Bernard GM, Gee M, Lumsden MD, Eichele K, Wasylishen RE. Practical aspects of modern routine solidstate multinuclear magnetic resonance spectroscopy: onedimensional experiments. Can. J. Anal. Sci. Spectrosc. 2001;46:46. 5. Medek A, Harwood JS, Frydman L. Multiple-Quantum Magic-Angle Spinning NMR: A New Method for the Study of Quadrupolar Nuclei in Solids. J. Am. Chem. Soc. 1995;117:12779. 6. Duer MJ (Ed). Solid State NMR Spectroscopy Principles and Applications. Blackwell Science: Oxford, 2002. 7. Gerstein BC, Dybowski CR. Transient Techniques in NMR of Solids. Academic press: London, 1985, p 213. 8. Brown SP, Spiess HW. Advanced Solid-State NMR Methods for the Elucidation of Structure and Dynamics of Molecular, Macromolecular, and Supramolecular Systems. Chem. Rev. 2001;101:4125. 9. Pawsey S, McCormick M, De Paul S, Graf R, Lee YS, Reven L, Spiess HW. 1 H Fast MAS NMR Studies of Hydrogen-Bonding Interactions in Self-Assembled Monolayers. J. Am. Chem. Soc. 2003;125:4174. 10. Buchanan GW. Nuclear Magnetic Resonance Studies of Crown Ethers. Prog. NMR Spectrosc. 1999;34:327. 11. Lee GSH, Ratcliffe CI, Ripmeester JA. A Solid State 31 P NMR Study of the Synthesis of Phosphorus Sulfides from PCl3 and H2 S in Microporous Materials. Can. J. Chem. 1998;76: 1660. 12. Nakayama H, Klug DD, Ratcliffe CI, Ripmeester JA. 23 Na MAS NMR Evidence for New Sodium Species in Metal-loaded Zeolites. J. Am. Chem. Soc. 1994;116:9777. 13. Nishikori S, Ratcliffe CI, Ripmeester JA. 113 Cd NMR Studies of Hofmann Type Clathrates and Related Compounds: Evidence for Two Room Temperature Orientational Glasses. Can. J. Chem. 1990;68:2270. 14. Curtis RD, Ratcliffe CI, Ripmeester JA. Structure and Ordering in Metal Cyanide Lattices: the Use of Doubly Labelled Cyanide (13 C-15 N) to Simplify the 13 C MAS NMR Spectrum. J. Chem. Soc. Chem. Commun. 1992:1800. 15. Collins MJ, Ratcliffe CI, Ripmeester JA. NMR Studies of Guest Species in Clathrate Hydrates: Lineshape Anisotropies, Chemical Shifts and the Determination of

Host–Guest Chemistry

17. 18.

19.

20. 21.

22.

23. 24. 25.

26.

27.

28.

29. 30.

31. 32. 33. 34. 35. 36. 37.

38. 39.

40. 41.

42.

43.

44.

45. 46.

47.

hyperquenched glassy clathrate hydrate forming solutions. J. Chem. Phys. 1999;110:6475. Munowitz M, Pines A. Multiple-Quantum Nuclear Magnetic Resonance Spectroscopy. Science 1986;233:525. He J, Ba Y, Ratcliffe CI, Ripmeester JA, Klug DD, Tse JS, Preston KF. Encapsulated Silicon Nanoclusters in Zeolite Y. J. Am. Chem. Soc. 1998;120:10697. Ratcliffe CI. Xenon NMR In: G. A. Webb (Ed). Annual Reports on NMR Spectroscopy, Vol. 36. Academic Press: London, 1998, p 123. Pietrass T. Optically Polarized 129 Xe in Magnetic Resonance Techniques. Magn. Reson. Rev. 2000;17:263. Cherubini A, Bifone A. Hyperpolarised Xenon in Biology. Prog. NMR Spectrosc. 2003;42:1. Ripmeester JA, Ratcliffe CI, Tse JS. The NMR of 129 Xe trapped in Clathrates and some other Solids. J. Chem. Soc. Farad Trans. I. 1988;84:3731. Terskikh VV, Moudrakovskii IL, Breeze SR, Lang S, Ratcliffe CI, Ripmeester JA, Sayari A. A General Correlation for the 129 Xe NMR Chemical Shift—Pore Size Relationship in Porous Silica Based Materials. Langmuir 2002;18: 5653. Ripmeester JA, Ratcliffe CI. The Anisotropic Chemical Shift of 129 Xe NMR in the Molecular Sieve AlPO-11: A Dynamic Averaging Model. J. Phys. Chem. 1995;99:619. Chmelka BF, Raftery D, McCormick AV, Menorval LC, Levine RD, Pines A. Measurement of Xenon Distribution Statistics in Na-A Zeolite Cavities. Phys. Rev. Lett. 1991;66:580. Moudrakovski IL, Ratcliffe CI, Ripmeester JA. 129 Xe NMR Study of Adsorption and Dynamics of Xenon in AgA Zeolite. J. Am. Chem. Soc. 1998;120:3123. Moudrakovski IL, Wang L-Q, Baumann T, Exarhos GJ, Ratcliffe CI, Ripmeester JA. Probing the Geometry and Interconnectivity of Pores in Organic Aerogels Using Hyperpolarized 129 Xe NMR Spectroscopy. J. Am. Chem. Soc. 2004;126:5052. Moudrakovski IL, Nossov A, Lang S, Breeze SR, Ratcliffe CI, Simard B, Santyr G, Ripmeester JA. Continuous Flow NMR with Hyperpolarized Xenon for the Characterization of Materials and Processes. Chem. Mater. 2000;12:1181; Enright GD, Udachin KA, Moudrakovski IL, Ripmeester JA. J. Am. Chem. Soc. 2003;125:9896. Moudrakovski IL, Sanchez AA, Ratcliffe CI, Ripmeester JA. Nucleation and Growth of Hydrates on Ice Surfaces: New Insights from 129 Xe NMR Experiments with Hyperpolarized Xenon. J. Phys. Chem. B. 2003;105:12338. Nossov AV, Soldatov DV, Ripmeester JA. In situ switching of sorbent functionality as monitored with hyperpolarized 129 Xe NMR spectroscopy. J. Am. Chem. Soc. 2001;123:3563. Soldatov DV, Moudrakovski IL Ripmeester JA. Peptides as Microporous Materials. Angew. Chem. Int. Ed. 2004;43:6308. Moudrakovski IL, Soldatov DV, Ripmeester JA. Sears DN, Jameson CJ. Xe NMR lineshapes in channels of peptide molecular crystals. Proc. Natl Acad. Sci. 2004;101: 17924. Moudrakovski IL, Ratcliffe CI, Ripmeester JA. 131 Xe, a New NMR Probe of Void Space in Solids. J. Am. Chem. Soc. 2001;123:2066.

Part I

16.

Cage Occupancy Ratios and Hydration Numbers. J. Phys. Chem. 1990;94:157. Lee F, Gabe E, Tse JS, Ripmeester JA. Crystal structure, CP/MAS 129 Xe and 13 C NMR of local ordering in Dianins compound clathrates. J. Am. Chem. Soc. 1988;110: 6014. He J, Klug DD, Uehara K, Preston KF, Ratcliffe CI, Tse JS. NMR and X-ray Spectroscopy of Sodium-Silicon Clathrates. J. Phys. Chem. B. 2001;105:3475. Sidhu PS, Enright GD, Udachin KA, Ripmeester JA. Structure and polymorphism in a pentamorphic guest-host material: a tris (5-acetyl-3-thienyl) methane (TATM) inclusion compound with 1,3-dichloropropane. Cryst. Growth Des. 2004;4:1240. Soldatov DV, Enright GD, Ripmeester JA. Polymorphism and pseudopolymorphism of the [Ni(4Methylpyridine)4(NCS)2] Werner complex, the compound that led to the concept of “Organic Zeolites”. Cryst. Growth Des. 2004;4:1185. Enright GD, Ratcliffe CI, Ripmeester JA. Crystal Structure and 13 C CP/MAS NMR of the p-Xylene Clathrate of Dianin’s Compound. Mol. Phys. 1999;97:1193. Brouwer EB, Enright GD, Ratcliffe CI, Ripmeester JA. Dynamic Molecular Recognition in Solids: a Synoptic Approach to Structure Determination in t-Butylcalix[4]areneToluene. Supramol. Chem. 1996;7:79. Enright GD, Brouwer EB, Udachin KA, Ratcliffe CI, Ripmeester JA. A re-examination of the low temperature crystal structure of the p-tert-butylcalix[4]arene toluene inclusion compound: Differences in spatial averaging with Cu and Mo K radiation. Acta Crystallogr. B. 2002;58: 1032. Brouwer EB, Ripmeester JA. Structural and Dynamic Properties of Solid Calixarenes. Adv. Supramol. Chem. 1998;5:121. Engelhardt G, Michel D. High-Resolution Solid-State NMR of Silicates and Zeolites. John Wiley & Sons: New York, 1987. Fyfe CA, Grondey H, Feng Y, Kokotailo GT. Naturalabundance two-dimensional silicon-29 MAS NMR investigation of the three-dimensional bonding connectivities in the zeolite catalyst ZSM-5. J. Am. Chem. Soc. 1990;112:8812. Brouwer DH, Kristiansen PE, Fyfe CA, Levitt MH. SymmetryBased 29 Si Dipolar Recoupling Magic Angle Spinning NMR Spectroscopy: A New Method for Investigating ThreeDimensional Structures of Zeolite Frameworks. J. Am. Chem. Soc. 2005;127:542. Hughes E., Jordan J, Gullion T. J. Structural Characterization of the [Cs(p-trt-butylcalix[4]arene -H)(MeCN)] GuestHost System by 13 C-133 Cs REDOR NMR. J. Phys. Chem. B2001;105:5887. Brouwer EB, Gougeon RDM, Hirschinger J, Udachin KA, Harris RK, Ripmeester JA. Intermolecular distance measurements in supramolecular solids: 13 C-19 F REDOR NMR spectroscopy of p-tert-butylcalix[4]arene-fluorobenzene. Phys. Chem. Chem. Phys. 1999;1:4043. Ba Y, Ratcliffe CI, Ripmeester JA. Double Resonance NMR Echo Spectroscopy: New Tools for Materials Characterization. Adv. Mater. 2000;12:603. Tulk CA, Ba Y, Klug DD, McLaurin G, Ripmeester JA. Evidence for phase separation during crystallization of

References 153

154 Part I

Chemistry

Part I

48. Kaiser LG, Meersmann T, Logan JW, Pines A. Visualization of gas flow and diffusion in porous media. Proc. Natl Acad. Sci. USA. 2000;97:2414. 49. Moudrakovski IL, Sanchez A, Ratcliffe CI, Ripmeester JA. Applications of Hyper-Polarized Xenon to Diffusion in Vycor Porous Glass. J. Phys. Chem. B. 2000;104:7306. 50. Moudrakovski IL, Lang S, Ratcliffe CI, Simard B, Santyr G, Ripmeester JA. Chemical Shift Imaging with Continuously Flowing Hyperpolarized Xenon for the Characterization of Materials. J. Magn. Reson. 2000;144:372. 51. Moudrakovski IL, McLaurin GE, Ratcliffe CI, Ripmeester JA. Methane and Carbon Dioxide Hydrate Formation in Water Droplets: Spatially Resolved Measurements from Magnetic Resonance Microimaging. J. Phys. Chem. B. 2004;108:17591. 52. Hoatson GL, Vold RL. 2 H-NMR Spectroscopy of Solids and Liquid Crystals. NMR Basic Principles and Progress. 1994;32:1. 53. Vold RR. Deuterium NMR Studies of Dynamics in Solids and Liquid Crystals. In: R Tycko (Ed). Nuclear Magnetic

54.

55.

56.

57.

Resonance Probes of Molecular Dynamics. Kluwer Academic: Dordrecht, 1994, Chap 2, p 27. Ratcliffe CI, Ripmeester JA, Buchanan GW, Denike JK. A Molecular Merry-Go-Round: Motion of the Large Macrocyclic Molecule 18-Crown-6 in its Solid Complexes Studied by 2 H NMR. J. Am. Chem. Soc. 1992;114:3294. Bull LM, Cheetham AK, Powell BM, Ripmeester JA, Ratcliffe CI. The Interaction of Sorbates with Acid Sites in Zeolite Catalysts: a Powder Neutron Diffraction and 2 H NMR Study of Benzene in H-SAPO-37. J. Am. Chem. Soc. 1995;117: 4328. Brouwer EB, Enright GD, Facey GA, Ratcliffe CI, Ripmeester JA. Weak Intermolecular Interactions: Structure and Dynamics of the Benzene and Pyridine ptert-Butylcalix[4]arene Inclusions. J. Phys. Chem. B. 1999;103:10604. Ba Y, Ratcliffe CI, Ripmeester JA. Kinetics of Water Molecular Reorientation in Ice and THF (Tetrahydrofuran) Clathrate Hydrate from Line Shape Analysis of 17 O SpinEcho NMR Spectra, in preparation.

Part I

Imaging

157

Elke Kossel, Bogdan Buhai, and Rainer Kimmich Universit¨at Ulm, Albert-Einstein-Allee 11, 89069 Ulm, Germany

Introduction

depends on the FOV and the number of data points N :

In this contribution, radio frequency and field gradient pulse sequences for the encoding of position, velocity, and acceleration will be described and explained. The applicability of the techniques will be demonstrated by presenting experimentally obtained velocity and acceleration maps of fluid flow in artificial pore spaces. Porous model objects fabricated on the basis of random percolation clusters are taken as a paradigm for networks in any sort of natural or technical pores or channel complexes. Numerically obtained velocity and acceleration maps will be compared to the experimental data to test the reliability of the methods.

x =

2π γ t Gˆ x N



As long as the frequency shifts caused by the external field gradient are much larger than chemical shielding or susceptibility-induced shifts, the frequency encoded information of the signal can directly be transformed into spatial information. The “field of view” (FOV) is defined as the spatial window, which can unambiguously be sampled by the pulse sequence: −

π γ t Gˆ x

T1ρ ; (iii) a simple combination of (i) and (ii) cases. As expected, in the standard CP experiment, beside the differences in the absolute intensity, very similar temporal evolution of magnetization is observed in each case. However, dramatic differences in the curvature of the TORQUE temporal evolution are visible. This alI lows an unambiguous recognition of the T IS /T1ρ ratio and assures the proper analysis of dynamic CP parameters in terms of structural and/or motional features. A couple of illustrative examples of such analysis in silica

Fig. 2. Magnetization transfer time dependence in standard CP (left) and TORQUE (right) experiments calculated using EquaI = 1.0 ms; (ii) tions (1) and (2) with (i) T IS = 0.5 ms, T1ρ I = 0.5 ms; (iii) equally weighted (i) plus (ii) T IS = 1.0 ms, T1ρ conditions.

gels and layered hydrous sodium silicates are shown below.

Silica Gels Silica gels are highly porous materials which play an important role in numerous applications such as catalysis or chromatographic separation and have been the subject of much NMR investigations for several years [9]. The high-resolution solid state 29 Si CP/MAS NMR spectra of silica gel show three peaks at −91.5, −101, and −110 ppm assigned, respectively, to three Q(2) , Q(3) , and Q(4) types of silicon environments. The results from the CP, and TORQUE experiments on a Fisher S-157 silica gel sample are presented in Figure 3. Assuming a simple monoexponential polarization transfer, the results of a fit of three CP curves using Equation (1) are TUP = [2.3, 2.6, 10.3] ms and TDOWN = [10.3, 13.4, 30] ms, for [Q(2) , Q(3) , Q(4) ] silicons, respectively. It is also observed that the T1ρ (1 H) relaxation curves are

Fig. 3. Time dependence of 29 Si magnetization for Q2 (), Q3 (
), and Q4 (s) sites of silica gel in the (A) indirect T1ρ (1 H) measurements; (B) TORQUE experiment with a total constant time of 20 ms; (c) standard cross-polarization 1 H→29 Si experiment. Insert: 29 Si CP/MAS spectrum of a silica gel sample (Fisher S-157) showing three peaks assigned to three types of silicon environment.

Part I

duration tSL followed by the CP transfer of variable duration tCP , the total time TTORQUE = tCP + tSL being kept constant. The TORQUE signal grows as a function of tCP according to

Silica Gels 199

200 Part I

Chemistry

Part I

identical for protons involved in the CP process and give the unique value T1ρ (1 H) = 10.3 ± 0.5 ms. This clearly indicates that the differences observed in the decreasing part of the CP curves do not correspond to different T1ρ of protons involved in the CP transfer on different sites. The value of 10.3 ms is equal to TDOWN for Q(2) I and to TUP for Q(4) . This suggests that T IS < T1ρ for Q(2) I IS (4) (2) whereas T > T1ρ holds for Q . For Q and Q(4) silicons, these two diametrically opposite situations are visualized immediately by the opposite curvatures of the TORQUE temporal dependence. For Q(3) site, the shape of the TORQUE curve proves the existence of at least I two different Q(3) species, first one having T IS < T1ρ , the I IS second one with T > T1ρ .

Layered Sodium Hydrous Silicates This class of materials, available only in microcrystalline form, has a two-dimensional layered structure, the negative charge of the silicate layer being compensated by sodium ions that are coordinated by the oxygen atoms of the intercalated water molecules. Hydrated sodium silicates are of rapidly growing industrial interest due to their high ion- or proton-exchange properties and new applications in catalysis and in the synthesis of composite mesoporous materials. Magadiite, the most frequently researched, has the idealized formula Na2 Si14 O29 · nH2 O (n = 8–10). 29 Si MAS NMR studies show the presence of Q(3) and Q(4) silicons in its basic layer structure. The experimental build-up curves obtained from standard CP and TORQUE experiments for Q(3) and Na+ are shown in Figure 4.

Fig. 4. Temporal evolution of 29 Si (left) and 23 Na (right) magnetization in the standard CP (top) and TORQUE (bottom) experiments for Q(3) and Na+ sites of magadiite spinning at 3 kHz. Solid lines correspond to the fitted curves.

I Assuming as usual that T IS < T1ρ , the standard CP curves can be fitted according to Equation (1), each of them being described by two pairs of time constants Tup and Tdown for their rising and decreasing parts, respectively. A simple model for which each site is characterized by a single set of Tup and Tdown values was found to be inadequate. However, the observed curvature in the TORQUE temporal dependence makes it immeI diately evident that it is essentially the T IS > T1ρ situation which takes place for Q(3) sites. For Na+ ions, the situation is even more complex, the TORQUE curve exhibits a pronounced S-shape form which means that both the fast and the slow CP regime are equally relevant for this CP dynamics. From the independent measurements I of relaxation in the rotating frames two distinct T1ρ values have been obtained for protons appearing at 3.8 and 15.2 ppm, respectively. It turns out that these two relaxation times are equal neither to Tdown nor to Tup found in the fitting procedure of CP build-up curves when asI I suming T IS < T1ρ . As the T1ρ values reflect two different proton environments existing in magadiite, a realistic model should include both relaxation parameters and assume at least two types of Q(3) as well as sodium sites being differently coupled to hydrogen species. Consequently, the CP and TORQUE build-up curves are each the weighted sums of two different contributions, each one given by Equations (1) and (2) for CP and TORQUE, respectively. A good agreement between experimental and calculated CP and TORQUE temporal dependencies is indeed observed for both species under such assumptions. The fitted T IS and corresponding proportions are given and discussed in structural terms elsewhere [10].

Silicon-29 SSNMR in Materials

1 22°C

I(a.u)

0,8 0,6

backwardsCP 29Si

1H(SiO− )

0,4 0,2 T1ρ(1H)relaxation

Fig. 5. Temporal evolution of 29 Si magnetization for the Q(3) signal during CP and T1ρ (1 H) experiments. All experiments were run with ω1S = ω1H − ωr Hartmann–Hahn condition and at a spinning frequency of 2.5 kHz. The proton rf power was 62.5 kHz and its carrier frequency was exactly that of hydrogenbonded protons at 15.98 ppm, though the proton offset corresponding to the water resonance at 3.8 ppm does not lead to any changes in CP dynamics of the nonoscillating component.

0 0

10000

20000

The ensemble of fitted dynamic parameters brings evidence that the long time decays of magnetization in the standard CP experiments result from the back flow of magnetization to the proton system. Very similar situation occurs in the case of layered sodium hydrated octosilicate [11] with the idealized formula Na8 {Si32 O64 (OH)8 }·32H2 O. The experimental CP and indirect T1ρ (1 H) curves of Q(3) -type resonance peak are shown in Figure 5. In this case, an initial CP rapid growth, with an oscillating behavior is followed by long time decay. However, the independent T1ρ (1 H) measurements show a rate of magnetization decay at least one order higher than the observed long time decay. This clearly indicates once again that the observed long time decrease of CP curves is not due to I T1ρ relaxation. Although a single, isotropic Q(3) resonance signal is observed, it is obvious that the CP curves must be interpreted as composed principally of two types of contributions, an oscillating and a nonoscillating component, the first one cross-polarizing under fast CP regime I (i.e. T IS < T1ρ ), the second evolving under the slow CP I . This implies that, in analogy to regime (i.e. T IS > T1ρ magadiite, octosilicate contains two types of structurally different Q(3) silicones present in hydrogen-bonded –Si– O−-H... O–Si– and –Si–O− − type sites having dramatically different abilities to cross-polarize and being sensitive to different mobilities of neighboring hydrous species. In fact, the direct proton T1ρ measurements show that the oscillating component relaxes with the rate of proton signal representing hydrogen-bonded silanols, while the nonoscillating component is mainly influenced by much more rapidly relaxing water molecules. Consequently, the final decrease of the CP curves in Figure 5 can only result from the backwards 29 Si→1 H flow of the magnetization to the proton reservoir.Unambiguous experimental proof of this is provided by the TORQUE experiment (Figure 6).

tcp(µs)

30000

Indeed, the TORQUE curve exhibits a clear S-shaped character which according to the discussion above makes it immediately evident that some of Q(3) silicons are couI pled to one species of protons under T IS < T1ρ , the others I . This proves to a second type of protons under T IS > T1ρ that hydrogen-bonded –Si–O−H... O–Si– and –Si–O− − type sites evolve, respectively, under fast and slow CP regimes. Finally, it is also worth pointing out that a single set of CSA principal values characterizes both types of Q(3) sites. This means that the 29 Si shielding tensor is mainly related to the Si–O bond character (including the lengths and interbond angle differences between terminal and bridging oxygens) in the SiO4 tetrahedron and is rather insensitive to the presence of protons in the second sphere of coordination.

Probing the Geometry of Strongly Hydrogen-Bonded Silanols Hydrogen bonds are the most important of all directional intermolecular interactions and play a central role in determining molecular conformation and aggregation, as well as the function and dynamics of a great number of systems ranging from inorganic to biological chemistry.In order to understand the physical and chemical properties of layered microporous materials as well as the role of hydrogen bonds in the aggregation and ordering of silicate layers, the correlation of such contacts with the spectroscopic response is highly desired. In sodium hydrous silicates, the nature of strong hydrogen bonding having ˚ and present at room an O... O distance of less than 2.60 A or higher temperature remains indeed the subject of considerable controversy. Both inter- and intralayer hydrogen bonding involving the silanol or water protons have been proposed. As the intercalation of polar molecules in

Part I

coherentCP 1H 29Si (HBQ3)

Probing the Geometry of Strongly Hydrogen-Bonded Silanols 201

202 Part I

Chemistry

Part I

Fig. 6. Experimental time dependence of 29 Si magnetization for the Q(3) signal during TORQUE and T1ρ (1 H) experiments. Solid lines correspond to the fitted curves with two components (i) and (ii) having following time constants: for the TORQUE experiment: (i) T IS = I = 2.54 ms; (ii) T SI = 20.0 0.7 ms, T1ρ I ms, T1ρ = 0.8 ms; for the T1ρ experiment: I = 0.8 ms; (ii) T I = 2.54 ms. (i) T1ρ 1ρ

layered materials can be dramatically controlled by the existence of interlayer hydrogen bonds, the appropriate recognition of the extent and the nature of hydrogen bonding present in these materials is of prime importance. To get this local geometric information, one can determine the internuclear Si... H distances and the orientation of the 29 Si chemical shift tensor in the hydrogen-bonded Q(3) type units by exploiting a simple experiment based on the CP inversion of the 29 Si spin magnetization used as a modulation of the slow magic-angle spinning chemical shift

spectrum [12]. The experiment starts with the classical CP procedure followed by a period during which the contact between protons and silicons is maintained but the phase of proton spin-locking irradiation is inverted. As shown in Figure 7, this leads to non-uniform dipolar modulation of the 29 Si CSA spinning sidebands recorded under high power proton decoupling. Such an effect gives evidence for largely coherent magnetization transfer within the silanol groups having a pronounced inhomogeneous character of the dipolar

δll δl b)

δll

Fig. 7. Dipolar modulated (t1 = 400 µs), natural abundance 29 Si NMR spectrum of slowly magic-angle spinning (νr = 357 Hz) octosilicate (left bottom). Asterisks indicate the spinning sidebands of Q(4) sites. Fitted spectrum of Q(3) sites along with its individual components (right bottom). Dipolar modulated subspectrum (a) represents as indicated the hydrogen-bonded Q(3) sites, the subspectrum (b) comes from Si–O− Q(3) type sites cross-polarizing from the water molecules.

−60

δ1

−80 −100 −120 −140 ppm

a)

−60

−80 −100 −120 −140 ppm

Silicon-29 SSNMR in Materials

Conclusions The CP measurements are usually analyzed assuming that the CP time T IS of magnetization transfer from the abundant I spins to the rare S spins is shorter than the relaxation time T1ρ in the rotating frame of the I spins (fast CP regime). Here, it was shown that the reverse situation (T IS >> T1ρI , slow CP regime) frequently occurs for the 1 H →29 Si transfer in commonly encountered inorganic materials. This fact must be clearly recognized to avoid a false structural image of investigated materials. The efficiency of the TORQUE experiment in visualizing the real CP regime or its possible mixed character has been underlined. The proper exploitation of the proton–silicon polarization transfer spin dynamics in fast and slow magicangle spinning experiments permits a deeper insight into structural and motional features of silicon-containing

materials. The analysis of the dipolar modulated 29 Si CSA spectrum yields straightforward geometric information on the hydrogen-bonded silanols, including the orientation of 29 Si CSA tensor in the molecular frame. The CP methods employed to obtain such information take advantage of a weakly dipolar coupled proton–proton network, largely disconnected from the heteronuclear dipolar couplings within the silanol groups. This leads to significant truncation of weak dipolar couplings from neighboring protons by the largely dominant flip-flop coupling term of the heteronuclear spin pairs. This in turn makes it possible to exploit the coherent magnetization exchange without applying homonuclear decoupling which itself eliminates any uncertainty about the heteronuclear scaling factor inherently connected with homonuclear decoupling. The presented strategy may be useful to obtain structural information in the related layered alkali metal silicates, silica gels, calcium silicate hydrates as well as in other classes of microporous materials.

References 1. Fyfe CA, Solid State NMR for Chemists. C. F. C. Press: Canada, 1983. 2. Engelhardt G, Michel D. High-Resolution Solid-State NMR of Silicates and Zeolites. John Wiley & Sons: Berlin, 1987. 3. Eckert H. Prog. Nucl. Magn. Reson. Spectrosc. 1992;24:159. 4. Colombet P, Grimmer AR, Zanni H, Sozzani P (Eds). Nuclear Magnetic Resonance Spectroscopy of Cement-Based Materials. Springer-Verlag: Berlin, 1998. 5. McArthur D, Hahn EL, Waldstaet RE. Phys. Rev. 1969;188:609. 6. Mehring M. High-Resolution NMR in Solids. NMR Basic Principles and Progress. Springer-Verlag: Berlin, 1983. 7. Klur I, Jacquinot JF, Brunet F, Charpentier T, Virlet J, Schneider C, Tekely P. J. Phys. Chem. B 2000;104:10162. 8. Tekely P, G´erardy V, Palmas P, Canet D, Retournard A. Solid State NMR 1995;4:361. 9. Maciel GE. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance. John Wiley & Sons: Chichester, UK, 1996, p 4370. 10. Gardiennet C, Tekely P. J. Phys. Chem. B 2002;106, 8928. 11. Gardiennet C, Marica F, Fyfe CA, Tekely P. J. Chem. Phys. 2005;122:054705. 12. Gardiennet C, Marica F, Assfeld X, Tekely P. Angew. Chem. Int. Ed. 2004;43:3565.

Part I

system, the observed difference in the dipolar oscillation frequency of different spinning sidebands resulting from variation of the orientation-dependent dipolar coupling. More interesting in the context of this work, the dipolar modulated spinning sidebands contain all the desired information on the hereronuclear distance as well as the magnitude and orientation of the principal elements of the chemical shielding tensor in the molecular frame. In order to reproduce the observed dipolar modulated envelope of Q(3) spinning sidebands in Figure 6, the presence of two different components representing two types of Q(3) sites has to be assumed. Indeed, as discussed above, although a single isotropic Q(3) resonance signal is observed, two types of Q(3) tetrahedra, hydrogen-bonded silanols and Si–O− type sites need to be distinguished by their different abilities to cross-polarize. As can be seen in Figure 7, the calculated spectrum is in excellent agreement with the experimental envelope and phase features of the Q(3) family of spinning sidebands. The simulations show that dipolar modulated envelope of spinning sidebands is very sensitive to small changes of rSi...H distances and of their polar coordinates in the chemical shielding principal axis frame [11]. The results clearly support the intralayer character of strongly hydrogen-bonded silanol groups in a bridging albeit not symmetric position between neighboring tetrahedra.

References 203

205

Feng Deng, Jun Yang, and Chaohui Ye State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Institute of Physics and Mathematics, the Chinese Academy of Science, Wuhan 430071, P. R. China

Solid state NMR spectroscopy has become a powerful tool for investigation of the solid surface of various heterogeneous catalysts [1–4], such as zeolite, metal oxide, and solid heteropoly acid, which are widely used in petrochemical industry. Identification and characterization of the active centers, reaction intermediates, and products are essential for understanding reaction mechanisms occurring on the surface of heterogeneous catalysts. Compared to X-ray diffraction (XRD) which is determined by long-range orderings and periodicities, solid state NMR is more sensitive to local orderings and geometries, thus providing a more detailed description of the local structure, especially for powder samples. Multinuclear magic angle spinning (MAS) NMR, especially 1 H MAS NMR, probe molecule techniques, double-resonance techniques as well as two dimension correlation techniques have been employed to reveal the detailed structure of the active sites on heterogeneous catalysts. In addition, in situ MAS NMR technique [3,4] has been developed as an indispensable tool for investigation of the heterogeneous catalytic reaction mechanisms. Although many surface properties of heterogeneous catalysts can be investigated by solid state NMR spectroscopy, two main topics will be discussed in this section: surface acidity and catalytic reaction of heterogeneous catalysts.

Surface Acidity of Heterogeneous Catalysts The surface acidity is described by the following three properties: (1) the type of acid sites (Br¨onsted or Lewis site); (2) the acid strength, which can be defined, for a Br¨onsted site, as the ability of the surface hydroxyl groups to protonate an adsorbed molecule; (3) the concentration of acid sites accessible to probe molecules. 1 H MAS NMR can resolve various hydroxyl groups that may acts as proton donators (Br¨onsted acid sites) on the surface of heterogeneous catalysts. In the case of zeolites [1,2], 1 H MAS NMR signals consist of non-acidic SiOH groups at chemical shifts of δ = 1.2–2.2 ppm, extraframework AlOH groups at δ ≈ 3 ppm, acidic bridging SiOHAl groups at δ = 3.6–5.2 ppm, and residual ammoGraham A. Webb (ed.), Modern Magnetic Resonance, 205–211.
C 2008 Springer.

nium ions at δ = 6.5–7.0 ppm. For an ammonium-free zeolite, adsorption of a small amount of water molecules on Lewis acid site also gives rise to a 1 H signal at ca. 7.0 ppm. Hydrogen bonds of the surface OH groups with neighboring oxygen atoms or probe molecules will leads to a downfield shift of 1–20 ppm. For example, in a layered sodium disilicate material [5], isolated SiOH groups gives rise to 1 H resonances at 0–3 ppm, while inter-layer hydrogen-bonded SiOH groups correspond to a 1 H signal at 14 ppm, and strongly hydrogen-bonded silanols with the proton bonded to non-bridging oxygen at the same silicon atom shift the resonance position to 18.7 ppm. The advantage of 1 H MAS NMR over IR spectroscopy lies in the quantitative measurement of signal intensities, allowing an accurate determination of the OH concentrations [1]. Besides, some double resonance methods, such as 1 H/27 Al Transfer of Populations in Double Resonance (TRAPDOR) NMR technique [6], can correlate the various hydroxyl groups with the neighboring Al spins, while a spin echo pulse sequence is applied to proton and 27 Al is irradiated simultaneously during one of the echo period in the experiment. Under the 27 Al irradiation, the signals of protons that are strongly coupled with aluminum spins will be significantly suppressed while those that are not coupled with aluminum atoms remain unaffected. Therefore, the heteronuclear dipolar interactions between the two spins and the 1 H/27 Al internuclear distances can thus be extracted. As an example, Figure 1 shows the 1 H/27 Al TRAPDOR NMR of the ultrastable Y, the 2.2 ppm signal, which is due to non-acidic SiOH groups at the framework defects, is almost unaffected under 27 Al irradiation, while the signals at 4.3 and 5.2 ppm due to two types of the bridging OH groups and the signal at 3.0 ppm arising extra-framework AlOH groups that are all close to the Al atoms are significantly reduced [7]. Using various probe molecules, the surface acidity of heterogeneous catalysts can be well characterized. 15 Nenriched pyridine [8] and trimethylphosphine (TMP) [9] are two extensively used probe molecules for discriminating Br¨onsted and Lewis acid sites and quantitatively determining their concentrations. Both of the probe molecules give rise to large 15 N and 31 P chemical shift ranges of 100

Part I

Solid State NMR Characterization of Solid Surface of Heterogeneous Catalysts

206 Part I

Chemistry

Part I

4.3 5.2 6.8

3.0

2.2

a 6.8 2.2

b

c

20

0

10 ppm

−10

−20

Fig. 1. 1 H/27 Al TRAPDOR NMR spectra of ultrastable HY. (a) without 27 Al irraiation, (b) with 27 Al irradiation. (c) the different spectrum of (a) and (b) [7].

and 60 ppm, respectively. One important advantage of the TMP over pyridine is the relatively high NMR sensitivity of 31 P, which is very useful in the cases where the concentration of the acid sites is very low. When Br¨onsted or Lewis acid sites are present in zeolites, protonated adduct, TMPH+ , or Lewis-bound TMP complex are confirmed. The TMPH+ is characterized by a 31 P resonance at ca. −4 ppm and a JP−H coupling of approximately 500 Hz for zeolites, while the Lewis-bound TMP complexes give rise to resonances in the shift range from −32 to −58 ppm, and the physisorbed or weakly bound TMP has a resonance at ca. –60 ppm. 27 Al/31 P and 27 Al/1 H rotational echo double-resonance (REDOR) NMR methods [10] have been applied to measure Al-P and Al-HB (HB is the Br¨onsted acidic proton) distances in zeolite

˚ HY for the acid site-TMP complex of 3.95 and 2.8–3.1 A, ˚ was obtained by respectively. A P–HB distance of 1.40 A fitting the spinning sidebands in the 1 H MAS spectrum. As shown in Figure 2, combining the NMR results with ab initio calculations provides a more detailed description of the exact structure of TMP-Br¨onsted acid site complex formed in the zeolite. Various probe molecules are used for determination of the acid strength of surface OH groups. Deuterated pyridine [11] is one of probe molecules for this purpose. The formation of a hydrogen bond between pyridine and non-acidic silanol group (SiOH) shifts the 1 H MAS NMR signal position from 2 to ca. 10 ppm. In the case of acidic OH groups (Br¨onsted acid sites), the adsorption of pyridine results in 1 H NMR signals at chemical shifts in the

NMR Characterization of Heterogeneous Catalysts

Catalytic Reaction on the Surface of Heterogeneous Catalysts 207

Me H

P 1.4 Å

C

H 138 - 118° (118°)

H

3.95 (3.7) Å H

2.8 - 3.1 Å (2.9 Å)

O

O

O

Al

Fig. 2. Comparison of the NMR experimental and calculated distances. Calculated distances and angles obtained for the TMPH+ -Z− 2 cluster are given in parentheses. The dashed line indicates that an interaction between the two spins was observed experimentally, but the distance was not quantified [10].

range 12–19 ppm. The down-field signals result from a proton transfer to the probe molecule, forming pyridine ions. However, no quantitative correlation has been established between the acid strengths and the down-field shift of the 1 H signal. A more precise measurement of the acid strength of Br¨onsted and Lewis sites can be achieved with the aid of the 13 C chemical shift of the carbonyl atom of adsorbed 2-13 C-acetone [12,13]. The different degrees of interaction between the carbonyl oxygen of adsorbed acetone and the acid site result in different 13 C downfield shifts of the carbonyl carbon. By comparing the chemical shifts of 2-13 C-acetone adsorbed on various solid catalysts with the resonance position of the molecule in 100% H2 SO4 solution, Haw et al. [13] proposed that the solid acid strengths scale with the 13 C NMR isotopic chemical shift of adsorbed 2-13 C-acetone (Table 1). According to the scale, the acid strength of the bridging hydroxyl groups in zeolite HZSM-5 corresponds to that of 80% H2 SO4 solution.

Acetone SAPO-34 CF3 CH(OH)OCF3 HZSM-5 MgCl2 ZnCl2 AlBr3 100%H2 SO4 AlCl3 SbF5

208 217 221 223 221 230 243 244 245 250

heterogeneously catalyzed reactions [3, 4, 13–15]. The detection of the change of both active sites on the catalysts and species (such as reactants, products, and intermediates) adsorbed on the surface of catalyst in the process of the reaction can provide more direct information about what happens on the catalyst surface than that obtained by using off-line techniques, such as gas chromatograph (GC) and mass spectroscopy (MS). The large chemical shift of 13 C MAS spectra (more than 300 ppm) enables differentiation of various organic species by their characteristic resonances. For example, in the study of methanol to gasoline (MTG) reaction [16], Klinowski et al. had (i) identified 29 different adsorbed organic species and monitored their fate during the reaction; (ii) directly observed various kinds of shape selectivity in zeolite ZSM-5; (iii) discriminated mobile species from attached species. These results will assist in the design of shape-selective catalysts and provide a better understanding of the catalytic reaction. In situ 13 C NMR technique was also employed to study methane dehydroaromatization on Mo/HZSM-5 catalyst. Not only the products, such as benzene, ethane, and ethylene, but also the active phase Mo2 C were directly observed by 13 C NMR spectroscopy (Figure 3). The NMR results support the following reaction mechanism [17]: (1) During induction period MoO3 + CH4 → Mo2 C + CO + CO2 + H2 O + H2 (2) Formation of C2 (Mo2 C species as active center) CH4 → C2 H4 + C2 H6

Catalytic Reaction on the Surface of Heterogeneous Catalysts In situ solid state NMR spectroscopy has been demonstrated to be a very powerful method to study the

(3) Production of benzene (Br¨onsted acid sites as active center) C2 H4 → C6 H6 + H2

Part I

Table 1: 13 C MAS NMR isotopic chemical shift (in ppm) of carbonyl carbon of 2-13 C-acetone on (or in) different solid (or liquid) acids [12]

Me

208 Part I

Chemistry

Part I

1 pda

Mo2C powder

Mo2C C6H6

C2H4 C2H6

CP Mo2C

CO

973K for 30min

CO2

1pda

Machanism for Opening Trap Door and Driving Seal into Rotor

973K for 30min CH4

1pda 1pda

873K for 1h

RT

300

200

100

0

−100

chemical shift (ppm) Fig. 3. 13 C MAS NMR spectra of methane (13 C, 99%) reaction on 6Mo/HZSM-5 at different temperatures, which were acquired at room temperature using one pulse with 1 H decoupling (1pda) or 1 H–13 C cross polarization (CP). For comparison, 13 C MAS NMR spectrum of molybdenum carbide powder was also shown. Asterisks denote spinning sidebands. The signal at 112 ppm is due to background of spinning module [17].

Since NMR is a quantitative method, the concentration of the adsorbed species on the surface of catalysts can thus be directly obtained, which is very helpful for investigation of the reaction mechanism. The sensitivity of natural abundance 13 C surface species is usually not enough for 13 C MAS NMR detection. Although 1 H–13 C cross polarization experiment can be used to enhance the sensitivity of 13 C MAS spectra, 13 C isotope-enriched reactants are usually required for the in situ NMR study. In some cases, selectively labeled reactants are very much useful to identify the catalytic reaction pathway by monitoring the fate of the labeled atoms at specific sites in the process of the reaction [18]. In the earlier studies, in situ MAS NMR experiments were usually carried out under batch condition. The simple and very commonly used method employs a particularly symmetrical glass ampoule [19], which fits well into MAS rotor in order to achieve a stable high magic angle spinning rate at about 3–4 kHz. For the sample preparation of in situ MAS NMR measurements, the catalyst is packed into the glass ampoule and activated under a vacuum line at a specified temperature. A known amount of reactant in gas or liquid state is introduced onto the catalyst by freezing with liquid N2 , and then the ampoule is carefully sealed by flame. The reaction is allowed to occur in an oven at a specified temperature for a period of time and quenched with liquid N2 , and then the sealed ampoule is transferred to MAS rotor for NMR observation. Another method for sample preparation

35/25 Ball & Socket Joint Thermocouple Insert

trap door

Catalyst Bed

rotor cap MAS rotor

Fig. 4. Schematic drawing of a CAVERN device for the sample preparation of in situ MAS NMR measurement [13].

employs a specially designed device named CAVERN (CAVERN: cryogenic adsorption vessel enabling rotor nestling, Figure 4)[13], which allows the sample preparation in the MAS rotor with a gas-tight sealed cap. This device can be connected to a vacuum line. Activation of catalysts, adsorption of 13 C-enriched reactants, transfer of the loaded catalysts into the MAS rotor, and sealing of the MAS rotor are all carried out in this device,

Catalytic Reaction on the Surface of Heterogeneous Catalysts 209

NMR Characterization of Heterogeneous Catalysts

product flow

support

exhaust tube rotor cap

rotor

injection tube

catalyst bed

Fig. 5. Schematic drawing of a MAS NMR rotor reactor with an injection equipment applied for in situ MAS NMR experiments under flow condition [14].

Part I

reactant flow

210 Part I

Chemistry

Part I

Multi-position valve with sample loops Injector valve A

Injector valve B

Reactor Mass flow controller

Quench vent

GC-MS

Helium gas

Nitrogen gas for cooling (77 K) Fig. 6. Schematic drawing of pulse-quench reactor coupled with GC and MS for in situ MAS NMR experiments under flow condition [24].

preventing the samples from the exposure of atmosphere. The sealed MAS rotor can be transferred into NMR probe. The reaction is allowed to take place at a specified temperature in NMR magnet for a period of time and then the temperature is allowed to return to room temperature for the in situ 13 C MAS measurement. This apparatus is suitable for the study of reaction that begins to occur at low temperature. For example, since ethylene is very active on HZSM-5 zeolite at the room temperature, the CAVERN device can realize the adsorption and transfer of the sample at the liquid N2 temperature, and the 13 C MAS NMR observation of the reaction from 77 to ca. 600 K in NMR magnet. The Haw’s group has done a large number of in situ 13 C MAS NMR studies with this device [13]. It is well known that heterogeneously catalyzed reactions are usually operated under the flow condition and the reaction products under the batch condition are different from those under the flow condition. Several groups attempt to develop in situ MAS NMR techniques for flow condition measurement [20–25]. Hunger et al. [14, 25] reported a device for real continuous-flow MAS measurement (Figure 5). In their device, the activated catalyst bed is pressed as a hollow cylinder in the MAS rotor and an injection tube is inserted into the hollow catalyst via a hole on the cap of rotor, which allows a continuous-flow introduction of reactants into the catalyst during the NMR experiment. The reaction products leave

the MAS NMR rotor continuously through an annular gap in the rotor cap. It is possible to couple the in situ MAS device directly with an on-line GC [4]. Another in situ MAS NMR technique introduced by the Haw’s group for the flow condition measurement includes a quenchreactor device coupled with GC and MS (Figure 6) [24]. A significant feature of the apparatus lies in that catalytic reactions can be quenched with cryogenically cooled nitrogen within a few hundred milliseconds. The catalyst loaded with reactants and products is then transferred to the NMR rotor in a glove box at room temperature prior to 13 C MAS NMR measurement. Figure 7 shows the pulse-quench 13 C NMR spectra [26] of ethylene on HZSM-5 zeolite at 623 K with the reaction time varying from 0.5 to 16 s. The most prominent peaks in the spectrum obtained for the 0.5 s reaction are all almost due to cyclopentenyl cation. As the catalyst ages for 2–4 s, signals from the carbenium ion decrease with a commensurate increase in the signals due to toluene. With further aging in the flow reactor, signals due to toluene and other organic species diminish, and after 16 s only a modest amount of the carbenium ion remains in the catalyst bed. A semilog fit of the decrease of the cyclopentenyl cation over time yielded an approximate half-life time of 6 s at 623 K. In the last two decades, various catalytic reactions have been studied by in situ MAS NMR spectroscopy.

NMR Characterization of Heterogeneous Catalysts

References 211

References

Fig. 7. 13 C MAS NMR spectra of ethylene-13 C2 adsorbed on zeolite HZSM-5 at 623 K for various reaction time [26].

The formation of alkoxy species, such as methoxy groups (δ = 58 ppm), ethoxy groups (δ = 68 ppm), and isopropoxy groups (δ = 87 ppm), have been observed by 13 C NAS NMR following the adsorptions of the olefines and alcohols onto acidic H-ZSM-5 and H–Y zeolites. These species are confirmed to act as reactive components that play an important role in the course of the reaction [4]. Methanol to hydrocabon conversion has been extensively investigated by in situ MAS NMR under either batch or flow condition, and the experimental results shed insight on the mechanism of reaction [26,27]. In the reaction of ethylene, methanol, acetone on acidic zeolite, alkyl-substituted, cyclic structure carbenium ion have been observed under batch /flow condition by in situ 13 C MAS NMR [13, 26], and it was proposed that these carbenium ions and related neutral species might function as

1. Pfeifer H, Ernst H. Annu. Rep. NMR spectrosc. 1993;28: 91. 2. Hunger M. Catal. Rev. -Sci. Eng., 1997;39:345. 3. Haw JF (Ed.). In-situ Spectroscopy in Heterogeneous Catalysis, Wiley/VCH: Weinheim, 2002. 4. Hunger M, Weitkamp J. Angew. Chem. Int. Ed. 2001;40: 2954. 5. Ai X, Deng F, Dong J, Chen L, Ye C. J. Phys. Chem. B. 2002;106:9237. 6. Grey CP, Vega AJ. J. Am. Chem. Soc. 1995;117:8232. 7. Deng F, Yue Y, C Ye C. Solid State NMR. 1998;10:151. 8. Haw JF, Chuang S, Hawkins BL, Maciel GE, J. Am. Chem. Soc. 105, 7206 (1983). 9. Lunsford JH, Rothwell WP, Shen W, J. Am. Chem. Soc. 1985;107:1540. 10. Kao H-M, Liu H, Jiang J-C, Lin S, Grey CP. J. Phys. Chem. Bio. 2000;104:4923. 11. Hunger M. Solid State NMR 1996;6:1. 12. Biaglow AI, Gorte RJ, White D. J. Catal. 1994;148:779; Barich DH, Nicholas JB, Xu T, Haw JF. J. Am. Chem. Soc. 1998;120:12342. 13. Haw JF, Nicholas JB, Xu T, Beck LW, Ferguson DB. Acc. Chem. Res. 1996;29:259. 14. Hunger M. Catal. Today 2004;97:3. 15. Derouane EG, He H, Derouane SB, Lambert D, Ivanova I. J. Mol. Catal. A. 2000;158:5. 16. Anderson MW, Klinowski J, Nature 1989;339:200; Anderson MW, Klinowski J, J. Am. Chem. Soc. 1990;112:10. 17. Yang J, Ma D, Deng F, Luo Q, Zhang M, Bao X, Ye C. Chem. Commun. 2002;24:3046. 18. Ivanova II, Brunel D, Nagy JB, Derouane EG. J. Mol. Catal. A. 1995;95:243. 19. Carpenter TA, Klinowski J, Tennakoon DTB, Smith CJ, Edwards DC. J. Magn. Reson. 1986;68:561. 20. Haddix GW, Reimer JA, Bell AT. J. Catal. 1987;106:111. 21. Ernst H, Freude T, Mildner T. Chem. Phys. Lett. 1994;229: 291. 22. Ferguson DB, Haw JF. Anal. Chem. 1995;67:3342. 23. Haake M, Pines A, Reimer JA, Seydoux R. J. Am. Chem. Soc. 1997;119:11712. 24. Haw JF, Goguen PW, Xu T, Skloss TW, Song W, Wang Z, Angew. Chem. 1998;110:993; Angew. Chem. Int. Ed. 1998;37:948. 25. Hunger M, Horvath T. J. Chem. Soc. Chem. Commun. 1995;14:1423. 26. Haw JF, Nicholas JB, Song WG, Deng F, Wang ZK, Xu T, Heneghan CS. J. Am. Chem. Soc. 2000;122:4763. 27. Seiler M, Schenk U, Hunger M. Catal. Lett. 1999;62:139.

Part I

reaction centers (hydrocarbon pool species) for the conversion of methanol to hydrocarbon.

Part I

Isotope Labeling

215

Shin-ya Ohki1 and Masatsune Kainosho2 1 Japan

Advanced Institute of Science and Technology (JAIST), Ishikawa 923-1292, Japan; and 2 CREST-JST and Department of Chemistry, Graduate School of Science, Tokyo Metropolitan University, Tokyo 192-0397, Japan

Introduction The existence of stable isotopes, such as 2 H, 13 C, and 15 N, is a blessing from nature for protein NMR spectroscopy, because protons, carbons, and nitrogens are the major components of proteins. Thus, protein NMR has deeply benefited from these stable isotopes. Since their natural abundance is very low, selective enrichment and/or depletion of these nuclei for incorporation into proteins have/has been desired. These skills are called stable isotope labeling, which is an old technique that is undergoing renewal in protein NMR spectroscopy. Various stable-isotope-labeling methods are illustrated in Figure 1. Labeling methods can be generally classified into positive and negative. The former methods use NMR active nuclei, such as 13 C and 15 N, meaning that this labeling enables the monitoring of only the labeled sites in molecules [1]. The most famous example of the latter category is deuteration. The replacement of 1 H by 2 H can erase undesired peaks in the 1 H NMR spectra [2,3]. Thus, in other words, positive and negative mean “visible” and “invisible,” respectively, for NMR spectroscopy. Such labeling methods were already proposed even in earlier one-dimensional (1D) NMR studies, before the strategy for the three-dimensional (3D) structure determination of proteins was established. In another context, the labeling methods can be classified as selective and uniform. Examples of the former are the introduction of 13 C and/or 15 N into certain site(s) in protein samples. The latter labeling strategy, “uniform labeling,” is the preparation of protein samples in which all of the carbon and nitrogen atoms are labeled with stable isotopes, 13 C and 15 N, respectively. Then, all of the carbon, nitrogen, and proton atoms in the proteins become visible in NMR experiments. This was first reported with the adoption of new pulse sequences for the separation of peaks into multiple dimensions [4,5]. Nowadays, uniform labeling and multidimensional NMR measurements are standard for the structure determination of proteins smaller than ∼20 kDa.

Graham A. Webb (ed.), Modern Magnetic Resonance, 215–222.
C 2008 Springer.

The demands for NMR have become more complicated lately: structure determination of large molecules, quick structure determinations of proteins with moderate molecular weights, determination of protein structures at higher resolution with high accuracy and precision, identification of ligand-binding sites on the surface of proteins, detailed studies of molecular dynamics, structural transitions, etc. To satisfy these requests, numerous stable isotope techniques have been proposed as an extension of the methods mentioned above. For all of the cases, the key is how 2 H, 13 C, and 15 N are placed in the protein samples, and their concepts can simply be characterized using combinations of the four words: positive, negative, selective, and uniform. In this chapter, the stable isotope techniques developed in the past several years will be reviewed, and a novel labeling approach of the post-genomic era will be described.

Positive Labeling (Use of 13 C and 15 N) About 20 years ago, the strategy to determine the 3D structures of proteins by solution NMR was established [6]. With parallel progress in homonuclear 1 H twodimensional (2D) experimental techniques and computational algorithms, the structure determination of small polypeptides in solution became routine. For proteins larger than ∼10 kDa, however, the cross peaks are crowded in the 2D 1 H NMR spectra, so their structures are extremely difficult to determine. To investigate larger molecules, a new labeling technique was proposed in 1990 [4,5]. In that method, all of the carbon and nitrogen atoms in the protein sample are labeled with NMR active stable isotopes by a method called uniform labeling or 13 C, 15 N-double labeling. Then, all of the proton, carbon, and nitrogen atoms in the uniformly labeled protein become detectable with NMR. The employment of 1 H, 13 C, and 15 N enables the use of heteronuclear one-bond or two-bond spin–spin couplings in pulse sequences, and thus their through-bond correlations can be monitored. These experimental data

Part I

Recent Developments in Stable-Isotope-Aided Methods for Protein NMR Spectroscopy

216 Part I

Chemistry

Part I

a

C N H

H

C N H

H C N C

H

C H N

N

+

C N H

H C N C

H

C H N

C H

N H

H

N

+

H C N C

H

C H N

C H

N H

H

N

+.....

N

+.....

N

+.....

C H

N H

b 13

C N

H

H

13

C

H

C H N

N

+

H

13

C

H

C H N

N

+

H

H C N C

H

C H N

C H

N H

C N

H

H C N

C H

N H

C N

H

H C N

C H

N H

c C H

15

N H

H C N C

H

C H N

H 2

2

H N

C

N H

13

13

H

13

C

C H 15N N H

13

H

2

C H

N

+

H

13 13

13

C H 15N 15

N 2

+

N H

13

H

H C N

C

2

H

13

15

C H

2

H N

N

+

13

H

13

+.....

N

+.....

H

C 15N

C

C H 15N 15

2

N

C 15N H

H 13

H

C

N H

C 15N

C

2H

C

C 15N H

H

C H

C N H

H

C

N H

13

15

2

H N

H

C H N N H

H C N

H

C 15N

C

15

2

+

+

N H

H C N

2H

C

C 15N H

H 13

N

N

15

C

C H

C N H

H

C H

H

C H N N H

H C N C

C

+

N H

H C N

2H

C N

d

N

15

C

C H

N H

e

C H

N H

13

H

15

C H

Fig. 1. Cartoons showing protein molecules with various stable-isotope-labeling patterns. Each yellow circle indicates a protein molecule. The small circles labeled H, C, and N indicate atoms. NMR observable and unobservable atoms are colored red and cyan, respectively. (a) Unlabeled, (b) selectively positive-labeled with 13 C, (c) selectively positive-labeled with 15 N, (d) selectively negativelabeled with 2 H, (e) uniformly 13 C, 15 N-labeled, (f) random-fractionally uniformly deuterated, and uniformly 13 C, 15 N-labeled, (g) site-specifically protonated, but otherwise uniformly 2 H, 13 C, 15 N-labeled, (h) protein–protein complex (uniformly labeled monomer and unlabeled one), and (i) segmental labeling (N-terminal half of the protein labeled with 13 C and 15 N ). (See also Plate 25 on page XXVII in the Color Plate Section.)

Positive Labeling (Use of 13 C and 15 N) 217

Developments in Stable-Isotope-Aided Methods

13 2

C 15N

H

H

13 13

13

2

13

H

2

13

2

C H N 15

15

13

N 2H

15

+

N

2

13 13

15

N

+

2

13 13

H 13

C

H

13

C 15

13

N

C

H

H 15

15

N H

13

i 13

13

C

N

N

H

13

N

2

C 15

15

13

N H

C

H

15

N N

H 13

C

H

2

N

N 2H

13

C 2H

13

H

2

H

2

13

C H 15N N 2H

+.....

N

H

C 15N

C

15

15

C H

C 15N

H 13

+

2

C H 15N 15

15

C 15N

13

2

H

15

N

+.....

C 2H

C H C

N

H

H

H

13

13

H

C

N

C

15

+

13

C

H

N

H H

H 15

N

N

H

13

C 15N

13

C

C

C

N 2H

H

N

H

H N-term.

15

13

C H N 15

N

2

15

C 2H

C 15N H

H

15

C 15N H

H

C 2H

C

h 13

H

C 15N

H

13 2

C 15N

13

N H

13

H

13

C 2H 15N

H 2

H

C

15

C 15N

C

C 15N H

H

C 2H

C 15N

H

13

H

13

N H

H

C 15N

C H 15N

g

13

13

C

15

2

2

Part I

f

N

H

N

C

C

N

N

H

H H

H C

H

N

C-term.

C H

C

Fig. 1. (Continued) (See also Plate 25 on page XXVIII in the Color Plate Section.)

can provide the chemical shifts for all of the 1 H, 13 C, and 15 N in proteins unambiguously. Various J coupling constants are also obtainable with such samples, and they provide angle information for the protein backbone and side chains. Furthermore, the novel pulse sequences using 13 C or 15 N can separate crowded 1 H 2D NOESY spectra into several planes with the chemical shifts of the hetero nuclei attached to 1 H, thus enabling the identification of numerous NOE peaks for structure calculation. The strategy has expanded the possibility of NMR structure determination for proteins smaller than ∼20 kDa.

Uniform labeling has also been adapted to protein complexes. Several labeling approaches have been proposed to study protein-protein complexes and symmetrical oligomers. The most popular method is mixing labeled and unlabeled components, which yields a complex in which one subunit is labeled with 13 C and 15 N, and the other is unlabeled. The binding surface is identified by intermolecular NOEs using isotope-filtered NMR experiments [7]. In the early 1990s, experimental methods in molecular biology were widely adopted by NMR laboratories, and uniform labeling instantly become a standard technique.

218 Part I

Chemistry

Part I

Various vectors and cells are commercially available today, and thus it has become routine to prepare expression systems for proteins from a wide variety of sources. Commonly, bacteria, Phicia pastoris, and insect or mammalian cells are used for NMR sample preparation. The uniform labeling is generally achieved by the addition of labeled chemicals into the growth medium. For the 13 Cprecursors, 13 C-labeled glucose, 13 C-acetic acid, or 13 Cmethanol is frequently used, with 15 N-labeled NH4 Cl or NH4 SO4 as the 15 N-precursors. In an alternative method, a 13 C, 15 N-labeled amino acid mixture can be included in the medium. Amino acid type selective labeling to simplify NMR spectra has been used as a powerful tool to study local conformations. This method can be applied to large proteins, and thus it is still useful in current research. For this case, 13 C and 15 N are employed as labels, and an expression or chemical synthesis method is needed to prepare the protein samples. When we use Escherichia coli to express the NMR samples, it is easy to label the Ile, Leu, Val, Phe, and Tyr residues. However, for several amino acids including Glu, Gln, Asp, and Asn, the selective labeling is not achieved with standard E. coli strain such as BL21, because of isotope scrambling and dilution during expression. To overcome these problems, a new set of genetically engineered E. coli strains with lesions in the biosynthetic pathways of certain amino acids has been developed [8]. The use of these extraordinary E. coli systems is one of the solutions to achieve amino acid type selective labeling. Another choice for sample preparation is to use cell-free protein synthesis. A cell-free system has the potential for robust isotope labeling without isotope scrambling and dilution [9]. The system can also be used for toxic proteins and membrane proteins [10], which are difficult to express in bacteria, and it can be extended to incorporate non-natural amino acids containing spin or fluorescent labels [11]. For large proteins, an attractive labeling method, segmental labeling, has been proposed. This method is used to prepare a protein in which a part is labeled, and is based on peptide splicing reactions with inteins. Inteins are inserted amino acid sequences that splice themselves out after translation [12]. In the first demonstration, the isotopiclabeled N-terminal (or C-terminal) half of a protein was ligated to the unlabeled C-terminal (or N-terminal) half [13]. As a modification of the original method, labeling at the central region of a protein is also possible [14].

Negative Labeling (Use of 2 H) The application of multidimensional NMR experiments for larger proteins (> 20 kDa) often yields poor NMR spectra, due to two factors. One is severe line broadening

and the other is overlapping of numerous peaks. The former is caused by the shorter spin–spin relaxation time (T2 ) and the latter is due to the number of protons. In the past decade, great progress has been made to overcome these problems. The breakthrough was achieved with new pulse sequences based on the transverse relaxationoptimized spectroscopy (TROSY) technique and a combination of 2 H negative labeling and 13 C, 15 N positive labeling. The TROSY principle was originally found as a modification of 1 H–15 N HSQC (heteronuclear single quantum coherence spectroscopy) on biomolecules. The critical feature of TROSY is that heteronuclear one-bond 1 H–X (X = 13 C, 15 N) splitting should not be decoupled. Then, each peak is split into four in the 2Dplane. The line widths of the doublet components in both dimensions are different, because they have different relaxation mechanisms. In the original TROSY experiment, only the sharpest component, which shows the longest T2 , was observed by a phase cycling scheme that cancels the three broader peaks of the multiplet [15,16]. The higher field magnets (800– 900 MHz) that are currently available are suitable for obtaining the maximum TROSY effect, and thus the NH detection period in many triple resonance pulse sequences has been rewritten, based on the TROSY technique, for larger proteins [17]. Negative labeling, using of 2 H, has been employed to reduce the number of peaks in 1 H NMR spectra. This was recognized as a powerful technique even in earlier 1D NMR [2,3]. Moreover, the use of 2 H yields another benefit, in that it strengthens and sharpens the signals in NMR spectra. Since 2 H has a significantly lower gyromagnetic ratio as compared with 1 H (γ [2 H]/γ [1 H] = 0.15), the use of 2 H can contribute toward eliminating the 1 H– 1 H dipolar and 1 H–X (X = 13 C, 15 N) heteronuclear spin relaxation pathways, resulting in a longer T2 . The first experiments using 2 H labeling combined with triple resonance multidimensional NMR experiments were reported in 1993 [18]. The effect of 2 H in 13 C, 15 N-labeled proteins is very successful, and thus many applications using this labeling scheme have been published [19–21]. A recent study has shown that the triple labeling method coupled with TROSY-based NMR analysis can work even for a 900 kDa protein complex [22]. In earlier application of triple-resonance multidimensional experiments with 2 H decoupling, the uniformly random fractional incorporation of 2 H into the nonexchangeable 1 H sites in proteins was employed. In general, the degree of the 2 H labeling level affects the quality of the NMR spectra, so a higher level of random uniform 2 H labeling yields sharper signals and thus is better for main chain assignments. However, the absence of 1 H at non-exchangeable sites is disadvantageous to side chain analysis, especially for the observation of 1 H–1 H NOEs for structure determination. Thus, 50–90% uniformly

Developments in Stable-Isotope-Aided Methods

Negative Labeling (Use of 2 H) 219

Part I

Fig. 2. A typical structure of each stereo-arrayed-isotope-labeled (SAIL) amino acid. There are various other isotopomers, which may be useful for NMR applications.

220 Part I

Chemistry

Part I Fig. 3. Two-dimensional HCCH-TOCSY of an 18.2 kDa protein, EPPIb (S. Ohki, T. Hayano, T. Terauchi, M. Kainosho, unpublished data). The sample contains with uniformly 13 C, 15 N-labeled Gln, [ul-13 C,15 N]-Gln, (a) and SAIL-Gln (b), respectively. The intraresidue connectivity for each residue is shown by a dotted line with the residue number.

random fractional 2 H labeling is frequently employed for analysis, but this has considerable problems related to isotopomers. For example, each methyl group of Ala, Leu, Ile, Val, and Met contains isotopomers, i.e. CH3 , CH2 D, CHD2 , and CD3 . Three of the four isotopomers give NMR signals, and they appear at slightly different chemical shifts due to the isotope shift. Then, the methyl region signals become crowded, which hinders extensive analysis. However, the methyl group is interesting as a probe for protein dynamics. In some cases, 13 CH2 D is monitored, but the topic of protein dynamics is beyond the scope of this review. Although an optimized sample preparation method and pulse sequences to observe one isotopomer in the sample solution have been reported [23], the signal intensity is lower than expected, because the actual number of detectable molecules is much

less than the total sample concentration. If one obtains fine filtered NMR spectra using such pulse techniques, then the number of 1 H to be analyzed increases with the molecular weight, and thus the amount of effort is never reduced. To improve the spectral complexity, alternative labeling methods have been proposed by using 2 H, 13 C, and 15 N. The method is amino acid type selective labeling in deuterated proteins. In other words, the method is selective protonation. In an earlier report of this strategy, the labeled protein sample was expressed in minimal medium containing 95% D2 O, 2 H-labeled glucose, 15 NH4 SO4 , and 1 H/13 C/15 N-{Ile, Leu, Val} amino acids [24]. The samples gave very clear 1 H–13 C HSQC and NOESY for these hydrophobic residues. The aliphatic– aliphatic NOEs combined with the 1 HN–1 HN NOEs can

Developments in Stable-Isotope-Aided Methods

Acknowledgment 221

Part I

Fig. 4. Simulation of structure determination for proteins labeled with SAIL amino acids (S. Ohki, M. Kainosho, unpublished data). (a) Ribbon model of cystatin A (PDB code; 1CYV) determined by NMR experiments. Structures (b) and (c) were simulated based on the coordinate. (b) Structures using all NOEs theoretically observed, and (c) structures using NOEs expected for the SAILed protein. (See also Plate 26 on page XXVIII in the Color Plate Section.)

be subjected to structure calculations; however, only the global fold is available. Although several 13 C precursors, such as 13 C pyruvate [23,25,26] or [2-13 C]glycerol [27], were examined for the selective labeling, further structural information, such as residual dipolar couplings (RDC), was needed to determine the high-resolution NMR structures of large proteins [28]. Recently, a novel labeling method termed stereoarrayed isotope labeling (SAIL) has been developed [29]. In this labeling method, the 2 H labeling sites and the occupancy are controlled at an extremely high level. The arrayed deuteration sites in proteins are designed to suppress redundant structural information [30]. The SAIL amino acids, shown in Figure 2, are chemically and enzymatically synthesized. Then, these amino acid compounds are incorporated into the cell-free synthesis system for sample protein preparation. The SAILed proteins have ∼50% protons as compared to fully protonated proteins, and the SAILed protein molecules in the sample solution represent only one isotopomer. Thus, the NMR spectra are simplified, with very narrow signals. Figure 3 indicates an example of the NMR data. In both samples, only Gln residues were labeled with uniformly [ul-13 C, 15 N]Gln (Figure 3a) or SAIL-Gln (Figure 3b). Obviously, the SAIL method gives much better NMR spectrum. Furthermore, a simulation of the structure calculation indicates that obtainable NOEs must be sufficient to solve the structure at high resolution and accuracy (Figure 4). The application of SAIL method promises to relieve the limitation of molecular weight for NMR analyses and to contribute to high-throughput structure determination in the postgenomic era.

Concluding Remarks The recent progress in stable-isotope-labeling strategies has provided opportunities for NMR studies of a wide range of proteins and their complexes. The advance of methodologies for sample preparation, including conventional expression systems, cell-free systems, and chemical synthesis, will continuously propose various labeling strategies. In concert with improved instruments, novel experimental techniques, and faster computing, the stableisotope-labeling will become increasingly significant for studying larger biological systems by NMR in the future.

Acknowledgment The SAIL method that was briefly introduced in this chapter has been developed in the CREST project supported by JST.

References 1. Jardetzky O, Roberts GCK. NMR in Molecular Biology. Academic Press: New York, 1981. 2. Butter TB. Proton magnetic resonance fully deuterated except for 1 H-leucine side chains. Science. 1968;161:795– 98. 3. Markley JL, Putter I, Jardetzky O. High-resolution nuclear magnetic resonance spectra of selectively deuterated staphylococcal nuclease. Science. 1968;161:1249–51. 4. Kay LE, Ikura M, Tschudin R, Bax A. Three-dimensional triple-resonance NMR spectroscopy of isotopically enriched proteins. J. Magn. Reson. 1990;89:496–514.

222 Part I

Chemistry

Part I

5. Ikura M, Kay LE, Bax A. A novel approach for sequential assignment of 1 H, 13 C, and 15 N spectra of proteins: heteronuclear triple-resonance three-dimensional NMR spectroscopy: application to calmodulin. Biochemistry. 1990;29:4659–67. 6. W¨uthrich K. NMR of Proteins and Nucleic Acids. John Wiley & Sons: New York, 1986. 7. Folkers PJM, Folmer RHA, Konings RNH, Hilbers CW. Overcoming the ambiguity problem encountered in the analysis of nuclear Overhauser magnetic resonance spectra of symmetric dimer proteins. J. Am. Chem. Soc. 1993;115:3798– 99. 8. Waugh DS. Genetic tools for selective labeling of proteins with α−15 N-labeled amino acids. J. Biomol. NMR. 1996;8:184–192. 9. Torizawa T, Shimizu M, Taoka M, Miyano H, Kainosho M. Efficient production of isotopically labeled proteins by cell-free synthesis: A practical protocol. J. Biomol. NMR. 2004;30:311–25; and references cited therein. 10. Berrier C, Park KH, Abes S, Bibonne A, Betton JM, Ghazi A. Cell-free synthesis of a functional ion channnel in the absence of a membrane and in the presence of detergent. Biochemistry. 2004;43:12585–91. 11. Rothschild KJ, Gite S. t-RNA-mediated protein engineering. Curr. Opin. Biotechnol. 1999;10:64–70. 12. Perler FB. Protein splicing of inteins and hedgehog autoproteolysis: structure, function, and evolution. Cell. 1998;92:1–4. 13. Yamazaki T, Otomo T, Oda N, Kyogoku Y, Uegaki K, Ito N, Ishino Y, Nakamura H. Segmental isotope labeling for protein NMR using peptide splicing. J. Am. Chem. Soc. 1998;120:5591–92. 14. Otomo T, Ito N, Kyogoku Y, Yamazaki T. NMR observation of selected segments in a larger protein: central-segment isotope labeling through intein-mediated ligation. Biochemistry. 1999;38:16040–44. 15. Pervushin K, Riek R, Wider G, W¨uthrich K. Attenuated T2 relaxation by mutual cancellation of dipole–dipole coupling and chemical shift anisotropy indicates an avenue to NMR structures of large biological macromolecules in solution. Proc. Natl. Acad. Sci. U.S.A. 1997;94:12366–71. 16. Pervushin K, Riek R, Wider G W¨uthrich K. Transverse relaxation-optimized spectroscopy (TROSY) for NMR studies of aromatic spin systems in 13 C-labeled proteins. J. Am. Chem. Soc. 1998;120:6394–400. 17. Salzmann M, Pervusin K, Wider G, Senn H, W¨uthrich K. TROSY in triple-resonance experiments: new perspectives for sequential NMR assignment of large proteins. Proc. Natl. Acad. Sci. U.S.A. 1998;95:13585–90.

18. Grzesiek S, Anglister J, Ren H, Bax A. 13 C line narrowing by 2 H decoupling in 2 H/13 C/15 N-enriched proteins— application to triple-resonance 4D J-connectivity of sequential amides. J. Am. Chem. Soc. 1993;115:4369–70. 19. Yamazaki T, Lee W, Arrowsmith CH, Muhandiram DR, Kay LE. A suite of triple-resonance NMR experiments for backbone assignment of 15 N, 13 C, 2 H-labeled proteins with highsensitivity. J. Am. Chem. Soc. 1994;116:11655–66. 20. Shan X, Gardner KH, Muhandiram DR, Rao NS, Arrowsmith CH, Kay LE. Assignment of 15 N, 13 Cα, 13 Cβ, and HN resonances in an 15 N, 13 C, 2 H labeled 64 kDa trp repressor– operator complex using triple-resonance NMR spectroscopy and 2 H-decoupling. J. Am. Chem. Soc. 1996;118;6570–79. 21. Garrett DS, Seok YJ, Liao DI, Peterkofsky A, Gronenborn AM, Clore GM. Solution structure of the 30 kDa N-terminal domain of enzyme I of the Escherichia coli phosphoenolpyruvate: sugar phosphotransferase system by multidimensional NMR. Biochemistry. 1997;36;2517–30. 22. Flaux J, Bertelsen EB, Horwich AL, W¨uthrich K. NMR analysis of a 900K GroEL–GroES complex. Nature. 2002;418:207– 11. 23. Ishima R, Louis JM, Torchia DA. Optimized labeling of 13 CHD methyl isotopomers in perdeuterated proteins: po2 tential advantages for 13 C relaxation studies of methyl dynamics of larger proteins. J. Biomol. NMR. 2001;21:167– 71. 24. Metzler WJ, Wittekind M, Goldfarb V, Mueller L, Farmer II BT. Incorporation of 1 H/13 C/15 N-{Ile, Leu, Val} into a perdeuterated, 15 N-labeled protein: potential in structure determination of large proteins by NMR. J. Am. Chem. Soc. 1996;118:6800–1. 25. Rosen MK, Gardner KH, Willis RC, Parris WE, Pawson T, Kay LE. Selective methyl group protonation of perdeuterated proteins. J. Mol. Biol. 1996;263:627–36. 26. Lee AL, Urbauer JL, Wand AJ. Improved labeling strategy for 13 C relaxation measurements of methyl groups in proteins. J. Biomol. NMR. 1997;9:437–40. 27. LeMaster DM, Kushlan DM. Dynamical mapping of E. coli thioredoxin via 13 C NMR relaxation analysis. J. Am. Chem. Soc. 1995;118:9255–64. 28. Delaglio F, Kontaxis G, Bax A. Protein structure determination using molecular fragment replacement and NMR dipolar couplings. J. Am. Chem. Soc. 2000;122:2142–43. 29. Kainosho M. The SAIL method for protein NMR spectroscopy. XXIst ICMRBS, Hyderabad, India, 2005. p 46. 30. Terauchi T, Ohki, S, Kainosho M. Developing a new approach for high-throughput, high-accuracy NMR structural analyses of genomic proteins. Protein nucleic acid enzyme. 1998;47:1045–51.

223

Yoshiki Yamaguchi1,2 and Koichi Kato1,2,3,4 1 Nagoya

City University, Nagoya, Japan; 2 CREST/JST, Saitama, Japan; 3 Institute for Molecular Science, Okazaki, Japan; and 4 Genomic Sciences Center, RIKEN Yokohama Institute, Yokohama, Japan

Introduction Recent advances in structural biology have made possible the high-throughput structural determination of proteins, which is reflected in the very rapid growth of Protein Data Bank content. In structural proteomics, recombinant proteins used for structural determination by NMR spectroscopy and X-ray crystallography are conventionally produced by use of bacterial expression systems or recently by cell-free protein expression systems and therefore do not possess carbohydrate moieties. However, many of the proteins in the living systems are covalently linked to carbohydrate moieties, which mediate molecular recognition involved in cell–cell communication, contribute to solubility and structural integrity of proteins, and determine the fates of glycoproteins in cells, i.e. folding, transport, and degradation via interactions with a variety of intra-cellular lectins [1,2]. Although the biological importance of glycans expressed on proteins has been widely recognized, little is known about their specific roles from the structural aspect. This deficiency in our knowledge is largely due to the lack of an appropriate methodology to deal with glycoproteins as targets of structural biology. The carbohydrate moieties of glycoproteins generally exhibit microheterogeneities and possess a significant degree of freedom in internal motion, which hampers crystallization or interpretation of electron density [3,4]. NMR spectroscopy can potentially provide us with information on structure and dynamics of glycoproteins in solution. However, there are few reports of structural determination of glycoproteins by NMR spectroscopy [5–7]. The desirable methods for NMR structural biology of glycoproteins and carbohydrate–protein interactions are 1) Production of a large amount of isotopically labeled glycoprotein by appropriate expression systems. 2) Determination of a covalent structure of target glycoprotein including carbohydrate moieties.

Graham A. Webb (ed.), Modern Magnetic Resonance, 223–229.
C 2008 Springer.

3) Preparation of a large amount of isotopically labeled oligosaccharides that can be used as ligands.

Three-Dimensional HPLC Mapping Prior to NMR analyses of glycoproteins, it is essential to obtain information concerning the covalent structures of their carbohydrate moieties. Takahashi and coworkers have established a method to identify asparagine-linked oligosaccharides rapidly on inspection of HPLC elution profiles of their pyridylamino (PA)-derivatives [8]. On the basis of the combination of the retention time data on the three kinds of HPLC columns, i.e. anion exchange, ODS, and amide–silica columns, the elution map of the 500 different PA-oligosaccharides has been established. Based on this three-dimensional HPLC map combined with mass spectrometric data, we have recently made GALAXY (http://www.glycoanalysis.info/), a web application that greatly facilitates NMR structural biology of glycoproteins [9]. The HPLC method is also useful for isolation of PA-oligosaccharides discriminating isomeric structures [10], which can be used as ligand for NMR analyses of carbohydrate–protein interactions.

Stable Isotope Labeling of Glycoproteins The authors have been developing a systematic method for isotope labeling of glycoproteins for NMR analyses using immunoglobulin G (IgG) as model system [11,12]. The Fc portion of IgG possesses one conserved glycosylation site at Asn-297 in each of the two heavy chains, where biantennary complex-type oligosaccharides (Figure 1) are expressed. These carbohydrate chains are essential for the binding to effector molecules such as Fcγ rceptors [13,14]. The carbohydrate chains exhibit microheterogeneities resulting from the presence or absence of the non-reducing terminal galactose (Gal), core fucose (Fuc), and bisecting N -acetylglucosamine (GlcNAc)

Part I

Structural Glycobiology by Stable-isotope-assisted NMR Spectroscopy

224 Part I

Chemistry

Part I

Fig. 1. The structures of the glycans attached to the Fc portion of IgG.

residues depending upon physiological and pathological states [15,16]. For example, agalacosylation of serum IgG is associated with a variety of diseases such as rheumatoid arthritis [17]. Figure 2 illustrates the scheme of strategy for stable isotope labeling of the Fc glycans. For incorporation of the labeled precursors into the glycoprotein, we use two alternative methods. One is metabolic labeling via biosynthesis pathway of mammalian cells. The other is in vitro

labeling by use of enzymatic glycosylation onto isolated glycoproteins.

In vitro Labeling of Sugar Chains Enzymatic attachment of isotopically labeled sugar onto the carbohydrate moiety is a conventional method of selective isotope labeling of the non-reducing terminal sugar

Fig. 2. The scheme of stable isotope labeling of the Fc glycoprotein and the glycopeptide derived therefrom.

NMR Structural Glycobiology

Stable Isotope Labeling of Glycoproteins 225

Part I

Fig. .3 1 H–13 C HSQC of galactosyl Fc in which the Gal residues are fully (A) or partially (B) labeled with 13 C by using UDP[1-13 C]Gal and (C) 2D HCCH-COSY spectrum of galactosyl Fc in which the Gal residues are fully labeled with 13 C by using UDP-[u-13 C6 ]Gal.

residues such as Gal and sialic acids. Figure 2 shows the scheme of in vitro labeling of the terminal Gal residues [12,18]. [1-13 C]Gal can be converted to UDP-[1-13 C]Gal through the successive enzymatic reactions using galactokinase and galactose-1-phosphate uridyl transferase and then attached by using galactosyltransferase onto the nonreducing ends of the Fc carbohydrate chains, which are degalactosylated in advance. The Fc preparation gave two HSQC peaks originating from the anomeric groups of the terminal Gal residues, i.e. Gal-6 and Gal-6 (Figure 3A). Under the mild reaction condition using small amount of unlabeled UDP-Gal, galactosylation occurs fully and partially at the mannose (Man) α1-6 and Manα1-3 branches, respectively. The unoccupied Manα1-3 branches of this Fc preparation can be fully galacosylated by use of enough amount of UDP-[13 C]Gal, giving rise to Fc labeled with

C exclusively at Gal-6 . HSQC spectrum of this Fc preparation gave a single anomeric peak originating from Gal-6 (Figure 3B) and therefore led us to assign the peaks shown in Figure 3A to each of the Gal residues. Starting from the anomeric peaks thus assigned, intra-residue scalar connectivities were identified by HCCH-COSY spectrum of the isotopically labeled Fc prepared by using UDP-[1-13 C6 ]Gal (Figure 3C). It should be noted that the peaks originating from Gal6 gave much narrower peak than those from Gal-6 , indicating that Gal-6 is much more mobile than Gal-6 . This is consistent with the crystallographic data of Fc, which shows that Gal-6 makes contacts with an inner surface of the CH 2 domain, while the Manα1-3 branch protrude to a space between the CH 2 domains [19]. Hence, 13 C resonances can be useful probes to provide us with information 13

226 Part I

Chemistry

Part I Fig. .4 1 H–13 C ct-HSQC spectrum observed for the glycopeptide metabolically labeled with [u-13 C6 ]Glc (A) and 1 H–13 C HSQC spectra of agalactosyl Fc metabolically labeled with [u-13 C6 ]Glc (B), [u-13 C6 , 2 H7 ]Glc (C), and [1-13 C]GlcN (D). F, fucose; GN, N -acetylglucosamine; M, mannose.

on dynamics of the carbohydrate moieties of glycoproteins at atomic resolution. Similar technique can be applied for isotope labeling of the terminal sialic acid residue on galactosylated ovalbumin [20].

Metabolic Labeling of Sugar Chains The major drawback of in vitro labeling method is that it can only be applied to the NMR analyses of nonreducing terminal residues of glycans in glycoproteins. By contrast, metabolic labeling can be applicable for the observation of NMR signals originating from all the sugar residues. For expression of isotopically labeled glycoproteins subjected to NMR analyses, mammalian [5,21], plant [22], yeast [23,24], insect cells [25], and cellular slime mold [26] have so far been available. The authors have developed protocols of metabolic labeling

of IgG glycoproteins by cultivating hybridoma cells in a serum-free medium that contains isotopically labeled amino acids and/or sugars [12,27]. Since glucose can be metabolically converted to all of the sugar residues in biosynthetic pathways, isotope labeling using [u-13 C6 ]glucose (Glc) as a metabolic precursor results in uniform 13 C labeling of their carbohydrate moieties of glycoproteins. Figures 4A and B compare a part of an HSQC spectrum of uniformly 13 C-labeled Fc (agalactosyl form) thus prepared with that of a glycopeptide derived from it by V8 protease digestion. In this spectral region, the peaks originating from the CH groups of carbohydrate moieties are observed. The HSQC peaks originating from the glycopeptide can be unambiguously assigned since their 1 H and 13 C chemical shifts are in agreement with those of isolated oligosaccharides [28]. On the other hand, significant differences were observed in the chemical shifts of most of the peaks between Fc and

NMR Structural Glycobiology

Carbohydrate–Protein Interactions Stable isotope labeling of sugar chains is also useful for NMR analyses of carbohydrate–protein interactions.

Chemically or enzymatically 13 C-labeled oligosaccharides have been used for NMR analyses of interactions of oligosaccharides with their cognate proteins [7,29– 31]. The authors used glycopeptides derived from isotopically labeled Fc glycoproteins as ligands. The Fc fragment metabolically labeled with [u-13 C6 ]Glc was digested by V8 protease and trypsin and the isolated glycopeptide was subjected to galactosidase and hexosaminidase treatments for trimming of its carbohydrate moiety giving rise to Manα1-6(Manα1-3)Manβ1-4GlcNAcβ14(Fucα1-6)GlcNAcβ1-peptide. Figure 5 shows HMQC-NOESY spectrum of the 13 Clabeled glycopeptide in association with the sugar-binding domain (SBD) of Fbs1, a substrate-binding component of sugar-recognizing ubiquitin ligase SCFFbs1 [32,33]. Intermolecular NOE connectivities between the anomeric proton of the innermost GlcNAc residue and the aromatic ring of Tyr-279 as well as inter-residue NOE connectivity within the carbohydrate moiety were observed in the spectrum. These data provide us with information of conformation and binding mode of the glycan in association with the SBD of Fbs1.

Concluding Remarks Conformational analyses of carbohydrate moieties covalently attached to or non-covalently interacting with a protein are very important for obtaining unique knowledge that has never been possible with liberated oligosaccharides and provide information regarding the structural basis of functions of glycans and of rational design of sugar mimics. By stable isotope labeling of glycans, it becomes feasible to elucidate the conformation and dynamics of glycans attached to proteins based on NMR parameters, i.e. chemical shifts, NOEs, or relaxation rates. At higher magnetic field, it becomes possible to observe residual dipolar couplings of isotopically labeled glycoprotein molecules weakly oriented in the presence of ordering media [34,35]. Structural glycobiology is an unexplored field beyond structural genomics. Stable-isotope-assisted NMR spectroscopy will open up a new avenue in this field and greatly contribute to decoding the glycocodes.

Acknowledgments We wish to acknowledge Dr. Yoji Arata and Dr. Ichio Shimada and colleagues (The University of Tokyo) for the project of NMR analyses of IgG. We acknowledge our fruitful collaboration with the laboratory of Dr. Keiji Tanaka and Dr. Yukiko Yoshida (Tokyo Metropolitan Institute of Medical Science) on Fbs1. This work was supported in part by CREST/JST, by Research on Health

Part I

the isolated glycopeptide, indicating that the glycans are surrounded by a different environment when they are built in Fc. Especially, two anomeric peaks exhibit pronounced low frequency 1 H chemical shift values (1:50 molar ratio) the helix axis changes its tilt angle from about 95◦ to approximately 125◦ , with the C-terminus pointing toward the bilayer interior. This tilted “T-state” represents a novel feature of antimicrobial peptides, which is distinct from a membrane inserted I-state. At intermediate concentration, PGLa is in exchange between the S- and T-states in the timescale of the NMR experiment. In both states the peptide molecules undergo fast rotation around the membrane normal in liquid crystalline bilayers, hence large peptide aggregates do not form. Very likely the obliquely tilted T-state represents an antiparallel dimmer of PGLa that is formed in the membrane at increasing concentration.

Conclusions It is clearly demonstrated that oriented bilayer media can give useful information on structure, orientation, and dynamics of biologically active peptides which are strongly bound to the membranes. Spontaneously oriented bilayer system such as MOVS is shown to be an excellent media to study membrane associated peptides because they show excellent magnetic alignments of molecules bound to the membranes if one carefully prepare the sample. Since the magnetic field will be going higher by the development of higher field magnet, the magnetic alignment

NMR of Oriented Bilayer Systems

References 1. Cornall BA, Separovic F, Baldassi AT, Smith R. Biophys. J. 1988;53:67. 2. Ketchem RR, Hu W, Cross TA. Science. 1993;261:1457. 3. Marassi FM, Ramamoothy A, Opella SJ. Proc. Natl. Acad. Sci. U.S.A. 1997;94:8551. 4. Qin X, Miran PA, Pidgeon C. Biochim. Biophys. Acta. 1993;1147:59. 5. Seelig F, Borle F, Cross TA. Biochim Biophys. Acta 1985;814:195. 6. Scholz F, Helfrich W. Biophys. J. 1984;45:589. 7. Spyer JB, Spipada PK, Das Gupta SK, Shipley GG, Griffin RG. Biophys. J. 1987;51:687. 8. Brumm T, M¨ops C, Dolainsky C, Br¨uckner S, Bayerl TM. Biophys. J. 1992;61:1018. 9. Dempsey CE, Watts A. Biochemistry. 1987;26:5803. 10. Dempsey CE, Sternberg B. Biochim. Biophys. Acta. 1991;1061:175. 11. Pott T, Dufourc EJ. Biophys. J. 1995;68:965. 12. Naito A, Nagao T, Norisada K, Mizuno T, Tuzi S, Saitˆo H. Biophys. J. 2000;78:2405. 13. Sanders CR, Prestegard JH. Biophys. J. 1990;58:447. 14. Sanders CR, Schwonek JP. Biochemistry. 1992;31:8898. 15. Sanders CA, Landis GC. Biochemistry. 1995;34:4030. 16. Boroske E, Helfrich W. Biophys. J. 1978;24:863. 17. Volt RR, Prosser RS. J. Magn. Reson. 1996;B113:267.

18. Bolze T, Fujisara T, Nagao T, Norisada K, Saitˆo H, Naito A. Chem. Phys. Lett. 2000;329:215. 19. Bax A, Tjandra N. J. Biomol. NMR. 1997;10:289. 20. Prosser RS, Hunt SA, DiNatale JA, Volt RR. J. Am. Chem. Soc. 1996;118:269. 21. Prosser RS, Hwang JS, Vold RR. Biophys. J. 1998;74:2405. 22. Naito A, Nagao T, Obata M, Sindo Y, Okamoto M, Yokoyama S, Tuzi S, Saitˆo H. Biochim. Biophys. Acta. 2002;1558: 34. 23. Toraya S, Nagao T, Obata M, Izumi S, Tuzi S, Saito H, Naito A. Biophys. J. 2005;89:3214. 24. Marassi FM. Concepts Magn. Reson. 2002;14:212. 25. Glaubitz C, Watts A. J. Magn. Reson. 1998;130:305. 26. Sizun C, Bechinger B. J. Am. Chem. Soc. 2002;124:1146. 27. Glaubitz C, Burnett IJ, Gr¨obner G, Mason AJ, Watts A. J. Am. Chem. Soc. 1999;121:5787. 28. Middleton DA, Ahmed A, Glaubitz C, Watts A. J. Magn. Reson. 2000;147:366. 29. Mason AJ, Grage SL, Straus SK, Glaubitz C, Watts A. Biophys. J. 2004;86:1610. 30. Toraya S, Nishimura K, Naito A. Biophys. J. 2004;87:3323. 31. Wu CH, Ramamoothy A, Opella SJ. J. Magn. Reson. 1994;A109:270. 32. Marassi FM, Opella SJ. J. Magn. Reson. 2000;144:150. 33. Marassi FM, Ma C, Gesell JJ, Opella SJ. J. Magn. Reson. 2000;144:156. 34. Wang J, Denny J, Tian C, Kim S, Mo Y, Kovacs F, Song Z, Nishimura K, Gan Z, Fu R, Quine JR, Cross TA. J. Magn. Reson. 2000;144:162. 35. Marassi FM, Opella SJ. Protein Sci. 2003;12:403. 36. Zeri AC, Mesleh MF, Nevzorov AA, Opella SJ. Proc. Natl. Acad. Sci. 2003;100:6458. 37. Uezono T, Toraya S, Obata M, Nishimura K, Tuzi S, Saitˆo H, Naito A. J. Mol. Struct. 2005;749:13. 38. Kimura S, Naito A, Tuzi S, Saitˆo H. Biopolymers. 2002;63:122. 39. Glaser RW, Sachse C, Durr UH, Wadhwani P, Afonin S, Strandberg E, Ulrich A. Biophys. J. 2005;88:3392.

Part I

of molecule bound to membrane will be much promising in the future. Mechanically aligned bilayer system will be a very good media since one can adjust the orientation to get more information on the membrane bound molecules. The combination of glass aligned sample with the MAS provide high-resolution signals in the oriented systems to provide more information of the structure of membrane associated biologically active molecules.

References 247

249

Michael F. Brown1,2 , Silvia Lope-Piedrafita3 , Gary V. Martinez1 , and Horia I. Petrache4 1 Department

of Chemistry, University of Arizona, Tucson, AZ 85721, USA; of Physics, University of Arizona, Tucson, AZ 85721, USA; 3 Department of Radiology, University of Arizona, Tucson, AZ 85724, USA; and 4 Laboratory of Physical and Structural Biology, National Institutes of Health, NICHD, Bethesda, MD 20892, USA 2 Department

Solid-state NMR spectroscopy is widely applicable to the investigation of non-crystalline or amorphous materials, e.g. polymers, glasses, protein precipitates, and membrane proteins. Rather than being mainly an alternative to X-ray crystallography, solid-state NMR is virtually unique among current analytical and spectroscopic methodologies in that it provides both structural and dynamical information at an atomically resolved level. In solid-state NMR, the structural information is obtained from the static or motionally averaged coupling tensors due to dipolar, chemical shift, or quadrupolar interactions [1,2]. Corresponding dynamical information is acquired from the tensor fluctuations, which depend on the meansquared amplitudes and rates of the motions and affect the NMR lineshapes and relaxation times. For these reasons, solid-state NMR is finding increasing applicability in the chemistry of materials, structural biology, and genomics research, and this trend can be expected to continue well into the future. One area of solid-state NMR spectroscopy that has proven fruitful with regard to the investigation of membranes is 2 H NMR spectroscopy. Previous detailed reviews of 2 H NMR as applied to membrane lipids are available [3–7]. Recently, 2 H NMR has been used to investigate raft-like lipid mixtures implicated in membrane signaling functions [8–10], and moreover 2 H NMR studies of membrane proteins [11–15] and DNA fibers [16] have also been conducted. The present chapter is focused on the liquid-crystalline state of membrane lipids as investigated from combined 2 H NMR lineshape and relaxation studies. A related aspect entails the correspondence of 2 H NMR studies to molecular dynamics simulations [17]. The salient aspects of 2 H NMR are that it enables both membrane lipids and membrane proteins to be studied by substitution of 2 H for 1 H; the structural data are highly complementary to X-ray [18–21] and neutron diffraction studies [22,23], and virtually unique information regarding the functional dynamics of membrane constituents can be acquired.

Graham A. Webb (ed.), Modern Magnetic Resonance, 249–260.
C 2008 Springer.

Equilibrium and Dynamical Properties of Membrane Lipids are Studied by Solid-State Deuterium NMR Phospholipid bilayers are classified as smectic A lyotropic liquid crystals, and an illustration of the liquid-crystalline lamellar phase is shown in Figure 1. The hydrophobic effect leads to a sequestering of the nonpolar acyl chains within the bilayer interior, whereas the polar head groups interact with water at the membrane surface. The nanostructure of a membrane lipid aggregate is the result of a delicate balance of forces acting at the level of the polar head groups and hydrocarbon regions of the membrane [24–27]. Representative glycerophospholipids are depicted in Figure 2, in which the polar head groups differ in their size, capacity for hydrogen bonding, and charge, whereas the nonpolar acyl chains vary in their length and degree of unsaturation. The phase equilibria of phosphatidylcholines in excess water include three regions as temperature increases, a lamellar gel phase with tiled chains (L β ), an intermediate ripple phase (Pβ ), and a lamellar liquid-crystalline phase (L α) [25]. Other types of phospholipid nanostructures are possible, for instance unsaturated phosphatidylethanolamines form the reverse hexagonal (HII ) phase, and cubic phases can also be present [25]. Moreover, when cholesterol is present lipid mixtures can form condensed complexes [28], microdomains [29], or undergo phase separation [30] which may be associated with rafts and caveloae in cellular membranes [31]. An important feature of 2 H NMR spectroscopy is that one introduces site-specific 2 H labels, corresponding to the individual C–2 H bonds, and in this way obtains atomically resolved information for liquid-crystalline systems. In liquid-crystalline membranes, the residual quadrupolar couplings correspond to the segmental order parameters of the flexible molecules—they can be directly measured as experimental observables. Moreover, the nuclear spin relaxation rates can be determined, e.g. the relaxation of

Part I

Solid-State Deuterium NMR Spectroscopy of Membranes

250 Part I

Chemistry

Part I Fig. 1. Nanostructure of a lipid bilayer in the fluid, liquidcrystalline (L α) phase. Reprinted with permission from Ref. c 2002 American Chemical Society. (See also Plate 28 [54].
on page XXIX in the Color Plate Section.)

Zeeman order (R1Z ) or quadrupolar order (R1Q ), which depend on the molecular mobility. By combining 2 H NMR order parameter measurements with relaxation studies, one can probe the structural fluctuations of fluid membrane lipids that give rise to averaging of the coupling tensors in solid-state NMR spectroscopy.

Deuterium NMR Spectroscopy Allows Direct Observation of Coupling Tensors Related to Molecular Structure and Dynamics Besides the Zeeman interaction of the nuclear spin with the external magnetic field, additional perturbations are

Fig. 2. Chemical structures of representative glycerophospholipids. The polar head groups vary in their size, capacity for hydrogen bonding, and charge. Representative examples are indicated for the zwitterionic head groups phosphocholine (PC) and phosphethanolamine (PE), and the anionic head group phosphoserine (PS). The non-polar acyl chains vary in their length and degree and position of unsaturation.

due to magnetic interactions (dipolar coupling, chemical shift) and electric interactions (quadrupolar coupling). These couplings provide a wealth of information regarding both the structure and dynamics of biomolecular systems. Generally speaking the principal values and principal axis systems (PAS) of the various coupling tensors yield structural knowledge, whereas their fluctuations give rise to spectral transitions, and are related to the dynamics of the system of interest. Deuterium NMR spectroscopy is particularly valuable as an illustration of the principles of solid-state NMR as applied to molecular solids, liquid crystals, and biomembranes [32]. This is because a single coupling is very large—the electric quadrupolar interaction dominates over the magnetic dipolar couplings of the 2 H and 1 H nuclei, as well as the 2 H chemical shifts. The 2 H nucleus has a spin of I = 1, and hence there are three Zeeman energy levels corresponding to the projection of the nuclear spin angular momentum, with eigenstates |m >= |0 , |±1 given by the Hamiltonian Hˆ Z . According to quantum mechanics, transitions between the adjacent spin energy levels are allowed giving two single-quantum nuclear spin transitions. In 2 H NMR the degeneracy is removed due to the coupling of the quadrupole moment of the 2 H nucleus with the electric field gradient (EFG) of the C–2 H bond, as given by the Hamiltonian Hˆ Q . (An electric quadrupole interacts with an EFG analogously to the interaction of an electric dipole with an electric field.) This is illustrated in Figure 3, part (a), together with a representative 2 H NMR spectrum of a solid polymer, PMMA-d8 , as shown in part (b) which will be discussed subsequently. A general prescription for calculating the 2 H NMR transition frequencies and spectral lineshapes is the

Solid-State Deuterium NMR Spectroscopy of Membranes

Molecular Structures and Motions are Revealed by Deuterium NMR Lineshapes 251

Part I

Fig. 3. (a) Energy levels and resonance lines in 2 H NMR spectroscopy. The Zeeman Hamiltonian Hˆ Z is perturbed by the quadrupolar Hamiltonian Hˆ Q giving an unequal spacing of the nuclear spin energy levels, indicated by |m where m = 0, ±1. The quadrupolar splitting ν Q is the difference in the frequencies (ν ± Q ) of the single-quantum transitions, and is due to the perturbing interaction of the 2 H nuclear quadrupole moment with the EFG of the C–2 H bond. (b) Representative 2 H NMR spectrum of an unoriented powder sample of deuterated plexiglass, PMMA-d8 . The contributions from the C2 H2 groups differ from those of the C2 H3 groups, which undergo rapid threefold motion on the NMR timescale (cf. the text).

following. First one starts with the perturbing Hamiltonian; next Schr¨odinger’s equation is solved to obtain the energy levels; and lastly one introduces the spectroscopic selection rules to calculate the frequencies of the spectral lines. This gives as a final result for the quadrupolar frequencies (ν ± Q ) that 3 ηQ  (2) (2) (PL ) − √ D−20 (PL ) νQ± = ± χQ D00 4 6

(2) (PL ) . + D20

(1)

Here χ Q ≡ e2 qQ/h is the static quadrupolar coupling constant, ηQ is the corresponding asymmetry parameter of the EFG tensor, and PL ≡ (α PL , β PL , γ PL ) are the Euler angles relating the PAS of the EFG tensor (P) and the laboratory frame (L). The experimentally observed 2 H NMR quadrupolar splitting (Figure 3) is then given by the difference in the frequencies of the spectral lines, νQ ≡ νQ+ − νQ− . One should note that the development is also applicable to other second-rank tensors; for instance the magnetic dipolar interaction and the chemical shift [1,2,32].

Molecular Structures and Motions are Revealed by Deuterium NMR Lineshapes Measurement of the deuterium (2 H) NMR lineshapes yields knowledge of the average structure through the principal values of the coupling tensor, as well as the PAS. For the sake of illustration, let us first consider a static oriented sample, e.g. a single crystal in the absence of motions. The crystal can be rotated with respect to

the laboratory frame, giving discontinuities in the NMR spectrum, which correspond to the main external magnetic field aligned along each of the three principal axes of the coupling tensor. The case of an aligned dispersion of phospholipid bilayers deposited on a planar surface is exactly analogous. Here one has a residual or effective coupling tensor, which is pre-averaged by the motions of the flexible lipid molecules in the L α state, but otherwise the transformation under rotations is identical. In either case, the principal axes and principal values of the static coupling tensor, or the residual tensor in the presence of motions, can be obtained from the rotation pattern according to Equation (1). But often one has a polycrystalline sample with a random or spherical distribution of the various C–2 H bond orientations. A powder (or powder-type) spectrum is then obtained, from which one can “read off” the principal values of the coupling tensor directly from the spectral discontinuities [1]. In this case a drawback is that the orientation of the PAS of the coupling tensor within the crystal frame is unavailable, since the spectral discontinuities correspond to the laboratory system. Returning to Figure 3, an experimental 2 H NMR spectrum of a randomly oriented, powder-type sample of deuterated plexiglass, PMMA-d8 , is shown in part (b). Here the outer splitting (±60 kHz) of the powder pattern is due to the C2 H2 groups of PMMA-d8 . For the C2 H2 groups, motion is essentially absent on the 2 H NMR timescale, and the static coupling tensor is observed. The experimental 2 H NMR splitting (due to the large peaks) represents the θ = 90◦ orientation, for which −3χ Q /4 = −127.5 kHz in the case of immobile methylene groups. (Weaker shoulders are also evident, corresponding to the θ = 0◦ orientation with a splitting of 3χ Q /2 = 255 kHz.) On the other hand, the central component (±20 kHz) of the

252 Part I

Chemistry

Part I

2 H NMR spectrum is due to the methyl groups, which are rapidly rotating in the solid state. The threefold rotation about the methyl axes means that the static coupling tensor is averaged to yield a residual coupling tensor, which is axially symmetric (ηeff Q = 0), and whose largest principal value (χ eff Q ) is correspondingly reduced by a factor of −1/3. Hence, for the θ = 90◦ orientation, the C2 H3 splitting is (−3χ Q /4)(−1/3) = 42.5 kHz in good agreement with the experimental spectrum. (The weaker shoulders correspond to the θ = 0◦ orientation with a splitting of χ Q /2 = −85.0 kHz.) According to this example, one can essentially “read off” the coupling parameters, and hence the types of motions, directly from the experimental 2 H NMR spectrum [1]. In passing, we note that rather different, uniaxial powder-pattern lineshapes are observed for certain membrane proteins, such as bacteriorhodopsin [13,33] or rhodopsin [15], and also for nucleic acid fibers [16]. From such 2 H NMR lineshape investigations, one is able to extract information about the molecular structure, as well as the disorder of the sample in terms of the appropriate distribution functions [15,34]. Our next example involves the case of membrane lipid bilayers, where rapid axial averaging occurs about the normal to the membrane film surface, referred to as the director axis. For membranes in the fluid state, the quadrupolar splittings are due to the orientational order parameters of the individual C–2 H-labeled groups, leading to a profile as a function of acyl position. The segmental order parameter SCD describes the amplitudes of the angular excursions of the C–2 H-labeled groups and is given by [6]:

$ % (2) SCD ≡ D00 (0, βPD , 0) = P2 (cos βPD ) , =

 1 3 cos2 βPD − 1 . 2

(2a) (2b)

(2) (PD ) is a Wigner rotation maIn the above formula, D00 trix element, P2 (x) is the second Legendre polynomial where x ≡ cosβ PD , and β PD is the time-dependent angle between the C–2 H bond axis and the director axis (perpendicular to the surface of the membrane). The angular brackets mean an average over all the motions faster than the inverse of the anisotropy in the static quadrupolar coupling ( Na+ > K+ > Tl+ . The binding constants determined for the alkali cations are in agreement with those obtained with 13 C NMR spectroscopy of 13 CO labeled gramicidin A [13]. The binding constants for divalent cations were found to be much larger than those for the monovalent cations. This binding study of the gramicidin channel explains the selectivity of transport for the monovalent cations and why the divalent cations are not transported. The kinetic activation enthalpy for the transport of Li+ , Na+ , and K+ has been determined for gramicidin A and its analogs using the magnetization inversion transfer (MIT) technique [16]. If a membrane impermeable chemical shift reagent, such as [Dy(P3 O10 )2 ]−7 is added to an aqueous salt solution of large unilamellar vesicles with incorporated gramicidin, the internal and external pools of the NMR active cations (7 Li+ ; 23 Na+ or 39 K+ ) can be distinguished by there individual NMR signals. The MIT experiment allows one to obtain the kinetic rate constant for the transfer of magnetization for one cation aqueous pool to the other. When the MIT experiment is performed as a function of temperature, the rate constants can then be used to determine the activation enthalpy for the transport process. The activation enthalpy of transport through the gramicidin A channel was found to increase in the order of cations: 39 K+ (4.2 kcal/mol− ), 23 Na+ (5.4 kcal/mol− ), and 7 Li+ (7.2 kcal/mol− ). The dynamic nature of the gramicidin channel has been the subject of considerable interest. For example, the 15 N spin-lattice relaxation time of the nitrogen atom at the Leu-4 position has been used to investigate the local dynamics about the Ala-3/Leu-4 linkage [17,18]. The NMR results of the experiments suggest a correlation between the local dynamics and ion transport through the channel. The backbone dynamics of gramicidin A in bilayers have been studied using low temperature solid-state 15 N NMR spectroscopy [19]. A 1 H T1 and T2 study of the tryptophan indole NH of gramicidin analogs incorporated into SDS micelles showed a systematic decrease in the overall motion of the indole ring from the

Biological Ion Channels

used to investigate structure and function using a variety of NMR techniques [40,41].

References 1. Hinton JF, Webb GA (Eds). Annual Reports on NMR Spectroscopy. Academic Press Limited: London, 1999, p. 89. 2. Bystrov VF, Gavilov YD, Ivanov VT, Ovchinnikov YA. Eur. J. Biochem. 1977; 8:63. 3. Townsley LE, Tucker WA, Sham S, Hinton JF. Biochemistry. 2001;40:11676. 4. Sham SS, Shobana S, Townsley LE, Jordan JB, Fernandez JQ, Andersen OS, Greathouse DV, Hinton JF. Biochemistry. 2003;42:1401. 5. Jordan JB, Easton PL, Hinton JF. Biophys. J. 2005;88:224. 6. Katchem RR, Hu W, Cross TA. Science. 1993;261:1457. 7. Cornell BA, Separovic F, Baldassi A, Smith R. Biophys. J. 1988;53:67. 8. Killian JA, Taylor MJ, Koeppe RE. Biochemistry. 1992;31:11283. 9. Bouchard M, Davis JH, Auger M. Biophys. J. 1995;69:1917. 10. Urry DW. Proc. Natl. Acad. Sci. U.S.A. 1971;68:676. 11. Hinton JF. J. Magn. Reson. B. 1996;112:26. 12. Smith R, Thomas DE, Atkins AR, Separovic F, Cornell BA. Biochim. Biophys. Acta. 1990;1029:161. 13. Urry DW, Walker JT, Trapane TL. J. Membr. Biol. 1982;69:225. 14. Separovic F, Gehrmann J, Milne T, Cornell BA, Lin SY, Smith R. Biophys. J. 1994;67:1495. 15. Hinton JF, Fernandez JQ, Shungu D, Millett FS. Biophys. J. 1989;55:327. 16. Hinton JF, Easton PL, Newkirk K, Shungu DC. Biochim. Biophys. Acta. 1993;1146:191. 17. Hu W, Cross TA. Biochemistry. 1995;34:14147. 18. North CL, Cross TA. J. Magn. Reson. B. 1993;101:35. 19. Lazo ND, Hu W, Cross TA. J. Magn. Reson. B. 1995;107: 43. 20. Mo Y, Cross TA, Nerdal W. Biophys. J. 2004;86:2837. 21. Easton PL, Hinton JF, Newkirk DK. Biophys. J. 1990;57:297. 22. McKim S, Hinton JF. Biochim. Biophys. Acta. 1993;1153:315. 23. Franklin JC, Ellena JF, Jayaasinche S, Kelsh LP, Cafisco DS. Biochemistry. 1994;33:4036. 24. Brachais L, Davoust D, Molle G. Int. J. Peptide Protein Res. 1995;45:164. 25. North CL, Barranger-Mathys M, Cafisco DS. Biophys. J. 1995;69:2392. 26. Lam Y, Morton CJ, Separovic F. Eur. Biophys. J. 2002;31:383. 27. Lam Y, Wassall SR, Morton CJ, Smith R, Separovic F. Biophys. J. 2001;81:2752. 28. Lauterwein J, Brown LR, Wutherich K. Biochim. Biophys. Acta. 1980;622:219. 29. Pott T, Dufourc EJ. Biophys. J. 1995;68:965. 30. Bechinger B, Zasloff M, Opella SJ. Biophys. J. 1998;74:981. 31. Hirsh DJ, Hammer J, Maloy WL, Blazyk J, Schaefer J. Biochemistry. 1996;335:12733.

Part I

Trp-15 (at the aqueous interface) to Trp-9 (at the interior of the micelles) for all analogs. There are other applications of NMR spectroscopy for studying various aspects of the gramicidin channel and the interaction of the channel with a membrane environment. Solid-state NMR has been used to investigate the closed state of the gramicidin channel in lipid bilayers [20]. The kinetic activation parameters for the incorporation of gramicidin analogs into vesicles as a channel have been determined using 23 Na NMR spectroscopy [5, 21]. The differential photochemical degradation of the four tryptophan residues in gramicidin A has been studied using 1 H two-dimensional NMR spectroscopy [22]. Another type of small, naturally occurring peptide, the peptaibols, has been used as an ion channel model. These channels consist of a bundle of transmembrane helices surrounding a central core. Alamethicin, a 20-residue linear peptide, is the most thoroughly studied member of this class of model channels. A number of NMR studies of alamethicin in SDS micelles [23] and in methanol and aqueous methanol solution [24] have been conducted to determine the conformation of the monomers within the bundle. It appears that the N-terminal region is of an α-helical nature with several 3.010 segments in the C-terminal region. Solid-state 15 N NMR results were found to be consistent with an α-helical conformation inserted along the bilayer normal [25]. NMR studies of other peptaibols also indicate the characteristic α-helical conformation. Other naturally occurring peptides, such as melittin, magainin, cecropin, and pardaxin, form bundles that produce a central channel. There have been many NMR studies of melittin in solution, in micelles, in bilayers, and interacting with lipid membranes [26–29]. The structure of the monomer appears to be that of an α-helix. NMR investigations of magainin [30,31], cecropin [32,33], and pardaxin [34, 35] show that, in general, these peptides also form α-helical monomers that assemble into a structure that has a central pore or channel. Ligand-gated ion channels provide efficient communication between cells of the central nervous system. At the molecular biochemical level, the nicotinic acetylcholine receptor is one of the best-characterized membrane proteins and serves as a paradigm for a family of ligand-gated ion channels [36]. Of the transmembrane segments, M1, M2, M3, and M4, five M2 helices form the central ion channel or pore. Solid-state 15 N NMR experiments of the labeled M2 segment in bilayers have shown that the helical segment is perpendicular to the plane of the bilayer [37]. The M2 protein from the influenza A protein functions as an ion channel. Solid-state 15 N NMR results with the M2 protein from the influenza A virus suggest that this tetrameric protein is in a left-handed, four-helix bundle [38,39]. Peptide mimics of protein channels have been

References 287

288 Part I

Chemistry

Part I

32. Srisailam S, Kumar TKS, Arun kumar AI, Leung KW, Yu C, Chen HM. Eur. J. Biochem. 2001;268:4278. 33. Marassi FM, Opella SJ, Juvvadi P, Merrifield RB. Biophys. J. 1999;77:3152. 34. Zagorski MG, Norman DG, Barrow CJ, Iwashita T, Tachibana K, Patel DJ. Biochemistry. 1991;30:8009. 35. Porcelli F, Buck B, Lee D-K, Hallock KJ, Ramamoothy A, Veglia GJ. Biol. Chem. 2004;279:45815. 36. Dani JA, Mayer ML. Curr. Opin. Neurobiol. 1995;5: 350.

37. Bechinger B, Kim Y, Chirlian LE, Gesell J, Neumann JM, Montal M, Tomich J, Zasloff M, Opella SJ. J. Biol. NMR 1991;1:167. 38. Kovacs FA, Cross TA. Biophys. J. 1997;73:2511. 39. Tian C, Tobler K, Lamb RA, Pinto L, Cross TA. Biochemistry. 2002;41:11294. 40. Doak DG, Mulvey D, Kawaguchi K, Villalain J, Campbell ID. J. Mol. Biol. 1996;258:672. 41. Esposito G, Dumy P, Varma V, Mutter M, Bodenhausen G. Biopolymers. 1997;41:27.

Part I

Membrane Proteins

291

Hazime Saitˆo Department of Life Science, Himeji Institute of Technology, Harima Science Garden City, Hyogo 678-1297, Japan and Center for Quantum Life Sciences, Hiroshima University, Higashi-Hiroshima 739-8526, Japan

Introduction Integral membrane proteins, traversing the membrane once or several times as α-helices, play crucial roles in maintaining various activities of cells such as transport of appropriate molecules into or out of the cell, catalysis of chemical reaction, and receiving and transducing chemical signals from the cell environment. Naturally, biological activity of such proteins may depend upon their conformations and dynamics regulated by specific lipid– protein and/or protein–protein interactions as structural determinants, as studied by analysis of 2D assembly of bacteriorhodopsin (bR) as a typical membrane protein [1]. bR is active as a proton pump and considered as a prototype of a variety of G-protein coupled receptors, consisting of seven transmembrane α-helices. Interestingly, the bR structure is far from static at ambient temperature in spite of currently available 3D structural models revealed by crystallography at low temperature but flexible even in the 2D crystal, especially at the loops and N- or C-terminal residues fully exposed to aqueous phase, and undergoing fluctuation motions with correlation times of the order of 10−4 –10−5 and 10−8 s, respectively, as revealed by recent site-directed solid-state 13 C NMR[2–6]. Well-resolved 13 C NMR signals are fully visible from 2D crystalline 13 C-labeled [3-13 C]Ala-[2,3,7] or [1-13 C]Val-labeled bR[3,5] recorded by both crosspolarization-magic angle spinning (CP-MAS) and dipolar decoupled-magic angle spinning (DD-MAS) techniques. Inherent motional fluctuation of the transmembrane αhelices of bR monomer, however, turns out to be accelerated by two orders of magnitude in the lipid bilayer in the absence of specific protein–protein interactions, from their correlation times of the order of 10−2 s in 2D crystal [2,3,8] to 10−4 –10−5 s, [9–13] in the monomer. Accordingly, 13 C NMR signals from several residues in the transmembrane α-helices and loops could be suppressed due to the failure of attempted peak-narrowing caused by interference of the motional fluctuation frequency with the frequency of proton decoupling or MAS [2,3,14–16], although the functional unit responsible for the photocycle is the monomer itself rather than the trimeric form found in the 2D crystal [17,18]. In this case, uniform Graham A. Webb (ed.), Modern Magnetic Resonance, 291–297.
C 2008 Springer.

13

C-labeling is not always favorable for solid-state NMR, because 13 C NMR study of densely 13 C-labeled proteins such as [1,2,3-13 C3 ]Ala-labeled bR could be substantially broadened in the presence of such intermediate and slow motions, due to the accelerated relaxation rate through a number of homonuclear 13 C–13 C dipolar interactions and scalar J couplings [16]. We demonstrate here how the site-directed 13 C NMR approach is useful to reveal conformational features of intact membrane proteins with emphasis on their surface structures and dynamics at ambient temperature, as revealed by 13 C NMR studies on bR from the 2D crystal and monomer and various membrane proteins active as signal transducers and enzyme, expressed from E. coli and present as monomer in lipid bilayers.

Conformation-Dependent 13 C Chemical Shifts It has been demonstrated that Cα and C=O 13 C chemical shifts of a variety of polypeptides taking the α-helix form are displaced to high frequencies by 3.5–8.0 ppm with respect to those of the β-sheet form, while the Cβ signals of peptides taking the α-helix form are displaced to low frequencies by 3.4–5.2 ppm as compared with those of the β-sheet form [2,3,19]. In addition, it is possible to distinguish even the following six kinds of local secondary structures, right-handed and left-handed α-helices, αII -helix, collagen-like triple helix, silk I and β-sheet forms, besides random coil form, with reference to the conformation-dependent 13 C chemical shifts of Ala residues (Table 1). 13 C NMR peaks from the α-helices in membrane proteins, however, are more widely spread than the expected values of the conformation-dependent displacement of the peaks from solid polypeptides. In fact, several 13 C NMR peaks from the α-helical residues resonate at their lowest (Cβ) and highest (Cα and C=O peaks) boundary peak positions with those of random coil form in the presence of intermediate or low frequency motions, but their peak positions are distinct from those of the loop and β-sheet form [2,3,10]. In such case, the observed distribution of the chemical shifts may deviate greatly from their expected peak positions from the distribution of the torsion angles

Part I

Site-Directed NMR Studies on Membrane Proteins

292 Part I

Chemistry

Part I

Table 1: Conformation-dependent 13 C chemical shifts of Ala residues (ppm from TMS)

Cα Cβ C=O

αI -helix (αR -helix)

αII -helix

αL -helix

β-sheet

Collagen-like triple helix

Silk I

Random coil

52.4 14.9 176.4

53.2 15.8 178.4

49.1 14.9 172.9

48.2 19.9 171.8

48.3 17.6 173.1

50.5 16.6 177.1

50.1 16.9 175.2

Source: Adapted from Refs. [3,10].

determined by X-ray diffraction [20]. Nevertheless, the conformation-dependent displacement of 13 C peaks are very useful as a structural constraint to predict the local structure of membrane proteins. A possibility of the conformation-dependent 15 N chemical shifts, however, may be obscured, because 15 N chemical shifts are influenced by both the local conformation and the primary structure or probably by the higher order structure [21].

with suppressed signals from residues located near the ˚´ by accelerated transverse surface (within ca. 8.7 A) relaxation due to surface-bound Mn2+ [20]. Site-specific assignment of 13 C NMR signals has been attempted for [1-13 C]Val- [3,5], Pro- [24], Trp-, and Ile-[5] labeled bR.

Site-Directed Assignment of 13 C NMR Signals Figure 1 illustrates the 13 C DD-MAS and CP-MAS NMR spectra of fully hydrated [3-13 C]Ala-labeled bR in 2D crystal (MW 26 kDa) at ambient temperature [2,3]. Twelve Ala Cβ 13 C NMR peaks are resolved in the CP-MAS NMR (bottom) among 22 Ala residues present in the transmembrane α-helices and loops. The three intense 13 C NMR signals from membrane surface (gray; top) are noteworthy in the DD-MAS NMR spectrum (consisting of contribution from total 29 Ala residues) and are unambiguously assigned to Ala 228 and 233 (C-terminal α-helix), Ala 240, 244–246 (C-terminal tail taking random coil), and Ala 235 (corner at the C-terminal α-helix) from the upper to the lower field with reference to the conformation-dependent 13 C chemical shifts [2,3,19] together with their absence after enzymatic cleavage by papain [22]. Naturally, these 13 C NMR signals are suppressed in the CP-MAS NMR, because the C-terminal α-helix and its tail undergo fluctuation motions with correlation times of the order of 10−6 and 10−8 s, respectively [10]. The assigned peaks indicated at the individual peaks are obtained in view of selectively reduced 13 C NMR peak intensity of relevant mutant in which an individual Ala residue is replaced by other amino acid residue (for instance, A196G, A126V, A215G, etc.) as compared with that of wild type as illustrated in Figure 2 [3,10] provided that global conformational change is not induced as in D85N [23]. Such a 13 C NMR peak from the transmembrane α-helices can be identified as a single Ala residue by the difference 13 C NMR spectrum between a wild type and a mutant, together

Fig. 1. 13 C DD-MAS (A) and CP-MAS (B) NMR spectra of [3-13 C]Ala-labeled bacteriorhodopsin. The 13 C NMR signals from the C-terminal residues are in gray.

Site-Directed NMR on Membrane Proteins

Dynamic Aspect of Membrane Proteins 293

Part I

Fig. 2. Comparison of the 13 C CP-MAS NMR spectra of [313 C]Ala-labeled bacteriorhodopsin (A) and A126G (B) and A196G mutants (C).

Dynamic Aspect of Membrane Proteins Surprisingly, 13 C NMR signals of bR labeled with certain [1-13 C] amino acid residues are not always fully visible from the loops and transmembrane α-helices by both CP-MAS and DD-MAS techniques, even if 2D crystalline preparations were examined [5,16]. In contrast, 13 C NMR signals from the N- and C-terminal regions with the correlation time shorter than 10−8 s can be observed by DD-MAS NMR only [2,3,5]. Indeed, 13 C NMR signals of [1-13 C]Gly-, Ala-, Leu-, Phe-, and Trp-labeled bR are partially or almost completely suppressed from residues located at the membrane surfaces, and the signals of which can be conveniently estimated by means of accelerated transverse relaxation effect from the surfacebound Mn2+ ions [5]. Indeed, the relative proportions of 13 C-labeled residues from the negatively charged surface ˚´ from fully visible 13 C NMR peaks residues (within 8.7 A) 13 13 of [3- C]Ala-, [1- C]Val-, and Ile-bR[5,20] were consistent with the expected numbers of such residues available from the secondary structure. In contrast, we found that the relative contributions of the surface areas estimated by

Fig. 3. Schematic representation of the location of the Cterminal α-helix (helix G protruding from the membrane surface), its interaction with the C-D and E-F loops (dotted lines) leading to the cytoplasmic surface complex and their correlation times. Note that the correlation times for the transmembrane α-helices differ substantially between preparations of 2D crystal or monomer.

this procedure are substantially lower than the expected numbers of residues from the secondary structure for 13 C NMR spectra of [1-13 C]Gly-, Ala-, Leu-, Phe-, and TrpbR from 2D crystalline purple membranes [5]. This means that these 13 C NMR signals from the surface area are partially or completely suppressed as a result of failure of the attempted peak-narrowing by interference of incoherent low frequency fluctuation motion (104 Hz) with the coherent frequency of MAS [15]. This kind of peak suppression for fully hydrated bR can be utilized as an invaluable means to evaluate in situ protein dynamics with the local correlation time of the order of 10−4 s in the 2D crystal as schematically illustrated in Figure 3, although this phenomenon is obviously a serious disadvantage as viewed from choice of a suitable 13 C-labeled amino acid residue. One should also anticipate that 13 C NMR signals of fully hydrated, monomeric [1-13 C]Gly-, Ala-, Leu-, Phe-, and Trp-labeled membrane proteins [5,9–13] in lipid bilayers are almost completely suppressed, because even the transmembrane α-helices are able to acquire accelerated fluctuation motions in the absence of specific protein–protein interactions essential

294 Part I

Chemistry

Part I

for trimeric structure in the 2D crystals (Figure 3). This is because the lowest energy minimum for Gly and the others in the conformation map may be shallow and the backbone dynamics of the last four amino acid residues represented by the Cα–CβH2 –Z system could be coupled with a possible rotational motion of the χ1 angle around the Cα–Cβ bond. On the other hand, such a conformational space leading to fluctuation motion may be limited to a very narrow area for Val or Ile residues with a bulky side chain at Cα. For [3-13 C]Ala-labeled bR in either 2D crystal or monomeric state in lipid bilayer, peak suppression could occur when incoherent fluctuation motion is interfered with frequency of proton decoupling (correlation time being ca. 10−5 s) [14]. Obviously, intermediate or slow fluctuation motions with correlation times in the order of 10−4 –10−5 s may be well related to individual biological functions, because flexibility in the loops and/or transmembrane α-helices may be essential for initiating transport of proton or ion, receiving external signals, binding substrate, etc. for their respective biological functions such as proton pump, signal transduction, enzymatic activity, etc. In spite of 2D crystalline bR, the fluctuation frequency is spontaneously increased from 102 Hz in the ground state to the order of 104 Hz at the M-like state as a result of a modified retinal–protein interaction owing to proton transfer from the Schiff base to Asp 85 [23].

Surface Structures The surface structure of bR is still obscured, or inconsistent, among a variety of the 3D structures so far revealed by cryoelectron microscopy and X-ray diffraction studies at low temperature [25–27]. This arises because it can be easily altered by a variety of intrinsic or environmental factors such as the manner of crystallization either in the 2D or in the 3D crystals, temperature, pH, ionic strength, crystallographic contact, etc [2,3]. Instead, the 13 C NMR approach proved to be a very suitable means to reveal its surface structure in relation to biological activity at ambient temperature. In particular, 13 C NMR studies revealed that the C-terminal residues, 226–235, participated in the formation of the C-terminal α-helix as viewed from their peak positions [16,22], with reference to the conformation-dependent 13 C chemical shifts [2,3,19]. Only a part of this α-helix, however, was visible by X-ray diffraction [26], owing to the presence of motions with correlation times in the order of 10−6 s detected at ambient temperature, as judged from the carbon spin–lattice relaxation times, T1C , and spin–spin relaxation times, T2C , under CP-MAS conditions [16]. Yonebayashi et al. examined the 13 C NMR spectra of [3-13 C]Ala-labeled bR and its mutants while varying a

variety of environmental or intrinsic factors such as ionic strength, temperature, pH, truncation of the C-terminal α-helix, and site-directed mutation at cytoplasmic loops [28]. For instance, increased ionic strength from 10 to 100 mM NaCl causes simultaneous changes of the high frequency displacement of Ala 103 signal of the C-D loop and the reduced peak intensity of the C-terminal α-helix [28]. This finding together with other similar changes caused by temperature and pH variations leads to the conclusion that the cytoplasmic loops and the C-terminal αhelix are not present independently but are held together to form cytoplasmic surface complex stabilized by salt bridges and/or cation-mediated linkages of a variety of side chains as schematically indicated by dotted lines in Figure 3. Indeed, 13 C NMR signals from such loops are suppressed by accelerated fluctuation motion with a correlation time of the order of 10−5 s and the 13 C chemical shift of the C-terminal α-helix was displaced to low frequency, when blue membranes were prepared by either complete removal of surface-bound cations (deionized blue) or neutralization of surface charge by lowered pH to 1.2 (acid blue) [28,29]. Further, partial neutralization of Glu and Asp residues at the extracellular side such as E194Q/E204Q (2 Glu), E9Q/E194Q/E204Q (3 Glu), and E9Q/E74Q/E194Q/E204Q (4 Glu) caused global fluctuation motions at these loop regions as well as the disorganized trimeric form [30]. The cytoplasmic surface complex in which the C-terminal α-helix is probably tilted toward the direction of the B- and F-helices seems to prevent unnecessary fluctuations of the helices for efficient proton uptake during the photocycle [28]. It appears that such surface structure is disrupted at a low temperature or in the M-like state. This view is consistent with the previous data for “the proton binding cluster” consisting of Asp 104, Glu 160, and Glu 234.

Site-Directed 13 C NMR on Membrane Proteins Present as Monomers Most of reconstituted membrane proteins may be present as monomer in lipid bilayers at ambient temperature in the absence of certain endogeneous lipid molecules essential for specific lipid–protein and protein–protein interactions, as manifested for bR in the 2D crystalline assembly as purple membrane [31,32]. Therefore, it seems to be very important to clarify how the present site-directed 13 C NMR approach is useful to reveal conformation and dynamics of reconstituted membrane proteins as Pharaonis phoborhodopsin ( ppR; sensory rhodopsin II), its cognate transducer ( pHtrII), and diacylglycerol kinase (DGK) which are overexpressed by E. coli [11–13]. In such cases, use of proteins labeled by [3-13 C]Ala is more preferable than [1-13 C]Val, because 13 C NMR signals of the latter preparations were substantially suppressed in a similar

Site-Directed NMR on Membrane Proteins

Site-Directed 13 C NMR on Membrane Proteins Present as Monomers 295

Part I

CP-MAS

* DD-MAS +pHtrII

-pHtrII

Fig. 4. Comparison of the 13 C CP-MAS (left) and DD-MAS (right) NMR spectra of [3-13 C]Ala-labeled phoborhodopsin ( ppR) in the presence (A) or absence (B) of its truncated transducer, pHtrII (1–159). The asterisked peak of the CP-MAS NMR spectrum in the upper trace indicates the peak of the C-terminal α-helix.

manner as encountered for monomeric bR as pointed out already [11,13].

ppR and pHtrII ppR is a retinal protein as a photoreceptor from N. pharaonis, consisting of seven transmembrane α-helices as in bR. Its cognate transducer pHtrII consists of two transmembrane α-helices and yields signaling for negative phototaxis activated by receiving incoming light by ppR through tightly formed complex with ppR. Spreads of the 13 C chemical shifts for reconstituted [3-13 C]Alalabeled ppR in egg PC bilayers and their relative peak intensities (Figure 4) are very similar to those of bR because of taking similar secondary structures for both proteins, in spite of their sequence homology of 27% [11]. The intense 13 C NMR peak at 15.9 ppm, ascribable to the Cterminal α-helix protruding from the membrane surface as found for bR on the basis of the conformation-dependent 13 C chemical shifts described already, is clearly visible in the DD-MAS spectrum (Figure 4B, right) but almost completely suppressed (Figure 4B, left) in the CP-MAS NMR because of fluctuation motion with a frequency of 105 Hz [6,11]. This peak, however, is made visible by the CP-MAS spectrum (Figure 4A, left: asterisked peak) by the complex formation with pHtrII (1–159) due to the

lowered fluctuation frequency in the C-terminal α-helix (104 Hz). This finding indicates that mutual interactions among the extended TM1- and TM2-helices of pHtrII (1–159) beyond the surface and the C-terminal α-helix of ppR play an important role for stabilization of the ppR– pHtrII complex. The intense high frequency αII -helical 13 C DD-MAS NMR peaks of [3-13 C]Ala-labeled pHtrII (1–159) resonate at 16.6 and 16.3 ppm [12] and ascribed to the coiledcoil portion protruding from the membrane surface, with reference to the conformation-dependent displacement of peaks [2,3,19]. These peaks were almost completely suppressed by CP-MAS NMR regardless of the presence or absence of ppR or by DD-MAS NMR in the absence of ppR. Surprisingly, this is caused by increased fluctuation frequency in the C-terminal α-helix from 105 Hz in the uncomplexed state to >106 Hz in the complexed state. This means that the transducers alone are in an aggregated or clustered state but the ppR– pHTrII complex is not aggregated.

Diacylglycerol Kinase DGK from E. coli is a small, 121 amino acid, membranebound enzyme to catalyze the conversion of diacylglycerol and MgATP to phosphatic acid and MgADP. It

296 Part I

Chemistry

Part I

is believed to be assembled into a trimer to be active as a catalytic unit, each consisting of three transmembrane α-helices, together with additional two amphipathic α-helices located at the membrane surface [33]. Yamaguchi et al. recorded 13 C NMR spectra of [3-13 C]Ala-, [1-13 C]Val-labeled E. coli DGK reconstituted in POPC and DPPC bilayers using CP-MAS and DD-MAS methods [13]. Surprisingly, 13 C NMR spectra of [3-13 C]Alalabeled DGK in lipid bilayer were broadened to yield rather featureless peaks at the physiological temperature of the liquid crystalline phase. It is also noted that 13 C NMR spectra of [1-13 C]Val-labeled DGK were completely suppressed at temperatures corresponding to the liquid crystalline phase. Such a suppression of peaks is obviously caused by interference of motional frequency with the frequency of the MAS or proton decoupling (104 –105 Hz), under the physiological condition exhibiting enzymatic activity. In the gel phase lipid, however, up to six distinct 13 C NMR signals were well resolved due to lowered fluctuation frequency (105 and ∼104 Hz, respectively [22]. The protein structures including those fluctuations related to the flexible nature of the membrane would be important to understand functions of the membrane-binding proteins under the physiological conditions. A solid-state NMR study of protein–lipid vesicle complexes provides information about the dynamic structures of the membrane-binding proteins characteristic to those of the anisotropic environments of the water-lipid bilayer interfaces under the physiological conditions. If the three dimensional structures of proteins in crystal or in solution are solved by X-ray diffraction studies or by solution NMR studies, as for the case of PLC-δ1, the information provided by solid-state NMR studies can be interpreted in detail by taking these high-resolution structural models into account.

Application of the Solid-State NMR on the PLC-δ1 PH Domain [23] As shown in Figure 1, rat PLC-δ1 is a 85 kDa protein consists of functionally and structurally distinguishable five domains; N-terminal PH domain, EF-hand domain, X domain, Y domain, and C-terminal C2 domain [24].

Solid-State NMR of Membrane-Binding Protein

Part I

Among these domains, the PH, EF-hand, and C2 domains are common structural modules of proteins included in the cellular signal transduction systems [25]. The X and Y domains form an active site of phospholipase activity of PLC-δ1. Inter-domain and membrane–domain interactions of these proteins provide intracellular cross-talking networks that support traffics of information in the living cells. The PH domain motif is a small and stable structure which consists of about 110 amino acid residues, and found in over 150 proteins involved in the cellular signal transduction pathway and the cytoskeltal reorganization [26, 27]. A number of the PH domains are suggested to mediate protein–lipid interactions by selective bindings to phosphorylated-inositol groups of inositol-phospholipids such as PIP2 [28–31]. Proteins such as the heterotrimeric G-protein and the protein kinase C are also reported to be specific ligands of some PH domains [29]. As mentioned above, the N-terminal PH domain of PLC-δ1 is known to have high affinities to PIP2 and IP3 , and regulates the membrane localization and the activity of PLC-δ1 [32, 33]. By applying solid-state 13 C NMR, structural alterations of the PH domain of PLC-δ1 during the membrane localization could directly be detected. Since the three dimansional structure of the PLC-δ1 PH domain-IP3 complex have been determined by an X-ray diffraction study [24], the structure of the water-soluble PH domain-IP3 complex could be utilized as a template to interpret the structural changes detected by solid-state NMR during the membrane localization of the PH domain. Figure 2 shows the solid-state 13 C NMR spectra of the 13 C-labelled methyl groups of alanine residues introduced into the PLC-δ1 PH domain. Figure 2A and B show the spectra of the complex of the PH domain and the phosphatidylcholine (PC) vesicles containing 5% of PIP2 measured by the cross polarization-magic angle spinning (CP-MAS) technique and the dipolar decoupled-magic angle spinning (DD-MAS) method, respectively. Vertical bars at the bottom of the spectra indicate the chemical shifts of the 13 C NMR signals of Ala residues in the PH domain–IP3 complex in solution. In order to reproduce the dynamic property of the plasma membrane under the physiological condition, the lipid vesicles are suspended in buffers at neutral pH, and enclosed in the air-tight solidstate NMR rotor to prevent an evaporation of water. The PLC-δ1 PH domain contains five Ala residues, Ala21, Ala88, Ala112, Ala116, and Ala118, as shown in the Figure 3A. Assignments of the signals in the solidstate NMR spectra to the individual Ala residues could be determined by site-specific replacements of the alanine residues by other amino acid residues, provided that no conformational changes are induced by resplacement of alanine with glycine, valine, or leucine. The assignment of the peaks was carried out by detecting the disappearance of a signal induced by a replacement of each alanine residue. The assignments of the Ala signals of the PLC-δ1 PH domain are shown at the top of the spectra in Figure 2.

Application of the Solid-State NMR on the PLC-δ1 PH Domain 301

Fig. 2. High-resolution solid-state NMR spectra of the [3-13 C]Ala labeled PLC-δ1 PH domains forming complex with PC/PIP2 vesicles obtained by (A) the CPMAS and (B) the DDMAS method. Assignments of the individual signals are shown at the top of the spectra. The vertical bars at the bottom of the spectra indicate the chemical shifts of the [3-13 C]Ala signals for the PLC-δ1 PH domain-IP3 complex in solution.

The three dimensional structure of the PLC-δ1 PH domain, forming complex with IP3, consists of β sandwich core containing seven β-strands, three α-helices located at the N-terminus, C-terminus, and the loop between β5- and β6-strands (β5/β6 loop), and loops connecting the β-strands (Figure 3A) as determined by an X-ray diffraction study [24]. The β1/β2, β3/β4, and β6/β7 loops form the specific ligand-binding site. The chemical shift displacements of the methyl carbons of the Ala residues in the PLC-δ1 PH domain shown in Figure 2 reflect the presence of a variety of torsion angles of the Ala residues in these higher-order structures. The conformation-dependent 13 C chemical shifts of Ala Cβ carbons have been investigated and reported by solidstate NMR studies on polypeptides and structural proteins [34–36] and by the quantum mechanical calculations of model peptides [34–37]. Ala21 located at the N-terminal α-helix, and Ala116 and Ala118 located at the C-terminal

302 Part I

Chemistry

Part I

Fig. 3. (A) A schematic representation of the three-dimensional structure of the PLC-δ1 PH domain [24]. The α-helices and β-sheets are indicated by cylinders and arrows, respectively. Ala residues are indicated by open circles. IP3 is shown by a cpk model. (B) The amphipathic α2-helix viewed from the C-terminus. Hydrophilic and hydrophobic residues are shown by grey and black circles, respectively. (C) A model of the conformational change of the PLC-δ1 PH domain induced at the membrane surface.

α-helix in the model structure are resonated between 14.8 and 16.1 ppm. Since the chemical shifts of these residues are identical to the chemical shifts of the PH domain forming complex with IP3 , it can be concluded that the conformations of the N- and C-terminal α-helices forming the hydrophilic face of the domain located at the opposite side of the membrane-binding surface do not change during the membrane localization. In contrast, significant changes in the chemical shifts of Ala88 and Ala112 as shown by arrows and kinked lines in Figure 2 indicate that the conformations of C-termini of the α2-helix and β7-strand are altered by the membrane localization of the PH domain. Changes of the chemical shift of Ala88, which is included in the α2-helix and that of Ala112, which is located at the C-terminus of the β7-strand flanking with the C-terminus of the β5/β6 loop, indicate the

conformational changes of the α2-helix and the β5/β6 loop at the membrane surface. As shown in Figure 3B, the α2-helix has an highly amphipathic structure. In the model structure of the PH domain-IP3 complex, the α2helix faces its hydrophobic surface to the hydrophobic surface of the β-sandwich core consists of β5-, β6-, and β7-strands. Considering the proposed orientation of the amphipathic α-helical peptides at the interface region of the lipid bilayer, the α2-helix is expected to be located at the interface region of the membrane, facing the hydrophilic surface to the aqueous phase and the hydrophobic surface to the hydrophobic core of the lipid bilayer. The expected conformational changes of the α2-helix and the β5/β6 loop at the membrane interface are likely to provide more typical α-helix and β-sheet structures for Ala88 and Ala112, respectively. The structural changes

Solid-State NMR of Membrane-Binding Protein

References 1. Grobler JA, Hurley JH. Biochemistry 1998;37:5020. 2. Kim YH, Park TJ, Lee YH, Baek KJ, Suh PG, Ryu SH, Kim KT. J. Biol. Chem. 1999;274:26127. 3. Lomasney JW, Cheng HF, Wang LP, Kuan Y, Liu S, Fesik SW, King K. J. Biol. Chem. 1996;271:25316. 4. Auge S, Bersch B, Tropis M, Milon A. Biopolymers. 2000;54:297. 5. Wang Z, Jones JD, Rizo J, Gierasch LM. Biochemistry. 1993;32:13991. 6. Matsunaga TO, Collins N, Ramaswami V, Yamamura SH, O’Brien DF, Hruby VJ. Biochemistry. 1993;32:13180.

7. Milon A, Miyazawa T, Higashijima T. Biochemistry. 1990;29:65. 8. Wakamatsu K, Okada A, Suzuki M, Higashijima T. Masui Y, Sakakibara S, Miyazawa T. Eur. J. Biochem. 1986;154:607. 9. Kutateladze TG, Capelluto DG, Ferguson CG, Cheever ML, Kutateladze AG, Prestwich GD, Overduin M. J. Biol. Chem. 2004;279:3050. 10. Muller G. FEBS Lett. 2002;531:81. 11. Pike LJ, J. Lipid Res. 2003;44:655. 12. Caroni P. Embo J. 2001;20:4332. 13. Vereb G, Szollosi J, Matko J, Nagy P, Farkas T, Vigh L, Matyus L, Waldmann TA, Damjanovich S. Proc. Natl. Acad. Sci. U S A. 2003;100:8053. 14. White SH, Ladokhin AS, Jayasinghe S, Hristova K. J. Biol. Chem. 2001;276:32395. 15. Raudino A, Mauzerall D. Biophys. J. 1986;50:441. 16. Hristova K, Wimley WC, Mishra VK, Anantharamiah GM, Segrest JP, White SH. J. Mol. Biol. 1999;290:99. 17. Zakharov SD, Lindeberg M, Griko Y, Salamon Z, Tollin G, Prendergast FG. Cramer WA. Proc. Natl. Acad. Sci. U S A. 1998;95:4282. 18. Elkins P, Bunker A, Cramer WA, Stauffacher CV. Structure 1997;5:443. 19. Lindeberg M, Zakharov SD, Cramer WA. J. Mol. Biol. 2000;295:679. 20. Boguslavsky V, Rebecchi M, Morris AJ, Jhon DY, Rhee SG, McLaughlin S. Biochemistry. 1994;33:3032. 21. Pastor RW, Venable RM, Feller SE. Acc. Chem. Res. 2002;35:438. 22. Yamaguchi S, Tuzi S, Yonebayashi K, Naito A, Needleman R, Lanyi JK, Saito H. J. Biochem. (Tokyo) 2001;129:373. 23. Tuzi S, Uekama N, Okada M, Yamaguchi S, Saito H, Yagisawa H. J. Biol. Chem. 2003;278:28019. 24. Ferguson KM, Lemmon MA, Schlessinger J, Sigler PB. Cell. 1995;83:1037. 25. DiNitto JP, Cronin TC, Lambright DG. Sci. STKE 2003;2003:re16. 26. Yao L, Janmey P, Frigeri LG, Han W, Fujita J, Kawakami Y, Apgar JR, Kawakami T. J. Biol. Chem. 1999;274:19752. 27. Lemmon MA, Ferguson KM, Abrams CS. FEBS Lett. 2002;513:71. 28. Lemmon MA, Ferguson KM. Biochem. J. 2000;350 Pt 1:1. 29. Maffucci T, Falasca M. FEBS Lett. 2001;506:173. 30. Hirata M, Kanematsu T, Takeuchi H, Yagisawa H. Jpn. J. Pharmacol. 1998;76:255. 31. Kavran JM, Klein DE, Lee A, Falasca M, Isakoff SJ, Skolnik EY, Lemmon MA. J. Biol. Chem. 1998;273:30497. 32. Guo Y, Philip F, Scarlata S. J. Biol. Chem. 2003;278:29995. 33. Lemmon MA, Ferguson KM, O’Brien R, Sigler PB, Schlessinger J. Proc. Natl. Acad. Sci. U S A 1995;92:10472. 34. Saito H, Tuzi S, Naito A. Annu. Rep. NMR Spectrosc. 1998;36:79. 35. Saito H, Ando I. Annu. Rep. NMR Spectrosc. 1989;21: 209. 36. Saito H. Magn. Reson. Chem. 1986;24:835. 37. Asakawa N, Kurosu H, Ando I. J. Mol. Struct. 1994;323:279. 38. Huang S, Lifshitz L, Patki-Kamath V, Tuft R, Fogarty K, Czech MP, Mol. Cell. Biol. 2004;24:9102. 39. Fadok VA. Henson PM. Curr. Biol. 2003;13:R655. 40. Frasch SC, Henson PM, Nagaosa K, Fessler MB, Borregaard N Bratton DL. J. Biol. Chem. 2004;279:17625.

Part I

include formations of the typical α-helix and β-sheet type hydrogen bonds of the residues that are missing in the model structure of the PH domain-IP3 complex. These conformational changes are consistent with the directions of the chemical shift displacements of Ala88 and Ala112 induced by the formation of the PH domain–vesicle complexes. A model of the conformational changes of the PLC-δ1 PH domain at the membrane surface expected from the solid-state NMR study is illustrated in Figure 3C. The structure of the PH domain at the membrane surface is also found to be remarkably affected by the lipid composition of the membrane. For instance, the abovementioned conformational alteration of the PLC-δ1 PH domain induced at the surface of the PC/PIP2 membrane are found to be suppressed at the negatively charged membrane surface containing acidic phospholipids, such as phosphatidylserine (PS). The solid-state NMR spectra of the PH domain binding to the PC/PS/PIP2 membrane indicate that the conformation of the PH domain is identical to that of the PH domain forming a complex with IP3 in solution. Moreover, a drastic increase in the mobility of the PH domain at the surface of the PC/PS/PIP2 membrane is also detected from the changes of the relaxation parameters of the solid-state NMR spectroscopy. That the structure and the mobility of the PLC-δ1 PH domain depend on lipid composition of the target membrane may provide molecular mechanisms for the regulation of the PLC-δ1 function; changes in the local lipid composition in response to a variety of physiological reactions in the cell [38–40]. Modification of the protein structure and dynamics induced at the water-lipid bilayer interface as observed for the PLC-δ1 PH domain would also occur for other lipidbinding domains and proteins that are localized at the surfaces of the cellular membranes. The high-resolution solid-state NMR provides an unique method to investigate structural characteristics of the membrane-binding proteins that take part in important cellular functions mediated by changes in the structure, composition, and dynamics of the intracellular and plasma membranes.

References 303

305

John D. Gehman and Frances Separovic School of Chemistry, University of Melbourne, Melbourne, VIC 3010 Australia

Membranes are commonly perceived as little more than a simple canonical bilayer formed by the obvious orientation-preference of the constituent amphiphilic lipid molecules. Hydrophilic head groups, such as the zwitterionic phosphatidylcholine (PC), line the interface with aqueous environments on both sides of the membrane, while the aliphatic fatty acid chains of varying lengths and degree of unsaturation meet tail-to-tail to fill the region flanked by the head groups. The common perception of membrane-associated peptides and proteins is that they generally span the membrane with simple secondary structures, usually α-helices, but include the occasional β-sheet structure. These secondary structures are frequently regarded as trivial anchors to be removed so that the more interesting soluble domains may be studied by conventional approaches. The classic “fluid-mosaic” model [1] suggests that membranes are simply a two-dimensional analog of a solution: lipids and membrane-associated protein rotate about single axes (normal to the bilayer surface) and translate across the membrane plane in a similar way to water and soluble proteins in three-dimensional solution. As well, study of membrane-associated proteins typically focuses almost exclusively on the protein—the membrane lipid is the negative space, akin to the buffer in which soluble protein is studied, while all the light is cast upon the protein. Closer inspection of membrane structure, however, suggests that a more balanced view of membranes and membrane-associated protein is often necessary and more rewarding. Lipids have complex phases and phase transitions which depend upon particular lipid properties, temperature, hydration levels, and, especially relevant— the protein composition of the mixture. For those lipids commonly employed in model membrane studies, longerchain fatty acids tend to persist in the lamellar gel-phase Lβ as temperature is increased, particularly at low hydration levels. Conversely, shorter fatty acid chains and higher hydration levels tend to allow transition to the lamellar liquid crystalline phase Lα at lower temperatures. Higher temperatures and different lipid geometries than those typically employed for model membrane studies, lead to additional phase transitions into Graham A. Webb (ed.), Modern Magnetic Resonance, 305–311.
C 2008 Springer.

cubic (Q I and Q II ) and hexagonal phases (HI and HII ) [2]. Membrane-associated peptides and proteins include cell signaling receptors, immune response factors, ion channels, cell adhesion elements, toxins, and metabolic and photosynthetic components, many of which are also exploited by viruses to recognize and gain entry into cells. The literature is replete with hyperbole of the suitability of solid-state NMR for study of proteins in membrane environment: solid-state NMR does not require long-range ordering of molecules as is required for crystal diffraction work, and is not subject to the same hydrodynamic restraints that liquid-state NMR requires of molecules in solution. These are actually strengths rather than mere lack of weakness: less stringent sample structure allows protein and lipid molecules to be studied under conditions much closer to their natural membrane environments, and being independent of rapid global molecular rotation reintroduces the sensitivity of orientation-dependent parameters to local molecular structure and motion. This sensitivity is based in large part upon the (3 cos2 θ − 1) dependence, which is treated under perturbation theory as first order corrections (H ) to the Zeeman energy Hamiltonian (Hz = m z γ B0 ). θ is the angle subtended by the relevant vector within the sample and the applied magnetic field B0 , and (3 cos2 θ − 1) is also known as the proportional Legendre polynomial P2 (θ ) or the spherical harmonic function Y02 :   HT ≈ Hz + H Y02 . Conveniently for NMR studies of phospholipid membranes, 31 P is 100% naturally abundant, is more sensitive than carbon, and phospholipids typically have just one 31 P nucleus per molecule in the head group. Also advantageous is the infinitesimal natural abundance of 2 H, as small amounts of 2 H-enriched lipids may be selectively added to a sample and observed against virtually zero background. Together, wideline 31 P and 2 H NMR can be used to report on relative motional differences or changes in conformation of the hydrophilic head group at the aqueous interface and the hydrophobic aliphatic tails within the lipid bilayer, respectively, upon addition of protein

Part I

Solid-State NMR of Membrane-Active Proteins and Peptides

306 Part I

Chemistry

Part I

and other lipid components. At an extreme, changes in sample composition actually introduce new lipid phases that may compromise protein conformation and dynamics and/or may be relevant to the function of the protein being explored (as is the case for cytolytic toxins). Similarly, despite relatively low intensity, selective enrichment of 13 C and 15 N in protein can be used against otherwise very low natural abundance to permit measurement of several structural aspects of membrane proteins: 13 C and 15 N NMR can help to define protein structure as well as orientation within the membrane.

Chemical Shift Anisotropy (CSA) For spin I = 1/2 nuclei, which include 31 P, 1 H, 13 C, and 15 N, Y02 modulates the interaction ( Hσ ) between the applied magnetic field and the anisotropic chemical shift tensor, specifically the anisotropic component. Each individual nucleus may be thought of as an individual crystallite, and each crystallite is most simply described by its principal axis system (PAS)—the orientation in which the chemical shift anisotropy (CSA) tensor is diagonal, PAS PAS with characteristic components σPAS · Each x , σ y , σz crystallite contributes signal intensity to the powder spectrum at frequency νσ according to its orientation relative to B0 , which in the laboratory frame is simply (0,0,B0 ). The frequency at which a given crystallite contributes to the spectrum is given by [3]  v σ = γ B0 σxPAS cos2 α sin2 β  + σ yPAS sin2 α sin2 β + σzPAS cos2 β where, for all basis set vector lengths normalized to unity and expressed in laboratory frame coordinates, β is the angle between the PAS z-axis and B0   β = cos−1 z PAS · B0 , and, provided β = 0, α is the angle between the PAS yaxis and the vector orthogonal to both the PAS z-axis and B0 . The direction of rotation by α is fixed by forcing the rotation axis to point in the same direction as the PAS +z-axis:  ⎞ ⎛ z PAS × B0 ⎠ α(β =0) = cos−1 ⎝ yPAS · sin β   ⎞ ⎛  z PAS × B0   ⎠ , = sin−1 ⎝ yPAS ×  sin β  

which simplifies to  α(β =0) = cos

−1

xPAS · B0 sin β



 −1

= sin

yPAS · B0 sin β

 .

α, β, and (although irrelevant here) γ are also commonly known as the Euler angles [4]. For the simple case considered here in which the PAS axis system is transformed directly to the laboratory frame, β and the Y02 angle θ are the same. When all individual crystallites are ordered identically, as in a macroscopic crystal, a single peak for each resonance is observed according to the orientation of the crystal relative to the magnetic field (e.g. Jones et al. [5]). More commonly, when crystallites are distributed randomly over all orientations, as for a powdered sample, a characteristic powder pattern is observed in which the discontinuities in the lineshape reveal the three PAS components. When the individual crystallite experiences rapid motion, whether by experimental mechanical spinning, or by molecular motion within the sample, time-averaged chemical shift values are observed. For the common case when the sample is mechanically spun about 54.74◦ (the “magic angle,” where Y02 = 0), the anisotropic component of the chemical shift tensor averages to zero, and the time-averaged value observed is the isotropic chemical shift; local magnetic environmental differences aside, this is the same value observed in solution NMR. When the orientation of the crystallite varies owing to molecular motion, as for a lipid rotating and translating within a model membrane, averaging of the CSA will depend on the range, axes, and frequency of motion. The 31 P CSA of the head group at the rigid-lattice limit is realized with anhydrous gel-phase lipid samples, which have principal tensor values of approximately −98, −35, and 133 ppm (in traceless form1 ) for PC [6], a lipid head group class most commonly used in model membranes. Hydration of the lipid permits head group rotation about the glycerol-carbon/phosphate-oxygen and phosphateoxygen/phosphorous bonds, and the static CSA is partially averaged to apparent principal tensor values of approximately −82, −27, and 109 ppm (in traceless form) for PC [6] (Figure 1). Further deviations from this lineshape are indicative of the lipid phase. Above the gel-phase transition temperature, lipids experience rapid axial rotation as well 1

Traceless form expresses principal chemical shift tensor values such that their average, the isotropic chemical shift, is zero. Adding the isotropic chemical shift to these principal CSA values gives reference-specific chemical shift values.

Solid-State NMR of Membranes

as translational motion across the lipid surface. In the liquid crystalline lamellar phase of unoriented samples, though the bilayer normal vectors would be distributed randomly just as for regular powder samples, lateral translation of lipid molecules on the NMR timescale causes negligible deviation in lipid long-axis orientation. Hence in this phase the effect of rapid axial rotation alone is observed, where the 31 P CSA tensor is motionally averaged so that it appears axially symmetric, and only two different principal tensor values are needed, labeled σ⊥ and σ|| with reference to the axis of rotation about the bilayer normal. Translational diffusion of lipids in the hexagonal phase, however, does serve to reorient the lipid rotation axis on the NMR timescale, and hence a further averaging of the CSA tensor is observed, which is generally manifest as a narrowing and reversal of the asymmetric liquid crystalline or bilayer phase lineshape. Finally, a number of other phases where lipids rapidly reorient on the NMR timescale yields a symmetric and relatively narrow peak about a single principal value— the isotropic chemical shift. Most relevant for biological work are micellar and small vesicle phases, where the

entire lipid structure is small enough to rotate quickly; indeed it is these phases that are used for solution NMR experiments of lipid-associated proteins. 31 P wideline NMR alone can provide a simple diagnostic, which may, for example, indicate whether temperature, hydration, or lipid characteristics and membrane composition need to be adjusted to maintain lamellar phase or vesicle structure upon addition of protein [7,8]. Another aspect of the CSA property of nuclei is explored by using aligned membrane samples. Lipid bilayers can be formed with parallel surfaces by layering hydrated phospholipids between glass plates. Similar to NMR of single crystals, by forcing a common orientation relative to the applied magnetic field B0 , a single chemical shift value may be observed. While this can be useful for lipid samples using 31 P [9], it is particularly useful for the determination of backbone orientation using carbonyl 13 C and amide 15 N chemical shifts of membrane-associated peptides and protein. The principal 13 C CSA values for peptide backbone carbonyl are approximately −75, −3, and 78 ppm (in traceless form), with the intermediate -3 ppm value being the most variable and aligned approximately 10◦ off the 13 C=O bond (on the opposite side of the 13 C–N peptide bond) in the peptide plane, and with the 78 ppm axis perpendicular to the peptide plane [10–13]. Hence, when a given carbonyl orientation and membrane bilayer normal vector orientation are such that the peptide plane is perpendicular to B0 , a maximum chemical shift value is observed. Other orientations yield Y02 -attenuated chemical shift values. This property can be combined together with consideration of molecular motion and secondary structure to help discriminate between different possible models for peptide and protein association with lipid bilayers, as applied to gramicidin A [14,15] (Figure 2) and melittin [16]. Similarly, principal shift tensor values for 15 N peptide backbone amides are approximately −60, −41, and 101 ppm (in traceless form), with the 101-ppm component lying 15◦ –20◦ off the 15 N–H bond vector and in the peptide plane [13,17,18]. When the fixed relationship between a particular backbone 15 N amide and an oriented membrane is such that the 15 N–H bond vector lies approximately parallel to B0 , a maximum of chemical shift is observed. From a collection of other chemical shift observations the orientation of each 15 N–H bond vector can be deduced from the chemical shift as a function of the angle of the oriented bilayer normal to B0 . This information can be used similarly to the above, for example, to determine whether an α-helix inserts into membranes (where 15 N– H bond vectors lie parallel to bilayer normal vectors), or lies parallel to the bilayer surface on membranes [9,19].

Part I

Fig. 1. Simulated lineshapes for static 31 P NMR of various lipid phases, in order of decreasing linewidth: static limit, monohydrated, liquid lamellar, hexagonal, and isotropic. Relatively intensities are only qualitatively correct. (See also Plate 34 on page XXXI in the Color Plate Section.)

Chemical Shift Anisotropy (CSA) 307

308 Part I

Chemistry

Part I

Fig. 2. (Left) Representation of the Ala3 carbonyl orientation of gramicidin A in a β6.3 -helical structure for the molecular and PAS frame relative to an axis of rotation parallel to the lipid bilayer normal. Expected CSA with fast axial rotation in this orientation is −17 ppm. (Right) Measurement of carbonyl chemical shift for Ala3 as a function of the angle of the bilayer normal to B0 . The experimental CSA is approximately −16 ppm. (Adapted from Cornell et al. [14].)

Quadrupolar Coupling Spin I > 1/2 nuclei have an additional interaction (HQ ) with the nuclear electric field gradient, manifest as a nonzero quadrupole moment H ≈ Hσ + HQ . In the case of the spin I = 1 nucleus 2 H bound to carbon, simplifications can be made of the anisotropic properties involved such that v Q (kHz) = 127.5 · (3 cos2 θ − 1), where the underlying tensor is oriented such that θ is the angle between the applied magnetic field and the C–2 H bond vector [20], and rapid axial rotation about the C–2 H bond reduces the static coupling constant by a factor of two, so that the quadrupolar splitting magnitude varies between 127.5 and −63.75 kHz at 0◦ and 90◦ , respectively [2]. Owing to a probability distribution of θ proportional to sin θ , the θ = 90◦ orientation is most probable, and θ = 0◦ is least probable. Hence, the characteristic powder pattern—a “Pake” pattern [21]—of a static ensemble of C–2 H bond vectors is symmetric, according to v = v σ ± 1/2νQ . The lineshape has greater intensity where the quadrupolar splitting is reduced (the “90◦ edge”), lower intensity where the quadrupolar splitting is greatest, and the intensity overlaps at v σ for both m I = +1 − 0 and m I = 0 − (− 1) transitions at θ = 54.74◦ . As for 31 P chemical shift, molecular motion which is rapid on a timescale relative to the quadrupole splitting causes variations in the angle θ and serves to time-average the quadrupolar splitting constant to a smaller-magnitude

effective splitting at each orientation. In the fatty acid chains of membrane lipids, carbon–carbon bond rotation is the principal molecular motion, and serves to attenuate the maximum potential splitting. Such an attenuation is attributed to disorder among the lipid tails, and is characterized in practice by an order parameter S, which is proportional to Y02 for the apparent quadrupole splitting, S = 12 3 cos2 θ − 1 , expressed to indicate that the splitting is a time-averaged value [22]. A fully deuterated fatty acid acyl chain produces a spectrum that is a sum of Pake patterns, one for each C–2 H position along the chain (e.g. Separovic and Gawrisch [23], Figure 3A). With sufficient resolution, the signal from each carbon position can be distinguished at the most intense portion of the lineshape at θ = 90◦ . These complex spectra may be simplified by “de-Pake-ing” [24]—a numerical procedure for deriving the spectrum for a single angle θ from the powder pattern (Figure 3B). While strict interpretation of spectra can be difficult without systematically deuterating each position individually [25], it is generally assumed that for the typically resolved 90◦ intensities, the outermost to innermost intensity of the 2 H spectrum corresponds to the sequential chain positions from the first CD2 positions after the carbonyl ester to the terminal methyl group deuterons. This indicates that the chain positions closest to the bilayer interface are most ordered, as they contribute the intensity at the wings of the spectrum with greatest quadrupole splitting. Positions become less ordered with subsequent steps further down the acyl chain [25]. Increasing disorder at each chain position causes the chain to protrude

Solid-State NMR of Membranes

less far from the membrane interface than it would in an extended conformation. Therein, greater acyl chain order can be interpreted as an increase in membrane thickness [23,26]. Hence, changes in quadrupolar splittings can indicate changes in membrane structure and dynamics upon addition of peptide or protein. 31

P and 2 H NMR of Lipids

Numerous examples demonstrate that together 31 P and 2 H NMR can provide important insight into membrane structure given the independent reporting of the relative motion and phase state of the head groups and hydrophobic tails. At one extreme, we have shown that Core Peptide from a transmembrane sequence of Tcell antigen receptor, known to be inhibitory of immune response, impacts upon 31 P and 2 H spectra of lipid bilayers only at unreasonably high peptide concentrations, supporting the supposition that its activity must be other than a consequence of membrane disruption [27]. At another extreme, the sphingomyelin-dependent cytolytic sea anemone protein equinatoxin II (EqtII) was shown to cause significant changes in model membranes at very low concentrations [28]. In this study, PC lipid head group motion increased upon addition of either 10% sphingomyelin or 0.1% cytolytic EqtII protein,

but appeared similar to PC alone when both are added to model membranes (although the lipid phase transition temperatures increase). 2 H spectra of a deuterated PC for the same samples show a similar trend: overall linewidth (demarcated by the 0◦ edges of the highest ordered carbons of the acyl chain) decreases, reflecting greater disorder upon addition of sphingomyelin or EqtII individually, but higher order is observed when both are added. These two observations (together with relaxation data) were interpreted as suggesting that EqtII and sphingomyelin significantly impact membrane dynamics independently, but when combined tend to preferentially segregate out together from bulk PC lipid, leaving the PC dominated 31 P signal and 2 H signal to appear similar to pure PC. A higher protein concentration (0.4%) and higher temperature were also shown to promote an additional phase with an isotropic spectral component in both 31 P and 2 H spectra. These are likely to be very small unilamellar vesicles as seen by cryo-electron microscopy, and are sphingomyelin-enriched as shown by magic-angle spinning 31 P spectra of phases separated by centrifugation. Lying between these extreme effects is the example of the αM1 transmembrane helix of nicotinic acetylcholine receptor added to PC bilayers. 31 P NMR [8] was used to show that high protein concentration and longer incubation times promoted the formation of an isotropic lipid phase. Of the conceivable lipid phases possibly giving rise to an isotropic signal, it was deduced that smaller, more rapidly tumbling vesicles were formed since 2 H NMR spectra did not show spectral averaging of the lipids. Furthermore, slightly higher order of the acyl chains, indicated by larger quadrupolar splittings, suggested a greater average membrane thickness. Addition of cholesterol prevented formation of the 31 P isotropic phase, while addition of the anesthetic halothane promoted conformational changes in the polypeptide due to changes in bilayer properties.

Dipolar (Re)-Coupling Magnetically active nuclei also directly interact with one another in a distance-dependent manner, a property that can be exploited for structure determination of membrane protein complexes. The most significant terms in the dipolar coupling Hamiltonian are also a function of Y02 . These are: (i) the zero-frequency term, “Iz Sz ” in productoperator parlance,which expresses the direct impact of the spin state of one nucleus upon another and (ii) the difference frequency/zero quantum term, “I+ S− + I− S+ ”, which drives mutual antiparallel “spin flips” between nuclei. While the Iz Sz term is significant for all nuclear pairs, the I+ S− + I− S+ term is significant only for nuclear pairs for which the orientation-dependent resonance

Part I

Fig. 3. 2 H spectrum of (2 H31 )-palmitoyl-oleoyl-PC (singlechain deuterated POPC) and natural abundance dioleoylphosphatidylethanolamine (DOPE) (5:1 molar ratio) bilayers: (A) powder pattern from unoriented dispersion; and (B) de-Paked half spectra, calculated for 0◦ orientation from A. (Adapted from Separovic and Gawrisch [23].)

Dipolar (Re)-Coupling 309

310 Part I

Chemistry

Part I

frequency difference ω is sufficiently small, principally homonuclei in a typical magnetic field. Dipolar interactions usually complicate a spectrum; experiments that include sample-labeling schemes to provide for such complicating interactions are generally also spun at a speed about the magic angle that is several times faster than the strength of the dipolar interaction. The Y02 dependent dipolar coupling is thus averaged to zero as for the anisotropic component of chemical shift and the first order quadrupolar coupling for spin >1/2 nuclei. Further conjuring lies in the techniques to selectively reintroduce the dipolar coupling under magic-angle spinning conditions. One technique, rotational resonance, (RR or R2 ), is so simple it may actually be achieved accidentally with an inopportune choice of magic-angle spinning speed ωr for an isotopically (often 13 C) labeled sample. When the condition ω = n × ωr is met for small integer n and chemical shift difference ω = |ω I − ω S |, the nuclear spin pair I, S are recoupled through the I+ S− + I− S+ term, and effectually trade nuclear polarization [29] at a distance-dependent rate [30]. The technique can be employed for somewhat qualitative determination of protein complex models [31,32]. Strict analysis of the data requires attention to several complicating physical factors [30,33], and has been employed ˚ for a to achieve distance determinations of 2.7 ± 0.2 A 31 P–31 P nuclear pair [34], as well as distances as long as ˚ with identical or better accuracy for 13 C–13 C nuclear 5A pairs in melittin measured in a membrane environment [7].

Conclusion Exploitation of the (3 cos2 θ − 1) dependence of anisotropy in the chemical shift, quadrupolar and dipolar interactions with magnetically active nuclei provides a rich source of information for both proteins and, importantly, lipids in membrane systems. Solid-state NMR is well suited to measure these anisotropies, and is becoming geometrically more important as the focus of biomedical research increasingly spotlights membrane proteins. While straightforward approaches require sometimes tedious isotopic labeling of samples, and data interpretation should not be assumed to be trivial, the experiments are relatively easy to perform and do not require extraordinarily high field strengths as do many other currently emerging and increasingly important NMR approaches. Although currently evolving at an accelerated pace, existing biological solid-state NMR techniques can also give

information on both the membrane and the associated protein.

References 1. Singer SJ, Nicholson G. Science. 1972;175:720. 2. Epand R. Lipid Polymorphism and Membrane Properties. Academic Press: San Diego, 1997. 3. Mehring M. Principles of High Resolution NMR in Solids. Springer-Verlag: New York, 1983. 4. Rose ME. Elementary Theory of Angular Momentum. Wiley: New York, 1957. 5. Jones GP, Cornell BA, Horn E, Tiekink ERT. J. Crystallogr. Spectrosc. Res. 1989;19(4):715. 6. Seelig J. Biochim. Biophys. Acta. 1978;515:105. 7. Lam YH, Wassall SR, Morton CJ, Smith R, Separovic F. Biophys. J. 2001;81(5):2752. 8. de Planque MRR, Rijkers DTS, Liskamp RMJ, Separovic F. Magn. Reson. Chem. 2004;42(2):148. 9. Balla MS, Bowie JH, Separovic F. Eur. Biophys. J. Biophys. Lett. 2004;33(2):109. 10. Stark RE, Jelinski LW, Ruben DJ, Torchia DA, Griffin RG. J. Magn. Reson. 1983;55:266. 11. Separovic F, Smith R, Yannoni CS, Cornell BA. J. Am. Chem. Soc. 1990;112:8324. 12. Oas TG, Hartzell CJ, McMahon TJ, Drobny GP, Dahlquist FW. J. Am. Chem. Soc. 1987;109(20);5956. 13. Hartzell CJ, Whitfield M, Oas TG, Drobny GP. J. Am. Chem. Soc. 1987;109(20):5966. 14. Cornell BA, Separovic F, Baldassi AJ, Smith R. Biophys. J. 198;53:67. 15. Smith R, Thomas DE, Separovic F, Atkins AR. Cornell BA. Biophys. J. 1989;56:307. 16. Smith R, Separovic F, Milne TJ, Whittaker A, Bennett FM, Cornell BA, Makriyannis A. J. Mol. Biol. 1994;241: 456. 17. Wu CH, Ramamoorthy A, Gierasch LM, Opella SJ. J. Am. Chem. Soc. 1995;117:6148. 18. Oas TG, Hartzell CJ, Dahlquist FW, Drobny GP. J. Am. Chem. Soc. 1987;109(20):5962. 19. McDonnell PA, Shon K, Kim Y, Opella SJ. J. Mol. Biol. 1993;233:447. 20. Schmidt-Rohr K, Spiess HW. Multidimensional Solid State NMR and Polymers: Academic Press, 1994. 21. Pake GE. J. Chem. Phys. 1948;16(4):327. 22. Seelig J, Niederberger W. J. Am. Chem. Soc. 1974;96(7):2069. 23. Separovic F, Gawrisch K. Biophys. J. 1996;71(1): 274. 24. Sternin E, Bloom M, MacKay AL. J. Magn. Reson. 1983;55:274. 25. Seelig A, Seelig J. Biochemistry. 1974;13(23):4839. 26. Douliez JP, Leonard A, Dufourc EJ. J. Phys. Chem. 1996;100(47):18450. 27. Ali M, De Planque MRR, Huynh NT, Manolios N, Separovic F. Lett. Peptide Sci. 2001;8(3–5):227. 28. Bonev BB, Lam YH, Anderluh G, Watts A, Norton RS, Separovic F, Biophys. J. 2003;84(4):2382.

Solid-State NMR of Membranes

32. Lam YH, Morton CJ, Separovic F. Eur. Biophys. J. Biophys. Lett. 2002;31(5):383. 33. Costa PR, Sun B, Griffin RG. J. Am. Chem. Soc. 1997;119:10821. 34. McDermott AE, Creuzet F, Griffin RG, Zawadzke LE, Ye Q-Z, Walsh CT. Biochemistry. 1990;29:5767.

Part I

29. Andrew ER, Bradbury A, Eades RG, Wynn VT. Phys. Lett. 1963;4(2):99. 30. Levitt MH, Raleigh DP, Creuzet F, Griffin R.G. J. Chem. Phys. 1990;92:6347. 31. Lam YH, Nguyen V, Fakaris E, Separovic F. J. Protein Chem. 2000;19(6):529.

References 311

313

Gary A. Lorigan Department of Chemistry and Biochemistry, Miami University, Oxford, OH 45056, USA

Membrane proteins (which make up approximately one-third of the total number of known proteins) are responsible for many important properties and functions of biological systems: they transport ions and molecules across the membrane, they act as receptors, and they have roles in the assembly, fusion, and structure of cells and viruses. Despite the abundance and clear importance of membrane-associated molecules, very little information about these systems exists. A plethora of membrane proteins are intimately associated with cardiovascular function and disease. Structural studies of these membrane proteins represent one of the final frontiers in structural biology. X-ray crystallography is the premiere technique that is used to elucidate structural information of biologically significant protein systems. However, this technique has not been very successful in providing structural information about membrane protein systems. The hydrophobic surfaces associated with membrane-bound protein systems make the crystallization process extremely difficult. Although researchers are making progress with X-ray techniques, still only a handful of membrane protein structures have been obtained via X-ray crystallography [1–5]. Alternatively, solution NMR spectroscopy, solid-state NMR spectroscopy, and EPR spectroscopy are powerful techniques that can be used to provide structural, orientational, and dynamic information about membrane protein systems in lipid bilayers [6–10]. This short review chapter will analyze some of the magnetic resonance techniques that have been used to investigate the integral membrane protein phospholamban (PLB).

Phospholamban The contraction/relaxation cycle intimately associated with cardiac muscle cells is regulated by cytosolic levels of Ca2+ ions. In order for cardiac muscle cells to relax after a contraction, Ca2+ ions must be rapidly transferred between the cytosol and the cardiac sarcoplasmic reticulum (SERCA) lumen. The transfer of Ca2+ ions is performed by the Ca-ATPase of the SERCA [11–14]. This unique pumping mechanism is activated by the cyclic Graham A. Webb (ed.), Modern Magnetic Resonance, 313–318.
C 2008 Springer.

AMP- and calmodulin-dependent phosphorylation of the integral membrane protein PLB [15–17]. Dephosphorylated PLB inhibits SERCA ATPase activity and stops the flow of Ca2+ ions. Conversely, when PLB is phosphorylated this inhibition is relieved, and Ca2+ ions are transferred through the membrane. This unique process controls the heartbeat of the cardiac cycle. The rate and extent of myocardial contraction is determined by the flow rate of Ca2+ ions into the myoplasm. Recent studies have suggested that an abnormal relaxation of cardiac muscle cells can induce heart failure, due to abnormalities in Ca2+ transients and decreases in Ca-ATPase concentrations [18,19]. PLB is a small (52 amino acid) type II membrane protein and shares many characteristics with some of the larger mammalian ion channels [20]. The size of PLB makes it an ideal candidate to investigate with both NMR and EPR spectroscopy. Determining the structure of PLB and its interaction with lipid bilayers is central to understanding its regulatory role. The full three-dimensional structure of WT-PLB in either the phosphorylated or dephosphorylated states has not been determined in a phospholipid bilayer. Sequence homology studies have indicated that the protein consists of three domains: a hydrophilic amphipathic Nterminus (1–21) MDKVQYLTR SAIRRASTIEMP section, a hinge or β-sheet region QQARQNLQN (22–30), and a hydrophobic (31–52) C-terminus LFINFCLILICLLLICIIVMLL segment that spans the bilayer. WTPLB is believed to consist of a homopentameric cluster, which retains activity when reconstituted into lipid bilayers with Ca-ATPase [21,22]. The structural characteristics of PLB have been investigated with several biophysical spectroscopic techniques including: CD spectroscopy, EPR spectroscopy, IR spectroscopy, and NMR spectroscopy in a variety of different environments (organic solvents, micelle, and phospholipid bilayer). CD and solution NMR studies carried out on the hydrophilic cytoplasmic domain of PLB in organic solvents have identified a partial α-helical structure [23–27]. The α-helical secondary structure of the cytoplasmic domain has been confirmed with CD and IR and extended further to include the transmembrane segments of PLB [28,29]. The structure of PLB may change when the full-length protein

Part I

Magnetic Resonance Spectroscopic Studies of the Integral Membrane Protein Phospholamban

314 Part I

Chemistry

Part I Fig. 1. (Left) Structure of PLB inserted into a phospholipid bilayer. (Right) Additional structural model of PLB inside a phospholipid bilayer. The structures are shown as monomers for clarity. WT-PLB is believed to exist as a pentamer. (See also Plate 35 on page XXXI in the Color Plate Section.)

is placed into a proper functional lipid environment such as a phospholipid lipid bilayer. Also, the structure may be modified by its interaction with Ca-ATPase which is required for regulatory function. Two structural models have been proposed for PLB based upon FTIR, NMR, and computer data/modeling (Figure 1). Tatulian et al. have indicated that PLB consists of two disjointed helices: a transmembrane helix that is parallel with the bilayer normal and a tilted helix that extends outside the membrane (Figure 1 (left side)) [30]. The two helices are connected by a small intervening βsheet/unstructured region. Another model (not shown) proposes that PLB is a continuous α-helix in which both the transmembrane and cytosolic elements are oriented at a tilt angle of 28◦ ± 6◦ with respect to a DMPC lipid bilayer [28]. According to this model, PLB is one long straight α-helical structure in which the hydrophobic portion is located within the membrane and the hydrophilic region lies outside the membrane. Analysis of all the different biophysical studies of WT-PLB, indicates that the structure and helix orientation of phosphorylated and dephosphorylated pentameric PLB with respect to the membrane is under debate. Phosphorylation serves as the regulatory switch for PLB. It confers protease resistance, indicating a possible structural change in PLB [29]. Upon β-adrenergic stimulation, PLB is phosphorylated at sites Ser-16 and Thr-17 concomitant with the flow of Ca2+ ions [13]. Once again, a discrepancy exists in the literature as to whether a change occurs in the secondary structure of phosphorylated PLB. Specifically, fluorescence, FTIR, and solution NMR measurements have observed secondary structural changes on the full-length and segmented portions of PLB, while computer modeling and additional FTIR and CD studies have not [11,12,26,29–31]. Recent solution NMR studies

on the cytoplasmic portion of PLB (residues 1–36) in trifluoroethanol have indicated that phosphorylation does not adversely affect the structure of the C-terminus between residues 21 and 36, and that phosphorylated PLB has more loose helical packing than the nonphosphorylated version of the protein.

Solid-State NMR Spectroscopic Studies of PLB Several research groups are studying the structural and dynamic properties of PLB utilizing NMR spectroscopy [14,23,32–44]. The structural properties of PLB have been studied utilizing the rotational echo double resonance (REDOR) and the rotational resonance solid-state NMR techniques. Solid-state NMR spectroscopic studies on PLB utilizing the rotational resonance method have indicated that the sequences Pro21-Ala24 and Leu42-Leu44 adopt an α-helical structure in pure lipid bilayers, in the presence and absence of Ca-ATPase [39]. Additional REDOR NMR experiments have revealed that the sequence Ala24Gln26 switches from an α-helix in pure lipid membranes to a more extended structure in the presence of SERCA [39,44]. The data gleaned from this study suggest that the Ca2+ -ATPase has a long-range effect on the structure of PLB around residue 25, which promotes the functional association of the two proteins. Additional rotational resonance NMR data have shown that internuclear 13 C distances between Leu7 and Ala11 in the cytoplasmic region, between Pro21 and Ala24 in the juxtamembrane region, and between Leu42 and Cys46 in the transmembrane domain of PLB all consist of an α-helical secondary structure [41]. REDOR experiments agree that the secondary structure is α-helical in the region of Pro21 and that there are no large conformational changes upon phosphoryla-

Magnetic Resonance Spectroscopic Studies of the PLB

Part I

(A) Leu28

Solid-State NMR Spectroscopic Studies of PLB 315

(C) Leu51

(B) Leu39

60 °C

45 °C

25 °C

60 °C

60 °C

45 °C

45 °C

25 °C

25 °C

0 °C

0 °C

0 °C

−25 °C

−25 °C

−25 °C

40

20 2

0 −20 H (kHz)

−40

40

20 2

0 −20 H (kHz)

−40

40

20 2

0 −20 H (kHz)

−40

Fig. 2. 2 H NMR powder pattern spectra of L-leucine-5,5,5-d3 incorporated at specific sites of TM-PLB and inserted into POPC phospholipid bilayers. 2 H NMR spectra are shown for (A) CD3 -Leu28 PLB, (B) CD3 -Leu39 PLB, and (C) CD3 -Leu51 PLB incorporated into POPC phospholipid bilayers at lipid/peptide molar ratios of 25:1.

tion [41]. The data indicate that PLB exists as homogenous α-helical pentamer. Additionally, the side chain and backbone dynamic properties of PLB have been investigated utilizing solidstate NMR spectroscopy. Figure 2 shows the solid-state 2 H NMR powder pattern spectra of specific labeled TM-PLB (transmembrane section of PLB consisting of residues Ala24-Leu52) samples incorporated into unoriented 1palmitoyl-2-oleoyl-phosphocholine (POPC) bilayers as a function of temperature [33]. 2 H solid-state NMR spectra of 2 H-labeled leucine (deuterated at one terminal methyl group) incorporated at different sites (CD3 -Leu28,

CD3 -Leu39, and CD3 -Leu51) along the TM-PLB peptide exhibited line shapes characteristic of either methyl group reorientation about the Cγ–Cδ bond axis, or by additional librational motion about the Cα–Cβand Cβ–Cγ bond axes. The 2 H NMR line shapes of all −CD3 labeled leucines are very similar below 0 ◦ C, indicating that all the residues are located inside the lipid bilayer. At higher temperatures, all three labeled leucine residues undergo rapid reorientation about the Cα–Cβ, Cβ–Cγ and Cγ–Cδ bond axes as indicated by 2 H line shape simulations and reduced quadrupolar splittings. At all the temperatures studied, the 2 H NMR spectra indicated that the Leu51

Chemistry

Magnetic Resonance Spectroscopic Studies of the AFA-PLB Monomer Recently, some exciting new NMR and EPR spectroscopic results have been obtained on a mutated version of PLB (AFA-PLB) that predominantly exists as a monomer [27,36,39,41,45,48,49]. AFA-PLB is obtained by mutating the three Cys residues (36, 41, and 46) to Ala, Phe, and Ala. This mutated PLB monomer has been shown to be fully functional [50]. One of the biggest breakthroughs has been the solution NMR structure of AFA-PLB by the Veglia research group. His research group has determined an “L-shaped” α-helical structure of the mutated monomeric version of PLB in dodecylphosphocholine micelles [38]. Figure 3 shows a well-resolved HSQC solution NMR spectrum of uniformly 15 N-labeled AFA-PLB with assignments. The spectrum was kindly provided by Dr. Gianluigi Veglia. The structure on the left hand side of Figure 1 illustrates the “L-shape” structure of AFA-PLB inserted into a phospholipid bilayer. Additional solid-state NMR research conducted on site-specific 15 N-labeled AFA-PLB has shown that the monomeric form of PLB has one component that is nearly transmembrane (hydrophobic segment, residues 31–52) and the amphipathic segment lies on the surface of the membrane [51,52]. The solid-state NMR data coupled with molecular dynamic simulations estimates that the monomeric transmembrane helix makes an approximate 10◦ angle with respect to the bilayer normal. NMR and EPR structural dynamic studies have indicated that PLB involves functionally important transitions

S16

111

T17

T8

N30 N27

113 S10

115 Shift (ppm)

Part I

side chain has less motion than Leu39 or Leu28 which is attributed to its incorporation in the pentameric PLB leucine zipper motif. The unique features associated with these spectra could be explained by the condition that the Leu51 side chain is involved in the so-called “knobsinto-holes” bonding arrangement in which the side chain of Leu51 from one α-helix is locked into the groove of a second α-helix (of the pentamer) so that the two α-helices are coiled around each other. Smith and co-workers have also reported that Leu44 is buried within the core of pentameric PLB and also observed the general bell-shaped characteristics in the line shape [45]. This type of structural arrangement is believed to help stabilize the structure of the pentamer [41]. 31 P NMR spectra of these samples indicate that TM-PLB is incorporated into phospholipid bilayers in the liquid crystalline (Lα) phase [46]. Additional studies have probed the interaction of PLB with the membrane utilizing spin-label EPR spectroscopy [40]. Finally, 15 N solid-state NMR studies have indicated that TM-PLB is transmembrane with respect to the membrane normal [47].

117

15N

316 Part I

119

L37 V49

121 E2

123

I47

M50

N30 I12

F32 I45 M51 N34 L42 I40 I38 Q29 N27 R25 V4 F35 I33 L43 Q23 Q26 Y6 L44 I48 L52 F41 R14/L39 K3 L7 R13 R9 A36 Q5 L31 I18 L28 A46 Q22 A15 M20 A24 E19 A11

8.9

8.5

8.1 1

7.7

7.3

6.9

H Shift (ppm)

Fig. 3. HSQC solution NMR spectrum of 15 N-labeled AFAPLB in dodecylphosphocholine micelles. The NMR spectrum was kindly provided by Dr. Gianluigi Veglia.

among potentially multiple structural states and that the structure of PLB is affected by its phosphorylation and its interaction with Ca-ATPase [49,51]. Primarily, the Thomas lab has studied backbone dynamics with EPR spectroscopy of specific-site 2,2,6,6,-tetramethylpiperidine-N -oxyl-4-amino-4-carboxylic acid (TOAC) attached at positions 0, 11, and 24 in the cytoplasmic domain or at position 46 in the transmembrane domain of AFA-PLB. The EPR spectrum of the AFA-PLB with a TOAC label at position 46 reveals a single broad component, indicating an ordered transmembrane helix. However, the cytoplasmic labels reveal two separate spectral EPR components. One of the components is disordered and nearly isotropic with dynamic motion on the ns timescale, while the second spectral component is more ordered and undergoing slower motion on the EPR timescale. The results indicate that the cytoplasmic domain of the AFA-PLB monomer is in dynamic equilibrium between an ordered confirmation (buried within the membrane) and a motionally dynamic form that is detached from the membrane and poised to interact with its regulatory target [50]. A model of AFA-PLB between these two states can be seen by examining the two structural

Magnetic Resonance Spectroscopic Studies of the PLB

Ack nowledgments GAL would like to acknowledge his research group and Professor Gianluigi Veglia for all of their help in preparing this review chapter. This work was supported by a National Science Foundation CAREER Award (CDE-0133433), an NIH Grant (GM60259-01), and an American Heart Association Scientist Development Grant (0130396). The 500 MHz Wide bore NMR Spectrometer was obtained from NSF Grant (#10116333).

References 1. Agre P, Lee MD, Devidas S, Guggino WB. Science. 1997;275(5305):1490. 2. MacKinnon R, Cohen SL, Kuo AL, Lee A, Chait BT. Science. 1998;280(5360):106. 3. MacKinnon R. Nature. 2002;416(6878):261. 4. Doyle DA, Cabral JM, Pfuetzner RA, Kuo AL, Gulbis JM, Cohen SL, Chait BT, MacKinnon R. Science. 1998;280(5360): 69. 5. Jiang YX, Lee A, Chen JY, Ruta V, Cadene M, Chait BT, MacKinnon R. Nature. 2003;423(6935):33. 6. Opella SJ. Nat. Struct. Biol. 1997;4(Suppl):845. 7. Opella SJ, Nevzorov A, Mesleh MF, Marassi FM. Biochem. Cell Biol. 2002;80(5):597. 8. Opella SJ, Marassi FM. Chem. Rev. 2004;104(8):3587. 9. Cross TA, Opella SJ. Curr. Opin. Struct. Biol. 1994;4(4):574. 10. Smith SO, Peersen OB. Annu. Rev. Biophys. Biomol. Struct. 1992;21:25. 11. Li M, Cornea RL, Autry JM, Jones LR, Thomas DD. Biochemistry. 1998;37(21):7869. 12. Cornea RL, Jones LR, Autry JM, Thomas DD. Biochemistry. 1997;36(10):2960. 13. Simmerman HKB, Jones LB. Physiol. Rev. 1998;78(4):921. 14. Li JH, Xiong YJ, Bigelow DJ, Squier TC. Biochemistry. 2004;43(2):455. 15. Kirchberger MA, Tada M, Katz AM. Rec. Adv. Cardiac. Struct. Metab. 1975;5:103. 16. James P, Inui M, Tada M, Chiesi M, Carafoli E. Nature. 1989;342:90. 17. Voss J, Jones LR, Thomas DD. Biophys. J. 1994;67(1):190. 18. Grossman W. In RG Johnson Jr, EG Kranias (Eds). Cardiac Sarcoplasmic Reticulum Function and Regulation of Contractility, Vol. 853. New York Academy of Sciences: New York, 1998, p 207.

19. Lehnart SE, Wolfgang S, Burkert P, Prestle J, Just H, Hasenfuss G. In RG Johnson Jr, EG Kranias (Eds). Cardiac Sarcoplasmic Reticulum Function and Regulation of Contractility, Vol. 853. New York Academy of Sciences: New York, 1998, p 220. 20. Tada M. Ann. N.Y. Acad. Sci. 1992;671:92. 21. Adams PD, Arkin IT, Engelman DM, Brunger AT. Nat. Struct. Biol. 1995;2(2):154. 22. Kovacs RJ, Nelson MT, Simmerman HBK, Jones LR. J. Biol. chem. 1988;263:18364. 23. Pollesello P, Annila A. Biophys. J. 2002;83(1):484. 24. Pollesello P, Annila A, Ovaska M. 1999;76(4):1784– 1795. 25. Hubbard JA, Maclachlan LK, Meenan E, Salter CJ, Reid DG, Lahouratate P, Humphries J, Stevens N, Bell D, Neville WA, Murray KJ, Darker JG. Mol. Membr. Biol. 1994;11: 263. 26. Mortishiresmith RJ, Pitzenberger SM, Burke CJ, Middaugh CR, Garsky VM, Johnson RG. Biochemistry. 1995;34(23):7603. 27. Lamberth S, Schmid H, Muenchbach M, Vorherr T, Krebs J, Carafoli E, Griesinger C. Helv. Chim. Acta. 2000;83(9): 2141. 28. Arkin IT, Rothman M, Ludlam CFC, Aimoto S, Engelman DM, Rothschild KJ, Smith SO. J. Mol. Biol. 1995;248(4): 824. 29. Arkin IT, Adams PD, Brunger AT, Smith SO, Engelman DM. Annu. Rev. Biophys. Biomol. Struct. 1997;26:157. 30. Tatulian SA, Jones LR, Reddy LG, Stokes DL, Tamm LK. Biochemistry. 1995;34(13):4448. 31. Ludlam CFC, Arkin IT, Liu XM, Rothman MS, Rath P, Aimoto S, Smith SO, Engelman DM, Rothschild KJ. Biophys. J. 1996;70(4):1728. 32. Mascioni A, Eggimann BL, Veglia G. Chem. Phys. Lipids. 2004;132(1):133. 33. Tiburu EK, Karp ES, Dave PC, Damodaran K, Lorigan GA. Biochemistry. 2004;43(44):13899. 34. Becker CFW, Strop P, Bass RB, Hansen KC, Locher KP, Ren G, Yeager M, Rees DC, Kochendoerfer GG. J. Mol. Biol. 2004;343(3):747. 35. Middleton DA, Hughes E, Madine J. J. Am. Chem. Soc. 2004;126(31):9478. 36. Metcalfe EE, Zamoon J, Thomas DD, Veglia G. Biophys. J. 2004;87(2):1205. 37. Dave PC, Tiburu EK, Damodaran K, Lorigan GA. Biophys. J. 2004;86(3):1564. 38. Zamoon J, Mascioni A, Thomas DD, Veglia G. Biophys. J. 2003;85(4):2589. 39. Hughes E, Middleton DA. J. Biol. Chem. 2003;278(23):20835. 40. Arora A, Williamson IM, Lee AG, Marsh D. Biochemistry. 2003;42(17):5151. 41. Smith SO, Kawakami T, Liu W, Ziliox M, Aimoto S. J. Mol. Biol. 2001;313(5):1139. 42. Sharma P, Patchell VB, Gao Y, Evans JS, Levine BA. Biochem. J. 2001;355:699. 43. Middleton DA, Ahmed Z, Glaubitz C, Watts A. J. Magn. Reson. 2000;147(2):366. 44. Ahmed Z, Reid DG, Watts A, Middleton DA. Biochim. Biophys. Acta Biomembr. 2000;1468(1–2):187.

Part I

configurations of PLB located within Figure 1. This TOAC EPR spin-label method provides excellent backbone dynamic information similar to 15 N amide solid-state NMR dynamic studies. Future NMR and EPR spectroscopic studies that probe the structural and dynamic properties of both the pentameric and monomeric forms of PLB and how they interact with SERCA and PKA are needed to help to better understand regulatory function.

References 317

318 Part I

Chemistry

Part I

45. Ying WW, Irvine SE, Beekman RA, Siminovitch DJ, Smith SO. J. Am. Chem. Soc. 2000;122(45):11125. 46. Dave PC, Tiburu EK, Damodaran K, Lorigan GA. Biophys. J. 2004;86:1564. 47. Tiburu EK, Dave PC, Damodaran K, Lorigan GA. Biochemistry. 2004;43(44):13899. 48. Li HM, Cocco MJ, Steitz TA, Engelman DM. Biochemistry. 2001;40(22):6636.

49. Kirby TL, Karim CB, Thomas DD. Biochemistry. 2004; 43(19):5842. 50. Karim CB, Kirby TL, Zhang ZW, Nesmelov Y, Thomas DD. Proc. Natl. Acad. Sci. U.S.A. 2004;101(40):14437. 51. Mascioni A, Karim C, Zamoon J, Thomas DD, Veglia G. J. Am. Chem. Soc. 2002;124(32):9392. 52. Mascioni A, Karim C, Barany G, Thomas DD, Veglia G. Biochemistry. 2002;41(2):475.

319

David A. Middleton Faculty of Life Sciences, University of Manchester, Sackville Street, P.O. Box 88, Manchester M60 1QD, UK

Abbreviations: REDOR, rotational echo double resonance; CP-MAS, cross-polarization magic-angle spinning; et-NOESY, exchange transferred nuclear Overhauser effect spectroscopy; STD, saturation transfer difference spectroscopy; DR, dipolar recoupling; waterLOGSY, water-ligand observation by gradient spectroscopy; MAOSS, magic angle oriented sample spinning; bR, bacteriorhodopsin; nAChR, nicotinic acetylcholine receptor.

Introduction The interactions between ligands and their receptors lie at the heart of many of the complex cascades of cellular events responsible for life and death, disease and therapy. The outcomes of these events depend upon the selectivity and affinity of natural agonists, antagonists, modulators, and inhibitors for their physiological targets, since it is only through specific interactions that the correct biological signals are generated and processed. Similarly, the therapeutic/toxicological ratios of synthetic drug compounds often hinge on their fidelity for a specific receptor sub-type against a background of closely related receptors. Recent technological advances in drug discovery have led to the wide availability of sophisticated methods for identifying natural or synthetic ligands of specific receptors with high throughput and sensitivity. When these methods are combined with combinatorial chemistry vast numbers of compounds can be screened for ligand activity in a fraction of the time taken 20 years ago [1]. Despite this level of progress, there remains a demand for lower-throughput techniques which can examine receptor–ligand interactions beyond the phenomenological and provide mechanistic information at the molecular level. The NMR community, in particular, has risen to the challenge presented by the revolution in drug discovery and the past decade has witnessed the arrival of many exciting new methods. The versatility of NMR makes it an attractive technique capable of addressing many aspects of drug discovery from screening weak interactions of Graham A. Webb (ed.), Modern Magnetic Resonance, 319–326.
C 2008 Springer.

ligands through to the structural assessment of receptor– ligand complexes. Over half of the targets for currently marketed drugs are proteins that function within the cell membrane [2] and the pipeline of drugs in clinical development suggests that this proportion will rise significantly in the future. There is, therefore, a compelling case for obtaining molecular level details about how ligands interact with membraneembedded receptors. Nevertheless, such information remains scarce owing to the insoluble nature of proteins in a lipid membrane environment, which has hampered crystallographic studies and, until recently, has precluded analysis by NMR. Recent progress in NMR hardware and methodology development has been astonishing, however, and the first details of how ligands interact with their receptors in a membrane environment are now being revealed to resolution unattainable by diffraction techniques. This review will give a brief account of some of the recent developments in NMR in both the solid-state and solution-state and comment on future prospects of these developments for drug discovery.

Choice of Technique The majority of the physiological and pharmacological ligands of relevance here are water-soluble small molecules that bind reversibly to the receptor embedded in what is effectively a solid support. The affinity of the ligand for the receptor is defined by the on-rate kon and off-rate koff of the ligand (Figure 1). If the association of a ligand with its receptor is diffusion controlled, kon is typically on the order of 107 M−1 s−1 and the dissociation constant (K D = koff /kon ) of the receptor–ligand complex is dependent largely on the off-rate. Low-affinity interactions (K D ≥ 1 mM) therefore usually involve rapid dissociation of the ligand from the binding site and the ligands are classified as weak. In such cases, solution NMR methods are advantageous because the ligand can be observed in solution whilst exploiting various physical properties of the ligand that are modulated by its transitory

Part I

NMR Studies of the Interactions Between Ligands and Membrane-Embedded Receptors: New Methods for Drug Discovery

Chemistry

Part I

Fig. 1. A schematic representation of the time-course of the interaction between a ligand L (red circles) and a membraneembedded receptor R (cups). The dissociation constant for the receptor–ligand complex K D is given by the ratio of the on-rate and off-rate, koff /kon . The graph shows the percentage of ligand molecules that are predicted to be bound to a receptor at equilibrium over a range of K D values, when the ligand and receptor are present in equimolar concentrations. The off-rate corresponding to each K D value is shown on the upper axis, based on the assumption that association is diffusion limited with an on-rate of 107 M−1 s−1 .

10−1 percentage site occupancy

320 Part I

koff (s−1) 101 103

membrane

105

100 80

L+R

60

k on

L.R

k off 40 20 0

10−8

10−6

association with the receptor (e.g. transferred NOEs, relaxation, or saturation). A corollary of this approach, which is implicit in the graph in Figure 1, is that the ligand must be present in large excess over the receptor in order to saturate the available binding sites. Such a large excess of ligand may promote non-specific binding, which can obscure the pharmacologically relevant interactions or lead to misinterpretation. Appropriate control experiments, usually involving competitive displacement, must therefore be designed to eliminate these effects. Higher affinity ligands (K D ≤ 10 µM) usually undergo slower dissociation from their targets (Figure 1) and cannot be detected directly by conventional solution NMR methods because the resonance lines are broadened as a result of the slow tumbling of the membrane assembly. In this motional regime, solid-state NMR is a more appropriate technique for detecting and characterizing the bound ligand directly. The graph in Figure 1 implies that, by adding an equimolar concentration of highaffinity ligand to a receptor, over 90% of the ligand will be bound. The sample can be frozen to eliminate interference from molecular dynamics and very precise structural data can be extracted with appropriate solid-state NMR techniques.

Solution NMR Methods In recent years a variety of 1 H NMR methods have been developed for screening weak ligand interactions with receptors as well as for determining sites of interaction, identifying ligands from compound mixtures, mapping interfacial sites and providing structural contraints [3].

10−4 KD (M)

10 −2

Some of the most notable amongst these are saturation transfer difference (STD) spectroscopy, water-ligand observations with gradient spectroscopy (waterLOGSY), and transferred NOE methods. In the STD approach, the 1 H magnetization from the protein target is irradiated at low power by a shaped pulse at a frequency well away from the ligand resonance. The saturation effect is, in turn, transferred to a ligand for the duration that it is associated with the binding site within the receptor. For a ligand in fast exchange between free and bound environments, the saturation is carried with the ligand when it dissociates from the receptor into the free state. If two spectra are obtained, one with and one without saturating pulses, the peaks from the ligand can be identified from the complex spectrum of the mixture [4]. This approach has been exploited for the analysis of carbohydrates to provide details of off-rates and binding constants, map the sites of ligands forming the interface with the receptor and identify ligands from mixtures of compounds [3]. The waterLOGSY technique is similar in concept but relies on the use of water to detect receptor-bound ligands, by generating negative water-ligand NOEs after saturating the water proton magnetization [5]. The bound conformations of fast exchanging ligands (i.e. koff > 1/T1 ) have been explored using exchange-transferred NOE (et-NOESY) experiments [6]. The strong NOES that develop when the ligand is bound to the receptor are transferred with the ligand to the free state from where they are measured. The experiments described above are valid only when the ligands bind weakly to the receptor and can be observed in solution. In the case of more tightly bound ligands, dissociation constants have been estimated indirectly with the development of relaxation based NMR

New Methods For Drug Discovery

Solid-State NMR Methods Solid-state NMR embraces a collection of diverse methods that take advantage of the spatial and dynamic properties of biomembranes to extract information about structure and dynamics. Such methods may exploit the positions of the spectral lines arising from the incomplete averaging of anisotropic interactions (chemical shielding, dipole–dipole, quadrupolar) within the membrane components. Alternatively, the technique of cross-polarization magic-angle spinning (CP-MAS) is used to eliminate or reduce the effects of restricted anisotropic motions and susceptibility effects in biomembranes to gain highresolution spectra from which site-specific information can be gained with the appropriate pulse sequences.

Sample Requirements In the analysis of biomembrane samples by CP-MAS NMR methods it is generally desirable that the receptor is fully functional and present in its native membrane or isolated and incorporated into a new lipid bilayer [11]. In both cases, the ligand is titrated into the hydrated membranes and the NMR sample is prepared as a gelatinous pellet by centrifugation to maximize the receptor density that can be loaded into the sample spinner. The receptor purity (and concentration) that can be achieved is often the key factor in determining which experiments are feasible. For example, Spooner et al. (see below) have reported various CP-MAS studies of the interactions between bacterial transport proteins and their natural substrates in native membranes in which the receptor of interest represents less than 60%, and as little as 20%, of the total membrane protein. Such studies were possible because of the excellent selectivity of the substrates for their receptors

and the high expression levels of protein achieved, which alleviated the requirement for difficult and inefficient purification procedures. In cases where the ligands are less specific or the protein of interest is in low abundance, however, it may be desirable to manipulate the receptor to achieve a higher level of purity. An alternative approach to CP-MAS involves measuring the orientation of anisotropic interactions (dipolar, chemical shielding, and quadrupolar) in the magnetic field under static (i.e. non-rotating) conditions to determine the alignment of bonds, functional groups or domains of the ligand relative to the receptor. Here, the hydrated membrane samples are laid down onto glass plates of dimensions appropriate for the NMR radiofrequency coil in such a way that the membrane components adopt a uniaxial alignment with respect to the magnetic field. Various methods for obtaining membrane alignment have been proposed, but most have achieved limited success in aligning receptors in native or reconstituted membranes. One apparently successful approach is the isopotential spin-dry ultracentrifugation technique, which produces good alignment whilst preserving biological integrity of the membrane sample [12]. A historical limitation of solid-state NMR has been its inability to exploit protons directly because the strong couplings between them lead to severe line broadening. Although new measures are being taken to eliminate proton line broadening, the observation of naturally rare nuclei (13 C, 15 N, 2 H, 19 F) in isotope enriched samples remains central to most biological solid-state NMR experiments.

Detection of Ligand Binding by CP-MAS NMR In conventional solid-state CP-MAS NMR, HartmannHahn cross-polarization serves to enhance the signal from the observed low-gamma nucleus and reduce recycle times, by transferring magnetization from protons which have higher sensitivity and more favorable T1 relaxation times. Spooner et al. [13] demonstrated an alternative use for CP-MAS in which interactions between 13 C labeled L-glucose and the bacterial sugar transporter GalP could be detected by Hartmann-Hahn cross-polarization in partially purified E. coli membrane samples at 4 ◦ C. Using CP-MAS it is possible to discriminate between free and bound substrate molecules because of the differences in the motional characteristics of substrate in the two environments. The CP-MAS method detects ligand binding only when the membrane sample is in a fluid state, allowing weak receptor–ligand interactions to be characterized under physiologically relevant conditions that are intractable to other methods of analysis. This development provided the incentive for further studies in which bacterial transporters were nitroxide spin-labeled at unique

Part I

experiments that measure the relaxation rates of a standard ligand of known affinity for a receptor before and after displacement with the ligand of interest [7]. The applications of these and other solution NMR methods to membrane embedded proteins have so far been limited to a few examples. Weak interactions between a cyclic peptide inhibitor of the membrane-spanning protein integrin incorporated into liposomes and living cells have been detected using STD NMR [8] and et-NOESY [9]. The latter technique was also used to determine the conformation of ligands bound to the muscarinic acetylcholine receptor and showed that the bound conformations of muscarine and methacholine were different from their conformations in solution [10]. The full potential of these methods remains to be tested, however, and it is anticipated that many more examples of applications to membrane receptor–ligand interactions will appear in the literature.

Solid-State NMR Methods 321

322 Part I

Chemistry

Part I

[13C]ouabain and Na+/K+-ATPase

[13C]glucuronide and GusB

O

O

HOOC HO

O

HO

OMe

HO HO H

O

*

HO

CH3

OH

H3C

OH

O CH3

O

*

H3C

200

150 13

100

50

0

200

C chemical shift (ppm) 1.0

100

50

0

C chemical shift (ppm)

0.6 0.4

KD ~ 4 mM

0.2

6 mM 3 mM

intensity

intensity

6 mM 3 mM

0.8

0.0 0.000

HD3

1.2

1.2 1.0

150 13

O

0.005

0.010

contact time (s)

0.8 0.6 0.4 0.2 0.0 0.000

KD < 600 nM 0.005

0.010

contact time (ms)

Fig. 2. A demonstration of how 13 C CP-MAS NMR is used to estimate the affinities of ligands for membrane-embedded receptors. Examples of spectra (top) and cross-polarization build-up curves (bottom) are shown for two cases in which the ligands have different affinities for their receptors. The spectrum and peak intensities for [13 C]methylglucuronide interacting with GusB in E. coli membranes is shown on the left. On the right are shown data for [13 C]acetonidoouabain interacting with Na+ /K+ -ATPase. In the graphs, the experimental peak intensity values at two ligand concentrations are shown as circles and squares and the lines represent the curves of best fit, corresponding to the K D values shown.

cysteine residues in order to locate residues close to the substrate binding site. It was shown that the strong magnetic dipole of the electron spin broadened the peaks from the bound substrates and permitted distances between the labeled residues and the bound ligand to be estimated [14–15]. Recently, the CP-MAS approach has been extended to provide quantitative information about ligand binding affinities for receptors in fluid membranes. In this new development, peak intensities from different concentrations of an isotope labeled ligand in the presence of a membrane receptor are measured at increasing crosspolarization contact times [16]. The build up of peak intensities at different ligand concentrations is highly sensitive to binding affinity (Figure 2) and the experimental data are compared with simulated curves to extract values of K D .

One advantage of this method is that the K D value can be checked independently by measuring the peak intensities after displacement of the labeled ligand by titration of unlabeled ligand. Another attractive feature of this method is that the cross-polarization procedure can be “tuned” to eliminate the signal from non-specifically bound ligand. To date, K D values have been measured for the interactions of glucuronide sugars with the bacterial transport protein GusB [16], ouabain analogs with the Na+ /K+ ATPase [17] (Figure 2) and the antidepressant drug trifluoperazine (TFP) with gastric H+ /K+ -ATPase [18]. In the latter work, the intrinsic 19 F signal of TFP was exploited to measure ligand binding affinities. All of the examples highlighted above rely upon being able to resolve the peaks for the labeled ligand from the natural abundance signals from the lipids and protein. In

New Methods For Drug Discovery

Structural Analysis of Ligands In the solid-state, structural information comes in the form of internuclear distances, torsional angles and bond orientations relative to a fixed reference frame. The data is extracted using variants of the CP-MAS experiment, from uniformly aligned membranes or using a combination of both methods called magic angle oriented sample spinning (MAOSS) [20]. In the case of CP-MAS, structural measurements rely upon a variety of dipolar recoupling (DR) experiments, which manipulate the nuclear spin systems to restore the weak, but structurally informative, dipolar interactions that are otherwise removed by sample rotation [21]. Many such experiments have been devised including rotational resonance NMR, which restores and measures homonuclear couplings, REDOR for heteronuclear couplings and a range of experiments to measure H–C–C–H, H–C–N–H, and N–C–C–N torsional angles [22]. Early applications of solid-state NMR to membrane protein–ligand interactions focused on the structure and orientation of the retinal chromophore within the proton pump bacteriorhodopsin (bR) of the bacterium H. salarium. Rotational resonance experiments on double 13 C labeled retinal in bR [23] confirmed the torsional angle defining the relative orientations of the β-ionone ring and the polyene chain. Deuterium NMR experiments on bR in aligned membranes revealed changes in quadrupolar tensor orientations for 2 H labels in a methyl group (position 19) of the retinal polyene chain after photoexcitation of bR to the M-state, which was consistent with isomerization about the C13–C14 double bond [24]. Further measurements of angles along the polyene chain using DR CP-MAS methods have resolved a slightly distorted conformation of retinal in the binding pocket and resolved ambiguities about bond configurations in the various photointermediates that were not evident from the crystal structures [25]. Recently, sophisticated multidimensional

DR experiments have been devised to examine dipolar contacts between the retinal Schiff base and residues lining the binding pocket of bR [26]. Similar methods have been applied to examine the structure of the retinal chromophore in the G-protein coupled receptor (GPCR) rhodopsin during the photocycle responsible for visual signal transduction. Studies of the orientation of retinal using MAOSS revealed significant changes in the orientation of the β-ionone ring in the first stages of the photoexcitation cycle before major protein conformational changes occur [27]. More recently, Smith and co-workers used two-dimensional dipolar-assisted rotational resonance experiments to identify couplings between 13 C labels in retinal and labels in Tyr, Gly, Ser, and Thr residues around the binding pocket [28]. It was shown that the transition of rhodopsin to the MII state involves the disruption of helix interactions in the transmembrane ˚ toward helix 5 and the domain as retinal moves some 5 A C20 methyl group of retinal moves toward extracellular loop 2. The studies of retinal in bR and rhodopsin are excellent demonstrations of how solid-state NMR is ideally suited to filling in details about ligand structure and orientation that are beyond the resolution of available crystal structures. A similar case is the interaction of acetylcholine with the nicotinic acetylcholine receptor (nAChR), the ligand-gated cation channel that mediates synaptic transmission. The highest resolution structure of ˚ and inadequate to visualize the ponAChR is at about 4 A sitions and structure of the natural ligand in the binding sites. Solid-state NMR studies of a derivative of the natural ligand, bromoacetylcholine, bound covalently to the receptor purified from torpedo electroplax have helped to provide clues about how the ligand interacts with the binding pocket [reviewed in Ref. 29]. Deuterium NMR experiments on uniformly aligned membranes containing [2 H]bromoacetylcholine indicated that the ligand is positioned in the binding site with the quaternary ammonium group facing outwards and oriented at about 40◦ with respect to the membrane normal. Further, CP-MAS spectra of the bound 13 C labeled ligand showed an up-field perturbation in chemical shift relative to the free ligand in solution which was consistent with a ring current effect of aromatic residues that are believed to line the binding pocket. The examples described above all benefit from the availability of receptor crystal structures of varying resolution with which to combine NMR measurements to draw conclusions about ligand binding. The following section will show examples of the less favorable, but unfortunately common, situation in which mechanistic information about ligand binding is obtained by solid-state NMR methods in the absence of any crystallographic coordinates for the receptor.

Part I

most of the published experiments on 13 C labeled substrates of bacterial transporters, it is by design that the peaks from the ligand lie in a region of the spectrum (90–110 ppm) uncontaminated by background signal. To extend the applicability of the CP-MAS method beyond this rather limited situation to the many cases where the peaks from the ligand and membrane inevitably overlap, it is desirable to remove the background signal from the membranes. One recent approach has been to express a bacterial protein in 12 C-enriched media, thereby minimizing the background signal from 13 C [19]. This is a rather expensive strategy, however, that will be most suitable in cases where high expression levels of protein can be achieved.

Solid-State NMR Methods 323

324 Part I

Chemistry

Part I

A Case Study: Solid-State NMR Investigations of Ion Pump Inhibitors The P-type ATPases are a class of large (>100 kDa) membrane-embedded enzymes that are found in higher eukaryotic organisms and some bacteria. The specific functions of these enzymes vary from cell to cell, but all are related by their ability to couple ATP hydrolysis to vectorial ionic transport across the cell membrane [33]. Ionic transport may be electroneutral (i.e. translocation of charge-equivalent ions) or electrogenic (establishing a transmembrane potential) and is coupled to conformational changes in the enzymes that reveal, in turn, high-affinity sites for ions on opposite sides of the membrane. Two members of this family are well-established targets for drugs. The cardiac Na+ /K+ ATPase controls the relaxation–contraction cycle of the heart and is the receptor for high-affinity cardiac glycoside inhibitors, collectively known as digitalis, which have been used for over 200 years in the treatment of congestive heart failure. The gastric H+ /K+ -ATPase is a proton pump responsible for secretion of acid into gastric glands, and

is a target for drugs in the treatment of gastric ulcer disease. Reversible inhibitors of the proton pump include the aryl-substituted imidazopyridines, which are reasonably potent (IC50 ∼ 1 µM) but have undesirable toxicological properties. Solid-state NMR experiments have revealed the first details about how these inhibitors interact with their targets despite high-resolution structures being unavailable for either of the two proteins. Information about the bound conformation and orientation of these drugs has given clues about how they achieve selectivity for their specific targets. Selectivity is a crucial property of these inhibitors because promiscuous binding to other ATPases could have catastrophic consequences for patients treated with these or related drugs. The NMR experiments form a component of the broader approach summarized in Figure 3 for the specific example of proton pump inhibitors. First, aryl-substituted imidazopyridines were isotope labeled at various sites with 13 C and 19 F and titrated into gastric membranes enriched in H+ /K+ -ATPase. Measurements of heteronuclear and homonuclear dipolar couplings using REDOR and rotational resonance NMR [30,31] have

Fig. 3. An overview of the strategy for modeling an aryl-substituted imidazopyridine inhibitor in its binding site within the H+ /K+ ATPase. The details of the strategy are described in the text. The model is shown alongside a model of ouabain in its site of action within the Na+ /K+ -ATPase, derived by a similar approach. (See also Plate 36 on page XXXI in the Color Plate Section.)

New Methods For Drug Discovery

Future Prospects The versatility of NMR in the characterization of receptor–ligand interactions, when viewed alongside the therapeutic importance of membrane-embedded proteins, suggests that this technique will continue to play a valuable role in drug discovery. This optimistic prognosis will only be realized, however, if developments in NMR hardware and methodology keep apace with the breathtaking progress seen in other aspects of the drug discovery process. With this in mind, the author speculates below on how NMR may continue to provide new and

groundbreaking information about membrane receptor– ligand interactions. (i) Solution state NMR remains underdeveloped as a technique for characterizing receptor–ligand interactions in biomembranes, yet many of the experiments are in place and simply await validation with the appropriate membrane samples. There is a clear opportunity to take advantage of the tremendous improvements in the sensitivity of solution NMR instrumentation in order to compensate for the often low quantities of membrane receptors that can be obtained. (ii) The examples of work on bR, rhodopsin, and the gastric proton pump inhibitors given in this brief review are good demonstrations of how solid-state NMR can provide precise structural constraints for ligands bound to receptors. If such information were obtained for current drug targets, this could help to guide medicinal chemistry toward structures with higher affinity or selectivity for their receptors. This approach has already been used with some success in the case of the aryl-substituted imidazopyridine inhibitors of the gastric proton pump (Figure 4, bottom). By chemically restraining the aryl group in a configuration close to that seen for the flexible parent inhibitor in the binding site, the new compound inhibited the pump with almost 100-fold higher potency. (iii) The major proportion of membrane-embedded drug targets are GPCRs [2], but NMR studies of their interactions with ligands have so far been limited to a few well-chosen cases. The NMR studies of retinal in the GPCR rhodopsin have shown that structural details can be obtained with high precision, but rhodopsin is a unique case amongst this class of proteins and the retinal chromophore is not typical of a GPCR agonist or antagonist of interest in pharmaceutical research. Recently Baldus and co-workers used

N

N

13

O

φ2

φ3 F

N

N

φ1

C H3

O

F

F

F F

flexible, IC50 = 10 µM

NMR conformation

Constrained, IC50 = 100 nM

Fig. 4. An example of how the structure of a receptor-bound ligand can provide information of relevance to drug discovery (refer to the text for details). (See also Plate 37 on page XXXII in the Color Plate Section.)

Part I

provided structural constraints defining the relative orientations of the aryl and imidazopyridine rings within the site of action, showing that the molecule is distorted from a planar configuration and exhibits a slightly curved profile. It has been possible to produce a model describing the location of the inhibitor within the luminal face of H+ /K+ -ATPase with the aid of published site directed mutagenesis (SDM) data and coordinates from the crystal structure of the homologous Ca2+ -ATPase. The sequence of H+ /K+ -ATPase was threaded onto the structure of the Ca2+ -ATPase and the spatial disposition of residues in the proton pump, known from SDM studies to be implicated in drug binding, were predicted. The inhibitor in its curved conformation was found to fit convincingly into a putative binding site bounded by residues and beneath the luminal surface of the protein (Figure 3). By contrast, similar solid-state NMR measurements of 13 C/19 F labeled digitalis analogs have suggested that the recognition site of the Na+ /K+ -ATPase lies at the membrane surface [32] with the functionally-redundant sugar group of ouabain facing away from the protein (Figure 3). These distinct binding characteristics may hold the clue as to why the two inhibitors have such remarkable selectivity for these closely related proteins.

Future Prospects 325

326 Part I

Chemistry

Part I

solid-state NMR to solve the structure of a fragment of the neurotensin peptide when bound to its receptor [34]. This approach is potentially very useful, as determining the pharmacophore structure of a natural peptide ligand could help to predict small molecule antagonists. In all three cases above, the future role of NMR in the analysis of membrane receptor–ligand interactions will be intimately linked to improvements in the methods for producing useful quantities of functional receptors in a suitable form for NMR studies.

References 1. Hajduk PJ, Burns DJ. Comb. Chem. High Throughput Screen. 2002;5:613. 2. Drews J. Science. 2000;287:1960. 3. Meyer B, peters T. Angew. Chem. Int. Ed. 2003;42:864. 4. Mayer M, Meyer B. Angew. Chem. Int. Ed. 1999;38:1784. 5. Dalvit C, Pevarello P, Tato M, Veronesi M, Vulpetti A, Sundstrom M. J. Biomol. NMR. 2000;18:65. 6. Post CB. Curr. Opin. Struct. Biol. 2003;13:581. 7. Dalvit C, Flocco M, Knapp S, Mostardini M, Perego R, Stockman BJ, Veronesi M, Varasi M. J. Am. Chem. Soc. 2002;124: 7702. 8. Claasen B, Axmann M, Meinecke R, Meyer B. J. Am. Chem. Soc. 2005;127:916. 9. Zhang L, Mattern R-H, Malaney TI, Pierschbacher MD. J. Am. Chem. Soc. 2002;124:2862. 10. Furukawa H, Hamada T, Hayashi MK, Haga T, Muto Y, Hirota H, Yokoyama S, Nagasawa K, Ishiguro M. Mol. Pharmacol. 2002;62:778. 11. Watts A, Burnett IJ, Middleton DA, Spooner PJR, Watts JA, Williamson PTF. Nat. Prod. Rep. 1999;16:419. 12. Gr¨obner G, Taylor A, Williamson PTF, Choi G, Glaubitz C, Watts JA, de Grip WJ, Watts A. Anal. Biochem. 1997;254: 132. 13. Spooner PJR, Rutherford N, Watts A, Henderson PJF. Proc. Natl. Acad. Sci. USA. 1994;91:3877.

14. Spooner PJR, Veenhoff L, Watts A, Poolman B. Biochemistry. 1999;38:9634. 15. Spooner PJR, O’Reilly WJ, Homans SW, Rutherford NG, Henderson PJF, Watts A. Biophys. J. 1998;75:2794. 16. Patching SG, Brough AR, Herbert RB, Rajakarier JA, Henderson PJF, Middleton DA. J. Am. Chem. Soc.2004;126:3072. 17. Boland MP. Ph.D. Thesis, University of Manchester, 2005. 18. Boland MP, Middleton DA. Magn. Reson. Chem. 2004;42: 204. 19. Patching SG, Herbert RB, O’Reilly J, Brough AR, Henderson PJF. J. Am. Chem. Soc.2004;126:86. 20. Glaubitz C, Watts A. J. Magn. Reson. 1998;130:305. 21. Dusold S, Sebald A. Annual Reports on NMR Spectroscopy. 2000;41:185–264. 22. Feng X, Lee YK, Sandstrom D, Eden M, Maisel H, Sebald A, Levitt MH. Chem. Phys. Lett. 1996;257:314. 23. Creuzet F, McDermott A, Gebhard R, van der Hoef K, SpijkerAssink MB, Herzfeld J, Lugtenburg J, Levitt MH, Griffin RG. Science 1991;251:783. 24. Ulrich AS, Wallat I, Heyn MP, Watts A. Nat. Struct. Biol. 1995;2:190. 25. Lansing JC, Hohwy M, Jaroniec CP, Creemers AFL, Lugtenburg J, Herzfeld J, Griffin RG. Biochemistry. 2002;41: 431. 26. Jaroniec CP, Lansing J, Tounge B, Belenky M, Herzfeld J, Griffin RG. J. Am. Chem. Soc. 2001;123:12929. 27. Gr¨obner G, Burnett IJ, Glaubitz C, Choi G, Mason AJ, Watts A. Nature. 2000;405:810. 28. Petal AB, Crocker E, Eilers M, Hirshfeld A, Shenes M, Smith SO. Proc. Natl. Acad. Sci. USA. 2004;101:10048. 29. Williamson PTF, Meier BH, Watts A. Eur. Biophys. J. 2004; 33:247. 30. Middleton DA, Robins R, Feng X, Levitt MH, Spiers ID, Schwalbe C, Reid DG, Watts A. 1997;410:269. 31. Watts JA, Watts A, Middleton DA. J. Biol. Chem. 2001;276: 43197. 32. Middleton DA, Rankin S, Esmann M, Watts A. Proc. Natl. Acad. Sci. USA. 2001;97:13602. 33. Watts, J.A. D.Phil. Thesis, University of Oxford, 2001. 34. Luca S, Sohal AK, Fillipov D, van Boom J, Grisshammer R, Baldus M. Proc. Natl. Acad. Sci. USA. 2003;100:10706.

327

H.J.M. de Groot Leiden Institute of Chemistry, Gorlaeus Laboratories, NL2300 RA Leiden, The Netherlands

Introduction In photosynthesis, light energy conversion proceeds in two steps [1]. First excitons are generated in antenna systems and subsequently charge separation takes place in reaction centers (RCs, Figure 1). To gain insight into the structural and functional properties of such active elements in photosynthesis, solid-state NMR is increasingly important. Here a number of examples of recent investigations are summarized, first structure– function studies of antennae and RCs, and second structure determination, including methodology development. To resolve molecular electron pumping mechanisms in bacterial RCs and to study the electronic structure of light-harvesting (LH) proteins, both global and specific assays of cofactors and protein side chains have been performed. Novel techniques allow a determination of structural models for self-assembled chlorophyll preparations in vitro and for the natural chlorosome antenna system. Finally, it is possible to perform sequence specific assignments of uniformly labeled complexes and to observe intermediate states in light-triggered reactions, produced by illumination of frozen samples in the spectrometer.

Structure–Function Studies of Antenna Systems and RCs The electronic ground states of the bacteriochlorophyll (BChl) a type B800 and type B850 (BChl molecules absorbing around 800 or 850 nm) in the LH2 complex of Rhodopseudomonas acidophila strain 10050 have recently been characterized by magic angle spinning (MAS) dipolar 13 C–13 C correlation NMR spectroscopy [2]. Extensive sets of isotropic 13 C NMR chemical shifts were obtained for the BChl in the LH2. Density functional theory calculations were performed to analyze the data in detail. By correction for the ring current shifts, the 13 C shift effects due to the interactions with the protein matrix can be resolved. The shift effects for the B800 and B850 are similar, and are attributed to a weak nonlinear dielectric response of the protein environment to the cofactor binding, in contrast with local effects due to interaction

Graham A. Webb (ed.), Modern Magnetic Resonance, 327–333.
C 2008 Springer.

with specific amino acid (AA) residues. In addition, the polarization of the electronic ground states induced by the protein environment is comparable for both cofactors and corresponds with a red shift of ∼30 nm relative to the monomeric BChl in solution. According to the NMR, the electronic coupling between the B850 cofactors due to macrocycle overlap is the predominant mechanism responsible for the color difference between the B800 and B850 cofactors. In another study, photosynthetic RCs of Rhodobacter sphaeroides R26 were reconstituted at the QA site with ubiquinone-10, selectively 13 C-enriched on positions 1, 2, 3, 4, and 3-Me [3]. RCs dispersed in LDAO detergent were studied with 13 C CP/MAS NMR spectroscopy at temperatures between 180 and 240 K, while RCs precipitated by removal of the detergent were investigated at ambient temperature and at temperatures down to 180 K. Electrostatic charge differences in QA induced by polarization from the protein are small, less than 0.02 electronic equivalent for any of the labeled positions. The 4-carbonyl signal indicates a rigid environment for this functionality, which contrasts with previous studies with FTIR that provided evidence for a strong perturbation and possibly dynamic disorder of this quinone functionality [4]. The QA site is slightly heterogeneous on the scale of the NMR as the observed line widths of the labels are between 150 and 300 Hz and inhomogeneous broadening is observed for the signals of positions 1, 2, and 3 upon cooling. For the 4-carbonyl only at sample temperatures below −255 K, a CP/MAS response can be observed at 183 ppm. The data indicate a difference between the dark adapted state monitored by NMR and the light adapted form that is probed by optical investigations. Various 15 N and 13 C CP/MAS NMR methods have been used to analyze BChl–histidine interactions and the electronic structure of histidine residues in RCs and antenna complexes [5,6]. For the LH2 complex of R. acidophila, the histidines were selectively labeled at both or one of the two nitrogen sites of the imidazole ring. The resonances of histidine nitrogens that are interacting with B850 BChl a have been assigned. Specific 15 N labeling confirmed that it is the τ -nitrogen of α-His30 and β-His31 that are ligated to Mg2+ of BChl cofactors. The π-nitrogens of these Mg2+ -bound histidines were

Part I

Photosynthetic Antennae and Reaction Centers

328 Part I

Chemistry

Part I

β α

Fig. 1. Self-organized photosynthetic complexes, which play key roles in the photosynthetic process. Left: Top view on the LH2 LH membrane protein complex. This 9-mer contains two α-helical segments in every monomeric unit. The transmembrane helices embed the BChl that performs the LH function. Center: Chlorosomes are oblong bodies inside a protein free antenna system that currently serves as one possible paradigm for light concentration in artificial photosynthesis research. With solid-state NMR it was possible to show that a chlorosome antenna contains tubular bilayers of self-aggregated BChl. Right: Photo excitations are transferred to a RC, which is brought into an excited state. Electron transport (ET) occurs from the special pair (DA /DB ) and via two additional chlorophyll molecules (BA /BB ), to a pheophytin (A ), and finally the quinones (QA , QB ) as indicated by the arrows. (See also Plate 38 on page XXXII in the Color Plate Section.)

found to be protonated and may be involved in hydrogen bond interactions. Comparison of the 2-D MAS NMR homonuclear (13 C–13 C) dipolar correlation spectrum of [13 C,15 N]-histidines in the LH2 complex with model systems in the solid state reveals two different classes of electronic structures for the histidines in the LH2. In terms of the 13 C isotropic shifts, one corresponds to the neutral form of histidine and the other resembles a positively charged histidine species [5,7]. 15 N–13 C double-CP/MAS NMR data provide evidence that the electronic structure of the histidines in the neutral BChl a/His complexes resembles the positive charge character form. While the isotropic shift of the 15 N ligated to the Mg2+ confirms a partial positive charge transfer, its anisotropy is essentially of the lone pair type. This provides evidence that the hybridization structure corresponding to the neutral form of the imidazole is capable of “buffering” a significant amount of positive charge. To study the active chlorophyll and pheophytin cofactors involved in the primary processes in photosynthesis, and their environment in the protein, photochemically induced dynamic nuclear polarization (photo-CIDNP) is the method of choice. The unmatched sensitivity of the photoCIDNP allows the detection of signals at natural abundance of the 13 C [8]. Recently we have reported photoCIDNP for the RCs of plant photosystems I and II (PS I, II) [9,10]. The light-enhanced NMR provides information on the electronic structure of the primary electron donors. The radical cation response from the bacterial RC shows at least four emissive center bands, indicating a symmetric spin density distribution over the entire

BChl macrocycle. In contrast, the data for the PS II reveal a pronounced asymmetry of the electronic spin density distribution within the P680•+ (chlorophyll molecules absorbing around 680 nm). PS II shows only a single broad and intense emissive signal and the spin density appears shifted compared to monomeric chlorophyll in solution. It leads to a first hypothesis as to how the planet can provide itself with the chemical potential to split water and generate an oxygen atmosphere using the Chl a macroaromatic cycle: a local electrostatic field can polarize the electronic charge and associated spin density and increase the redox potential of P680 by stabilizing the highest occupied molecular orbital, without a major change of color. In addition, RCs of wild-type R. sphaeroides were selectively 13 C-isotope labeled in BChl and bacteriopheophytin (Bphe), and 13 C solid-state CP/MAS NMR and photo-CIDNP were used to provide insight into the electronic structure of the primary electron donor and acceptor on the atomic and molecular levels [11,12]. The first 2D photo-CIDNP 13 C–13 C solid-state MAS NMR spectra reveal that negative polarization of the two BChl rings of the primary donor is involved in ground state tuning of the oxidation potential of these cofactors in the protein via local electrostatic interactions (Figure 2). In particular, the 13 C shifts show moderate differences in the electronic structure between the two BChl molecules of the special pair in the electronic ground state, which can be attributed to hydrogen bonding of one of the BChl molecules. The major fraction of the electron spin density is strongly delocalized over the two BChl molecules of the

Photosynthetic Antennae and RCs

Structure–Function Studies of Antenna Systems and RCs 329

Part I

Fig. 2. Contour plot of the “dark” 2-D RFDR CP/MAS spectrum (left) and the 2-D RFDR photo-CIDNP spectrum (right) of RCs containing 13 C labels in the cofactors. The data were recorded at ∼220 K with 12 and 5 kHz spinning frequencies and ∼5 ms RFDR mixing. The assignments of the correlations of the two BChl of the special pair (A, B), and the BPhe (C) are indicated with the dashed lines.

330 Part I

Chemistry

Part I

special pair and the photochemically active BPhe. A small fraction of the π -spin density is distributed over a fourth component, which is assigned to the accessory BChl. Comparison of the photo-CIDNP data with “dark” NMR spectra obtained in ultrahigh field indicates that putative structural changes of the special pair during the primary process of photosynthesis should be reversed upon charge recombination on the timescale of the photo-CIDNP experiment.

MAS NMR Structure Determination: Chlorosomes and LH2 The structure of photosynthetic membrane proteins and other membrane associated assemblies represents a solid matrix that is essential to the mechanism of biological function. In recent years MAS NMR has been used to gain in-depth understanding the structure of photosynthetic energy conversion systems. In photosynthesis, light is collected by LH antenna complexes that generally contain 50–200 chlorophylls in well-defined membrane bound protein structures. Photosynthetic green sulfur bacteria, however, contain chlorosomes (Figure 1). These are unique antenna systems among photosynthetic organisms since their LH pigments are organized without proteins. Already at an early stage the chlorosome of the green bacterium Chlorobium tepidum was studied by 1-D MAS NMR methods [13]. With 2-D correlation spectroscopy, it was shown that the BChl are self-organized into supramolecular aggregates by coordinative bonding and π –π stacking interactions [14]. Homo- and heteronuclear MAS NMR, along with comparison of BChls with different side chains lead to a model for the structure with two concentric tubes for the aggregated BChl in chlorosomes (Figure 3) [15]. Heteronuclear 2-D and 3-D MAS NMR dipolar correlation spectroscopy was applied to determine solid-state 1 H shifts for aggregated BChl c in uniformly 13 C-enriched chlorosomes. A complete assignment of 29 different observable resonances of the 61 protons of the aggregated BChl c in the intact chlorosomes was obtained. The 21-H, 32-H, and 31-H resonances are shifted upfield by −2.2, −1, and −3.3 ppm, respectively, relative to monomeric BChl c in solution, revealing parallel stacking of the BChl in the antenna. Although the resonances are inhomogeneously broadened and reveal considerable global structural heterogeneity, the 5-CH and the 7-Me responses are doubled, which provide evidence for the existence of at least two relatively well-defined structurally different arrangements. Ab initio quantum chemical modeling studies were performed to refine a model for the self-assembled BChl c with two different types of BChl stacks. The BChl in the stacks can

Fig. 3. The structure model of the chlorosomal antennae in Chlorobium tepidum derived from MAS NMR, consisting of bilayers of self-assembled BChl c.

adopt either anti- or syn-configuration of the coordinative bond, where anti and syn designate the relative orientation of the Mg–OH bond relative to the direction of the 17–171 bond. Based on the NMR data, a bilayer model for the tubular supra-structure of sheets of BChl c was proposed, from a homology modeling approach (Figure 2). As a spin-off mainly from the investigations into chlorosome LH antenna structures and ligand–protein interactions for membrane proteins, the construction and use of novel ultrahigh field (750 MHz) MAS NMR equipment was recently demonstrated [16]. The new technology represents a twofold increase of field strength with respect to previous practice, since biological MAS NMR was and is still often performed at 300–400 MHz 1 H frequency. The higher field increases the sensitivity by ∼60% and improves the range and resolution considerably [17]. To confront the new technique with real biomolecular targets from research in molecular structural biology and to explore the range of the new technology for structure determination of membrane proteins, the sequence specific assignments for the transmembrane

Photosynthetic Antennae and RCs

MAS NMR Structure Determination: Chlorosomes and LH2 331

Part I

Fig. 4. 13 C-isotope enrichment of the residues and BChl a cofactors in LH2. The green color shows the labeling pattern that is obtained by growing on [1,4-13 C]-succinic acid, while the labeling pattern obtained by growing on [2,3-13 C]-succinic acid is indicated in red. The small sections indicate a partial scrambling of isotopes. Ile and Leu are labeled due to the uptake from an AA nutrient source. (See also Plate 39 on page XXXII in the Color Plate Section.)

helices in the monomeric unit of the LH2 were recently obtained [18,19]. Here MAS NMR was used in combination with extensive and selective biosynthetic isotope labeling methods. For both the residues of the protein and for the cofactors distinct labeling patterns have been deduced with 2-D proton-driven spin diffusion (PDSD) solid-state NMR correlation spectroscopy for samples prepared from [1,4-13 C]-succinic acid, [2,3-13 C]-succinic or AA labeled media (Figure 4) [20]. All residues, except isoleucine and leucine, have been labeled almost homogeneously by the succinic acid precursor. Carbonyl carbons in the protein backbone were labeled by [1,4-13 C]-succinic acid, while the Cα and Cβ carbons of the residues were labeled by [2,3-13 C]-succinic acid. Leucine and isoleucine residues were labeled using a uniformly labeled AA mixture in the medium. The pattern labeling yields an increase of the resolution and less spectral crowding. The partial labeling technique in combination with conventional solid-state NMR methods at ultrahigh magnetic fields provides an attractive route to resolve chemical

shifts for a helical transmembrane protein structures. Assignments have been performed on the basis of 2-D PDSD 13 C–13 C correlation experiments with mixing times of 20 and 500 ms and band selective 13 C–15 N correlation spectroscopy on a series of site-specific biosynthetically labeled samples (Figure 5) [19]. The decreased line width and the reduced number of correlation signals of the selectively labeled samples with respect to the uniformly labeled samples enable to resolve the narrowly distributed correlation signals of the backbone carbons and nitrogens involved in the long α-helical transmembrane segments. Correlations between nearby residues and between residues and the labeled BChl a cofactors, provided by the 13 C–13 C correlation experiments using a 500 ms spin diffusion period, are used resolve the NMR responses for many residues in the protein complex. In this way it is demonstrated that MAS NMR methods combined with site-specific biosynthetic isotope labeling can be used for sequence specific assignment of a transmembrane protein complex.

332 Part I

Chemistry

Part I Fig. 5. In the upper panels two regions from homonuclear 13 C–13 C PDSD50 correlation spectra collected from 2,3-LH2 (red) and AA-LH2 (black) are shown. The region shown in the upper left panel contains cross peaks involving the aliphatic carbons and carbonyl carbons, while the upper right panel shows correlations between aliphatic carbons, present in the side chains of the AAs. In the left part, a few responses are observed for 2,3-LH2, belonging to H, Q, and E residues. The responses from AA-LH2 in the carbonyl area are from I, L, A, G, and V residues. The blue-coded spectrum in the carbonyl region comprises carbonyl responses from 1,2,3,4-LH2. In the upper right panel the aliphatic responses are shown. The dashed lines indicate correlations involving the αT38 and 4P residues for the 2,3-LH2, and correlations involving βI16 for the AA-LH2. The residues that are labeled via both nutrient sources are also indicated. In the middle pane, the aliphatic region of the NCACX spectra of 2,3-LH2 (red) and AA-LH2 (black) are shown. The data are aligned with the PDSD50 spectrum and correlations involving αT38, βI16, and the 4P residues are indicated with dashed lines for the two different samples. The responses of the G residues are indicated with a rectangular box. The NCA signals are aligned with the carbonyl area of the PDSD50 spectrum. Finally, in the lower panel the NCACX spectrum of a 1,2,3,4-LH2 sample is shown. (See also Plate 40 on page XXXIII in the Color Plate Section.)

Photosynthetic Antennae and RCs

1. Hoff AJ, Deisenhofer J. Phys. Rep. 1997;287:2. 2. van Gammeren AJ, Buda F, Hulsbergen FB, Kiihne S, Hollander JG, Egorova-Zachernyuk TA, Fraser NJ, Cogdell RJ, de Groot HJM. J. Am. Chem. Soc. 2005;127:3213. 3. van Liemt WBS, Boender GJ, Gast P, Hoff AJ, Lugtenburg J, de Groot HJM. Biochemistry. 1995;34:10229. 4. Brudler R, de Groot HJM, van Liemt WBS, Steggerda WF, Esmeijer R, Gast P, Hoff AJ, Lugtenburg J, Gerwert K. EMBO J. 1994;13:5523. 5. Alia, Matysik J, Soede-Huijbregts C, Baldus M, Raap J, Lugtenburg J, Gast P, van Gorkom HJ, Hoff AJ, de Groot HJM. J. Am. Chem. Soc. 2001;123:4803. 6. Zysmilich MG, McDermott AE. J. Am. Chem. Soc. 1996;118:5867. 7. van Gammeren AJ, Hulsbergen FB, Erkelens C, de Groot HJM. J. Biol. Inorg. Chem. 2004;9:109. 8. Zysmilich MG, McDermott AE. J. Am. Chem. Soc. 1994;116:8362. 9. Alia, Roy E, Gast P, van Gorkom HJ, de Groot HJM, Jeschke G, Matysik J. J. Am. Chem. Soc. 2004;126:12819. 10. Matysik J, Alia, Gast P, van Gorkom HJ, Hoff AJ, de Groot HJM. Proc. Natl. Acad. Sci. U.S.A. 2000;97:9865.

11. Egorova-Zachernyuk T, van Rossum B, Boender GJ, Franken E, Ashurst J, Raap J, Gast P, Hoff AJ, Oschkinat H, de Groot HJM. Biochemistry. 1997;36:7513. 12. Schulten EAM, Matysik J, Alia, Kiihne S, Raap J, Lugtenburg J, Gast P, Hoff AJ, de Groot HJM. Biochemistry. 2002;41:8708. 13. Nozawa T, Ohtomo K, Suzuki M, Nakagawa H, Shikama Y, Konami H, Wang ZY. Photosynth. Res. 1994;41:211. 14. Balaban TS, Holzwarth AR, Schaffner K, Boender GJ, de Groot HJM. Biochemistry. 1995;34:15259. 15. van Rossum B-J, Steensgaard DB, Mulder FM, Boender GJ, Schaffner K, Holzwarth AR, de Groot HJM. Biochemistry. 2001;40:1587. 16. Kiihne SR, de Groot HJM (Eds). Perspectives on Solid State NMR in Biology. Kluwer: Dordrecht, 2001. 17. van Rossum BJ, Boender GJ, de Groot HJM. J. Magn. Reson. A. 1996;120:274. 18. Egorova-Zachernyuk TA, Hollander J, Fraser N, Gast P, Hoff AJ, Cogdell R, de Groot HJM, Baldus M. J. Biomol. NMR. 2001;19:243. 19. van Gammeren J, Hulsbergen FB, Hollander JG, de Groot HJM. J. Biomol. NMR. 2005;31:279. 20. van Gammeren AJ, Hulsbergen FB, Hollander JG, de Groot HJM. J. Biomol. NMR. 2004;30:267.

Part I

References

References 333

335

Philip L. Yeagle and Arlene Albert Department of Molecular and Cell Biology, University of Connecticut, Storrs, CT 06269, USA

Introduction Since α-helices and turns (helix-turn-helix motif) are stabilized by short-range interactions, and since many membrane proteins are built around (transmembrane) helical bundles, much of the secondary structure of such membrane proteins can be captured in peptide fragments. Furthermore, if sufficient long-range point-to-point experimental distance constraints are available from the intact protein, a structure for the whole protein can be assembled from the structures of the peptide fragments. In this review, we will describe the basis for the first statement and give some examples of the second. The review will conclude with a brief look at the future of high-resolution NMR in the study of the structural biology of intact membrane proteins in detergent micelles.

Membrane Protein Structure—Current Status Our understanding of the structure of integral membrane proteins lags considerably our understanding of soluble protein structure. Less than 0.5% of the structures in the PDB represent integral membrane proteins, whereas genomic analysis indicates that 25–40% of proteins encoded by most genomes are membrane proteins [1]. Clearly there is a huge deficit of structural information on membrane proteins. The major source of this knowledge deficit lies in the difficulties in applying the most productive techniques in protein structure determination to membrane proteins. X-ray crystallography requires crystallization of membrane proteins and to date only about 80 structures of membrane proteins have been published using this approach in large part because of difficulties in crystallization. Membrane proteins have extensive hydrophobic surfaces and therefore are insoluble in water. Because of this insolubility, many of the standard techniques for crystallization will not work on membrane proteins. Powerful NMR techniques have been developed and exploited to determine structures of a variety of soluble proteins. These techniques in general require that the protein under study be stable and active in aqueous solution. Integral membrane proteins are not soluble in aqueous Graham A. Webb (ed.), Modern Magnetic Resonance, 335–343.
C 2008 Springer.

solution. These membrane proteins must be solubilized in detergent micelles. The protein-detergent micelles tend to be large, with molecular weights in excess of 50 kDa. The long rotational correlation times of such structures enhance dipolar and chemical shift anisotropy interactions and consequently broaden resonances, which inhibits the acquisition of high-resolution spectra. At this writing, structures from only one family of membrane proteins have been reported using high-resolution NMR data [2,3]. These structures are of porins, β-barrel proteins from the outer membrane of E. coli. Structures of the larger families of membrane proteins consisting of transmembrane bundles of α-helices have proven more difficult to solve. It is, however, quite feasible to solve structures of fragments of membrane proteins containing 20–45 residues. We will examine how such structure determinations on membrane protein fragments can provide insight into the secondary structure of membrane proteins consisting of helical bundles.

Peptides from Helices and Turns have Intrinsic Structures that can Provide Secondary Structure Information About the Parent Soluble Protein A growing body of data suggests that solution structures of peptides derived from some classes of proteins retain the secondary structure of the parent protein because of the dominance in α-helices and turns of short-range interactions [4] that can be captured in peptides. Studies on segments of soluble proteins forming α-helices show that peptides containing these sequences form α-helix in almost every case under some solution conditions [5–15]. Peptides representing segments that are turns in the native protein also show turns as peptides in solution [10–13,16–22]. In some cases, the entire sequence of a helical bundle protein has been incorporated in a series of peptides spanning that sequence and the individual peptides have reported the secondary structure of much of the native protein with fidelity [19,23–26]. The implication from these studies on peptides from soluble proteins is that peptide fragments from these proteins preserve much of the secondary structure of the intact protein.

Part I

Insight into Membrane Protein Structure from High-Resolution NMR

336 Part I

Chemistry

Part I

Therefore, since structure determination of intact membrane proteins is problematic, determination of structures of peptide fragments becomes an important alternative approach to structural information on membrane proteins.

Structures of Peptide Fragments from Membrane Proteins can Provide Secondary Structure Information More recently, analogous studies have been performed for membrane proteins. Membrane proteins containing transmembrane helical bundles can be thought of as a collection of transmembrane helices (TM) and connecting turns, both elements of secondary structure stabilized by short-range interactions. It is reasonable to hypothesize, therefore, that peptide fragments corresponding to TM or to connecting turns would display local stable structure characteristic of helix or turn, respectively. This hypothesis has now been tested many times on a number of membrane proteins and found to be correct. In what follows, NMR studies on peptide fragments of a variety of membrane proteins will be reviewed, in each case demonstrating the preservation of native secondary structure in the peptide fragments.

Human Erythrocyte Glycophorin Perhaps the earliest indication of this property of fragments from helical membrane proteins was work on glycophorin. The isolated protein was fragmented by trypsin hydrolysis. Each of the fragments was examined by circular dichroism (CD). When the CD of these individual fragments was summed, the CD of the whole protein was obtained, indicating that the secondary structure of the whole protein was largely preserved in the fragments [27]. Subsequent 1 H NMR studies revealed the α-helical nature of the transmembrane domain separate from the remainder of the protein [28], consistent with the CD studies.

Bacteriorhodopsin Whether useful information of membrane protein secondary structure could therefore be obtained from peptide fragments spanning the entire sequence of a membrane protein was tested on bacteriorhodopsin. Bacteriorhodopsin is a light-activated protein from Halobacterium salinarium (part of the purple membrane, a specialized patch in the membrane of this bacterium) that uses light energy to pump protons against a concentration gradient. The molecular weight of this protein is 24.5 kDa [29]. Bacteriorhodopsin is the prototypical transmembrane protein consisting of a bundle of seven hydrophobic

helices that constitutes the transmembrane portion of the protein with loops connecting each of the helices in the bundle. Therefore the dominant secondary structures of this protein are α-helices and turns. Both α-helices and turns are stabilized by short-range interactions (i to i + 4 or shorter), the internal hydrogen bonds. Therefore peptides with the sequence of one of the transmembrane helices or one of the turns could be expected to contain the hydrogen bonds characteristic of the corresponding secondary structure and thus stabilize the relevant secondary structure in the peptide. A number of three-dimensional structure determinations have been published for bacteriorhodopsin [30–34], one of the very few membrane proteins for which complete structures are available. Furthermore, experiments have shown that bacteriorhodopsin can be expressed in two separate pieces and the pieces will assemble properly in a membrane to re-form the protein [35–36], suggesting a subdomain character for the helices in the transmembrane region. The collection of this evidence encouraged investigation into the structural stability of fragments of bacteriorhodopsin. Several fragments of bacteriorhodopsin were synthesized by solid phase peptide synthesis and their three-dimensional structures determined [37–40]. Two dimensional, homonuclear NMR experiments were used on these unlabeled, hydrophobic peptides in organic solvent. The local stability of the helices was clearly observed. Peptide fragments from some of the helical regions of bacteriorhodopsin formed helices separate from the remainder of the protein. This question of localized stability of secondary structure was then explored in depth for the entire bacteriorhodopsin molecule. A series of overlapping peptides that spanned the sequence of the protein were synthesized. Each peptide encompassed either a (transmembrane) helical region or a turn flanked by two short helical regions (connecting two transmembrane helices). Structures of the peptides were determined by NMR in DMSO, a solvent with no propensity to stabilize any particular secondary structure and with a dielectric similar to the membrane interface at which initial folding of secondary structure of helical membrane proteins is proposed to occur [41]. All the peptides that encompassed a sequence corresponding to a transmembrane helix formed a helix in solution, except for the peptide for helix G in bacteriorhodopsin (which proved to be unstable in solution). All the peptides that corresponded to turns formed turns in solution, separate from the remainder of the protein [42,43]. Overlays of these fragments on the crystal structure showed good agreement between the structures of the peptide fragments and the X-ray crystal structure of the intact protein (see Figure 1). These results were interpreted in terms of local stability of secondary structure, such that crucial interactions

Insight into Membrane Protein Structure from High-Resolution NMR

Structures of Peptide Fragments from Membrane Proteins 337

Rhodopsin + light → Meta II (R∗ ) → Opsin + retinal

Fig. 1. Alpha carbon maps of the superposition of the backbone atoms of the peptide structures on the corresponding sequences in one crystal structure (2BRD) of bacteriorhodopsin. In each case, one member of the family of peptide structures, randomly chosen, was superimposed on the crystal structure. Similar results were obtained from superposition of these peptide structures on 1AP9. The superpositions were calculated using only the well-ordered portions of the peptide structures, as listed in Table 1. Inset: The ˚ of the superposition is plotted for each peptide as a rmsd (A) function of the sequence of bacteriorhodopsin. The horizontal line represents the average rmsd of superposition of 2BRD on 1AP9. (Reproduced from ref. [43] with permission.)

(hydrogen bonds) could form within the peptide, just as in the X-ray crystal structure of the intact protein. The agreement between the peptide structures and the structure of the intact protein suggested that structural studies of a series of overlapping peptides spanning the sequence of an α-helical membrane protein should provide valid information on the secondary structure of the protein (α-helices and turns).

Bovine Rhodopsin This information provided the basis for a study of the secondary structure of bovine rhodopsin, another protein built around a transmembrane bundle of 7 α-helices like

Expression studies suggested that rhodopsin was built of subdomains. Rhodopsin can be expressed as two independent bundles of TM, such as a set of 3 TM and a set of 4 TM, and these separately expressed helical bundles will spontaneously assemble in the membrane [47]. These studies led to the hypothesis stated above that a protein built around a transmembrane helical bundle can be dissected into peptide fragments that retain the secondary structure of the native protein. When studies based on that hypothesis were begun [48], no X-ray crystal structure of rhodopsin was available. Therefore an intense effort was dedicated to determine the secondary structure of this protein through fragments of rhodopsin corresponding to transmembrane helices or turns. Initially, just the cytoplasmic loops were studied in depth. These loops had been shown to be biologically active, inhibiting the interaction between this GPCR and its G protein [49,50]. These loops were soluble in water and showed CD characteristic of structure under conditions similar to those in which biological activity had been demonstrated. They were therefore ideal candidates to determine whether structure of a loop could exist separate from the rest of the protein. NMR studies were undertaken and the structures determined of these peptide fragments of rhodopsin [48,51–53]. All three cytoplasmic loops formed stable structures in solution as suggested by the CD and consistent with the observed biological activity. These studies encouraged an in-depth study of the whole protein. A complete set of overlapping peptides spanning the sequence of the protein was synthesized. The structures of the remaining fragments were determined by solution NMR techniques [15,54,56]. These peptide fragments in each case showed structure. Fragments from putative helices showed helical structures and fragments from turns showed turn structures. It became clear that most of the secondary structure of this helical membrane protein could be captured in these peptide fragments.

Part I

bacteriorhodopsin. The initial events of low-level visual transduction take place in the retinal rod cell on the retinal rod outer segment (ROS) disk membrane. The G-protein coupled receptor (GPCR), rhodopsin, is the major protein of the disk membrane, comprising 80–90% of the total disk membrane protein [44,45]. When light strikes the ROS and is absorbed by the photopigment, rhodopsin goes through a series of spectrally defined intermediates. The transition to the Metarhodopsin II (R∗ ) intermediate permits the activation of the G protein, transducin and initiates the cGMP cascade [46] that culminates in the hydrolysis of cGMP and closure of the plasma membrane Na+ channels. This results in a hyperpolarization of the plasma membrane and generates a signal at the synapse.

338 Part I

Chemistry

Part I

When a crystal structure became available [57], it was possible to verify that the secondary structure in the peptide fragments mimicked the secondary structure in the intact protein [58]. It subsequently became possible to utilize these structures of peptide fragments to assemble a structure for the whole protein. This will be discussed at the end of this review.

Lactose Permease The lac permease is a β-galactoside transport system of E. coli encoded in the lac operon (lacY). The first report of this activity was in 1955 [59]. Subsequently it was shown that this transport activity was driven by a proton gradient [60], in accord with the Mitchell hypothesis [61]. This protein was cloned [62] and sequenced [63] early in the study of such membrane proteins. It was purified and reconstituted in defined lipid bilayers and found to exhibit the same transport activity that had been measured in the native membrane [64]. The protein functions as a monomer in the membrane [65,66]. Cysteine scanning has shown that six amino acid residues are essential for the transport process: E126, R144, E269, R302, H322, and E325 [67]. The structure of this protein consists of a bundle of 12 TM. The X-ray crystal structure shows a pseudo symmetry of two bundles of 6 TM connected by loops [68]. Therefore the lactose permease is a good candidate for the same analysis of secondary structure described above for bacteriorhodopsin and rhodopsin. Such a project was begun before the crystal structure was reported. A series of peptides spanning the sequence of the protein were synthesized, each peptide encompassing a putative transmembrane helix or a turn connecting two TM. As in the case of the other membrane proteins, a set of structures were obtained from solution NMR studies demonstrating the short-range stability of elements of secondary structure, including turns [69] and helices [123].

Protein Fragments of Other Membrane Proteins Bacteriorhodopsin, rhodopsin, and lactose permease are the current examples of complete analysis of secondary structure using a series of peptides spanning the entire sequence of the membrane protein. However, a number of other examples have been reported of fragments from other integral membrane proteins that also show preservation of secondary structure in peptide fragments.

Saccharomyces Cerevisiae α-factor Receptor The Saccharomyces cerevisiae α-factor receptor is a GPCR of yeast involved in mating. Peptides correspond-

ing to all the TM of this 7TM transmembrane protein were synthesized and some of the loops. Structures were determined in organic solvent by NMR. All TM showed helical structures and one of the loop peptides was also structured [70–73].

Parathyroid Hormone Receptor Fragments of the parathyroid hormone receptor, a GPCR, have been studied by solution NMR techniques in detergent micelles. A peptide containing the amino acid sequence of the first extracellular loop was synthesized. The NMR structure revealed a short helix on each end of the peptide corresponding to portions of the TM on either side of the loop. The interior of the loop also contained an additional helix [74]. A peptide corresponding to the third cytoplasmic loop of this same receptor was synthesized and the NMR solution structure revealed a loop structure in the presence of detergent micelles [75,76].

The Human Cholecystokinin-2 Receptor The third extracellular loop of the cholecystokinin-2 receptor (residues 352–379), a GPCR, was synthesized and its structure determined in detergent (DPC) micelles by solution NMR. The ends of the two attached helices, 6 and 7, are seen as is the turn [77]. Interactions between this third extracellular loop and ligands have been probed by NMR and other techniques [77–80].

The Human Cannabinoid Receptor The human cannabinoid 1 receptor functions as a receptor for 9 tetrahydrocannabinol and is coupled to Gi/o . A peptide has been expressed of 44 residues containing the third cytoplasmic loop of this GPCR. The peptide is biologically active. There is evidence for helix at both ends of the peptide, corresponding to portions of the two connected transmembrane helices, and the peptide forms a turn in detergent micelles [81]. The putative helix 8 of the cannabinoid 2 receptor has been synthesized and the structure determined in DPC micelles and in DMSO. In both environments an α-helix was observed [82].

Bradykinin B2 Receptor The bradykinin B2 receptor is a GPCR. A peptide corresponding to the second intracellular loop of the bradykinin B2 receptor and containing 34 residues was synthesized and the structure determined by solution NMR. A helixturn-helix motif was observed, with portions of both attached transmembrane helices visible. In addition, the structure of a portion of the C-terminus was examined

Insight into Membrane Protein Structure from High-Resolution NMR

to the transmembrane domain of three forms of this ion channel, Shaker, ROMK1, and minK. The structures of these peptides were studied in solution by NMR and CD and found to be predominantly helical [91].

Rat Angiotensin II AT1A Receptor Human Red Cell Anion Transporter, Band 3 The rat angiotensin II AT1A receptor is a GPCR. The third cytoplasmic loop, the first extracellular loop and a portion of the carboxyl terminus of this receptor have been studied as peptides in solution with NMR. Two peptides spanning the third cytoplasmic loop show some of the attached transmembrane helices [85]. The peptide corresponding to the first extracellular loop forms a type 2 β turn [86], as do two of the turns in bovine rhodopsin [53]. The C-terminal peptide, corresponding to residues 300–320, form in part an amphipathic helix, perhaps corresponding to helix 8 observed in other GPCRs [87].

Tachykinin NK-1 Receptor A peptide corresponding to the 7th transmembrane domain of the tachykinin NK-1 receptor was synthesized and the structure determined in organic solvent by solution NMR. Evidence for the helical nature of this domain was found in DMSO [88].

β-adrenergic Receptor The β-adrenergic receptor is a GPCR. Peptides corresponding to the third intracellular loop of the turkey receptor (residues 284–295) were synthesized and studied in micelles by solution NMR. The C-terminal region of the peptide showed helical structure, likely corresponding to the beginning of TM 6. The putative helix 8 region of the human β-adrenergic receptor was examined with a peptide in detergent and in DMSO and found to be helical while in water the peptide was disordered [89].

The human red cell anion transporter, band 3, is one of two most abundant membrane proteins of the human erythrocyte membrane, involved in chloride transport. Two of the putative transmembrane segments of this protein, containing residues 405–424 and residues 436–456, were synthesized and their structure determined in trifluoroethanol. Predominantly α-helical structures were reported [92]. In addition, a peptide fragment corresponding to a loop on the cytoplasmic face of the protein connecting TM12 and TM13 was synthesized with 46 residues. The NMR solution structure showed a helix-turn-helix motif [93].

Phosphatidylglycerophosphate Synthase Phosphatidylglycerophosphate synthase from E. coli is an integral membrane protein. Peptides corresponding to two putative TM of this protein were synthesized and their structure determined (residues 6–25 and residues 149– 176). Two-dimensional 1 H NMR studies and CD studies revealed that these sequences were stable as helices in solution and in SDS micelles [94].

IsK Isk is a voltage-gated potassium channel with a single TM per monomer. A peptide was synthesized containing the putative transmembrane domain (residues 42–68). This peptide exhibited biological activity and solution NMR studies in organic solvent showed an α-helical structure [95].

EmrE, a Multidrug Resistance Protein EmrE, a multidrug resistance protein is a membrane protein from E. coli of 110 residues. A set of peptides partially spanning the sequence of the protein was synthesized and structures determined by NMR in SDS micelles. Two of the peptides, corresponding to predicted transmembrane segments, formed helices as expected and another formed a turn [90].

Potassium ion Channel The potassium ion channel is an oligomeric transmembrane structure. Peptides were synthesized corresponding

General Features of the Studies on Membrane Protein Fragments This survey of structure reports on fragments of membrane proteins reveals several common features. (1) The peptides are 20–40 residues in length (if the peptides are too short, then the secondary structure can be destabilized). (2) The peptides encompass either a TM or a turn in the sequence of the fragment. (3) For studies in organic solvents, solvents of intermediate dielectric, such as DMSO, are common and stable secondary structure is often found in such solvents. (4) High-resolution multidimensional NMR experiments are effective in determining

Part I

in an expressed, labeled peptide containing residues 309– 366 of the receptor. Evidence for helix 8 was observed as well as other structured regions [83,84].

General Features of the Studies on Membrane Protein Fragments 339

340 Part I

Chemistry

Part I

structure of such peptides in both organic solvent and in detergent micelles. In the most favorable circumstances, these kinds of studies can provide a nearly complete picture of the secondary structure of the membrane protein if the protein is built around a transmembrane helical bundle. Of course, not every fragment of a membrane protein selected as described above forms stable secondary structure. And in the case of β-barrels, such as the bacterial porins, these kinds of studies to determine secondary structure would not be useful. However, this survey does not reveal a case where the structure of the fragment reported incorrectly on the secondary structure of the protein. Rather the only cases that did not report the correct secondary structure were cases in which the fragment was disordered and thus provided no information on secondary structure. Therefore studies on the structures of nearly 100 of these fragments of membrane proteins have not yet misled the investigator as to the true structure in the protein from which they were derived.

How Sparse Long-Distance Experimental Constraints can be Combined with Fragment Structures to Build a Structure of the Intact Membrane Protein An interesting recent analysis demonstrated the ability to organize elements of secondary structure (α-helices) in three dimensions to mimic the structure of a native membrane protein, using a limited number of long-distance constraints [96]. The concept starts with helices as locally stable structures. In a protein like rhodopsin, the transmembrane domain consists of 7 TM in a bundle. If one assumes the structure of the helices (from, say, hydrophobicity plots), then one needs to define the density of long-distance constraints necessary to the proper organization of the helical bundle. In this study, employing just 27 experimental long-distance constraints was sufficient to define the three-dimensional organization of the 7 helices to mimic the structure of the transmembrane bundle. A conceptually similar (but different in approach) method was used previously with 7 ideal helices to build the bundle constituting the transmembrane domain of rhodopsin [97]. Again a limited number of long-distance constraints were used to organize the bundle. These more theoretical studies provide a basis for assembling a structure of a transmembrane protein from experimental structures of protein fragments, defined as above, and additional experimental long-distance constraints from the intact protein. Structures of two membrane proteins have been successfully built using such an approach.

Bacteriorhodopsin Bacteriorhodopsin was used as a test case to develop this approach to membrane protein structure. Bacteriorhodopsin is a seven transmembrane helical protein and several crystal structures are available. The structures of each in a series of overlapping peptides spanning the sequence of bacteriorhodopsin were determined using NMR. The solution structures of these peptides were found to have the same secondary structure as the corresponding regions of the X-ray crystal structure. These individual peptide structures, and the distance constraints obtained for them, were then used, with additional experimental long-distance constraints, to build the threedimensional structure for the whole protein. The resulting structure agreed with structures determined from electron and X-ray diffraction data. The B factors from X-ray and electron diffraction studies are high in most of the loops of all the bacteriorhodopsin structures. Therefore quantitative comparisons were made with the transmembrane domain, which has substantially lower B factors. Accordingly, the rmsd of superposition of the NMR based structure on the crystal structure 2brd is 2.9. The structure obtained by our method is in as close agreement with the X-ray structure as the X-ray structures are with each other. [43] (see Figure 2). This work demonstrated that a three-dimensional structure could be obtained for a membrane protein using the secondary structural information

Fig. 2. Structure of bacteriorhodopsin (blue), determined from experimental data as described in the text, superimposed on a crystal structure (cyan) of bacteriorhodopsin (1FBB). (See also Plate 41 on page XXXIV in the Color Plate Section.)

Insight into Membrane Protein Structure from High-Resolution NMR

Bovine Rhodopsin The same technique was used to determine a structure of rhodopsin. Structures of all the extramembraneous domains of rhodopsin and the structures of all the transmembrane helical domains were determined by twodimensional homonuclear 1 H NMR [15,43–48,51–56,98– 102]. The peptides typically exhibit well-defined structures in solution. Comparison of the peptide structures to the corresponding region in the crystal structure indicates good agreement. This work formed the basis of a threedimensional structure of rhodopsin that is in agreement with the crystal structure of rhodopsin in the dark-adapted state [57]. Furthermore, we were able to use the same approach to determine a structure for Meta II, the activated form of rhodopsin (and the only structure to date of an activated GPCR) [103]. The technique is described below. The solution structures of a series of overlapping peptides, which spanned the rhodopsin primary sequence, were determined by two-dimensional homonuclear 1 H NMR and linked into a construct corresponding to the entire sequence of rhodopsin [56]. The construct was originally built by superimposing the overlapping regions of the fragments to link one fragment to the next in the sequence. 11-cis retinal was added to K296 to make a specially defined amino acid. Experimental distance constraints were written into the mol2 file for this construct in SYBYL (Tripos). 3030 short-range NOE-derived experimental distance constraints were available from the NMR structure determinations on the individual peptides. Long-range constraints from independent experiments on intact dark-adapted rhodopsin were added. For example, site directed spin labeling put pairs of spin labels at specific sites and the dipolar interactions between the spin labels provided distance measurements [104–115]. Other experimental distance constraints were obtained from engineered disulfide crosslinking [47], engineered metal binding sites [116] and other experimental measures of site-to-site distances in the intact rhodopsin. The 11-cis retinal was constrained by the solid state NMR data of Watts, et al. [117]. The construct with the distance constraints was subjected to several cycles of simulated annealing (1000 fs at 1000 ◦ K followed by 1500 fs cooling to 200 ◦ K). The resulting compact structure determined strictly from experimental data (no modeling), showed a bundle of seven helices connected by six turns. This work demonstrates that a valid structure for membrane proteins built on helical bundles can be obtained from the secondary structures of the protein fragments and

selected long-range constraints. This three-dimensional structure of rhodopsin can be quantitatively superimposing this structure on the X-ray crystal structure. Good agreement with the crystal structure [57] is observed in the transmembrane region with an rmsd of 1.85. (It is only valid to compare this structure with the crystal structure in the transmembrane region because the crystal structure is poorly resolved in the cytoplasmic face. This poor resolution is typical of crystal structures of membrane proteins and is manifest in very high B values.

New High-Resolution NMR Studies on Intact Membrane Proteins The story of high-resolution NMR and membrane protein structure does not end here. Recent studies have exploited new TROSY [118] techniques and deuteration to obtain spectra and structures from several β-barrel porins from the outer membrane of E. coli [2,3,119]. While no complete structure of a membrane protein built on a helical bundle has been published at this writing, important progress is currently being made on helical bundles including bacteriorhodopsin [120] and diacylglycerol kinase [121]. Very helpful to this effort is the discovery of new detergents that produce improved NMR spectra of integral membrane proteins [122]. It is to be hoped that these early efforts presage a strong presence in the future of high-resolution NMR in the field of membrane protein structural biology.

References 1. Buchan DW, Shepherd AJ, Lee D, Pearl FM, Rison SC, Thornton JM, Orengo CA. Genome. Res. 2002;12:503. 2. Arora A, Abildgaard F, Bushweller JH, Tamm LK, Nat. Struct. Biol. 2001;8:334. 3. Fernandez C, Hilty C, Bonjour S, Adeishvili K, Pervushin K, Wuthrich K, FEBS Lett. 2001;504:173. 4. Yang A-S, Hitz B, Honig B, J. Mol. Biol. 1996;259:873. 5. Gao J, Li Y, Yan H, J. Biol. Chem. 1999;274:2971. 6. Ramirez-Alvarado R, Serrano L, Blanco FJ, Prot. Sci. 1997; 6:162. 7. Gegg CV, Bowers KE, Matthews CR, Prot. Sci. 1997;6:1885. 8. Hamada D, Kuroda Y, Tanaka T, Goto Y, J. Mol. Biol. 1995;254:737. 9. Callihan DE, Logan TM, J. Mol. Biol. 1999;285:2161. 10. Wilce JA, Salvatore D, Waade JD, Craik DJ, Eur. J. Biochem. 1999;262:586. 11. Jimenez MA, Evangelio JA, Aranda C, Lopez-Brauet A, Andreu D, Rico M, Lagos R, Andreu JM, Prot. Sci. 1999;8: 788. 12. Fan JS, Cheng HC, Zhang M, Biochem. Biophys. Res. Commun. 1998;253:621. 13. Cox JPL, Evans PA, Packman LC, Williams DH, Woolfson DN, J. Mol. Biol. 1993;234:483.

Part I

obtained from peptides representing protein sub-domains (helices and turns), coupled with experimental long-range distance constraints.

References 341

342 Part I

Chemistry

Part I

14. Hunt JF, Earnest TN, Bousche O, Kalghatgi K, Reilly K, Horvath C, Rothschild KJ, Engelman DM, Biochemistry 1997;36:15156. 15. Katragadda M, Chopra A, Bennett M, Alderfer JL, Yeagle PL, Albert AD, J. Pept. Res. 2001;58:79. 16. Chandrasekhar K, Profy AT, Dyson HJ, Biochemistry 1991;30:9187. 17. Ghiara JB, Stura EA, Stanfield RL, Profy AT, Wilson IA, Science 1994;264:82. 18. Blumenstein M, Matsueda GR, Timmons S, Hawiger J, Biochemistry 1992;31:10692. 19. Blanco FJ, Serrano L, Eur. J. Biochem. 1994;230:634. 20. Goudreau N, Cornille F, Duchesne M, Parker F, Tocqu´e B, Garbay C, Roques BP, Nat. Struct. Biol. 1994;1: 898. 21. Adler M, Sato MH, Nitecki DE, Lin J-H, Light DR, Morser J. J. Biol. Chem. 1995;270:23366. 22. Campbell AP, McInnes C, Hodges RS, Sykes BD, Biochemistry 1995;34:16255. 23. Behrends HW, Folkers G, Beck-Sickinger AG. Biopolymers 1997;41:213. 24. Reymond MT, Merutka G, Dyson HJ, Wright PE, Protein Sci. 1997;6:706. 25. Dyson HJ, Merutka G, Waltho JP, Lerner RA, Wright PE, J. Mol. Biol. 1992;226:795. 26. Padmanabhan S, Jimenez MA, Rico M, Protein Sci. 1999; 8:1675. 27. Schulte TH, Marchesi VT, Biochemistry 1979;18:275. 28. Lemmon MA, Flanagan JM, Hunt JF, Adair BD, Bormann B-J, Dempsey CE, Engelman DM. J. Biol. Chem. 1992;267:7683. 29. Reynolds JA, Stoeckenius W. Proc. Natl. Acad. Sci. U.S.A. 1977;74:2803. 30. Gouaux E. Structure 1998;6:5. 31. Grigorieff N, Ceska TA, Downing KH, Baldwin JM, Henderson R. J. Mol. Biol. 1996;259:393. 32. Henderson R, Baldwin JM, Ceska TA, Zemlin F, Beckmann E, Downing KH. J. Mol. Biol. 1990;213:899. 33. Luecke H, Schobert B, Richter H-T, Cartailler J-P, Lanyi JJK. J. Mol. Biol. 1999;291:899. 34. Pebay-Peyroula E, Rummel G, Rosenbusch JP, Landau EM. Science 1997;277:1676. 35. Kahn TW, Engelman DM. Biochemstry 1992;31:6144. 36. Marti T. J. Biol. Chem. 1998;273:9312. 37. Barsukov IL, Nolde DE, Lomize AL, Arseniev AS, Eur. J. Biochem. 1992;206:665. 38. Lomize AL, Pervushin KV, Arseniev AS. J. Biomol. NMR 1992;2:361. 39. Pervushin KV, Orekhov VY, Popov AI, Musina LY, Arseniev AS. Eur. J. Biochem. 1994;219:571. 40. Sobol AG, Arseniev AS, Abdulaeva GV, LYu M, Bystrov VF, J. Bio. NMR 1992;2:161. 41. White SH, Wimley WC, Ann. Rev. Biophys. Biomol. Struct. 1999;28:319. 42. Katragadda M, Alderfer JL, Yeagle PL. Biochim. Biophys. Acta 2000;1466:1. 43. Katragadda M, Alderfer JL, Yeagle PL. Biophys. J. 2001;81:1029. 44. Papermaster D, Dreyer W. Biochemistry 1974;13:2438. 45. Smith HG, Stubbs GW, Litman BJ. Exp. Eye Res. 1975; 20:211.

46. Bennett N, Michel-Villay M, K¨uhn H. Eur. J. Biochem. 1982; 127:97. 47. Yu H, Kono M, McKee TD, Oprian DD. Biochemistry 1995;34:14963. 48. Yeagle PL, Alderfer JL, Albert AD. Nat. Struct. Biol. 1995; 2:832. 49. Hamm HE, Deretic D, Arendt A, Hargrave PA, Koenig B, Hofmann KP. Science 1988;241:832. 50. Konig B, Arendt A, McDowell JH, Kahlert M, Hargrave PA, Hofmann KP. Proc. Natl. Acad. Sci. U.S.A. 1989;86:6878. 51. Yeagle PL, Alderfer JL, Albert AD. Biochemistry 1995; 34:14621. 52. Yeagle PL, Alderfer JL, Albert AD. Molecular Vision 2, http://www.molvis.org/molvis/v2/p12/ (1996). 53. Yeagle PL, Alderfer JL, Albert AD. Biochemistry 1997; 36:3864. 54. Yeagle PL, Salloum A, Chopra A, Bhawsar N, Ali L, Kuzmanovski G, Alderfer JL, Albert AD. J. Pept. Res. 2000; 55:455. 55. Chopra A, Yeagle PL, Alderfer JA, Albert A. Biochim. Biophys. Acta 2000;1463:1. 56. Yeagle PL, Danis C, Choi G, Alderfer JL, Albert AD, Molecular Vision, www.molvis.org/molvis/v6/a17/, (2000). 57. Palczewski K, Kumasaka T, Hori T, Behnke CA, Motoshima H, Fox BA, Le Trong I, Teller DC, Okada T, Stenkamp RE, Yamamoto M, Miyano M. Science 2000;289:739. 58. Yeagle PL, Choi G, Albert AD. Biochemistry 2001;40: 11932. 59. Cohen GNR, Rickenberg HV. Compte Rendu 1955;240:466. 60. Kaback HR. J. Cell Physiol. 1976;89:575. 61. Mitchell P. Biochem. Soc. Symp. 1963;22:142. 62. Teather RM, Muller-Hill B, Abrutsch U, Aichele G, Overath P. Mol. Gen. Genet. 1978;159:239. 63. Buchel DE, Gronenborn B, Muller-Hill B. Nature 1980;283: 541. 64. Matsushita K, Patel L, Gennis RB, Kaback HR. Proc. Natl. Acad. Sci. U.S.A. 1983;80:4889. 65. Dornmair K, Corni AF, Wright JK, Jahnig F. EMBO J. 1985;4:3633. 66. Costello MJ, Escaig J, Matsushita K, Vitanen PV, Menick DR, Kaback HR. J. Biol. Chem. 1987;262:17072. 67. Frillingos S, Sahin-Toth M, Wu J, Kaback HR. FASEB J. 1998;12:1281. 68. Abramson J, Smirnova I, Kasho V, Verner G, Kaback HR, Iwata S. Science 2003;301:610. 69. Bennett M, Yeagle JA, Maciejewski M, Ocampo J, Yeagle PL. Biochemistry 2004;43:12829. 70. Arshava B, Liu SF, Jiang H, Breslav M, Becker JM, Naider F. Biopolymers 1998;46:343. 71. Xie HB, Ding FX, Schreiber D, Eng G, Liu SF, Arshava B, Arevalo E, Becker JM, Naider F. Biochemistry 2000; 39:15462. 72. Valentine KG, Liu SF, Marassi FM, Veglia G, Opella SJ, Ding FX, Wang SH, Arshava B, Becker JM, Naider F. Biopolymers 2001;59:243. 73. Arshava B, Taran I, Xie H, Becker JM, Naider F. Biopolymers 2002;64:161. 74. Piserchio A, Bisello A, Rosenblatt M, Chorev M, Mierke DF. Biochemistry 2000;39:8153. 75. Mierke DF, Royo M, Pelligrini M, Sun H, Chorev M. J. Am. Chem. Soc. 1996;118:8998.

Insight into Membrane Protein Structure from High-Resolution NMR

102. Albert AD, Watts A, Spooner P, Groebner G, Young J, Yeagle PL. Biochim. Biophys. Acta 1997;1328:74. 103. Choi G, Landin J, Galan JF, Birge RR, Albert AD, Yeagle PL. Biochemistry 2002;41:7318. 104. Farahbakhsh ZT, Ridge KD, Khorana HG, Hubbell WL. Biochemistry 1995;34:8812. 105. Altenbach C, Yang K, Farrens DL, Farahbakhsh ZT, Khorana HG, Hubbell WL. Biochemistry 1996;35:12470. 106. Farrens DL, Altenbach C, Yang K, Hubbell WL, Khorana HG. Science 1996;274:768. 107. Yang K, Farrens DL, Altenbach C, Farahbakhsh ZT, Hubbell WL, Khorana HG. Biochemistry 1996;35:14040. 108. Yang K, Farrens DL, Hubbell WL, Khorana HG. Biochemistry 1996;35:12464. 109. Cai K, Langen R, Hubbell WL, Khorana GH. Proc. Natl. Acad. Sci. U.S.A. 1997;94:14267. 110. Altenbach C, Cai K, Khorana HG, Hubbell WL. Biochemistry 1999;38:7931. 111. Klein-Seetharaman J, Hwa J, Cai K, Altenbach C, Hubbell WL, Khorana HG. Biochemistry 1999;38:7938. 112. Cai K, Klein-Seetharaman J, Farrens D, Zhang C, Altenbach C, Hubbell WL, Khorana HG. Biochemistry 1999;38:7925. 113. Altenbach C, Klein-Seetharaman J, Hwa J, Khorana HG, Hubbell WL. Biochemistry 1999;38:7945. 114. Altenbach C, Cai K, Khorana HG, Hubbell WL, Biochemistry 1999;38:7931. 115. Altenbach C, Cai K, Klein-Seetharaman J, Khorana HG, Hubbell WL. Biochemistry 2001;40:15483. 116. Sheikh SP, Zvyaga TA, Lichtarge O, Sakmar TP, Bourne HR. Nature 1996;383:347. 117. Grobner G, Burnett IJ, Glaubitz C, Choi G, Mason AJ, Watts A. Nature 2000;405:810. 118. Salzmann M, Pervushin K, Wider G, Senn H, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 1998;95:13585. 119. Fernandez C, Adeishvili K, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 2001;98:2358. 120. Schubert M, Kolbe M, Kessler B, Oesterhelt D, Schmieder P, Chembiochem 2002;3:1019. 121. Oxenoid K, Kim HJ, Jacob J, Sonnichsen FD, Sanders CR. J. Am. Chem. Soc. 2004;126:5048. 122. Krueger-Koplin RD, Sorgen PL, Krueger-Koplin ST, RiveraTorres IO, Cahill SM, Hicks DB, Grinius L, Krulwich TA, Girvin ME. J. Biomol. NMR 2004;28:43. 123. Bennett M, D’Rozario R, Sansom M, Yeagle PL. Biochemistry. 2006;45: in press.

Part I

76. Pellegrini M, Royo M, Chorev M, Mierke DF. Biopolymers 1996;40:653. 77. Giragossian C, Mierke DF. Biochemistry 2001;40:3804. 78. Pellegrini M, Mierke DF. Biopolymers 1999;51:208. 79. Giragossian C, Mierke DF. Biochemistry 2002;41:4560. 80. Giragossian C, Mierke DF. Life Sci. 2003;73:705. 81. Ulfers AL, McMurry JL, Kendall DA, Mierke DF. Biochemistry 2002;41:11344. 82. Choi G, Landin J, Xie XQ. J. Pept. Res. 2002;60:169. 83. Piserchio A, Prado GN, Zhang R, Yu J, Taylor L, Polgar P, Mierke DF. Biopolymers 2002;63:239. 84. Piserchio A, Zelesky V, Yu J, Taylor L, Polgar P, Mierke DF. Biopolymers. 2005;80:367. 85. Franzoni L, Nicastro G, Pertinhez TA, Oliveira E, Nakaie CR, Paiva AC, Schreier S, Spisni A. J. Biol. Chem. 1999; 274:227. 86. Nicastro G, Peri F, Franzoni L, de Chiara C, Sartor G, Spisni A. J. Pept. Sci. 2003;9:229. 87. Franzoni L, Nicastro G, Pertinhez TA, Tato M, Nakaie CR, Paiva AC, Schreier S, Spisni A. J. Biol. Chem. 1997;272: 9734. 88. Berlose J, Convert O, Brunissen A, Chassaing G, Lavielle S. FEBS Lett 1994;225:827. 89. Katragadda M, Maciejewski MW, Yeagle PL. Biochim. Biophys. Acta. 2004;1663:74. 90. Venkatraman J, Nagana Gowda GA, Balaram P. Biochemistry 2002;41:6631. 91. Haris PI. Biosci. Rep. 1988;18:299. 92. Gargaro AR, Bloomberg GB, Dempsey CE, Murray M, Tanner MJ. Eur. J. Biochem. 1994;221:445. 93. Askin D, Bloomberg GB, Chambers EJ, Tanner MJ. Biochemistry 1998;37:11670. 94. Morein S, Trouard TP, Hauksson JB, Rilfors L, Arvidson G, Lindblom G. Eur. J. Biochem. 1996;241:489. 95. Aggeli A, Bannister ML, Bell M, Boden N, Findlay JB, Hunter M, Knowles PF, Yang JC. Biochemistry 1998; 37:8121. 96. Sale K, Faulon JL, Gray GA, Schoeniger JS, Young MM. Protein Sci 2004;13:2613. 97. Herzyk P, Hubbard RE. J. Mol. Biol. 1998;281:741. 98. Albert AD, Yeagle PL. Methods Enzymol. 2000;315:107. 99. Yeagle PL, Albert AD. Biochem. Soc. Trans. 1998;26:520. 100. Yeagle PL, Alderfer JL, Albert AD. Biochemistry 1997;36: 9649. 101. Albert AD, Yeagle PL. Behav. Brain Sci. 1995;18:469.

References 343

Part I

New Developments

347

Ray Freeman1 and Eriks Kupˇce2 1 Jesus 2 Varian

Multidimensional NMR spectroscopy [1–3] has proved extremely productive, particularly for the elucidation of the structure of biomolecules such as proteins. The essence of the technique is to allow the nuclear spins to evolve in one or more consecutive time intervals before passing on phase or intensity information to the detection stage where the spectrometer receiver is active. In the traditional implementation, the motion of the nuclear spins is monitored systematically and independently in all n evolution dimensions (t1 , t2 , . . . tn ) at sampling rates that satisfy the Nyquist condition, and for durations that guarantee adequate resolution. This generates a well-digitized n + 1-dimensional time-domain data array, and repeated Fourier transformation converts this into an n + 1-dimensional spectrum in frequency space (F1 , F2 , . . . Fn+1 ). For years this methodology was accepted as a perfectly acceptable modus operandi. Although the duration of the measurement is long, particularly for high-dimensional spectra, the signal-to-noise ratio increases as the square root of the measurement time in the same manner as in multiscan averaging. But as NMR spectrometers become intrinsically more sensitive with the advent of higher magnetic fields and cryogenic receiver coils, often it is time that is the limiting factor rather than sensitivity. Long experimental durations impose an upper practical limit on the dimensionality, a slow throughput of spectra, and an inability to study time-dependent phenomena or unstable materials. The question naturally arises—are there more economical ways to sample n-dimensional evolution space? The answer of course is yes. Standard multidimensional NMR methodology commonly sets operating conditions that are less than ideal, simply to keep the experimental duration within reasonable bounds. Although sampling rates should normally satisfy the Nyquist condition to ensure that NMR frequencies are not aliased, sometimes a controlled “folding” of high-frequency signals may have to be tolerated in the quest for speed. The maximum length of the evolution time, t1 (max), determines the achievable resolution in the corresponding frequency dimension. A common expedient is to use a shorter value of t1 (max), thereby accepting Graham A. Webb (ed.), Modern Magnetic Resonance, 347–352.
C 2008 Springer.

College, Cambridge, UK Ltd, Eynsham, Oxford, UK

less-than-optimum resolution. Often this data set is artificially extended by linear prediction, thus avoiding truncation artifacts. But these are stop-gap measures that merely mitigate a fundamental problem crying out for more drastic solutions. This chapter examines alternative schemes for sampling evolution space. The main focus is on threedimensional spectroscopy (n = 2) although the principles apply equally well to multidimensional experiments. The aim is to discover sampling modes that are comprehensive enough to derive the full spectrum, but sufficiently economical to permit fast acquisition. Since typical three-dimensional NMR spectra are relatively sparsely populated with correlation peaks, the prognosis is optimistic. Note that any fast-acquisition mode must inevitably sacrifice signal-to-noise ratio, since signals are accumulated over a short experimental duration. Almost all fast-acquisition modes are doomed to fail in circumstances of poor sensitivity. In what follows it is implicitly assumed that the intrinsic sensitivity is adequate. Four approaches to the speed problem are described. The first (“filter diagonalization”) uses a fitting procedure in the time domain. Its key feature is the ability to compensate for sparse sampling in the evolution dimensions by fine sampling during acquisition. The second (“spatially encoded single-scan NMR”) is able to monitor all the evolution steps simultaneously by storing this information as NMR responses in different slices of the sample. The third (“Hadamard encoding”) avoids time-domain evolution entirely, using direct excitation of selected responses in the frequency domain. The fourth (“projection– reconstruction”) shortens the experimental duration by coupling evolution dimensions together. Fourier transformation generates plane projections that can be used to reconstruct the three-dimensional spectrum.

The Filter Diagonalization Method This novel technique [4–6] boasts a significant speed advantage for three-dimensional spectroscopy by monitoring the two evolution dimensions (t1 and t2 ) with very

Part I

Fast Multidimensional NMR: New Ways to Explore Evolution Space

348 Part I

Chemistry

Part I

few data points, but with comprehensive sampling of the acquisition dimension (t3 ), since this incurs no significant time penalty. The crux of the concept is that in the resulting computed spectrum, the resolution in all three frequency dimensions depends only on the volume of the experimental three-dimensional time-domain array, rather than on the number of samples in any particular dimension. Fine digitization in the t3 dimension compensates for incomplete sampling in t1 and t2 . The method abandons Fourier transformation in favor of fitting the time-domain data as a set of exponentially decaying sinusoids, in a manner similar to the better-known linear prediction method [7]. Repeated application of a time auto-correlation operator U is used to compute successive time-domain data points, thus creating a model for the NMR response that can be fitted to the experimental response. The aim is to diagonalize the operator U , giving eigenvalues that represent the line frequencies and widths, and eigenvectors that yield the amplitudes and phases. If this diagonalization were to be applied to the entire data set, the computation would be enormous, but the trick is to break it down into a set of much smaller bites of a rather large cherry. The key is to choose a set of basis functions that correspond to a set of localized (but overlapping) segments in frequency space. Because

resonances that are far apart in frequency show negligible interference effects, far off-diagonal matrix elements of the operator U can be safely neglected. A limited set of “local” or “filtered” diagonalizations are performed, neglecting basis functions beyond the boundaries of the segment under consideration. The problem then becomes tractable and the frequency segments can be assembled into a complete spectrum. Since linear algebra is involved, the fitting process is not hindered by the usual problems of false minima. However, in the presence of appreciable noise, or in very crowded regions, the procedure can become unpredictable. An illustration of the filter diagonalization technique is provided by the two-dimensional constant-time heteronuclear single-quantum correlation (HSQC) spectrum of a 1 mM aqueous solution of ubiquitin [6]. The conventional Fourier transform spectrum (Figure 1a) is compared with that derived by the filter diagonalization method (Figure 1b) where the number of samples in the evolution dimension has been reduced approximately sixfold. This was achieved by reduction of the constant-time parameter from 26.4 to 4.25 ms. Despite this reduction, the resolution in the conventional and fitted spectra is comparable, and only in the very crowded region are there any significant discrepancies.

Fig. 1. Part of the 500 MHz two-dimensional HSQC spectrum of ubiquitin. (a) The conventional Fourier transform mode with the constant-time parameter (CT) set to 26.4 ms. (b) The spectrum fitted by means of the filter diagonalization method with CT shortened to 4.25 ms, requiring only 4 min of data collection. Spectra courtesy of A.J. Shaka.

Fast Multidimensional NMR

Hadamard Encoding 349

Part I

Fig. 2. Schematic representation of the spatially encoded single-scan experiment. The active sample volume is divided into slices by selective excitation in an intense field gradient. Three representative slices are considered here. Two chemical sites A and B build up different phase handicaps during evolution. After a non-selective mixing pulse, these phase handicaps are unwound in a refocusing gradient, site B forming a spin echo earlier than site A. This “spin echo spectrum” has essentially the form of the true NMR spectrum, but is obtained in a very short time, allowing the cycle to be repeated many times during a scan of less than a second.

Spatially Encoded Single-Scan NMR One may think of this as the creation of a set of “pigeon holes” to store the evolution information, achieved in practice by selective excitation in an intense applied magnetic field gradient, thus dividing the sample into a set of thin parallel slices, excited sequentially [8–10]. The NMR signal evolves for a different interval in each slice (Figure 2). After the usual non-selective mixing pulse, the responses are detected by the application of a refocusing field gradient. (In practice both gradients are bipolar pairs.) Responses from different chemical sites, having evolved for different periods, come to a focus at different times, creating a sequence of gradient-recalled echoes (Figure 2). This “spin echo spectrum” broadly resembles the form of the true NMR spectrum in the F1 frequency domain, but it is acquired in a very short time, typically 500 µs. This is the key to the speed factor— many acquisitions of the F1 spectrum can be nested within the conventional acquisition scan of less than a second. A typical application might be a two-dimensional proton COSY or TOCSY spectrum. The F1 spectra are repeatedly recorded as a function of the acquisition parameter t2 as correlation effects gradually build up. Only a single stage of Fourier transformation is required to generate the two-dimensional spectrum. The main drawback of the technique is its relatively poor sensitivity—engendered

partly by the short accumulation time and by additional noise contributions caused by the wide frequency bandwidth employed during the spatial-encoding stage. The division of the sample into slices during evolution does not itself diminish the signal strength since NMR responses from the entire sample are brought to a focus at the same instant. Such ultrafast sampling opens up new possibilities for studying unstable materials, time-dependent phenomena, rapid screening of spectral “libraries”, and flow techniques such as hyphenated liquid chromatography–NMR [11].

Hadamard Encoding This approach abandons the conventional step-by-step exploration of evolution space, exciting the chemical sites with selective radio frequency pulses directly in the frequency domain [12–16]. Because NMR spectra are sparse, considerable time can be saved in this manner. The required NMR frequencies are obtained in a prior one-dimensional measurement that uses little spectrometer time. If the selective excitations are performed one at a time [17–20], the rate of data collection is slow, and the sensitivity is consequently poor—the famous “multiplex advantage” is lost. However by encoding the excitations (plus or minus) according to a Hadamard matrix, all the sites can be excited simultaneously, thereby restoring

350 Part I

Chemistry

Part I

the multiplex advantage. The individual responses are separated by a decoding scheme based on the same matrix. Hadamard matrices [21] are higher-order versions of the simple add–subtract matrix ++ +−

F1 (ppm) 169 171 172 173 174

+ + + + + + + +

+ + + + − − − −

+ + − − + + − −

+ + − − − − + +

+ − + − + − + −

+ − + − − + − +

+ − − + + − − +

+ − − + − + + −

Eight scans are performed with the senses of the eight selective radio frequency pulses encoded according to the rows of this matrix. Combining the eight resulting composite free induction decays according to the columns of this matrix allows the individual responses to be extracted one at a time. The speed advantage of the multiple excitation method is given by N /Q, where Q is the order of the Hadamard matrix and N is the number of increments in the evolution dimension of the corresponding conventional experiment. The number of irradiated chemical sites S must be less than or equal to Q. Note that the operator may choose to set S less than the actual number of chemical sites, selecting only sites of particular interest, thereby generating a partial spectrum. This feature can be extremely useful for converting global isotopic enrichment (in 15 N or 13 C) to essentially specific enrichment, something that is very expensive to achieve chemically. The main drawback of this method is the requirement that Q scans be completed. For three- or four-dimensional spectroscopy, the intermediate radio frequency pulses are also made selective, and encoded according to the appropriate Hadamard matrices. The speed advantages in each evolution dimension are then multiplicative. An illustration of the speed of the Hadamard method is shown in Figure 3. The sample was a 0.3 mM aqueous solution of agitoxin, a 4 kDa protein uniformly enriched in 13 C and 15 N. The 700 MHz conventional threedimensional HNCO spectrum required 20 h and 43 min of data collection. A projection of this spectrum onto the C–H plane is shown in Figure 3A. Figure 3B demonstrates how the Hadamard technique can record a subspectrum from only seven selected sites, as if the sample had been selectively (rather than globally) enriched in 13 C.

Cys-35

168 170

that is widely used in physical science. As an example, the Hadamard matrix of order eight may be written:

A

175 176

Lys-27

Cys-8 Arg-24 Ser-11 Thr-9 Met-29 Met-23 His-34 Arg-31 Gly-26 Thr-36 Cys-28 Cys-18 Phe-25 Asn-30 Ile-15 Val-2 Ser-7 Lys-38 Val-6 Lys-19 Lys-32 Asn-5 Cys-14 Ile-4 Ala-21 Gly-22 Lys-16 Cys-33 Gly-13

177

Gly-10

178 179

Asn-20

180 9.5

F1 (ppm)

9.0

8.5

8.0 7.5 F3 (ppm)

7.0

6.5

6.0

6.5

6.0

B

168 169 170 Thr-9

171

Arg-24 Met-23

172 173

Lys-19

174

Cys-33

Gly-22

175 176

Gly-13

177 178 179 180 9.5

9.0

8.5

8.0 7.5 F3 (ppm)

7.0

Fig. 3. Projection of the 700 MHz three-dimensional HNCO spectrum of agitoxin onto the C–H plane. (A) The conventional Fourier transform spectrum. Seven residues were selected at random, indicated by arrows. (B) The corresponding Hadamard spectrum of these seven residues, obtained more than 200 times faster.

The Hadamard matrix of order eight was used, requiring eight scans, thus speeding up acquisition by a factor of more than 200. The smaller the number of sites selected for excitation, the smaller the required Hadamard matrix, and the faster the measurement.

Fast Multidimensional NMR

Projection–Reconstruction 351

E38

V39

F40

L41

W42

E43

G44

S45

A46

118.6

121.5

119.9

121.9

119.4

128.6

127.2

109.8

114.5

124.8

9.39

9.20

8.68

8.19

8.16

8.89

9.11

8.92

7.47

8.31

Part I

F37 F1 (C-13, ppm) 46 48 50 52 54 56 58 60 62 64 66 68 70

F2 (N-15, ppm)

F3 (H-1, ppm) Fig. 4. Strip plots from the three-dimensional HNCA spectrum of nuclease A inhibitor mapping a chain of nine residues correlated through their 15 N and 13 C resonances. The projection–reconstruction technique shortens the experimental duration by an order of magnitude.

Projection–Reconstruction The “accordion” experiment [22,23] and subsequent extensions [24–27] save time by coupling the incrementation of two evolution parameters t1 and t2 , rather than scanning them separately. The evolving signal is recorded along a skew section through the time-domain data at an angle α given by tanα = t2 /t1 . Compared with systematic sampling of all N 2 elements of evolution space, this saves a factor of approximately N in spectrometer time (although slightly faster sampling is required along the skew axis). Fourier transformation of these signals generates a projection of the three-dimensional spectrum onto a plane inclined at the same angle α. With the standard quadrature detection in each evolution dimension, the technique gives two projections inclined at ±α. Thus by choosing the relative rates of incrementation, a projection can be obtained on a suitably tilted plane in frequency space.

It is well known from X-ray tomography [28] that an image of a three-dimensional object can be reconstructed from a set of projections taken at different angles of incidence. NMR spectra present a much more promising case, for they are made up of relatively sparse, discrete resonances, whereas the absorption in a physiological sample is continuous. As a result, an NMR spectrum can be reconstructed from a quite small number of projections recorded at different angles [29–33]. For threedimensional spectroscopy the initial projections are those on the orthogonal F1 F3 and F2 F3 planes which are often used during the setting up procedure. The information content of these two projections is insufficient to reconstruct the three-dimensional spectrum, but it defines all conceivable positions for the cross-peaks, as if every resonance in the F1 F3 plane were correlated with every resonance in the F2 F3 plane. The actual cross-peaks are identified by imposing further constraints based on tilted projections, since the final three-dimensional array

352 Part I

Chemistry

Part I

must be compatible with all the measured projections. If the signal-to-noise ratio is only marginal, projections are recorded at several different tilt angles and the inverse Radon transform [34] is employed for the reconstruction. The 800 MHz three-dimensional HNCA spectrum of a 3 mM aqueous solution of the 143-residue nuclease A inhibitor [35] serves as an illustration of the projection– reconstruction technique [33]. The strip plots of Figure 4 show 15 N–13 C correlations for a chain running between residues 37 and 46. Based only on projections recorded at 0◦ , ±30◦ , ±60◦ , and 90◦ , the measurements were completed in 1 h, compared with an estimate of 11 h for the conventional mode.

Acknowledgments The authors thank Lucio Frydman for several illuminating discussions, Gerhard Wagner for the sample of agitoxin, Robert London for the sample of nuclease A inhibitor, A.J. Shaka for permission to reproduce Figure 1, and the Journal of Biomolecular NMR for permission to reproduce Figure 3.

References 1. Jeener J. Amp`ere International Summer School, Basko Polje, Yugoslavia, 1971. 2. Aue WP, Bartholdi E, Ernst RR. J. Chem. Phys. 1976;64:2999. 3. Bax A. Two-Dimensional Nuclear Magnetic Resonance in Liquids. Delft, The Netherlands: Delft University Press, 1982. 4. Hu H, De Angelis AA, Mandelshtam VA, Shaka AJ. J. Magn. Reson. 2000;144:357. 5. Chen J, Mandelshtam VA, Shaka AJ. J. Magn. Reson. 2000;146:363.

6. Chen J, De Angelis AA, Mandelshtam VA, Shaka AJ. J. Magn. Reson. 2003;161:74. 7. Barkhuijsen H, DeBeer R, Bov´ee WMMJ, van Ormondt D. J. Magn. Reson. 1985;61:465. 8. Frydman L, Scherf T, Lupulescu A. Proc. Natl. Acad. Sci. U.S.A. 2002;99:15859. 9. Frydman L, Lupulescu A, Scherf T. J. Am. Chem. Soc. 2003;125:9204. 10. Shrot Y, Frydman L. J. Am. Chem. Soc. 2002;125:11385. 11. Shapira B, Karton A, Aronzon D, Frydman L. J. Am. Chem. Soc. 2004;126:1262. 12. Kupˇce E, Freeman R. J. Magn. Reson. 2003;162:158. 13. Kupˇce E, Freeman R. J. Magn. Reson. 2003;162:300. 14. Kupˇce E, Freeman R. J. Magn. Reson. 2003;163:56. 15. Kupˇce E, Freeman R. J. Biomol. NMR. 2003;25:349. 16. Kupˇce E, Nishida T, Freeman R. Prog. NMR Spectrosc. 2003;42:95. 17. Kupˇce E, Freeman R. J. Magn. Reson. A. 1993;102:122. 18. Kupˇce E, Freeman R. J. Magn. Reson. A. 1993;105:234. 19. Kupˇce E, Freeman R. J. Magn. Reson. A. 1993;105:310. 20. Blechta V, Freeman R. Chem. Phys. Lett. 1993;215:341. 21. Hadamard J. Bull. Sci. Math. 1893;17:240. 22. Bodenhausen G, Ernst RR. J. Magn. Reson. 1981;45:367. 23. Bodenhausen G, Ernst RR. J. Am. Chem. Soc. 1982;104:1304. 24. Ding K, Gronenborn AM. J. Magn. Reson. 2002;156:262. 25. Kim S, Szyperski T. J. Am. Chem. Soc. 2003;125:1383. 26. Kim S, Szyperski T. J. Biomol. NMR. 2004;28:117. 27. Kozminski W, Zhukov I. J. Biomol. NMR. 2003;26:157. 28. Hounsfield GN. Brit. J. Radiol. 1973;46:1016. 29. Kupˇce E, Freeman R. J. Biomol. NMR. 2003;27:383. 30. Kupˇce E, Freeman R. J. Am. Chem. Soc. 2003;125: 13958. 31. Kupˇce E, Freeman R. J. Biomol. NMR. 2004;28:391. 32. Kupˇce E, Freeman R. Concepts Magn. Reson. 2004;22A:4. 33. Kupˇce E, Freeman R. J. Am. Chem. Soc. 2004;126:6429. 34. Deans SR. The Radon Transform and Some of its Applications. Wiley: New York, 1983. 35. Kirby TW, DeRose EF, Mueller GA, Meiss G, Pingoud A, London RE. J. Mol. Biol. 2002;320:771–82.

353

P.J.M. van Bentum and A.P.M. Kentgens Institute for Molecules and Materials, Radboud University Nijmegen, 6525ED Nijmegen, The Netherlands

Abstract Nuclear magnetic resonance (NMR) has become the method of choice for many types of applications. Still, sensitivity is a limiting factor in the applicability of NMR, leading to long measurement times in advanced multidimensional experiments, and becoming prohibitive when very limited sample quantities are available. This low sensitivity is mostly an intrinsic consequence of the low energy scale of the nuclear moment in a static field, when compared to other thermodynamic energies like kB T . The commercial developments are mostly aimed at an increase in the static field and simultaneously a reduction of the noise using cryocooled detection coils. Current research shows a number of interesting developments toward the enhancement of the nuclear polarization by optical pumping or by transfer from the electronic bath in dynamic nuclear polarization (DNP) experiments. A more technological approach is based on the miniaturization of the RF coils. In the next decade, one may expect the advent of the “lab on a chip” with in situ chemical processing and NMR analysis capabilities. A brave new method to improve detection sensitivity is based on very sensitive micromechanical force detectors. Recently, it was demonstrated that the low-temperature force detection sensitivity is sufficient to detect the magnetic moment of a single (electron) spin. These developments show that the NMR detection limits in terms of absolute sensitivity or imaging resolution are still open to significant improvements.

Sensitivity Issues in NMR Spectroscopy NMR and magnetic resonance imaging (MRI) have had a tremendous impact on research in physics, chemistry, biology, and medicine. This success is rather surprising, given the fact that even at the highest possible fields the nuclear Zeeman splitting, and thus the electromagnetic radiation to pump the transitions, remains much smaller than the thermal energy kB T at room temperature. Hence, the equilibrium magnetization that can be manipulated and detected in an NMR experiment is many orders of magnitude smaller compared to e.g. electron spin resonance (ESR). In optical, infrared or mass spectroscopy one is used to work with near quantum efficiency detectors, and Graham A. Webb (ed.), Modern Magnetic Resonance, 353–361.
C 2008 Springer.

a single photon or mass fragment can be detected with a reasonable signal-to-noise ratio. In radio-frequency detectors, this is not the case. In spectroscopy, typical sample sizes are of the order of 10–100 mm3 and one typically requires more than 1016 nuclei in the sample. It is clear that there are many research topics that would benefit from even a modest improvement in sensitivity or resolution. An improvement in sensitivity by a factor of 10 would bring the data acquisition times down from days to minutes and would allow for example online quality control in chemical or pharmaceutical production. In chemical synthesis, one would be able to reduce the volume of the reactants to microliters and be able to map out the full phase diagram in an in-situ “lab on a chip” procedure. Apart from the gain in time there is also an environmental advantage because the flow of waste chemicals is substantially reduced. Gradual improvements of equipment and measurement procedures have pushed up the sensitivity of NMR by nearly a factor of 10 in the last decade. Current research shows a number of interesting new or renewed developments toward sensitivity enhancements through polarization transfer or detector optimization. In the following, we will first summarize the options available to increase the effective magnetization of a given sample. Mostly, these are based on thermodynamic approaches, including higher magnetic fields, lower temperatures, and transfer of magnetization. This can be for example the transfer of magnetization from protons to low-gamma nuclei in a cross-polarization (CP) experiment. Also, polarization transfer from electrons in dynamic nuclear polarization (DNP) or from optically polarized nuclei, like Xe and 3 He, is feasible. In some cases, one can pump the molecules directly, and combined with an optical detection full spin polarization can be reached, at least for electron spins. In half-integer quadrupolar spins systems generally only the central transition of the multiplet is observed as the other transitions are much more strongly affected by the quadrupolar interaction. In this case, one can transfer polarization from the satellite transitions to this central peak using adiabatic passages. In the second part, we will review the basic physical laws that determine the inductive response of a traditional RF coil. We will discuss the options available to optimize the inductive detection for mass limited samples such as

Part I

High-Sensitivity NMR Probe Systems

354 Part I

Chemistry

Part I

small crystals or thin surface layers. In the last part of this contribution, we will discuss approaches that move away from the traditional inductive detections, like for example the detection of the magnetic force in a magnetic resonance force microscopy (MRFM) scanning probe setup.

Thermodynamics For a nuclear spin system, the magnetization (or total magnetic moment per unit volume) is given by Curie’s law in the limit for high temperatures: M = N γ 2h¯ 2 I (I + 1)

B0 3k B T

where N is the number of spins per unit volume, γ the gyromagnetic ratio, I the spin quantum number, B0 the static field, T the temperature, and h¯ and kB are Planck’s and Boltzmann’s constant, respectively. For a give nucleus studied at room temperature, one can only manipulate the concentration N or the static field B0 . Technological progress in high-homogeneity magnets has pushed the field limit to 22.3 T (950 MHz). This is near the limit that can be produced with conventional niobium-based superconductors. With high Tc superconductor insert coils, 1 GHz can be reached, but the technological problems in reliability and reproducibility are still substantial. On the other hand, if the samples allow measurements at low temperatures, a substantial gain can be achieved. The equilibrium magnetization at very low temperatures is given by: M = N γ h¯ I, since now only the lowest level is occupied. For protons in a field of 14 T at 300 K, the equilibrium magnetization is thus of the order of 5 × 10−5 of the low temperature fully polarized limit. For very small nano-crystals, it therefore pays off to go to special facilities that employ installations with a very high B/T ratio. In various high magnetic field facilities, one typically can reach a base temperature of 40 mK at a static field of 40 T, in which case the low temperature limit is well satisfied. An early example of this approach is given by Gonen et al. in 1989 [1]. They showed that it is possible to detect the 13 C signal of a chemisorbed CO layer on SnO2 oxidizing catalyst. Despite the low number of spins in their sample, the low-temperature population enhancement allowed a fully resolved spectrum in a single scan. Note that the spin–lattice relaxation time T1 may become very long at low temperatures, so this approach is mostly relevant for one-dimensional experiments where no signal averaging or phase cycling is necessary. Also, one often needs to study the material at ambient temperatures, for example

in liquid solution, and cryogenic cooling of the sample is not an option.

Polarization Transfer There are various options to increase the magnetic polarization of the nuclei under study. In typical NMR sequences, one can transfer the polarization from abundant protons with a much larger Zeeman splitting to the low-gamma nuclei such as 13 C in a CP experiment. The enhancement factor is typically of the order of the gyromagnetic ratios, and is thus substantial for very lowgamma nuclei. In liquid NMR, many variations of the nuclear overhauser effect (NOE) [2] are used to transfer polarization to for example the 13 C nuclei. In general, all these methods rely on the fact that the total (coupled) spin system tries to keep overall thermal equilibrium when one of the transitions is saturated by RF radiation. For example in distortionless enhancement by polarization transfer (DEPT) type experiments, one 1 H transition is saturated, leading to a strong enhancement of the 13 C resonance (Figure 1). Another option to beat the thermodynamic equilibrium is to use the hyperfine coupling between electrons and nuclei to transfer effective magnetization from the electrons to the nuclei. Since the electron magnetic moment is much larger than all nuclear moments, one can achieve an appreciable electronic polarization even at ambient conditions. With the method of DNP, one basically saturates the electron spin system with high-power microwaves at the ESR. In the so-called solid effect DNP [3], one selectively excites the four-level system of the coupled two-spin state (see fig. 2). Effective polarization between the electron and nuclear bath is exchanged in the zero and double quantum transitions, which may have quite different probabilities. In theory, this might give a magnetization enhancement by as much as the gyromagnetic ratio between the electron and the nuclei (γe /γp = 658). In the solid effect scheme, it is required that the ESR line width is much smaller than the nuclear Larmor frequency. This is not easy to realize in practice and generally the homogeneous and inhomogeneous line broadening is considerable. However, by off-resonance excitation of the ESR transition one can still lower the nuclear spin temperature in a so-called thermal mixing scheme [4]. In practice, an enhancement by a factor of 400 has been demonstrated. The main problem is to find a suitable paramagnetic species that couples to the nucleus but does not disturb the local environment that is studied in the NMR experiment. Also, there is a substantial problem in moving to higher fields, where suitable sub-millimeter sources (above 500 GHz) are lacking. It should be noted, however, that good progress is made by the group of Griffin who have demonstrated the feasibility of DNP-enhanced

High-Sensitivity NMR

|− −>

|− +>

|− +>

|+ −>

|+ −>

|+ +>

|+ +>

Fig. 1. Illustration of the polarization transfer in distortionless enhancement by polarization transfer (DEPT) type experiments for a system of two coupled spins (1 H and 13 C). The corresponding NMR resonance signals are indicated by red (13 C) and blue (1 H), where the linewidth symbolizes the relative intensity. The left side represents the system in thermal equilibrium. If the proton resonance is saturated by the external RF field, indicated by the black arrow, then the spin temperature of the carbon subsystem is cooled, leading to a stronger resonance signal.

magic-angle spinning (MAS) NMR spectroscopy for biologically relevant materials at fields up to 9 T [5]. In high energy physics, fully polarized solid proton or deuterium targets are used to study the spin structure of the nucleon with muon scattering. In this case, a combination of cryogenic cooling (to 40 mK) and DNP is used to create a nearly 100% polarization. Such an approach was also used by Golman and coworkers [6] in order to polarize liquids, which can subsequently be used for metabolic studies in MRI.

|-e +p> |-e -p> ∆m=2 ∆m=0 |+e +p> |+e -p> Fig. 2. Illustration of the solid effect dynamic nuclear polarization (DNP). The blue arrows represent the electron spin resonance (ESR) transitions, while the red arrows refer to the nuclear resonance. The cross transitions represented by black arrows are the double quantum (flip–flip) and the zero quantum (flip–flop) transitions that transfer effective polarization from the electron to the nuclear system. Under proper conditions, this can give a nearly 100% polarization of the nuclei.

Half-integer quadrupolar spin systems have multiple spin levels without coupling to other nuclear or electron spins. In spectra of powdered samples, commonly, only the central −1/2 ↔ +1/2 transition is observed, while all other transitions are broadened over a MHz-wide frequency range. It is possible to use either pulses or adiabatic sweeps to invert the population of two adjacent levels [7,8,9]. If this is done simultaneously for both satellite transitions, for example in a so-called double frequency sweep [10] or fast amplitude-modulated pulse train [11], one can increase the population difference between the central levels to that of the outer levels in the multiplet. The procedure is schematically illustrated in Figure 3. For nuclei with spin I = 3/2, this can give a threefold increase in intensity for the central transition, and thus a factor 9 reduction in measurement time to obtain the same signal-to-noise ratio [11,12]. For higher spin nuclei like 27 Al with spin 5/2, one can get an enhancement up to 5 depending on the details of the adiabatic sweep method. With repetitive sweeps one can transfer basically all the polarization and enhancements near the theoretical maximum are observed in practice [13]. A final trick to induce a nuclear magnetization above the thermodynamic equilibrium is based on the selection rules in optical transitions. In an alkali metal vapor one spin selectively pumps an atomic transition and by spinexchange collision this magnetization in transferred to a noble gas of 3 He or 129 Xe atoms [14,15]. This optical pumping can lead to a nearly 100% polarization of the nuclear moment, and because of the very long relaxation times the gas can be transported to the experiment. In medical (lung inhalation) or catalysis (zeolites), the effective

Part I

|− −>

Polarization Transfer 355

356 Part I

Chemistry

Part I Fig. 3. Illustration of the signal enhancement in a quadrupolar nucleus with spin 3/2. On the left, we have the situation in thermal equilibrium with the experimental spectrum below showing three resonances as expected. The resonances corresponding to the lower and upper transition are considerably broadened. As shown on the right, the population levels can be inverted using adiabatic sweeps. Under proper conditions, this may give a factor 3 enhancement of the intensity in the central transition, and thus nearly a factor of 10 gain in measurement time.

surface area is big enough to allow a reasonable enhancement of the magnetization of the relevant nuclei in the sample. For example it was recently demonstrated by Jansch et al. [16] that a single monolayer of hyperpolarized Xe adsorbed on an Ir(111) single crystal surface could be measured successfully by NMR spectroscopy. In this case, the large Knight shift that was observed indicates a large overlap of the conduction electron wave function at the Fermi level of the metal with the Xe atomic states at the surface. Very interesting from a chemical point of view is the proposition of Bowers and Weitekamp [17–19] to use parahydrogen ( p-H2 ) in hydrogenation reactions, allowing the detection of reaction products and intermediates with high sensitivity. p-H2 exhibits a pure nuclear singlet state and is stable even in liquid solutions. Due to the symmetry breakdown of the parahydrogen during hydrogenation reaction typical polarization patterns are observed in the NMR spectra of the hydrogenation products. In a simple AX-spin system, the |αβ and |βα spin functions are overpopulated relative to a normal Boltzmann distribution. This gives signal enhancements of some orders of magnitude but typically ranges around a factor of a few

thousand. The method is extremely sensitive for the detection of short-lived intermediates and reaction products in small concentrations [20]. Note that one is not restricted to the noble gasses for efficient optical nuclear polarization (ONP). As was first suggested by Kastler in 1952, any two-level spin system can be used for optical pumping [21]. The procedure is illustrated in Figure 4. Because the optical photon in the beam of circularly polarized light carries an angular momentum of one, the selection rules dictate that only transitions from the ground state with J = −1/2 to the excited state with J = +1/2 are allowed. Relaxation paths generally leave the angular momentum untouched and the electrons in the excited state relax to the ground state and populate the upper spin level. If there is a spin flip during the relaxation, the electron ends up in the original level, and the pumping scheme can start all over again until virtually all electrons end up in the inverted spin ground state. Since the optical pumping depends on light intensity, one uses powerful laser systems or UV light sources, and a nearly 100% polarization of the electron spins can be achieved in rather short times (typically in the microsecond range) [22,23].

High-Sensitivity NMR

|g > Fig. 4. Illustration of the optical nuclear polarization (ONP) process. With circular polarized light, one can spin selectively pump the transition from the electronic ground state to an excited level. After relaxation, this gives a fast polarization of the ground state electron population. As in the case of NOE, this polarization is then transferred to the nuclear system.

As in the case of DNP, the electron polarization is transferred to the nuclei by the hyperfine coupling and by spin diffusion, polarization is transferred to the molecule under study. Theoretically this could achieve a polarization enhancement of about 4 orders of magnitude. In practice, however, one is limited by the choice of suitable doping molecules, the efficiency of the spin diffusion on the T1 timescale etc. Also, the optical pumping is most efficient in the low field range and the sensitivity is offset by the effect of the lower B0 field. In some cases, one can shuttle the sample between low and high field positions, but this does not appear to be a practical solution for all experiments. In special topics, however, this method can be quite useful and for example the NMR spectrum of individual layers in semiconductor quantum wells can be resolved [24]. Note that in ONP, one typically uses optical methods to measure the magnetization as well (ODMR). Since optical photons can be detected with near quantum efficiency, this can be very sensitive and in fact already in 1993 two groups demonstrated that (electron) magnetic resonance of a single molecule can be detected [25,26]. None of the above methods is sufficiently versatile to be employed as a standard enhancement tool. However, in specific cases, these techniques can be very useful and more research is needed to widen the application area into mainstream NMR spectroscopy.

Optimized Detection Coil Design For solid-state NMR, the most common approach to detect the NMR signals is the inductive detection in a helical

coil wound around the sample cylinder. After a 90◦ pulse, the magnetization rotates in the laboratory frame at the Larmor frequency and the oscillating flux induces an EMF that can be measured in high-sensitivity digital quadrature detection. The noise of high-frequency components like oscillators, mixers, and IF amplifiers is nowadays at such a level that the effective noise in the signal is dominated by the resistive noise in the pickup coil. The signal-to-noise ratio in a typical NMR experiment can be written as: [27]  √  k0 (B1 /i)Vs ω0 1/ 2 M S k0 B1 Vs 2 ∝ √ , = N F 4kB T Rnoise  f i R where k0 is a scaling factor accounting for the RFinhomogeneity of the coil, B1 /i is the magnetic field induced in the RF coil per unit current, and VS is the sample volume. M is the magnetization as described before. The denominator describes the noise using the noise factor of the spectrometer (F), conductive losses of the coil, circuit, and sample (Rnoise ) for the spectral bandwidth  f . The main factor that saves the day for inductive detection is the Larmor frequency in the nominator because the time derivative of the magnetic flux through the coil scales with ω0 . As is clear from the right hand side of this equation, the rule of thumb is to make a detector coil with a homogeneous field (k0 = 1), a high field factor B1 /i, and a low resistance R. As usual there is no single solution to a multitude of problems. In high-resolution liquid NMR, the optimal configuration is that of a saddle coil. This geometry allows the best static field homogeneity and gives sample access along the bore axis of the magnet. However, it comes at a price, since the field factor is generally lower when compared with a helix. (If we consider a cylindrical volume with the length of the cylinder equal to the diameter, then the B1 /i field factor for a saddle coil is only 60% of the helix version). In addition, the length of wire is larger for the saddle coil, making the total sensitivity a factor of three less than the helix version. A similar argument is true for imaging experiments. To have a wide access bore, one cannot use helical coils oriented perpendicular to the field axis. On the other hand, a combination of orthogonal saddle coils can detect the full circular polarized magnetization of the processing spins gaining a factor equal to the square root of two. In MRI, one typically uses birdcage coils that are basically just a series of phase-shifted saddle coils. Also, because the B1 field gradient is mostly close to the wires or strips that form the saddle coil, one can move to cryogenically-cooled RF coils without losing too much effective space for the sample. In this case, a decrease in temperature (from 300 to 25 K) leads to a factor of 3–4 reduction in noise. Although the helical coil is theoretically more sensitive, this geometry is much more prone to susceptibility effects and unless special matching

Part I

|e >

Optimized Detection Coil Design 357

358 Part I

Chemistry

Part I

fluids are used to embed the coil, one sees deterioration in resolution, and for narrow lines therefore also a reduction in sensitivity. In solid-state NMR, this is less of a concern and in this field the helix largely dominates. Note, however, that in experiments where the B1 homogeneity is essential, one can use only part of the available volume. If we allow a 10% variation in B1 field, the effective volume is about half the space inside the coil. Given the fact that roughly one half of the magnetic field energy is anyway outside the coil, it is clear that none on the conventional geometries are ideally matched to the NMR problem. In order to have the best detection sensitivity, one needs to place the detection coil as close as possible to the sample. Put in another way, if we increase the sample volume V , we do not gain a linear increase in signal but as a result of the 1/R term in the inductive voltage, we see only a V 2/3 increase. If we have only a small amount of sample, we have no other option than to scale the detection coil in a proper way. This is the rationale behind various development efforts to produce very small microcoil probes using lithographic or micromachining methods [28–30]. The smallest helical coil that was reported for NMR spectroscopy had a diameter of about 20 µm. When compared with a traditional probe of a few millimeters diameter, this microcoil would give a better sensitivity by about 2 orders of magnitude. It was indeed demonstrated by Seeber et al. [31] that with this type of solenoidal microcoils one can attain a sensitivity sufficient to measure the NMR proton signal in a water sample of only 10 femtoliter, or 7 × 1011 spins with a signal-to-noise ratio of 1 in a single scan. Solid-state NMR probeheads using solenoid microcoils with an inner diameter of 300–400 µm have been implemented for the study of mass-limited solid samples [30]. The performance, in terms of sensitivity and RFcharacteristics, of these probeheads was demonstrated for 1 H, 31 P, and 27 Al in different model compounds. The sen√ sitivity is approximately 1014 spins/ Hz to get a signalto-noise ratio of 1 in a single scan. A specific advantage of microcoils for solid-state NMR applications is that they can generate extremely high RF-fields if implemented in appropriate circuits. Using RF powers in the hundreds of Watts range, RF fields up to 5 MHz were realized. This allows the excitation of spectra of nuclei whose resonance lines are dispersed over several MHz. This is particularly useful for quadrupolar nuclei experiencing large quadrupolar interactions. This type of microcoil circuitry also proved compatible with double tuning allowing the implementation of double resonance experiments as is shown for 13 C CP spectroscopy of glycine in Figure 5 [32]. One of the appealing aspects of lithographically produced RF coils is that it may fulfill the promise of a “lab on a chip” in which the material synthesis and analysis are integrated on a single chip [33,34]. It is even possible to integrate the NMR RF source and data acquisition

Fig. 5. Static 13 C cross-polarization spectrum of a 25% 13 C2 enriched glycine sample, containing 6 microcrystals (100 µm)3 measured in a dual-channel microcoil probe, averaged in 5000 scans. The inset shows the microcoil configuration with the helix embedded in the center of a low loss capacitor to form the resonant LC circuit [32]. (See also Plate 42 on page XXXIV in the Color Plate Section.)

electronics on the same chip. An additional advantage of the high sensitivity and small volume of the microcoil approach is that one can multiplex the excitation and data acquisition over many different samples in an efficient way. This allows a very high throughput of the instrument (Figure 6) [35].

Magnetic Resonance Force Microscopy A relatively recent method for improving the detection sensitivity of NMR is based on mechanical detection. Although the ideas with respect to the mechanical detection of the magnetic resonance phenomenon were proposed as early as 1964 and its concepts proven in 1967 for electron spins [36], mechanical detection of the nuclear magnetic moment was only achieved much later. The first method that was successfully applied to this end is called magnetic resonance force microscopy as proposed by Sidles in 1991 [37]. It was demonstrated for magnetic resonance detection of electrons by Rugar and coworkers in 1992 [38] and for protons in 1994 [39]. It uses a mechanical cantilever as is known from atomic force microscopy (AFM) to detect forces exerted on a spin system in a very inhomogeneous magnetic field. The deflection of the cantilever can be measured very accurately with sub-angstrom resolution by optical methods. In the most common situation where both the field gradient and the modulated component of the magnetic moment are pointed along the z-axis, we can write the time dependent force F(t) on the sample as: ∂ B0 Fz (t) = Mz (t) dV . ∂z V

High-Sensitivity NMR

Magnetic Resonance Force Microscopy 359

Part I

Fig. 6. COSY spectra acquired in a probe with eight microcoils of about 300 µm diameter, resulting in a typical sample volume of 35 nl. The small space required for these capillary sample tubes allows one to measure multiple compounds quasi simultaneously. Each sample (10 mM solution in D2 O) was loaded into the coil via the attached teflon tubes. (A) sucrose, (B) galactose, (C) arginine, (D) chloroquine, (E) cysteine, (F) caffeine, (G) fructose, and (H) glycine. Results are shown corresponding to an averaging of eight scans (reproduced from ref. [35]).

Furthermore, as a result of the presence of the magnetic field gradient, the Larmor resonance condition, ω0 = γ B0 , varies over the sample, i.e. becomes spatially dependent. Thus, we have the option to selectively excite slices from the sample through variation of the irradiation frequency or by altering the position of the magnetic field gradient source. In nuclear MRFM, the gradient field is usually brought about by introducing a small magnetic particle in an otherwise homogeneous magnetic field (B0 ). Since this external B0 magnetic field is very strong, the magnetic particle will be completely saturated. The main reason for the high sensitivity in the MRFM experiment is the fact that the gradient is introduced by a microscopic particle that is well matched to the small size of the sample. Also, the current densities that mimic

the magnetization of a saturated Ferro magnet are much higher than what is common in electrical circuits. Finally, the friction losses in a mechanical resonator are generally much smaller than the resistive losses in a coil, leading to higher Q values of the resonator. For protons, a breakeven point is typically reached for samples of a few hundred micrometers in size. In smaller samples, the MRFM method will be more sensitive. For a 10 µm sample in a hypothetical RF coil of 50 µm diameter, the inductive detection limit will be typically of the order of 1012 spins in a bandwidth of 1 Hz, corresponding to a concentration of about 1 mol/l. For the same sample size, a mechanical detection at room temperature may give an order of magnitude better sensitivity. The ultimate goal of MRFM is to improve detection sensitivity to the single nuclear

360 Part I

Chemistry

Part I Fig. 7. Configuration of the single-spin MRFM experiment. The magnetic tip at the end of an ultrasensitive silicon cantilever is positioned approximately 125 nm above a polished SiO2 sample containing a low density of unpaired electron spins. The resonant slice represents those points in the sample where the field from the magnetic tip (plus an external field) matches the condition for magnetic resonance. As the cantilever vibrates, the resonant slice swings back and forth through the sample causing cyclic adiabatic inversion of the spin. The cyclic spin inversion causes a slight shift of the cantilever frequency owing to the magnetic force exerted by the spin on the tip. Spins as deep as 100 nm below the sample surface can be probed (reproduced from ref. [40]). (See also Plate 43 on page XXXIV in the Color Plate Section.)

spin level, and thus enable three-dimensional imaging of macromolecules (for example, proteins) with atomic resolution. MRFM has also been proposed as a qubit readout device for spin-based quantum computers. A breakthrough was established recently when Rugar et al. reported the first successful detection of an individual electron spin by MRFM [40]. The spatial resolution that can be achieved in this setup is of the order of 25 nm as demonstrated for an unpaired spin in silicon dioxide (Figure 7). Despite the impressive detection sensitivity, a few words of caution are in order. First, the experiment is done at very low temperatures where the mechanical noise of the cantilever is very small. At ambient temperatures the noise increases substantially. Combined with the thermodynamic population and the much lower moment of the nucleus as compared to the electron spin, one cannot hope to achieve such ultimate sensitivity in a routine NMR experiment. Moreover, since the very strong local gradient is essential in the detection mechanism, it is not straightforward to do spectroscopy with any significant resolution. On the other hand, it should be able to improve the sensitivity and thus the spatial resolution in a magnetic resonance microscopy experiment by several orders of magnitude. The main challenge in the field is to find suitable methods to combine the superior detection sensitivity

of mechanical detection with the high frequency of operation, preferably at the Larmor frequency and without the loss in spectral resolution caused by the static field gradient. A possible solution in this direction was proposed by Weitekamp [41]. In this so-called BOOMERANG configuration, a compensating gradient ring is positioned around the magnetic particle on the cantilever, in such a way that the sample sees a nearly homogenous field (allowing spectroscopy) while the force between cantilever and sample remains. In conclusion, there are no strict physical laws that prohibit further improvement of sensitivity and there are several paths that lead toward sensitivity optimization in NMR. In fact, there are at least two proven methods to detect a single electron spin and it is conceivable that with further improvements we will see single nucleus detection. Even for the traditional inductive detection, there is no fixed limit and for specific cases we can expect an increase in sensitivity by several orders of magnitude. The bigger challenge is to find ways to improve the sensitivity without compromises to the wealth of information that can be obtained with modern pulsed NMR techniques. The quest for ultimate sensitivity without this in mind may become rather academic. For the near future, the detection method of choice depends very much on the sample shape and characteristics. For small solid-state samples with broad resonance lines, the optimum configuration is probably that of a suitably matched microcoil. In this case, one profits both from the high-sensitivity and the high excitation fields. This configuration can also be used for micrometer scale imaging where one combines the high sensitivity with the ability to produce the large pulsed field gradients that are needed to obtain the required spatial resolution. In the quest for the ultimate imaging resolution, the MRFM technique is clearly at the forefront and true chemical (quadrupolar) contrast may be possible down to the 10–100 nm scale. This method is particularly suitable for low field applications and low-gamma nuclei. For the ultimateresolution liquid state NMR, the present method of choice is the cryocooled saddle coil inductive detection. The relatively low concentrations in for example protein solutions do not allow a downscaling to very small coil sizes. In addition, the susceptibility issues connected with (micro)coils very close to the sample are not easily solved and parts per billion resolutions are still to be demonstrated. There are options to restore the full resolution even in inhomogeneous static fields, but these generally lead to lower sensitivities. Sensitivity enhancement methods based on DNP are still in the research stage, with as the main bottleneck the absence of suitable mm-wave sources. Another niche in NMR spectroscopy is the study of thin surface layers. Methods to achieve a reasonable sensitivity for this configuration are still very much in its

High-Sensitivity NMR

References 1. Gonen O, Kuhns PL, Waugh JS, Fraissard JP. J. Phys. Chem. 1989;93:504. 2. Overhauser AW. Phys. Rev. 1953;92:411. 3. Abragam A. The Principles of Nuclear Magnetism. Clarendon Press: Oxford, 1961. 4. Farrar CT, Hall DA, Gerfen GJ, Inati SJ, Griffin RG. J. Chem. Phys. 2001;114:4922. 5. Bajaj VS, Farrar CT, Mastovsky I, Vieregg J, Bryant J, Elena B, Kreischer KE, Temkin RJ, Griffin RG. J. Magn. Reson. 2003;160:85. 6. Ardenkjaer-Larsen JH, Fridlund B, Gram A, Hansson G, Hansson L, Lerche MH, Servin R, Thaning M, Golman K. Proc. Nat. Acad. Sci. U.S.A. 2003;100:10158. 7. Pound RV. Phys. Rev. 1950;79:685. 8. Vega S, Naor Y. J. Chem. Phys. 1981;75:75. 9. Haase J, Conradi MS. Chem. Phys. Lett. 1993;209:287. 10. Kentgens APM, Verhagen R. Chem. Phys. Lett. 1999;300:435. 11. Madhu PK, Goldbourt A, Frydman L, Vega S. Chem. Phys. Lett. 1999;307:41. 12. Iuga D, Schafer H, Verhagen R, Kentgens APM. J. Magn. Reson. 2000;147:192.

13. Kentgens APM, van Eck ERH, Ajithkumar TG, Anupold T, Past J, Reinhold A, Samoson A. J. Mag. Res. 2006;178:66. 14. Pietrass T, Gaede HC. Adv. Mater. 1995;7:826. 15. Happer W, Miron E, Schaeffer S, Schreiber D, Vanwijngaarden WA, Zeng X. Phys. Rev. A. 1984;29:3092. 16. Jansch HJ, Gerhard P, Koch M, Stahl D. Chem. Phys. Lett. 2003;372:325. 17. Bowers CR, Weitekamp DP. Phys. Rev. Lett. 1986;57:2645. 18. Bowers CR, Weitekamp DP. J. Am. Chem. Soc. 1987;109:5541. 19. Carson PJ, Bowers CR, Weitekamp DP. J. Am. Chem. Soc. 2001;123:11821. 20. Grant DM, Harris RK. Advances in NMR, Vol.9, Encyclopaedia of NMR. Wiley: 2002, p. 598. 21. Brossel J, Kastler A, Winter J. J. de Physique et Le Radium. 1952;13:668. 22. Buntkowsky G, Hoffmann W, Vieth HM. Appl. Magn. Reson. 1999;17:489. 23. Suter D. J. Magn. Reson. 1992;99:495. 24. Eickhoff M, Suter D. J. Magn. Reson. 2004;166:69. 25. Kohler J, Disselhorst JAJM, Donckers MCJM, Groenen EJJ, Schmidt J, Moerner WE. Nature. 1993;363:242. 26. Moerner WE, Kador L. Phys. Rev. Lett. 1989;62:2535. 27. Hoult DI, Richards RE. J. Magn. Reson. 1976;24:71. 28. Webb AG. Prog. Nucl. Magn. Reson. Spectrosc. 1997; 31:1. 29. Minard KR, Wind RA. ConceptsMagn. Reson. 2001;13: 128. 30. Yamauchi K, Janssen JWG, Kentgens APM. J. Magn. Reson. 2004;167:87. 31. Seeber DA, Cooper RL, Ciobanu L, Pennington CH. Rev. Sci. Instrum. 2001;72:2171. 32. Poor B, van Eck ERH, Janssen JWG, van Bentum PJM, Kentgens APM. 2005; [to be published]. 33. Massin C, Boero C, Vincent F, Abenhaim J, Besse PA, Popovic RS. Sensors and Actuators A-Physical. 2002;97:280. 34. Massin C, Vincent F, Homsy A, Ehrmann K, Boero G, Besse PA, Daridon A, Verpoorte E, de Rooij NF, Popovic RS. J. Magn. Reson. 2003;164:242. 35. Wang H, Ciobanu L, Edison AS, Webb AG. J. Magn. Reson. 2004;170:206. 36. Alzetta G, Arimondo E, Ascoli C, Gozzini A. Nuovo Cimento B. 1967;52:392. 37. Sidles JA. Appl. Phys. Lett. 1991;582854. 38. Rugar D, Yannoni CS, Sidles JA. Nature. 1992;360:563. 39. Rugar D, Zuger O, Hoen S, Yannoni CS, Vieth HM, Kendrick RD. Science. 1994;264:1560. 40. Rugar D, Budakian R, Mamin HJ, Chui BW Nature. 2004;430:329. 41. Leskowitz GM, Madsen LA, Weitekamp DP. Solid State Nucl. Magn. Reson. 1998;11:73.

Part I

infancy, although impressive results have been reported from ONP experiments. There have been some efforts to use ex situ methods where the sample is in the projected field outside the actual magnet enclosure. Although the sensitivity can be quite reasonable, the inhomogeneity of the field precludes spectroscopy with any sensible resolution. Indeed, the main challenge in the field is not so much the search for new physical methods for sensitivity enhancement but to find a practical way to incorporate these methods without compromises to the versatility of modern NMR pulse sequences and with the full spectral resolution that is possible in the modern high field magnets. At present, the low-noise cryoprobe systems have become widely available in many labs. Despite their considerable costs, it is considered to be essential for an improved throughput in liquid NMR research. For online screening or combinatorial chemistry applications, it will become relevant to implement high-sensitivity microcoil probes. Although the susceptibility broadening of the present designs still puts some restraints on the applicability, this does not seem to be an intrinsic problem and the technology for a “lab on a chip” NMR implementation is not far away.

References 361

363

B.C. Gerstein1 and H. Kimura2 1 Department

of Chemistry, Iowa State University, Ames, IA 50011-3111, USA of Chemistry, University of Tsukuba, Tsukuba 305-8571, Japan

2 Department

Introduction

Theory

Combined rotation and multiple pulse spectroscopy (CRAMPS) [1] is one of a number of techniques for narrowing NMR spectra in solids, the broadening of which in this case is predominantly associated with:

Coherent Averaging; Average Hamiltonians in Spin Space

(a) homogeneous homonuclear dipolar interactions in ensembles of spin 1/2 nuclei (e.g. 1 H in poly(ethylene) or 19 F in Teflon, but not 23 Na in NaCl) and (b) shielding anisotropies. Rotations in both co-ordinate space [via magic angle spinning (MAS)] and spin space [via radio frequency (rf) pulses] are combined to achieve narrowed spectra. The CRAMPS technique utilizes single quantum coherence and is not used for narrowing broadened spectra of quadrupolar nuclei in solids. The basic ideas are that: (a) Coherent averaging [2] is used to attenuate dipolar interactions via resonant cyclic, and periodic multiple pulse excitations over cycle times short compared to the inverse of the homogeneous dipolar coupling, f D ≡ ωD /2π and (b) MAS, with spinning frequency, f MAS , small compared to the cycle times of the multiple pulse sequences (but see the comments under section “Magic angle spinning”), is used to average shielding anisotropies to their isotropic values. The details of the basic theory and techniques for achieving narrowed lines of spin 1/2 nuclei in one-dimensional experiments utilizing: (a) multiple pulse decoupling and (b) in the limit where spinning frequencies are small compared to multiple pulse cycle times have been presented [3] and reviewed [4].

Graham A. Webb (ed.), Modern Magnetic Resonance, 363–371.
C 2008 Springer.

The time-dependent Schr¨odinger equation is H | = i

d | dt

(1)

With the density operator defined as ρ = | |, the bar implying an ensemble average, Equation (1) becomes i

dρ = [H, r ] dt

(2)

The expectation value of any observable, Oˆ is ˆ ˆ  O(t) = trρ(t) O(t)

(3)

The observable in a pulse NMR experiment is a signal proportional to the decay of magnetization with time, M(t), associated with an oscillating magnetic moment, proportional to the transverse component of angular momentum, I ± which in turn is proportional to the component of nuclear magnetization perpendicular to the applied static field; M ± (t) ∼ I ± (t) = trρ(t)I ±

(4)

Such a signal, M(t), observed after some type of excitation which removes the magnetization of the ensemble of spins from its equilibrium state polarized along the static field, is shown at the top of Figure 1. Here the ensemble of protons are in liquid water. The oscillating decays, designated as Mx,y (t) are a series of single points taken by a digital recorder at intervals, or “dwell” times, set by the

Part I

CRAMPS

364 Part I

Chemistry

Part I

M0 cos φ0

Mx(t) My(t) M0 sin φ0

4 ms first ten points of BR-24 f.i.d. 200 µs

in the ensemble under observation. Such observed points are shown in the center portion of Figure 1. The oscillating decay labeled “first 10 points of BR-24” is that obtained on linear, high density poly(ethylene) under the BR-24 homonuclear decoupling sequence [5] on a static sample. Shown at the top of the center portion of Figure 1 is the digitized decay of the voltage induced by the magnetization of protons in this sample under homonuclear decoupling. Below that decay, again in the center portion of Figure 1, and on the same timescale, is shown the signal obtained on the same sample after a single pulse excitation using a dwell of 0.5 µs. Note that this signal decays in about 10 µs due to dephasing of the magnetization associated with homogeneous dipolar interactions. The bottom plot in Figure 1 exhibits the shielding tensor of 1 H in this sample obtained from the Fourier transform of the points obtained under the BR-24. A physical picture which may be useful in understanding how rf pulses may be used to decouple two-body dipolar interactions from each other results from the fact that the dipolar Hamiltonian between spins i and j is of the form Hi,Dj = ωD (ri, j , θi, j )( Iˆi · Iˆj − 3Izi Iz j )

4

10 Chemical shift

16x10 −6

Fig. 1. Examples of the experimental observation of the decay of magnetization, M(t), with time, after various excitations. Top: decay of the magnetization of protons in H2 O after a single pulse excitation; absorption and dispersion both shown. While seemingly continuous, these are in fact a series of single points taken by a digital recorder at “dwell” times, set by the experimenter. Here, the dwell was 20 µs, and on the timescale of the photo shown in Figure 1, the plot appears to be continuous. Center: First 10 points of the decay of magnetization of protons in linear high density poly(ethylene) (with thanks to Dr. D.L. VanderHart for supplying the sample) observed under homonuclear decoupling by the BR-24 sequence. Below those points, again in the center portion is shown the signal obtained on the same sample, and on the same timescale, after a single pulse excitation. Bottom: The shielding tensor of this sample obtained from the Fourier transform of the points under the BR-24.

experimenter. In the case of the top scans in Figure 1, the dwell was 20 µs, and on the timescale of the photo shown there, the plot appears to be continuous. In the case of the 1-D CRAMPS experiment utilizing multiple pulse decoupling, the signal is observed in the windows between pulses in the cyclic and periodic (vide infra) excitations which are used to attenuate homogeneous dipolar interactions among the spin 1/2 nuclei

(5)

Note that the form of Hi,Dj is a product of spin, ( Iˆi · Iˆj − 3Izi Iz j ), and co-ordinate (ωD (ri, j , θ i, j )) space variables. The co-ordinate space portion of Hi,Dj , designated as the frequency ωD (ri, j ,θ i, j ), scales as the product of the magnetogyric ratios of the two nuclei involved, the inverse cube of the internuclear distances, ri, j , and the spherical harmonic (1 − 3cos2 θ i, j ). Here, θ i, j is the angle between the internuclear two-body dipolar vector, ri j , and the axis of quantization, which is set by the dominant static field, B0 . One easily sees two facts from Equation (5): First is that if the spins are forced to lie along the x, y, and z axes of the spin co-ordinate system for equal times, τ , then since the scalar product is invariant to rotation, the average over time of the term ( Iˆi · Iˆj − 3Izi Iz j ) becomes 3τ ( Iˆi · Iˆj − Iˆi · Iˆj ), or zero. Another way of visualizing this picture is that the spins, on time average, are aligned along the (1, 1, 1) axis of the spin co-ordinate system. In the former case, one uses finite pulses to align the spins along x, y, and z, the basic pulse sequence the “solid echo,” or “dipolar echo” sequence [6–8], one form of which may be expressed as (τ , 90x , τ , 90 y ,τ ). In the latter case, the spins are first aligned along the (1, 1, 1) axis of the spin co-ordinate system, and then spin-locked there, which is the Lee–Goldburg [9] technique recently resurrected and improved by Vega and co-workers [10]. The solid echo sequence is neither cyclic nor periodic, but has the charm that at time 3τ , the homonuclear dipolar interaction becomes severely attenuated as a perturbation on the Zeeman interaction.

Theory 365

CRAMPS

M(t = N tc ) ∼ I± (N tc ) ∼ trρ(N tc )I±

(6)

and may be expressed via the use of the Magnus

Expansion for the propagator associated with internal interactions, Uint (Ntc , 0). With the definitions idUint /dt = Hint · Uint

(7)

H˜ int (t) = Urf (t, 0)Hint Urf+ (t, 0)

(8)

idUrf (t)/dt = Hrf (t)Urf (t)

(9)

and

we arrive at the result (see Chapter 4 in Ref. [3]) 0 0 ˜ H¯ int tc + · · ·}] N ρ[exp{i tc + · · ·}] N . ρ(N tc ) = [exp{−i H¯ int

(10) ρ˜ is the density operator as manipulated by the internal, and rf propagators as stated in Equations (7)–(9); + (N tc , 0)ρ(t)Uint (N tc , 0)Urf (N tc , 0). ρ˜ = Urf+ (N tc , 0)Uint

(11) Since for cyclic and periodic pulse sequences the system is returned to its initial state at then end of each sequence, ρ˜ 0 is the average Hamilmay be identified as ρ(t = 0). H¯ int tonian, including in this case, the homonuclear dipolar interaction and the scaled shielding interaction.

tc

0 H¯ int = 1/tc

dt H¯ int (t)

(12)

0

and H¯ int (t) is the internal Hamiltonian, in this case dipolar plus shielding, as manipulated by the rf pulses [see Equation (8)]. In the quasi-static limit, f s  (1/tc ), under, e.g. the cyclic and periodic MREV-8 homonuclear decoupling sequence, at cycle times N tc = 12τ for appropriately short rf pulses, the time average of the magnitude of the dipolar 6t Hamiltonian, 1/tc 0 c dt H¯ D (t), becomes severely attenuated compared to |HD |. The condition that the rf pulses, with field strength |B1 | ≡ ω1 /γ , are able to “manipulate” HD , the magnitude of which scales as ωD , is that |ω1 |  |ωD |. This means that the cycle times tc must be short compared to (|ωD |)−1 . For tc  (2π/|ωD |), there is effective homonuclear decoupling. As we will see in section “Applications,” for cycle times greater than the inverse of the dipolar frequency, i.e. for tc  (2π/|ωD |), coherent averaging of HD is destroyed. This means that with appropriate adjustments of tc , experimental conditions such as receiver gain and cycle time, a minor, mobile species can be detected independently of the major, solid component, with negligible signal from the probe background, to an accuracy of ≤0.01

Part I

Second is the term (1 − 3cos2 θ i, j ) may also be averaged in time by motion of the sample. In particular, MAS may average this term to zero. The result is that under MAS with spinning speeds appropriately large compared to the magnitude of the interaction being averaged, the shielding tensor is averaged to its isotropic value, σ iso , and the dipolar tensor is averaged to its isotropic value of zero. But since the magnitude of the dipolar interaction, |Hi,Dj | can be of the order of 100 kHz and for protons, at least, the magnitude of the shielding anisotropy is generally less than 6 kHz at a proton frequency of 600 MHz, MAS at currently achievable frequencies of ≤50kHz rather easily averages shielding anisotropies of protons to their isotropic value, but does not completely average dipolar interactions in sufficiently rigid systems such as linear, high density poly(ethylene). CRAMPS, therefore, utilizing either pulse decoupling or Lee–Goldburg spinlocking, (it must be mentioned that the “MP” portion of the CRAMPS acronym means “Multiple Pulse, so perhaps it is not appropriate to place Lee–Goldberg decoupling under the term CRAMPS, but we are not going to invent new acronyms at this point) both homogeneous dipolar coupling and homogeneous shielding anisotropies are averaged to their isotropic values. As a variation on the theme of pulse decoupling, the Emsley group has recently developed the technique of random phase decoupling [11]. The technique uses an optimal search scheme on the phases of windowless sequences to maximize resolution and minimize the effects of rf field inhomogeneity, and offers what is perhaps the best resolution to date on strongly coupled solids such as alanine. Here we consider pulse decoupling with phases, which are not random. Attacking the spin space portion of Hi,Dj , the dipolar echo sequence, symmetrized to become cyclic, meaning that under this sequence the system is returned to its initial state (in the absence of relaxation) may be expressed as [12] (τ , 90x , τ , 90 y ,τ , τ , 90−y ,τ , 90−x , τ ). A general discussion of constructing symmetrized pulse sequences has been presented by Mansfield [13]. Then in the 2τ windows between the pulse decoupling sequence, the time decay, M(t), at times t = Ntc , with tc being the cycle time for the periodic and cyclic pulse sequences used to average internal interactions, and N = 1, 2, . . . , may be expressed in terms of the density operator at multiple pulse cycle times Ntc . The density operator, ρ(Ntc ), is proportional to the expectation value of a transverse component of nuclear angular momentum, I± (Ntc ) .

366 Part I

Chemistry

Part I

mol.% in the cases of some organic solids crystallized from solution (vide infra).

Scaling Under Pulse Sequences A phenomenon important to understanding how the data taken under a multiple pulse experiment may be compared with data from a single pulse excitation is that of scaling. In the presence of a magnetic field of B0 magnetic moment with magnetogyric ratio γ will precess at a frequency ω0 = γ |B0 |. Under series of resonant pulses in the rotating frame, the magnetic moments precess under the effective field. For example, under the cyclic and periodic flip-flop sequence, (τ, 90x , τ, 90−x ), the response of protons in water is shown in Figure 2. With an average Hamiltonian for an offset ω, under the flip-flop cycle being 0 H¯ flip -flop = (ω/2)(k − j),

(13)

with k and j being unit vectors along the z and y axes in the Zeeman frame of the spins, √ the effective field in the 2γ , so the scaling factor rotating frame is || = ω/ √ in this case is 1/ 2. This means that in the frame of observation, the magnetization under the flip-flop cycle will precess more slowly by roughly a factor of 0.7 than the response under single pulse excitation as seen in Figure 2.

Fig. 2. Time decay of protons in water under the flip-flop cycle (τ, 90x , τ, 90−x ), and a resonance offset ω, (left side of figure), and under the same offset in the absence of the pulse sequences (right side of the figure). With an average Hamiltonian for an offset ω, under the flip-flop 0 cycle being H¯ flip -flop = (ω/2)(k − j), the effective field in the rotating frame √ is || = 2ω/2, so√the scaling factor in this case is 1/ 2. This means that in the frame of observation, the rotating frame, the magnetization will precess more slowly by roughly a factor of 0.7 as seen in the figure.

√ Under the MREV-8, the scaling factor is 2/3 for perfectly short pulses with no phase errors. The details of the treatment for non-ideal pulses are given in Chapter 5 of Ref. [3]. The manner in which the scaling factor is experimentally determined is discussed here in section “Experimental”.

Magic Angle Spinning As shown by Lowe [14] and Andrew et al. [15], under physical rotation of a sample at an angle to the static field of 54.74◦ , the chemical shielding anisotropy and homonuclear dipolar interactions between spins 1/2 are averaged to their isotropic values at spinning frequencies, f rot sufficiently greater than the line-widths of these two interactions. This is to say that an interaction the coordinate space portion of which scales as (1 − 3cos2 θ ) can be averaged to its isotropic value by sufficiently fast spinning at the magic angle. Since the isotropic value of the dipolar Hamiltonian between spins 1/2 has zero value, in principle, CRAMPS is not needed to remove this interaction. In fact recent studies [16] on protons in mesoprous silicas with spinning frequencies ≥40 kHz have shown that pulse decoupling is unnecessary for high resolution NMR of 1 H in such samples. The development of relatively high spinning speeds [17] up to 70 kHz has therefore to a certain extent made the use of CRAMPS

CRAMPS

Two-dimensional Experiments If one wishes to achieve a higher resolution for NMR of strongly coupled spin 1/2 nuclei than is available from one-dimensional CRAMPS, e.g. in relatively rigid solids containing carbon bound to hydrogen, and is willing to give up the time needed for gathering of data in two dimensions, one ties the resolution of the proton lines to that of the carbon lines via 2-D HETCOR. Here, because the resolution of the proton signal is tied to that of the carbon, the proton signal being obtained from the t2 domain in the 2-D experiment, there is no reason to be concerned with ring-down and windows in which to observe the proton signal. In fact, with appropriately high sample spinning speeds, on relatively mobile samples, e.g. protons in some mesoporus silicas, there is no need to use pulse, random phase, or spin-lock decoupling at all. As stated above, adding a second dimension to any NMR experiment adds time to the acquisition of data. It is therefore to great advantage to achieve as high a sensitivity as possible with a maximum magnetic field. The sensitivity of detection is roughly proportional to the three-halves power of the static field. Along with higher fields comes, however, a relative problem. This is that to avoid the problem of sidebands without using some scheme such as rotor synchronization, which limits the

dwell to the inverse of the rotor period, the spinning speed must be maximized. The fact that the dipolar interaction re-focuses every 180◦ of the sample’s physical rotation has been used by Hafner and Spiess [18] to combine the timing of pulse sequences in such a manner that the dipolar interaction is attenuated to a certain extent by the spinning, and pulse sequences then used remove the residual broadening, and the results utilizing relatively slow spinning are recovered.

Experimental General We limit our discussion here to the CRAMPS experiment under the quasi-static limit, which is easily achieved, even with spinners capable of 70 kHz spinning speeds, if they are capable of stable spinning at 3–6 kHz. When the CRAMPS technique was initially performed, the stability of transmitters, of the pulse widths controlled by the pulse programmers, and of the phases in the rf unit, were such that quite careful tuning was required for a successful experiment. These factors seem to have led to the mythology that the CRAMPS technique was for the initiated only, and not worth taking the time. Fortunately, with the advent of modern spectrometers with relatively stable amplifiers, digital control of phases, and nanosecond control of pulse widths, obtaining CRAMPS spectra of protons even in the most rigid of strongly coupled systems (e.g. high-density linear poly(ethylene) and adipic acid; see the bottom of Figures 1 and 3) is now relatively easy if one has a reasonable grip on the experimental and theoretical foundations of the experiment, and quite reasonably possible even if one has not.

Spectrometer Requirements; the Probe and Receiver Modern spectrometers suitable for NMR of solids, e.g. those supplied by Bruker, JOEL, and Varian, have the capabilities for power, timing, phase control, pulse programming, and sample rotation suitable for the CRAMPS experiment. The place where care must be exercised is in the receiver and probe ring-down. To illustrate what is needed, consider an ensemble of protons in a solid in which the internuclear distance is similar to that in high-density poly(ethylene), leading to a homogeneous line-width of δω/2π ≈ 100 kHz (see the time decay in the central portion of Figure 1 taken under a single-shot excitation). Those data were taken with an MREV-8 cycle time of 21 µs, meaning that 12 τ = 21 µs, τ = 1.75 µs, and the data were accumulated in the

Part I

for simply narrowing proton spectra in many solids unnecessary. At the time of writing, probes achieving such spinning speeds are not generally available. But there still exist systems, e.g. high-density linear poly(ethylene), in which the dipolar frequency can be as high as 100 kHz, so fast MAS to achieve high-resolution proton spectra would not work in that case. We now become more explicit about the physics implied in the term “quasi-static” limit. In the case of CRAMPS, the movement of the sample can destroy the experiment if the dipolar interaction becomes timedependent on a scale similar to the cycle time of the multiple pulse sequence attacking H D . For example, a multiple pulse cycle time of 30 µs would imply that the spinning frequency f rot  (10−6 /30)s = 30 kHz. Practically, rotation speeds of 4–6 kHz are easily achieved and at static fields in which the shielding anisotropies of the protons studied are ≤10 ppm (which is 3 kHz at a resonant frequency of 300 MHz), the CRAMPS experiment works well. However, the development of commercial probes in which spinning speeds can routinely be in the neighborhood of 45 kHz has led to considerations of how to combine pulse decoupling with MAS outside of the quasistatic regime. The implications of this development are discussed in the next section.

Experimental 367

368 Part I

Chemistry

Part I Cramps

(f) 10

to take data for at least 0.2 µs at the end of the window of 2 τ . This means a total ring-down of 3.5 − 0.2 = 3.3 µs. So for a spectrometer operating at 300 MHz, a requisite probe Q is calculated as follows: 3.3 × 10−6 = 7Q/ f = 7 × 3.3 × 10−9 Q

0 ppm

⇒ Q = 140.

(14)

8.0 kHz

This is a reasonable Q for a solid-state probe, but unusual for a probe set to receive signals from liquids where the dwell is allowed to be of the order of ≥20 µs. In addition, the receiver must be protected during the high power pulses, and in such a manner that the voltage between each stage of the receiver must be arranged to ring-down so that there is ample time in the 2τ windows for each stage to receive signal. One such scheme is presented on page 222 of Ref. [3]. To the author’s knowledge all of the instrument producers of solid-state NMR spectrometers are able to meet the above requirements in receivers and probes.

(b)

6.0 kHz

Tuning to Maximize Resolution; Pulse Impurities

(a)

static

(e)

11.0 kHz

(d)

(c)

40

20

0

-20

-40

kHz

Fig. 3. Comparison of narrowing of proton spectra in the strongly coupled protons in adipic acid (HOOC–(CH2 )4 –COOH) under: (a) Single pulse excitation with dwell of 0.5 µs. (b) MAS AT 6.0 kHz (c) MAS AT 8.0 kHz (d) MAS AT 11.0 kHz (e) CRAMPS using the MREV-8 sequence. With thanks to Gary Maciel for the figure.

2 τ = 3.5 µs windows of the sequence. With pulses of 500 W, and pulse widths of 1 µs, into a probe with impedance 50 Ohms, implying a p–p voltage of about 160 V, the time needed to ring-down this voltage to 10−7 V is 3.3 µs. The value of 10−7 V is what one wishes to have to keep the receiver happy. This ring-down time is approximately 21 Q/3 f , where f is the resonant frequency in Hz and Q the quality factor of the probe. Q ∼ = ( f /2 f res ) where  f res is the width of the tuning curve of the probe at resonant frequency, f res . For example, at a resonant frequency of 300 MHz, a tuning curve with a half-width of 1 MHz would have Q ∼ = 150. We wish for the receiver to be able

One further point about probe tuning deals with what might be described the “art” (but is well understood analytically, as discussed in pp. 177–82 of Ref. [3]) deals with pulse impurities. A pulse of rf may be thought of as a continuous wave of rf multiplied by a window function. For pedagogy, we consider only a rectangular window, as illustrated in Figure 4. The frequency domain fingerprint of such a window function is a sine function, sinx/x, which contains many frequencies. The envelope also oscillates. Therefore, at the

Fig. 4. Response of protons in water under the flip-flop tuning cycle. Top: Properly tuned. Middle: Phase transient present. Bottom: Phase error present.

CRAMPS

Applications 369

“RINGDOWN” TIME CONSTANT Q τ∼ = 3f

RF GATE OPEN

RF GATE CLOSED z

M

y x

B1

Fig. 5. The phase-detected envelope of an rf pulse. Top: upper trace, the in-phase component. Lower trace, the out of phase component, with increased gain of the oscilloscope channel used for detection. Middle: Comparison of the response the signal under a “square-wave” envelope with that associated with ringup and ring-down. Note that both frequency and phase impurities are present during the transient periods after turn-on, and turnoff. Bottom: vector picture of the magnetization associated with the phase transients.

turn-on and turn-off periods of the pulses, there are both phase and frequency transients. The phase transients may be viewed as being orthogonal to the broadcast phase, i.e., an “x” pulse, will have “y” components during the ring-up and ring-down times. This idea and the experimental evidence are illustrated in Figure 5. It is necessary to minimize the cumulative effects of these “impurities” for maximum resolution under the CRAMPS experiment. One method is to slightly de-tune the probe to achieve maximum decay time for the sample under investigation. Because a slight de-tuning also affects the pulse widths, it is necessary to iterate the de-tuning with a check on pulse widths. Another, perhaps preferable method (personal communication Drs. Charles Bronnimann and Jim Frye, Varian, Inc., 12-29-04) is to place variable—length

Determination of the Scaling Factors As may be inferred from Figure 2, the Fourier transform of the decay shown there will yield peaks at two different frequencies. The difference between these immediately supplied the scaling factor under the flip-flop cycle. In the CRAMPS experiment, the difference in the frequencies of protons in water under a number of different offsets, generally 1 kHz apart, summed to produce a spectrum of peaks 1 kHz apart under an offset using a given homonuclear decoupling pulse sequence, compared to the experimental offset used to produce that signal immediately yields the scaling factor. It is an interesting fact that the scaling factor depends upon the offset [19], and this fact must be taken into account for each resonance detected under the CRAMPS experiment when applying a scaling factor to a given sample.

Applications CRAMPS in One Dimension After it became clear that pulse NMR could be a powerful tool for probing identities of protons in coals, the first applications of CRAMPS were the determinations of highresolution proton NMR of coals [20–25], and in polymers [26]. The CRAMPS spectra of 1 H in coals, which accompanied by high resolution of carbon in these systems, were used to infer aromaticity, and the average sizes of aromatic rings in the systems studied. The spectra of polymers, with varying cycle times, were used to infer amorphous fractions of polymers. Later applications were to determinations of structures of biological solids [27–32], and of environments of protons in silicas [33]. Finally, most recently at the time of writing [34], 1-D CRAMPS, with varying cycle times to selectively detect rigid and mobile portions of the sample, has been used to provide quantitative and qualitative determinations of solvents occluded during the process of crystallization in organic solids, to an accuracy of ≤0.01 mol.%. The fascinating part of this experiment is the lovely match between experiment and theory, in that if tc  (|ωD solid |)−1 , the solid

Part I

transmission lines in the nominally quarter-wave length portions of the transmission and receiving part of the circuitry (see Figures 5–24 in Ref. [13]) in order to maximally damp frequencies other than the carrier and vary the length to achieve maximum resolution. The procedure is described in “Commonly Run Solids NMR Experiments” which can be downloaded as a .pdf file from Varian’s web site, under User Pages “list of manuals title (alphabetical)” or “list of manuals by part number,” manual #0199907600A.

370 Part I

Chemistry

Part I

portion of the sample is detected. With the mobile impurity being of the order of 0.01 mol.%, this portion contributes negligibly to the observed signal under CRAMPS. D D If (|ωsolid |)−1  tc  (|ωmobile |)−1 , and the gain increased appropriately relative to the gain used to detect the major, rigid component, only the relatively mobile portion of the sample is detected. Since only signal of the sample inside the inductor contributes to the signal under CRAMPS, the observed signal of the mobile portion contains essentially no signal from the protons in the probe. Figure 6 illustrates how the observed signal under CRAMPS detects the D rigid component for tc  (|ωsolid |)−1 , changes gradually as tc is increased, and finally detects only the mobile por-

Gain = 32 τc = 36 µs (A) NS = 32 δ = 6.7

H3C

CH3

H3C

CH3

δ = 1.6

CH3

Ring

(B) Gain = 64

τc = 120 µs

D D tion when the inequality (|ωsolid |)−1  tc  (|ωmobile |)−1 is satisfied.

CRAMPS in Two Dimensions The addition of a second dimension to experiments involving homonuclear decoupling of protons, along with excitation of a second nucleus (excepting protons in the case of proton–proton HETCOR) has the usual disadvantage of added time for accumulation of signal. But the advantage is that, at least in heteronuclear HETCOR, the signal from protons (or fluorine) is observed in the t2 domain of the 2-D experiment. Therefore, there is no need to have the exacting conditions for ring-down and probe tuning optimizing the 1-D experiment. In addition, windowless sequences [11,35] may be used such that the limitations involved in the requirements of the quasi-static condition for pulse decoupling, when used with rotor-synchronized pulse sequences, allow for 2-D 13 C–1 H HETCOR with spinning speeds much higher than that used in the 1-D CRAMPS discussed above. As a final remark, we note that all techniques of detection described above utilize single quantum coherence for the narrowing and detection of spin 1/2 nuclei in solids. For narrowing and detection of half-spin quadrupolar nuclei, e.g. 23 Na and 27 Al, combinations of multiple pulse excitation and MAS utilizing multiple quantum coherence have been used [36].

Acknowledgments Gain = 4096 (C) τ = 204 µs c

δ = 1.6 δ = 3.2

(D)

δ = 4.5

Gain = 4096 τc = 288 µs NS = 4096

OH 9

8

7

6

5

CH 2 4

3

CH 3 2

1

0

ppm

Fig. 6. NMR Spectrum of durene crystallized from ethanol under CRAMPS using the MREV-8 sequence, with varying cycle time tc , and varying receiver gain. Spinning frequency is 5 kHz. At a cycle time of 288 µs, only the mobile portion, representing durene dissolved in ethanol, is seen. At a cycle time of 36 µs, only solid durene is observed. As the cycle time is increased from 36 to 288 µs, the line broadens, averaging of the dipolar interaction for the rigid portion is destroyed, and finally, only the mobile portion is detected.

The authors appreciate the invitation from Professors Ando, Saito, and Asakura to submit this article. We acknowledge the use of the facilities of the Chemistry Department at Iowa State University. Careful readings of the manuscript by Drs. Marek Pruski, Jim Frye, and Charles Bronniman were greatly appreciated. We are especially grateful to Drs. Bronnimann and Frye regarding information on minimizing the effects of phase transients on the CRAMPS experiment. The authors have attempted to fairly represent the work of all major contributors to this area of technology, but nevertheless make the apology regarding the excess of references to our own work which mirrors that which Thoreau made about his own writing. If I knew other’s lives as well as I do mine, I should write about them as well.

References 1. Taylor RE, Pembleton RG, Ryan LM, Gerstein BC. J. Chem. Phys. 1979;71:4541. 2. Haeberlen U. High Resolution in Solids: Selective Averaging. Advances in Magnetic Resonance (Suppl I). Academic Press: New York, 1976.

CRAMPS

22. 23.

24.

25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

(Chapter 2). In: Advances in Chemistry Series. Oxford University Press: New York, Vol. 192, 1981. Kamienski B, Pruski M, Gerstein BC, Peter H. J. Energy Fuels 1987;1:45–50. Gerstein BC, Pruski M, Michel D. Proton NMR spectroscopy of coals, cokes, and coal-derived liquids (Chapter 9). In: HLC Meuzelaar (Ed). Advances in Coal Spectroscopy. Plenum Press: New York, 1992. DeLaRosa L, Pruski M, Gersteinand BC. Quantitation of protons in the argonne premium coals by solid state 1 H NMR (Chapter 19). In: RE Botto, Y Sanada (Eds). Magnetic Resonance in Carbonaceous Solids. Advances in Chemistry Series 29. ACS Publishing: Washington, DC, 1993, pp 359–76. Snape CE, Axelson DE, Botto RE, Delpuech JJ, Tekely P, Gerstein BC, Pruski M, Maciel GE, Wilsonand MA. Fuel 1989;68:547. Pembleton RG, Wilson RC, Gersteinand BC. J. Chem. Phys. 1977;66:5133. Shoji A, Kimura H, Ozaki T, Sugisawa H, Deguchi K. J. Am. Chem. Soc. 1996;118:7604-7. Kimura H, Nakamura K, Eguchi A, Sugisawa H, Deguchi K, Ebisawa K, Suzuki E, Shoji J. J. Mol. Struct. 1998;447:247– 55. Kimura H, Ozaki T, Sugisawa H, Deguchi K, Shoji A. Macromolecules 1998;31:7398–403. Kimura H, Shoji A, Sugisawa H, Deguchi K, Naito A, Saito H. Macromolecules 2000;33:6627–9. Kimura H, Kishi S, Shoji A, Sugisawa H, Deguchi K. Macromolecules 2000;33:9682–7. Kishi S, Santos A, Ishi O, Ishikawa K, Kunieda S, Kimura H, Shoji A. J. Mol. Struct. 2003;649:155–67. Maciel G. Silica surfaces. In: Encyclopedia of NMR. John Wiley: Chichester, 1996. Gerstein BC, Kimura H. Appl. Magn. Reson. 2004;27:1. Burum DP. HETCOR in Organic Solids. In: Encyclopedia of NMR. John Wiley: Chichester, 1996. Amoureux J-P, Pruski M. Advances in MQMAS NMR. In: DM Grant, RK Harris (Ed). Encyclopedia of Nuclear Magnetic Resonance. Advances in NMR, Vol. 9. John Wiley & Sons Ltd.: Chichester, 2002, 226–51.

Part I

3. Gerstein BC and Dybowski CR. Transient Techniques in NMR of Solids: An Introduction to the Theory and Practice. Academic Press: Orlando, Fla., 1986. 4. Gerstein BC. CRAMPS. In: The Encyclopedia of NMR. John Wiley: Chichester, 1996. 5. Burum DP, Rhim W-K. J. Chem. Phys. 1979;71:944. 6. Lowe IJ, Bull. Am. Phys. Soc. 1957;2:344; Mansfield P. Phys. Lett. 1962;2:58 and Phys. Rev. 1965;127(A):961. 7. Ostroff ED and Waugh JS. Phys Rev. Lett. 1966;16:1097. 8. Mansfield P, Ware D. Phys. Lett. 1966;22:133. 9. Lee M, Goldburg WI. Phys. Rev. 1965;140:1261. 10. Vinogradov E, Madhu PK, Vega S. Chem. Phys. Lett. 1999;315:443. 11. Elena B, de Pa¨epe G, Emsley L, Chem. Phys. Lett. 2004;398:532 and references therein. 12. Waugh JS, Huber LM, Haeberlen U. Phys. Rev. Lett. 1968;20:180. 13. Emsley JW, Feeney J, Sutcliff LH (Ed). Progress in Nuclear Magnetic Resonance Spectroscopy. London. 1971;8:41. 14. Lowe IJ. Phys. Rev. Lett. 1959;2:285–7. 15. Andrew ER, Bradbury A, Eadesand RG. Nature (Lond.). 1958;182:1659. 16. Trebosc J, Wience J, Lin VS-Y, Pruski M. J. Am. Chem. Soc. 2005;127:7587. 17. Samoson A, Tuherm T, Past J, Reinhold A, Anup˜old T, Heinmaa I. New horizons for magic-angle spinning NMR. Top. Curr. Chem. 2004;246:15–31 (DOI 10.1007/b98647). 18. Hafner S, Spiess H. Advanced solid-state NMR spectroscopy of strongly dipolar coupled spins under fast magic angle spinning. Concepts Magn. Reson. 1998;10(1):99–128. 19. Shoji A, Kimura H, Sugisawa H. Structural studies of amino acids, polypeptides and proteins in the solid state by 1H CRAMPS NMR. In: GA Webb (Ed). Annual Reports on NMR Spectroscopy. Academic Press: London, 2001, 45, pp 69– 150. 20. Gerstein BC. Fingerprinting solid coals using pules and multiple pulse NMR (Chapter 5). In: Analytical Methods for Coal and Coal Products, Vol. 3. Academic Press: New York, 1980. 21. Gerstein BC, DuBois Murphy P, Ryan LM. A tentative identification of the size of polynuclear aromatic rings in coals

References 371

373

B. Bl¨umich and F. Casanova Institute of Technical Chemistry and Macromolecular Chemistry, RWTH Aachen University, Germany

Introduction The widespread use of NMR in materials testing is hampered by the fact that the object needs to be carried to the NMR equipment and needs to fit inside the magnet [1]. Both limitations are removed by mobile unilateral NMR at the expense of a lower and inhomogeneous magnetic polarization field [2]. Originally, open sensors were developed by the well-logging industries [3]. There, transverse relaxation decays are measured from the fluids in the porous formation with a spectrometer positioned inside the borehole. For well-logging sensors, the field gradient is minimized in a sweet spot or adjusted to a small value so as to eliminate signal attenuation from translational diffusion for short echo times [4–6]. Parallel to the development of the first well-logging tools, the first unilateral sensors were developed mostly for measuring moisture in soil, bridge decks, building materials, and food [7–12]. Some of these instruments employed electromagnets weighing several hundred kilograms. As long as solid materials including rubber are investigated even large field gradients can be tolerated as diffusion is absent. This is the idea behind the NMR-MOUSE, which operates with permanent magnets at frequencies between 10 and 20 MHz with magnetic field gradients up to 20 T/m (Figure 1) [13–16]. The NMR-MOUSE weights less than 1 kg, but is limited to depths typically less than 15 mm. With the commercialization of well-logging instruments and the availability of the NMR-MOUSE about 10 years ago, the NMR methods for use in inhomogeneous magnetic fields were systematically developed [17– 24]. A variety of open magnet geometries are currently being explored for portable use [25–32]. For investigation of large objects, open magnets are fitted with surface coils that provide a magnetic radiofrequency (rf) field B1 [15,16,33]. The volume outside the magnet, where B1 exhibits perpendicular components to the polarization field B0 , is the sensitive volume of the sensor. A unilateral NMR sensor essentially selects the signal of a pixel from the object, the size of which is defined by the sensitive volume. A recent development complementary to unilateral NMR devices is lightweight and comparably inexpensive, cylindrical magnets in the Halbach geometry constructed from many small blocks of permanent magnets [34,35]. Graham A. Webb (ed.), Modern Magnetic Resonance, 373–382.
C 2008 Springer.

Such magnets are suitable for studying pipe flow, geophysical drill cores [36], and plants at the site of the object.

Measurement Methods The open magnets used for unilateral NMR exhibit inhomogeneous magnetic polarization fields B0 . When positioned on a large object the response bandwidth is always much larger than the excitation bandwidth, and any excitation is selective [24], a situation encountered also in stray-field NMR imaging [18]. In addition, the surface coil provides an inhomogeneous rf field B1 , so that the flip angle of an rf excitation pulse varies across the sensitive volume. Most methods known from NMR in homogeneous fields can be adapted for use in inhomogeneous fields, but need to be reevaluated to account for selective excitation and the flip-angle distribution. Transverse relaxation decays are readily measured by Hahn echoes and CPMG echo trains (Figure 2a) [37–39]. Due to the flip angle distribution, Hahn and stimulated echoes are generated in a CPMG train, and different coherence pathways need to be discriminated [23]. One striking manifestation of this is, that the first echo in a CPMG train is always lower than the second [18–21]. But also, the echo envelope decays slower than in homogeneous fields, so that effective relaxation times T2eff are measured in inhomogeneous fields. Similar to the different variables that define the pixel amplitude in an NMR image [1], parameters and parameter-weighted spin densities can be extracted from the signal measured by the NMR-MOUSE (Figure 2b). Parameters are obtained by fitting the relaxation decay with a model function such as the stretched exponential function [40], or a parameter-weighted signal is obtained, for example, by forming the ratio of the signal at a given decay time with the signal amplitude at decay time zero. To improve the signal-to-noise ratio, several echo amplitudes are added in a defined time interval instead of just taking one amplitude value, for example, w(0, t1 , t2 , t3 ) = I (t2 , t3 )/I (0, t1 ). Furthermore, relaxation decays with sufficiently good signal quality can be inverted to distributions of relaxation times by a regularized inverse Laplace transformation (Figure 2c), a procedure which is routinely applied in NMR well logging

Part I

Mobile NMR

374 Part I

Chemistry

Part I

a)

c)

b)

d

Fig. 1. Portable NMR instrumentation. (a) Original NMR-MOUSE with a u-shaped magnet (Photo: Peter Bl¨umler). (b) Bar-magnet NMR-MOUSE. (c) Small-size portable spectrometer. (d) Halbach magnet with a mobile low-field spectrometer in a pilot’s case at the geothermal drilling site of RWTH Aachen. (Photo: Peter Winandy).

to facilitate data interpretation [3]. Parameters other than those referring to transverse relaxation can be measured as well but not directly in inhomogeneous fields. To this end the initial magnetization probed by a CPMG train is prepared for example, by a saturation or inversion recovery T1 filter [41–43], a multi-quantum filter [44,45], chemical shift [46,47], space encoding [48–51], flow encoding [52,53], or a diffusion filter (Figure 2d) [54–59], and the filter parameter is varied systematically in successive scans [55]. To improve the signal-tonoise ratio, successive echoes of the CPMG train employed for detection can be added. Single-shot encoding of diffusion and flow can be achieved by selection of suitable coherence pathways in multi-echo experiments [60,61]. Using such two-dimensional (2D) methods, a unilateral sensor can be employed for imaging (Figure 3) similar to a magnifying glass to measure, for example, an image of a defect in a textile-reinforced rubber hose (Figure 3c) and the axial velocity image of laminar water flow through a pipe (Figure 3e and f). In this case, the image and flow information is encoded in the detected magnetization by preparing the initial magnetization with pulsed linear gradient fields, which are generated with additional coils fitted to the unilateral sensor [48–53]. Images and profiles

involving the depth direction are constructed from several data sets acquired for subsequent slices through the object. For a long time it has been accepted that chemical shift cannot be measured in inhomogeneous B0 fields. However, this is not so, as has been demonstrated recently in high field with experimental 1D and 2D spectra [46,47]. The currently most successful approach to acquire chemical-shift-resolved spectra in inhomogeneous fields employs mixed echoes with dephasing and rephrasing evolutions in B1 and B0 fields with matched inhomogeneities (Figure 4a) [62]. The evolution in B0 depends on the chemical shift and the one in B1 does not. As a result, the amplitude of the mixed echo is modulated by the chemical shift. For example, a low-field 19 F NMR spectrum (Figure 4b) of a mixture of two perfluorinated solvents can already be measured with a unilateral sensor with a resolution better than 10 ppm in just 3 min [63]. However, the sample needs to be small and accurately positioned in the small region where the fields are matched. Unilateral NMR spectroscopy will be of most use for materials analysis, where high-resolution solid-state spectra of 1 H or even heteronuclei are needed, and it is a current challenge to develop adequate methods that include line narrowing [64,65].

2θ°x

2θ°x

2θ°x time

transmitter

t E/ 2

amplitude a

θ°y

a(0) exp{(t/T2eff)b/b} l (0, t1)

l (t2, t3)

exp{-t/T2eff}

receiver

0 0 t1

30 20 10

biexponential fit T2eff,short = (0.18 + − 0.01) ms T2eff,long = (1.75−+ 0.04) ms

tE

inverse Laplace

time t θ

0 10 20 30 40 50 60 time [ms]

θ

δ

1 0 0.01 0.1

c)

θ

CPMG

2

transformation 0

time t

t2 t3

b)

3 frequency

a) rel. echo amplitude [%]

tE

tE

δ ∆

10 1 T2eff [ms]

100 d)

preparation: diffusion

detection: relaxation

Fig. 2. Measurement methods for NMR in inhomogeneous fields. The excitation flip angle θ shows a distribution within the sensitive volume. In homogeneous fields, θ should be 90◦ . (a) Multi-echo sequence according to Carr, Purcell, Meiboom, and Gill (CPMG). The envelope of the echo maxima defines the decay of the transverse magnetization. (b) Analysis of transverse relaxation decays by either fitting model functions to obtain amplitudes and relaxation time constants or by computing relaxation-weighted amplitudes. (c) Processing of non-exponential relaxation data by inverse Laplace transformation for subsequent analysis of the distribution of relaxation times. (d) Pulse sequence for measuring a correlation map of diffusion and transverse relaxation. The initial magnetization detected by a CPMG echo train is diffusion encoded in a preparation period, and the experiment is repeated with a systematic variation of the encoding weight. The 2D correlation map is the 2D inverse Laplace transform of the experimental data.

Fig. 3. Unilateral imaging and flow NMR. The object to be investigated, a rubber pipe (a) or a tube with flowing water (d) is placed on the sensor, here a 36 kg magnet with a field of view of 40 × 40 × 20 mm3 . With pulsed gradient fields and phase encoding techniques, the image (c) of the defect in the tube (b) was obtained in 2 h with a spatial resolution of 0.7 mm in each dimension. Similarly, flow images (e, f) can be obtained, here for laminar water flow through a circular pipe (d). Images across depth are constructed from data acquired with multi-slice techniques.

376 Part I

Chemistry

b)

Fig. 4. Chemical-shift resolved spectroscopy in inhomogeneous fields. By matching the precession of magnetization in the inhomogeneous polarization field B0 to the precession in an inhomogeneous rf field B1 , chemical-shift resolved spectra can be obtained by unilateral NMR. (A) Principle of matching B0 and B1 profiles in space. (B) NMR spectra of fluorinated solvents acquired ex situ of the magnet.

Applications

the material [40]. The cross-link density correlates with the glass transition temperature Tg, which is determined in the physical testing laboratory on samples taken out of the production process. T2eff correlates well with Tg (Figure 5b), but needs to be measured at constant temperature [71] or extrapolated to a reference temperature and calibrated. As measurements by unilateral NMR are fast and non-destructive, T2eff and its spread can be followed at all steps during the production of rubber parts (Figure 5a). This has been demonstrated for the production of tires [72]. The spread of T2eff for five different tires and their intermediate products at different equivalent measurement positions were quantified in terms of the coefficient of variation and compared to the same quantity for the initial torque measured with a rheometer (RPA: rubber process analyzer) on samples drawn from

b) 1.0

Cis-BR/B I-BR

elastomer network

NR Cis-BR/A

cr s-

os

room temperature

lin

SBR

en

kd y sit

0.1

N-SBR

back-ground of the NMR-MOUSE

-100

-80

-60 -40 -20 -0 glass temperature [°C]

40

coefficient of variation [%]

The most attractive applications of portable NMR are with unilateral NMR in materials science. But portable NMR is also being used with earth field instruments to study sea ice in Antarctica [66–68], and a Halbach magnet has recently been used to study naturally wet rock core samples at a geological drilling site (see below) [36]. A few illustrative examples of portable NMR are given in the following. Soft matter can readily be studied by NMR as is demonstrated by the great success of medical imaging which maps biological soft matter. Rubber is synthetic soft matter from cross-linked macromolecules with a number of additives and fillers like carbon black. T2eff is sensitive to the technically important chemical cross-link density [69,70] but also to the homogeneity and state of

T2 [ms]

Part I

a)

30 20

RPA

NMR-MOUSE

10 0

3.4 IR

ch

at

rb

e st

20

a

m

ill

m

re

al

fin

ix

te

m

da

ru

t ex

lc.

vu un

e

e

tir

d ze

tir

i an

lc

vu

Fig. 5. Mobile NMR of rubber. (a) Large objects like tires can be investigated locally and non-destructively. (b) Transverse relaxation times measured near room temperature with the NMR-MOUSE and extrapolated to a reference temperature scale with the glass transition temperature and consequently with the cross-link density. (c) The spread of NMR parameters within a product and between products provides information about homogeneity and quality. It can be measured at all intermediate steps of the production process and at the final product, as demonstrated for the transverse relaxation times T2eff in the tire production.

Mobile NMR

Applications 377

Part I

Fig. 6. Semi-crystalline polymers. (a) Morphology of semi-crystalline polymers with chain-folded crystalline regions and disordered amorphous regions, and biexponential model fit function for separation of signal contributions. (b) Crystallinity of a PE pipe at different points around the circumference before deformation, after deformation, and after annealing. (c) Deformation of a PE water pipe. (d) 0 to 1 mm depth. The weighted spin density w(2, 9, 50, and 100) 2D crystallinity map acquired on a 1 cm2 raster inside a PE pipe at !9 ! !100 was computed from the echo maxima an according to 92 an 2 an + 50 an .

the same compounds (Figure 5c). The coefficient of variation decreases from one processing step to the next for the torque, but not for T2eff , demonstrating that rheometer and NMR measurements provide somewhat complementary information. Also the coefficients of variation of T2eff are larger showing a higher sensitivity of T2eff to material properties than the torque. Finally, torque measurements are destructive and cannot be done on the final product, whereas the NMR-MOUSE can be used for quality control there [40,73,74]. When testing tires with the NMRMOUSE, the presence of a steel belt is no obstacle. In fact even thin polymer coatings have been measured on steel sheets with the NMR-MOUSE [75]. Strained elastomers exhibit macroscopic molecular order, which can be probed with the u-shaped NMRMOUSE by the orientation dependence of the relaxation rate [76], an effect which is also observed for tendon [77]. Aging and fatigue also lead to changes in T2eff , and can consequently be studied non-invasively by unilateral NMR [71,78]. Leathery, semi-crystalline polymers like poly (ethylene) and poly(propylene) are less soft than rubber but still give excellent signal when short echo times like 25 µs are used. The transverse relaxation time T2eff,short of the chain-folded macromolecules in the crystalline

domains is considerably shorter than the T2eff,long of the coiled chains in the amorphous domains (Figure 6a). Consequently, both components can be discriminated in a biexponential fit of the transverse relaxation decay, and the relative amplitude of the short component defines the NMR crystallinity. This quantity has been determined at well-defined positions on the circumference of a PE water pipe (Figure 6b and c) [79]. Even in the new state, it varies from point to point due to shrinkage while cooling during fabrication (Figure 6b). When laying or repairing pipes, the water flow is stopped by squeezing the pipes. When applied for an extended time, such a deformation reduces the crystallinity due to strained amorphous chains creeping out of the crystalline domains. At the same time the order in the amorphous domains has initially been increased as the chains are strained. Upon annealing well below the glass temperature these, chains relax by the melting of small crystallites, so that the overall crystallinity decreases further. By measuring a 2D array of NMR crystallinity data, the inhomogeneities of a PE pipe can be depicted in a 2D map (Figure 6d). Another semi-crystalline polymer is cellulose. It is contained in wood and is the main component of paper. The state of wood and paper is inherently associated with the amount of bound water. The NMR-MOUSE has

378 Part I

Chemistry

Part I

been applied to characterize the degradation state of historic paper [80–82] and to map the moisture distribution in wood panels [83]. The original u-shaped NMR-MOUSE (Figure 1a) has the B0 field approximately parallel to the scanner surface. With decreasing rf frequency the sensitive volume is shifted away from the surface, and its shape flattens. To obtain an extremely flat shape, the magnet geometry needs to be somewhat modified. While the original NMRMOUSE collects signal from a slice about 1 mm thick near the surface, the optimized profile NMR-MOUSE collects signal from planar slices 30 µm thin and less (Figure 7a) [84]. To acquire high-resolution depth profiles, the sensitive slice can be shifted through the object by adjusting the distance between the NMR-MOUSE and the object with a precision lift of the NMR-MOUSE (Figure 7b). Even higher resolution can be obtained by Fourier transforming the echo acquired from the sensitive slice. Despite the thin sensitive volume, the sensitivity is good enough to acquire depth profiles of moisture in polymer sheets within a few minutes (Figure 7c and d). In this way the water uptake can be followed with a precision better than 0.2%. Even the water uptake from air can be studied (Figure 7d). Here the profile NMR-MOUSE provides information similar to that of the GARFIELD magnet [85,86], which has been designed with specially shaped pole shoes that ensure a

linear gradient field with a strong gradient in the magnet gap. But contrary to the Garfield design, the NMRMOUSE provides completely open access, for example, for in vivo skin studies and profiling of composites and adhesive layers [87–89]. Instead of shifting the sensitive volume mechanically through the object, it can also be shifted electrically by changing the field strength of the NMR-MOUSE [90] and by varying the rf excitation frequency on the expense of a changing shape of the sensitive volume. The characterization of moisture in soil and building materials are one of the earliest applications of unilateral NMR [8–12,40,57,91,92]. Absolute values of moisture per volume can readily be obtained, as the size of the sensitive volume is fixed and defined [93]. Pore-size distributions cannot as easily be obtained. There are two reasons: (1) The B0 field of the NMR-MOUSE is strongly inhomogeneous and the acquired transverse relaxation decay is attenuated from diffusive motion in the larger pores. As a consequence, the distribution of T2eff obtained by Laplace inversion of the CPMG decay appears compressed for larger T2eff , and T2eff is no longer proportional to the pore size in the fast diffusion limit. (2) The material to be investigated needs to be completely fluid saturated. This cannot be achieved for rock once it has been dried. For complete saturation, the CPMG signal amplitude or the integral of

Fig. 7. High-resolution depth profiling. (A) 1D depth cross-sections of the sensitive volumes of the original u-shaped NMR-MOUSE and the optimized profile NMR-MOUSE. (B) Mechanical precision lift for shifting the sensitive volume through the object. (C) Moisture profiles of a polymer sheet depicting the water uptake as a function of the wetting time. (D) Moisture profiles of a dry sheet and a sheet exposed to the humidity of air.

Mobile NMR

Acknowledgments 379

the relaxation time distribution should be proportional to the porosity. But for rocks soaked after drying it depends on the fluid conductivity (Figure 8a). So originally wet cores need to be measured to obtain reliable porosity values by NMR. This is one further area of application of portable NMR in geophysics. For porosities lower than 5% the sensitivity of the NMR-MOUSE is not good enough and a Halbach magnet with a more homogeneous B0 field and a larger sensitive volume need to be employed (Figure 1d). Its more homogeneous field permits the measurement of reliable relaxation time distributions, which, for example, in a drying study, reveal the preferential water loss from large pores with long relaxation times (Figure 8b) [36]. Nevertheless, the NMR-MOUSE has produced interesting results in a pilot study of the effect of stone conservation treatment conducted at the sandstone window frames of Paffendorf Castle near Cologne, Germany. The areas to be measured were partially wetted (Figure 8c) and differences in the relaxation time distributions of the untreated and treated frames reveal a more frequent occurrence of faster relaxation for the treated material. This promises, that unilateral NMR can be developed into a method to assess the success of stone conservation efforts non-invasively. A related study concerns the characterization of water and oil emulsions as model systems for food, where it has been demonstrated that the concentration of oil and water can be determined from the NMR-MOUSE signals [94].

Summary Portable unilateral NMR concerns methods and NMR devices, which are carried to the object under study. It is an offspring of well-logging NMR. For materials analysis strongly inhomogeneous magnetic fields can be employed. NMR in inhomogeneous fields is an active area of research, and in addition to NMR relaxation, it has recently been demonstrated that NMR images, flow profiles, and even spectra can be measured by unilateral NMR. Mobile unilateral sensors like the NMR-MOUSE are lightweight and inexpensive. The measurement is nondestructive, and arbitrarily large samples can be investigated in situ up to depths of some 10 mm. The NMRMOUSE can be employed for product development and quality control in a manufacturing environment. Other portable magnet geometries with better field homogeneity, such as the Halbach magnet, are also being explored for portable NMR. They can be used to study pipe flow, geophysical drill cores, and the like, where the completely open access provided by unilateral sensors is not required.

Acknowledgments This work has been conducted with support by Deutsche Forschungsgemeinschaft (DFG) and Bundesministerium

Part I

Fig. 8. Porous media. (a) Correlation of NMR porosities measured with the NMR-MOUSE and porosities of geological core samples with different fluid conductivities measured with a helium gas pycnometer. (b) Relaxation-time distribution functions of a core sample at different drying times. (c) Measurement of a partially wetted sandstone window frame in Paffendorf castle. (d) Relaxation time distributions for two frames, one not treated and the other one treated with a stone conservation agent.

380 Part I

Chemistry

Part I

f¨ur Bildung und Forschung (BMBF). The contributions of Juan Perlo, Kai Kremer, Sophia Anferova, Vladimir Anferov, Nicolae Goga, Vasilikis Demas, Alina Buda, Dan Demco, Peter Bl¨umler, Michael Adams, and Klaus Kupferschl¨ager are gratefully acknowledged.

References 1. Bl¨umich B. NMR Imaging of Materials, Clarendon Press, Oxford, 2000. 2. Matzkanin GA. A review of nondestructive characterization of composites using NMR. In: P H¨oller, V Hauck, G Dobmann, C Ruud, R Green (Eds). Nondestructive Characterization of Materials. Springer: Berlin, 1989, pp 655–69. 3. Coates GR, Xiao L, Prammer MG. NMR Logging Principles and Applications. Halliburton Energy Services: Houston, 1999. 4. Kleinberg RL. Well logging. In: DM Grant, RK Harris (Eds). Encyclopedia of NMR. Wiley-Liss: New York, 1996, pp 4960–69. 5. Cooper RK, Jackson JA. Remote (inside-out) NMR. I. Remote production of a region of homgeneous magnetic field. J. Magn. Reson. 1980;41:400–405. 6. Burnett LJ, Jackson JA. Remote (inside-out) NMR. II. Sensitivity of NMR detection for external samples. J. Magn. Reson. 1980;41:406–10. 7. Jackson JA, Burnett LJ, Harmon F. Remote (inside-out) NMR. III. Detection of nuclear magnetic resonance in a remotely produced region of homogeneous magnetic field. J. Magn. Reson. 1980;41:411–21. 8. Paetzold RF, Matzkanin GA, De Los Santos A. Surface water content measurement using pulse nuclear magentic resonance techniques. Soil Sci. 1985;49:537–40. 9. Paetzold RF, De Los Santos A, Matzkanin GA. Pulse nuclear magnetic resonance instrument for soil-water content measurement: sensor configurations. Soil Sci. 1987;51:287–90. 10. Hogan BJ. One-Sided NMR Sensor System Measures Soil/Concrete Moisture, Design News, May 5 (1986). 11. Rollwitz WL. Using radiofrequency spectroscopy in agricultural applications. Agric. Eng. 1985;66:12–14. 12. Nicholls CI, De Los Santos A. Hydrogen transient nuclear magnetic resonance for industrial moisture sensing. Drying Technol. 1991;9:849–73. 13. Eidmann G, Savelsberg R, Bl¨umler P, Bl¨umich B. The NMRMOUSE: a mobile universal surface explorer. J. Magn. Reson. A. 1996;122:104–9. 14. Bl¨umich B, Bl¨umler P, Eidmann G, Guthausen A, Haken R, Schmitz U, Saito K, Zimmer G. The NMR MOUSE: construction, excitation, and applications. Magn. Reson. Imaging. 1998;16:479–84. 15. Anferova S, Anferov V, Adams M, Bl¨umer P, Routley N, Hailu K, Kupferschl¨ager K, Mallet M, Schroeder G, Sharma S, Bl¨umich B. Construction of the NMR-MOUSE with short dead time. Magn. Reson. Eng. 2002;15:15–25. 16. Bl¨umich B, Anferov V, Anferova S, Klein M, Fechete R, Adams M, Casanova F. Simple NMR-MOUSE with a bar magnet. Concepts Magn. Reson. 2002;15:255–61.

17. Goelman G, Prammer MG. The CPMG pulse sequence in strong magnetic field gradients with application to oilwell logging. J. Magn. Reson. A. 1995;113:11–18. 18. McDonald PJ. Stray-field magnetic resonance imaging. Prog. Nucl. Magn. Reson. Spectrosc. 1997;30:69–99. 19. H¨urlimann MD, Griffin DD. Spin dynamics of Carr-PurcellMeiboom-Gill-like sequences in grossly inhomogeneous B0 and B1 fields and application to NMR well logging. J. Magn. Reson. 2000;143:120–35. 20. Balibanu F, Hailu K, Eymael R, Demco DE, Bl¨umich B. NMR in inhomogeneous magnetic fields. J. Magn. Reson. 2000;145:246–58. 21. H¨urlimann MD. Optimization of timing in the CarrPurcell-Meiboom-Gill sequence. Magn. Reson. Imaging. 2001;19:375–78. 22. H¨urlimann MD. Carr-Purcell sequences with composite pulses. J. Magn. Reson. 2001;152:109–23. 23. Song Y-Q. Categories of coherence pathways for the CPMG sequence. J. Magn. Reson. 2002;157:82–91. 24. Todica M, Fechete R, Bl¨umich B. Selective NMR excitation in strongly inhomogeneous magnetic fields. J. Magn. Reson. 2003;164:220–27. 25. Prado PJ. NMR handheld moisture sensor. Magn. Reson. Imaging. 2001;19:505–8. 26. Popella H, Henneberger G. Design and optimization of the magnetic resonance circuit of a mobile nuclear magnetic resonance device for magnetic resoance imaging. Compel. 2001;20:269–79. 27. Podol’skii A. Permanent-magnet assemblies for magnetic resonance imaging devices for various purposes. IEEE Trans. Magn. 2002;38:2–5. 28. Pulyer Y, Hrovat MI. An open magnet utilizing ferrorefraction current magnification. J. Magn. Reson. 2002;154:298–02. 29. Pulyer YM. Generation of remote homogeneous magnetic fields. IEEE Trans. Magn. 2002;38:1553–63. 30. Prado PJ. Single-sided imaging sensor. Magn. Reson. Imaging. 2003;21:397–400. 31. Hunter MW, Callaghan PT, Dyskra R, Eccles CD, Vamanan S. Design and construction of a portable one-sided access NMR probe, Abstract to the 7th International Conference on Magnetic Resonance in Porous Media, July 4–8, 2004, Paris, France. 32. Fukushima E, Jackson JA. Unilateral magnet having a remote uniform field region, US patent 6,489,872, 3 December 2002. 33. Bl¨umich B, Anferov V, Anferova S, Klein M, Fechete R. R An NMR-MOUSE
for analysis of thin objects. Macromol. Mater. Eng. 2003;288:312–17. 34. Halbach K. Design of permanent multipole magnets with oriented rare earth cobalt material. Nucl. Instr. Meth. 1980;169:1–10. 35. Raich H, Bl¨umler P. Design and construction of a dipolar Halbach array with a homogeneous field from identical barmagnets. Concepts Magn. Reson. 2004;23B:16–25. 36. Anferova S, Anferov V, Rata DG, Bl¨umich B, Arnold J, Clauser C, Bl¨umler P, Raich H. A mobile NMR device for measurements of porosity and pore size distributions of drilled core samples. Concepts Magn. Reson. 2004;23B: 26–32. 37. Hahn EL. Spin echoes. Phys. Rev. 1950;80:580–94.

Mobile NMR

59. Klein M, Fechete R, Demco DE, Bl¨umich B. Selfdiffusion measurements by a constant-relaxation method in strongly inhomogeneous magnetic fields. J. Magn. Reson. 2003;164:310–20. 60. Song Y-Q, H¨urlimann MD, Flaum C. A method for rapid characterization of diffusion. J. Magn. Reson. 2003;161:222– 33. 61. Song Y-Q, Scheven UM. An NMR technique for rapid measurement of flow. J. Magn. Reson. 2005;172:31–35. 62. Ardelean I, Kimmich R, Klemm A. The nutation and spin echo and its use for localized NMR. J. Magn. Reson. 2000;146:43–48 63. Perlo J, Demas V, Casanova F, Meriles CA, Reimer J, Pines A, Bl¨umich B. High-resolution NMR spectroscopy with a portable single-sided sensor. Science. 2005;308:1278. 64. Meriles CA, Sakellariou D, Pines A. Resolved magic-angle spinning of anisotropic samples in inhomogeneous fields. Chem. Phys. Lett. 2002;358:391–95. 65. Meriles CA, Sakellariou D, Moul´e A, Goldman M, Budinger TF, Pines A. High-resolution NMR of static samples by rotation of the magnetic field. J. Magn. Reson. 2004;169:13–18. 66. Callaghan PT, Eccles CD, Seymour JD. An earth’s field nuclear magnetic resonance apparatus suitable for pulsed gradient spin echo measurements of self-diffusion under Antarctic conditions. Rev. Sci. Instrum. 1997;68:4263– 270. 67. Callaghan PT, Eccles CD, Haskell TG, Langhorne PJ, Seymour JD. Earth’s field NMR in Antarctica: a pulsed gradient spin echo NMR study of restricted diffusion in sea ice. J. Magn. Reson. 1998;133:148–54. 68. Callaghan PT, Dykstra R, Eccles CD, Haskell TG, Seymour JD. A nuclear magnetic resonance study of Antarctic sea ice brine diffusivity. Cold Regions Sci. Tech. 1999;29:153–71. 69. Litvinov VM, De PP (Eds). Spectroscopy of Rubbers and Rubbery Materials. Rapra Technology Limited: Shawbury, 2002. 70. Herrmann V, Unseld K, Fuchs H-B, Bl¨umich B. Molecular dynamics of elastomers investigated by DMTA and the R NMR-MOUSE
. Colloid Polym. Sci. 2002;280:758–64. 71. Anferova S, Anferov V, Adams M, Fechete R, Schroeder G, Bl¨umich B. Thermo-oxidative aging of elastomers: a temperature control unit for operation with the R NMR-MOUSE
. Appl. Magn. Reson. 2004;27:361–370. 72. Goga N, Kremer K, Bl¨umich B. Qualit¨atskontrolle in der Reifenindustrie, Gummi Fasern Kunststoffe 2005;58:361– 70. 73. Bl¨umich B, Bruder M. Mobile NMR zur Qualit¨atskontrolle von Elastomerprodukten, Kautschuk Gummi Kunststoffe, 2003;56:90–94. 74. Bl¨umich B, Anferova S, Casanova F, Kremer K, Perlo J, Sharma S. Unilateral NMR: principles and applications to quality control of elastomer products. Kautschuk Gummi Kunststoffe. 2004;57:346–49. 75. Zimmer G, Guthausen A, Schmitz U, Saito K, Blu¨ mich B. Wheathering investigation of PVC coatings on iron sheets by the NMR MOUSE. Adv. Mater. 1997;9:987–89. 76. Hailu K, Fechete R, Demco DE, Bl¨umich B. Segmental anisotropy in strained elastomers detected with a portable NMR scanner. Solid State Nucl. Magn. Reson. 2002;22:327–43.

Part I

38. Carr HY, Purcell EM. Effects of diffusion on free precession in NMR experiments. Phys. Rev. 1954;94:630–38. 39. Meiboom S, Gill D. Compensation for pulse imperfections in Carr-Purcell NMR experiments. Rev. Sci. Instrum. 1958;29:688–91. 40. Bl¨umich B, Anferova S, Kremer K, Sharma S, Herrmann V, Segre A. Unilateral NMR for quality control: the NMRR MOUSE
. Spectroscopy. 2003;18:24–54. 41. Sezginer A, Kleinberg RL, Fukuhara M, Latour LL. Very rapid simultaneous measurement of nuclear magnetic resonance spin-lattice relaxation time and spin-spin relaxation time. J. Magn. Reson. 1991;92:504–27. 42. Guthausen A, Zimmer G, Bl¨umler P, Bl¨umich B. Analysis of polymer materials by surface NMR via the NMR MOUSE. J. Magn. Reson. 1998;130:1–7. 43. Song Y-Q, Venkataramanan L, H¨urlimann MD, Flaum M, Frulla P, Straley C. T 1 − T 2 correlation spectra obtained using a fast two-dimensional laplace inversion. J. Magn. Reson. 2002;154:261–68. 44. Wiesmath A, Filip C, Demco DE, Bl¨umich B. Doublequantum-filtered NMR signals in inhomogeneous magnetic fields. J. Magn. Reson. 2001;149:258–63. 45. Wiesmath A, Filip C, Demco DE, Bl¨umich B. NMR of multipolar spin states excited in strongly inhomogeneous magnetic fields. J. Magn. Reson. 2002;154:60–72. 46. Meriles CA, Sarkellariou D, Heise H, Moul´e AJ, Pines A. Approach to high-resolution ex situ NMR spectroscopy. Science. 2001;293:82–85. 47. Heise H, Sakellariou D, Meriles CA, Moul´e A, Pines A. Twodimensional high-resolution NMR spectra in matched B0 and B1 field gradients. J. Magn. Reson. 2002;156:146–51. 48. Prado PJ, Bl¨umich B, Schmitz U. One-dimensional imaging with a palm-size probe. J. Magn. Reson. 2000;144:200–06. 49. Casanova F, Bl¨umich B. Two-dimensional imaging with a single-sided NMR probe. J. Magn. Reson. 2003;163:38–45. 50. Casanova F, Perlo J, Bl¨umich B, Kremer K. Multi-echo imaging in highly inhomogeneous magnetic fields. J. Magn. Reson. 2004;166:76–81. 51. Perlo J, Casanova F, Bl¨umich B. 3D imaging with a single-sided sensor: an open tomograph. J. Magn. Reson. 2004;166:228–35. 52. Casanova F, Perlo J, Bl¨umich B. Velocity distributions remotely measured with a single-sided NMR sensor. J. Magn. Reson. 2004;171:124–30. 53. Perlo J, Casanova F, Bl¨umich B. Velocity imaging by ex situ NMR. J. Magn. Reson. 2005;173:254–258. 54. H¨urlimann MD. Diffusion and relaxation effects in general stray field NMR experiments. J. Magn. Reson. 2001;148:637–378. 55. H¨urlimann MD, Venkataramanan L. Quantitative measurement of two-dimensional distributions functions of diffusion and relaxation in grossly inhomogeneous fields. J. Magn. Reson. 2002;157:31–42. 56. Guthausen G, Guthausen A, Balibanu F, Eymael R, Hailu K, Schmitz U, Bl¨umich B. Soft-matter analysis by the NMRMOUSE. Macromol. Mater. Eng. 2000;276/277:25–37. 57. Casieri C, Bubici S, De Luca F. Self-diffusion coefficient by single-sided NMR. J. Magn. Reson. 2003;162:348–55. 58. Marko A, Wolter B. Diffusion studies in porous media with the “inside-out” technique. Magn. Reson. Imaging. 2003;21:363–64.

References 381

382 Part I

Chemistry

Part I

77. Haken R, Bl¨umich B. Anisotropy in tendon investigation in vivo by a porg NMR scanner, the NMR-MOUSE. J. Magn. Reson. 2000;144:195–99. 78. Cornelissen C, Schnettler A, Wiesmath A, Bl¨umich B. Ultraschall und Kernspinresonanz: Nicht-destruktive Verfahren zur Zustandsbewertung von Kabelsystemen, Das Magazin f¨ur die Energiewissenschaften ew 2002;101:38–43. 79. Bl¨umich B, Casanova F, Buda A, Kremer K, Wegener T. Anwendungen der mobilen NMR zur Zustandsbewertung von Bauteilen aus Polyethylen. 3R International. 2005;44:349–354. 80. Bl¨umich B, Anferova S, Sharma S, Segre A, Federici C. Degradation of historical paper: nondestructive analysis R by the NMR-MOUSE
. J. Magn. Res. 2003;161:204–9. 81. Proietti N, Capitani D, Pedemonte E, Bl¨umich B, Segre AL. Monitoring degradation in paper: non-invasive analysis by unilateral NMR. Part II, J. Magn. Reson. 2004;170:113–20. 82. Casieri C, Bubici S, Viola I, de Luca F. A low-resolution non-invaisve NMR investigation of ancient paper. Solid State Nucl. Magn. Reson. 2004;26:65–73. 83. Casieri C, Senni L, Romangnoli M, Santamaria U, de Luca F. Determination of moisture fraction in wood by mobile NMR device. J. Magn. Reson. 2004;171:364–72. 84. Perlo J, Casanova F, Bl¨umich B. Single-sided NMR profiler with microscopic resolution, manuscript in preparation. 85. Glover PM, Aptaker PS, Bowler JR, Ciampi E, McDonald PJ. A novel high-gradient permanent magnet for the profiling of planar films and coatings. J. Magn. Reson. 1999;139:90–97.

86. Godward G, Ciampi E. Cifelli M, McDonald PJ. Multidimensional imaging using combined stray field and pulsed gradients. J. Magn. Reson. 2002;155:92–99. 87. Hartwig A, Wolter B. NMR-Aufsatztechnik: Neue OnlineMethode zum zerst¨orungsfreien Pr¨ufen?, Adh¨asion kleben & dichten 2001;11:25–29. 88. A. Hartwig and Wolter B. Zerst¨orungsfreie NMRAufsatztechnik: Anwendungsbeispiele und M¨oglichkeiten, Adh¨asion kleben & dichten 2001;12:34–38. 89. Kremer K, K¨uhn H, Bl¨umich B, Seitzer J, Schmitz F-P. R u¨ bernimmt Qualit¨atskontrolle im KFZ-Bau, NMR-MOUSE
Adh¨asion kleben & dichten 2002;11:32–36. 90. Brown MCA, Verganelakis DA, Mallett MJD, Mitchell J, Bl¨umler P. Surface normal imaging with a handheld NMR device. J. Magn. Reson. 2004;169:308–12. 91. Dobmann G, Kroening M, Surkowa N, von Bernus J, Wolter B. The potential of nuclear magnetic resonance (NMR) to non-destructively characterize early-age concrete by an one-sided access (OSA) technique, NDE2002 predict, assure, improve, National Seminar of ISNT, Chennai, 7 December 2002, www.nde2002.org. 92. Sharma S, Casanova F, Wache W, Segre A, Bl¨umich B. Analysis of historical porous building materials by the R NMR-MOUSE
. Magn. Reson. Imaging. 2003;21:249–55. 93. Bl¨umich B, Anferova S, Pechnig R, Pape H, Arnold J, Clauser C. Mobile NMR for porosity analysis of drill core sections. J. Geophys. Eng. 2004;1:177–80. 94. Pedersen HT, Ablett S, Martin DR, Mallett MJD, Engelsen SB. J. Magn. Reson. 2003;165:49–58.

383

Paul T Callaghan MacDiarmid Institute for Advanced Materials and Nanotechnology, Victoria University of Wellington, New Zealand

One recent application of NMR concerns rheology [1,2], the study of the mechanical properties of fluids. This application has come to be known as “Rheo-NMR” [3–7]. Many interesting materials in their condensed phase possess both solid and liquid-like properties. These include high molecular mass polymers and elastomers, lyotropic and thermotropic liquid crystals, micellar surfactant phases, colloidal suspensions, foams, emulsions, micro-emulsions and bicontinuous phases. Such materials, strongly represented in the biological world, comprise what is often called “soft matter” or “complex fluids”. Complex fluids manifest both an elastic and a viscous response, and they generally possess “memory”, which means that the stress which they exhibit at any moment will depend on the history of prior deformation. They often exhibit non-linear flow behavior, which means that properties may change as the deformation increases, an effect which is generally attributed to molecular reorganisation. And they invariably possess a wide range of characteristic time scales, from the rapid (ps to ns) local Brownian motion of small molecules or molecular segments, to the very slow (ms to s) motions associated with the reorganisation or reorientation of large molecular assemblies or macromolecules. While rheology involves mechanical measurement of flow properties, the really interesting questions concern the molecular basis of these properties. Under flow, competition arises between the molecular organisational dynamics and the externally imposed deformation, with outcomes including conformational distortion [8], re-organisation of mesophase structure [9], doublevaluedness in the constitutive properties [10] banded flow [11], the driving of the material through nearby phase transitions [12], and soft glassy dynamics, the slow aging of a system as the structure reorganizes [13]. The interest in the molecular-mechanical link has led to the amalgamation of a number of spectroscopic and rheological techniques in which a flow or deformation cell is incorporated within the spectrometer detection system. Examples include the use of neutron scattering, light scattering, birefringence, and dichroism techniques [2]. The most recent addition, NMR, allows one to study materials which are optically opaque. The imaging capability of NMR means that it can be used to directly measure local Graham A. Webb (ed.), Modern Magnetic Resonance, 383–388.
C 2008 Springer.

velocity profiles and molecular densities. And the wideranging spectroscopic tools available to Rheo-NMR make it possible to measure molecular order and dynamics. Rheo-NMR based on micro-imaging approaches [14] allows the mapping of fluid velocity in small (mm to cm scale) deformation cells, the small volumes allowing the study of specialized materials. The velocimetry mode of micro-imaging generally employs a Pulsed Gradient Spin Echo (PGSE) sequence in which magnetic field gradientpulses define a wave vector domain, q, which imparts a phase shift to the spins depending directly on the motion of their parent molecules [14]. Inverse Fourier transformation of the signal with respect to q returns the local distribution of velocities, P(v), for each pixel of the image. A typical velocity image will take between seconds and several minutes to acquire, depending on signal-to-noise trade-offs. The upper limit to velocity is determined by image distortion or inflow-outflow effects and is typically 1 ms−1 . The lower limit is determined by Brownian motion, enabling velocity resolution on the order of 10 micrometer per second for small molecules such as water but down to 100 nms−1 for macromolecules or colloidal particles, as shown in Figure 1 which depicts the velocity field for a soft glassy material formed from a close packing of 370 nm diameter latex spheres [15]. This example exhibits both slip and yield stress behavior, indicating the value of such flow visualization. Rheo-NMR flow geometries [16,17] include coneand-plate cells, cylindrical Couette cells, four roll mills and bi-axial extension cells, these latter devices being used to produce purely extensional flow. All these deformational flow devices are driven by a drive shaft which sits in the bore of the magnet and which is turned by a steppermotor gearbox assembly mounted above the magnet bore. By contrast flow-though geometries include simple pipe as well as opposed jet systems. A typical Rheo-NMR kit is shown in Figure 2. In most materials there exists a monotonic relationship (“flow curve”) between the applied stress σ and the rate of strain, γ˙ [1]. In a rheological cell for which the stress is nearly uniform, such as the small angle cone-and-plate device, one would therefore expect a unique strain rate to occur at any applied stress. One of the first significant contributions of Rheo-NMR has been to show that

Part I

Rheo-NMR

Chemistry

Part I

Fig. 1. Velocity profile across a cylindrical Couette cell of inner cylinder ID 5 mm and outer cylinder ID 9.4 mm, for 0.48 volume fraction 370 nm diameter core-shell latex spheres suspension in water. The left hand arrow indicates the region of the annular gap where a yield stress point is apparent, dividing fluidized material from the glassy state. The right-hand arrow point to fluid within the center cylinder undergoing rigid body motion. Note the slip at the inner and outer walls (adapted from ref. [15]).

0.04 f = 3.2*10–3Hz

0.03 0.02 velocity (mm–1)

384 Part I

0.01 0 –0.01 –0.02 –0.03 –0.04

0

the uniform shear-rate assumption may be violated in the case of certain classes of fluids in which pathological flow properties are exhibited. Figure 3 shows velocity maps and associated shear-rate maps [18] obtained for the wormlike surfactant system, cetylpyridinium chloride/sodium

2 6 8 4 radial displacement (mm)

10

salicylate in water. While the velocity gradients show no deviation from uniformity at very low shear rates, above a certain critical value γ˙c a dramatic variation across the ◦ 6 cone gap is apparent in which a very high shear rate band exists at mid-gap.

RheoNMR Controller

Motor, gearbox, drive interface & drive adapter

Cell Kit

Drive shaft

Fig. 2. Set of Rheo-NMR cells and drive attachments for a 89 mm vertical bore NMR magnet (http://www.magritek.com). (See also Plate 44 on page XXXIV in the Color Plate Section.)

Rheo-NMR

Part I

a)

Rheo-NMR 385

25

20

15

20 0 600

10 500 400 300 tim e( 200 S) 100

b)

t en

0

em lac sp gap i d  s ial ros rad ac

velocity(mm/s)

40

5

0

60 50

30

time(s)

40

20 10 0

Fig. 4. (See also Plate 45 on page XXXV in the Color Plate Section.)

Fig. 3. Shear rate distribution for cetylpyridinium chloride/NaSal wormlike micelle solution at an apparent shear rate of 16 s−1 , well beyond the critical shear rate and in an unstable region of the flow curve (horizontal field of view 25 mm with vertical field of view smaller by a factor of 6) Also shown are experimental shear rate profiles along a line of approximately fixed radius (adapted from reference from reference 18).

This shear banding phenomena is believed to arise from an inflected flow curve that includes an unstable branch in which the stress declines with increasing shear rate. The fluid thus “phase separates” onto the upper and lower rising branches of the underlying flow curve and the proportions of each band will be as required to satisfy the average shear rate, in the manner of a lever rule. That the NMR results are consistent with this picture is clear in Figure 3 where a series of profiles show that as the gap apparent shear rate, is increased the high shear rate band expands in width at approximately constant maximum shear rate. More recent work has shown that these bands fluctuate extremely rapidly, as seen in the successive profiles taken, at one second intervals, in the Couette cell geometry of Figure 4 [19]. To achieve this time resolution, a high speed imaging sequence was employed. These NMR results have stimulated new theoretical models for

the coupling of flow to micro-scopic molecular order parameters. Rheo-NMR is also capable of investigating molecular order and alignment [20] through utilising internuclear dipole interactions or nuclear quadrupole interactions. It is through the use of such spectroscopic approaches that Rheo-NMR holds the promise of further linking mechanical and molecular properties. Figure 5 shows the result of a shearing study on a wormlike micelle system (20% CTAB/D2 O at 41 ◦ C) close to an isotropic-nematic transition [21,22]. The D2 O2 H NMR spectrum, is plotted as a function of radial position across the gap of a cylindrical Couette cell where the magnetic field is aligned with the vorticity axis. At the inner wall, where the stress is highest, a splitting is observed [21], indicative of a finite quadrupole interaction, while at the outer wall a single peak is observed. These data suggest the formation of a nematic phase at high stress and the transition to an isotropic phase, through a mixed phase region, at the region of low stress. Another intriguing correlation between shearing and molecular conformation concerns a random coil polymer melt being subjected to shear. Here the polymer chain suffers a biaxial deformation in which the principal axis of extension has a preferred orientation with respect to the

386 Part I

Chemistry

Part I

Couette gap

outer wall

inner wall

distance from center (mm) 9.46

outer wall

9.37 9.29 9.21 9.12 9.04 8.96 8.87 8.79 8.71 8.62 8.54 inner wall 10

5

0 frequencyoffset(Hz)

5

10

Fig. 5. 2 H NMR spectra obtained from 20% w/w CTAB/D2 O (41 ◦ C) at different positions across the annular gap of a cylindrical Couette cell and at an apparent shear rate of 20 s–1 . Near the inner wall, where the stress is highest, a quadrupole splitting is observed, consistent with an ordered phase, while near the outer wall the single peak of an isotropic phase is seen. In between a mixed phase region exists (adapted from reference 21).

hydrodynamic velocity direction, the deformation being described by means of an averaged segmental alignment tensor that may be evaluated using the Doi–Edwards formulation of entangled polymer dynamics [8]. Rheo-NMR has been used to obtain elements of the tensor in a high molecular weight polymethylsiloxane melt confined to a Couette cell of 0.5 mm gap [23,24]. In this work a small deuterated benzene probe undergoes steric interactions with the polymer segments and experiences an anisotropic

mean orientation. NMR micro-imaging is used to view the PDMS both to image the velocity distribution across the gap and to excite a desired region of the sample for spectroscopy experiments during steady-state shear. The deuteron NMR signal exhibits a scaled down quadrupole splitting proportional to the average value of the selected tensor element. Figure 6 shows the measured alignment tensor elements along with fits using the Doi–Edwards model, which is parameterized by the tube disengage-

Rheo-NMR

180y

t2

t1

t

Part I

90x

90-x

d-benzene in sheared PDMS

Suggested Reading 387

rf

scan of complete cell

45-x 45-x

G 80

Sxx

Doi-Edwards

polymer in the gap

splitting /Hz

60 40

Sxy 20 τd=215

0

ms

-20

Szz

0.5 mm gap

Syy

-40 0

10

20

30

40

50

60

70

80

90

100

shear rate /s-1 -200 hz

0

f1

200 Hz

Fig. 6. Quadrupole splittings, v, vs. Deborah number, obtained from deuterated benzene in 350 kdalton PDMS, using a selected region of the horizontal Couette cell in which the velocity direction (solid circles) and gradient direction (open circles) are respectively parallel to the magnetic field. The lines are fits using the Doi–Edwards model in which the horizontal axis is scaled to yield the tube disengagement time, τd (adapted from ref. [24]). (See also Plate 46 on page XXXV in the Color Plate Section.)

ment time associated with reptation. Also shown are the selected regions in which either the velocity direction or the velocity gradient (shear axis) is parallel to B0 . These examples provide a glimpse of possible applications of Rheo-NMR. While this is a very new field of research in which only a handful of groups presently participate, the potential exists for a substantial increase in Rheo-NMR research activity. Systems studied to date include polymer melts and semi-dilute solutions, thermotropic and lyotropic liquid crystals and liquid crystalline polymers, micellar solutions, food materials, and colloidal suspensions. The ability to combine velocimetry with localised spectroscopy, and the ability to access a wide range of molecular properties relating to organisation, orientation, and dynamics has enabled Rheo-NMR to provide a direct window on a variety of behaviors, including slip, shear-thinning, shear banding, yield stress behavior, nematic director alignment, and shear-induced mesophase reorganisation. The unique information available with this method suggests that it is likely to become an important tool in elucidating the in-

triguing rheological behavior of a wide range of complex fluids.

Suggested Reading 1. Barnes HA, Hutton JJ, Walters K. An Introduction to Rheology. Elsevier: Amsterdam, 1989. 2. Fuller GG. Optical Rheometry of Complex Fluids. Clarendon Press: Oxford, 1995. 3. Martins AF, Esnault P, Volino F. Phy. Rev. Lett. 1986;57: 1745. 4. Nakatani AI, Poliks MD, Samulski ET. Macromolecules 1990; 23:2686. 5. Xia Y, Callaghan PT. Macromolecules. 1991;24:4777. 6. Grabowski DA, Schimdt C. Macromolecules 1994;27:2632. 7. Callaghan PT. Rep. on Prog. in Phys. 1999;62:599. 8. Doi M, Edwards SF. The Theory of Polymer Dynamics. Oxford University Press: Oxford 1987. 9. Marrucci G. In: TCB McLeish (Ed.). Theoretical Challenges in the Dynamics of Complex Fluids. Kluwer Press: Dordrect, 1997, pp 141–158.

388 Part I

Chemistry

Part I

10. McLeish TCB, Ball R. A molecular approach to the Spurt effect in polymer melt flow. J. Polymer Science 1986;24: 1735–1745. 11. Cates ME, McLeish TCB, Marrucci G. The Rheology of Entangled Polymers at Very High Shear Rates. Europhys. Lett. 1993;21:451. 12. Helfand E, Fredrickson GH, Large fluctuations in polymersolutions under shear. Phys. Rev. Lett. 1989;62: 2468. 13. Durian DJ. Foam mechanics at the bubble scale. Phys. Rev. Lett. 1995;75:4780. 14. Callaghan PT, Principles of Nuclear Magnetic Resonance Microscopy. Oxford University Press: Oxford, 1991. 15. Wassenius H, Callaghan PT, Nanoscale NMR velocimetry. J. Magn. Reson. 2004;169: 250–256.

16. Britton MM, Callaghan PT, Kilfoil ML, Mair RW, Owens K. Applied Mag. Reson. 1998, 15 287. 17. Lukaschek M, Grabowski DA, Schmidt C Langmuir 1995;11: 3590. 18. Britton MM, Callaghan PT. Phys. Rev. Lett. 1997;78:4930. 19. L´opez-Gonz´alez MR, Photinos P, Holmes WM, Callaghan PT, Phys. Rev. Lett. 2004;93:268302–268305. 20. Siebert H, Grabowski DA, Schmidt C. Rheologica. Acta. 1997;36:618. 21. Fischer E, Callaghan PT. Europhy. Lett. 2000;50:803. 22. Decruppe JP, Cressely R, Makhoufli R, Cappelaere E. Colloid Polym. Sci. 1995;273:346. 23. Kilfoil ML, Callaghan PT, Macromolecules 2000;33:6828. 24. Cormier RJ, Kilfoil ML, Callaghan PT. Phys. Rev. 2001; E6405:1809.

389

Cecil Dybowski Department of Chemistry and Biochemistry, University of Delaware, Newark, Delaware 19716-2522 USA

Introduction In one sense of the word, every use of nuclear magnetic resonance is analysis, so to discuss analytical aspects of solid-state NMR spectroscopy is to discuss its myriad uses. Clearly, such a general approach is neither feasible nor appropriate in a short article, so one needs to focus the discussion. Roughly speaking, experiments involving NMR spectroscopy belong to one of two classes: (1) those that primarily focus on nuclear magnetic resonance as a process, with the goal of learning about or expanding its utility; and (2) experiments that involve NMR as a tool to address some question about the nature of a chemical or physical system and in which development of NMR technology is of secondary importance. While experiments in the first class may provide new information on samples (usually model materials), it is the latter sort of experiment that I make the theme of this chapter. It is important to emphasize that this dichotomy is not clear, and one must often include experiments on models to understand the nature of analysis with NMR spectroscopy.

Uses of Isotropic Shielding to Identify Materials A discussion of the analytical aspects of solid-state NMR spectroscopy involves an understanding of what is meant by “analytical.” To many NMR spectroscopists, the word connotes determination of molecular structure by a careful correlation of spectroscopic absorptions with expected functional groups. NMR spectroscopy is an excellent means to determine the details of molecular structure of a pure material by such experiments, whether the material is a solid or is in the liquid state. The long, successful history of organic structural analysis with liquid-state NMR spectroscopy exemplifies this analytical aspect of NMR spectroscopy. Analyses that specify the nature of functional groups in the solid state demonstrate a similar analytical methodology for addressing questions of molecular structure in the solid state. Of course, the information on molecular structure in the liquid state and the solid state may differ because of physical differences between the solidstate structure and the average structure detected with Graham A. Webb (ed.), Modern Magnetic Resonance, 389–394.
C 2008 Springer.

solution-state NMR spectroscopy. The classical example of this kind of analysis of a solid is found in the application of 29 Si MAS NMR spectroscopy to the examination of zeolite structure, exemplified by the work of the Exxon NMR group [1]. The isotropic position of NMR spectroscopic absorption depends strongly on the local environment of the nucleus, which allows one to assign the resonances in the NMR spectrum to specific silicon sites [2]. By examination of a wide variety of materials, it has been shown that the variable that most significantly affects the resonance position is the number of aluminum atoms in the nearby environment. In experiments that account for the effects of relaxation rates on line intensities, it is possible to use the intensities of lines from silicon in various environments to estimate the distribution of silicon in various sites. Such an NMR-derived distribution can then be compared with theoretical predictions of the distribution of silicon in the zeolite framework [3]. The concurrent examination of a zeolite by X-ray diffraction, where possible, and magic-angle spinning (MAS) NMR of a zeolite provides a synergy that often allows one to unambiguously assign structures. These techniques have been used repeatedly over many years to analyze zeolite structure and are a foundation of modern zeolite analysis. MAS NMR spectroscopy of spin-1/2 nuclei in solids is particularly relevant to organic materials. For example, an early application of MAS NMR to poly(phenylene oxide) revealed that in the solid state, the protonated aromatic in the carbon spectrum appears as a doublet, whereas in solution the carbon resonance is a singlet [4]. The explanation of this observation is that, in the solid state, the structure is locked on the NMR timescale, with the result that carbons that are nominally equivalent in solution become inequivalent in the solid state. Observation of such differences between solution- and solid-state spectroscopy points up the fundamental fact that chemical shielding depends on the details of structure, but it also provides an interesting use of solid-state NMR spectroscopy to investigate the existence of structural differences between the solution state and the solid state. An interesting use of MAS NMR spectroscopy is the application to archaeology. For example, the study of residues on pottery has given information on the nature of materials present on pots [5]. Similarly, the qualities

Part I

Analytical Aspects of Solid-State NMR Spectroscopy

390 Part I

Chemistry

Part I

of woods in certain archeological materials may be examined with solid-state NMR spectroscopy [6]. A wide variety of applications in this area can be envisioned in which a determination of the nature of substances resolves some question. An analytical application of the solid-state NMR technology derived from the sensitivity of the chemical shielding to local structure is the study of polymorphism in solids. Although two solid samples may be chemically identical, i.e. have the same atoms bonded in the same pattern, the constraints of packing in the solid state may produce structures that are different. If these structures result in different physical properties such as dissolution or accessibility, the differences may influence uses of the solid material. The ability to distinguish different polymorphic structures simply is an aid in formulation chemistry that has wide applicability. An example of this effect is seen in the differences in the spectra of solid 2,6-ditert-butylnaphthalene prepared by recrystallization from ethanol or acetone [7]. The carbon NMR spectra of the aromatic region of these materials clearly show that the two solids are distinct from each other. One area where the identification of polymorphic and pseudopolymorphic structures with NMR spectroscopy is tremendously effective is the analysis of pharmaceuticals [8]. For example, carbon NMR spectroscopy has been used effectively to identify several polymorphic structures of neotame and to specify the conditions under which each may be made. The fundamental principle behind interpretation of NMR spectra, be they of solids or solutions, is that there is an intimate connection between the local electronic structure and the resonance frequency. Early on, chemists correlated a wide variety of solution-state NMR isotropic chemical shift measurements in empirical rules, such as those given by Grant and Paul [9]. The effects of perturbation of the electronic structure (and thus the chemical shift) by appending groups to a center was given by a rationalized set of effect parameters. With these rules, one could predict with reasonable accuracy the isotropic positions of carbons in different chemical environments. Such rules work reasonably well for analysis of isotropic carbon chemical shifts in solid materials, but there can be discrepancies between solution and solid isotropic values due to differences in local environment, when comparing the solid to a solution of the same material [10]. Thus, using empirical rules based on solution-state NMR data to interpret solid-state data must be done with great care. A major consideration in determining isotropic chemical shifts in solid-state NMR spectroscopy is referencing. In the solution state, introduction of the reference material into the same solution as the material being analyzed allows a compensation for susceptibility shifts. In the solid state, one almost always uses external referencing to specify chemical shifts. As a result, there is an inherent

limit to accuracy of measurement of chemical shifts in the solid state, even though the measurement precision may be higher. There are two means to reference externally: (1) by mixing a small amount of the reference material (as a solid) with the material to be studied and (2) by substitution of a sample of the reference to calibrate the spectrometer, with the assumption that substitution changes none of the parameters of the spectrometer system. By far the most common method is the second method, and practitioners should realize that this method, in particular, does not account for susceptibility differences between the material and the reference, a potential source of systematic error in reporting chemical shifts determined with this method. In addition, for either method, one usually uses a secondary reference standard to make chemicalshift determinations. The two most commonly used reference standards for solid-state 13 C NMR spectroscopy are adamantane and hexamethylbenzene. According to Duncan’s compilation [11], the positions of the resonances for adamantane are 28.7 and 38.2 ppm relative to the position of an external tetramethylsilane (TMS) sample, and the aromatic resonance of hexamethylbenzene falls at 132.3 ppm relative to external TMS.

Uses of Shielding Tensors to Identify Materials While the isotropic chemical shielding is a principal parameter for specifying the identity of an unknown material, one of the advantages of analysis of materials in the solid state is the opportunity to observe all the tensor components of chemical shielding. Such additional information may provide a means to distinguish resonances that have identical, or nearly identical, isotropic chemical shieldings. The initial report of proton-enhanced carbon NMR spectroscopy, for example, showed the systematic dependence of the carbon chemical shielding elements of a carbonyl carbon, as exhibited in the powder pattern, on the nature of groups bound to the carbonyl center [12]. A flurry of activity to evaluate chemical-shielding tensors for simple situations resulted in reports of chemicalshielding tensors for carbons, protons, and other nuclei [11]. A particularly interesting example of the use of chemical-shielding tensor analysis to address a problem in solid-state chemistry is a study of the cadmium NMR chemical-shielding tensor elements of CdSO4 ·8H2 O [13]. For this material, there initially existed an apparent anomaly between the structure gleaned from NMR chemical-shielding anisotropy/Cd–O bond distance relationships and the bond distances determined earlier by more conventional scattering methods. The discrepancy was resolved by a redetermination of the structure that showed that the refined distance data were in agreement with the NMR correlations.

Analytical Aspects of Solid-State NMR Spectroscopy

the development of technology for ever-faster spinning, obtaining center-band maps is much easier than it has previously been, and analysis with the isotropic chemical shifts is now a nearly-routine technique for many nuclear species. In early work, Herzfeld and Berger demonstrated that an analysis of the relative intensity distribution among the sidebands of a resonance could be used to recover information on the chemical-shielding tensor elements [19]. The techniques involve a careful evaluation of relative intensities, which are then compared to the results of calculations, often presented as maps of relative intensity vs. certain parameters related to the tensor elements. (This can be done in several ways, but a common means is to compare them directly to theoretical maps of expected intensities of several sidebands to specify the values of parameters, from which the tensor elements are determined by a simple algebraic expression.) With this knowledge, it becomes possible to analyze the spectrum of a reasonably complex material containing multiple resonances to obtain both the isotropic chemical-shifts and the components of the chemical-shielding tensors for each site. In later work, several groups have demonstrated various means to create two-dimensional NMR spectra that correlate the isotropic chemical shift with the anisotropic chemical shielding [20,21]. The result of using these kinds of experiments is the availability of chemical-shielding tensor elements for nuclei (particularly carbon) in a wide range of chemical environments.

Using Quadrupolar Coupling to Identify Materials The presence of the quadrupolar coupling for spins with quantum numbers greater than 1/2 adds an extra dimension to the analysis of solid materials that contain these nuclei. The line shape for a quadrupolar nucleus is determined by the electric-field gradient at the nuclear site. Like the chemical shielding, the quadrupole coupling constant characterizes the local electronic state. A vast majority of nuclei in the period table have at least one quadrupolar isotope, so determining the quadrupolar coupling is a generally useful analytical tool for identifying chemical type in a host of different situations. The borosilicate glasses are representative of a system in which the study of the quadrupolar coupling can give information on local structure [22,23]. In inorganic systems, such as the Keggin ions, analysis of quadrupolar couplings can be used to analyze the kinds of sites quadrupolar nuclei like vanadium occupy [24]. Zeolites have been extensively studied with quadrupolar NMR spectroscopy, especially the aluminum centers. The studies of the quadrupolar coupling give information on site symmetry and structure [25]. Changes in the quadrupole coupling of aluminum in certain zeolites

Part I

A problem arises with analysis of chemical-shielding tensor elements of a complex material such as a multicarbon organic material: the powder patterns (from which the tensor elements are obtained) of carbons in samples examined without MAS generally overlap. Thus, analysis of powder patterns is generally restricted to materials in which there exist a limited number of unique chemical sites [14]. It is sometimes possible to use techniques like factor analysis to extract the tensor components in overlapping spectra [15]. In samples that contain a limited number of sites, there may be problems with the analysis, if the dispersion of the line is too large. Distortions in the powder-pattern line shape resulting from inadequate or non-uniform excitation, inherent limitations of bandwidth, or the effects of apodization may limit analysis of the chemical-shielding tensor elements directly from the spectrum. There are at least three ways to overcome this problem: (1) determination of a transfer function that predicts the nature of distortions and that is used as a parameter in fitting distorted spectra [14]; (2) determination of the spectrum is a point-by-point fashion [16]; or (3) by determining sections of the resonance line in single experiments, and then assembling the entire powder pattern from the overlap of these sections. Each has advantages and disadvantages, and the method to use depends on the spectral features. For example, to obtain spectra of platinum particles the resonant absorption of which was spread over tenths of a percent of the mean resonance frequency, it was necessary in studies of catalysts to use the second method to obtain a representation of the spectrum [17]. The use of the first method has provided chemical-shielding tensor elements for a variety of simple 207 Pb-containing materials [18], and the third method has been used to examine chemical shielding in lead-based oxides [16]. The dispersion of each resonance in the spectrum of a powdered solid material substantially lowers the apparent resolution of the spectrum, as compared to the spectrum of the material dissolved in solution. This loss of resolution was the impetus for the use of MAS as a means to simplify the 13 C spectroscopy of solids, mentioned above (MAS spectroscopy represents for many the quintessential solid-state NMR technique). Because spinning of this sort is a coherent modulation of the chemical-shielding interaction, the spectroscopic band is split into a center band and a series of sidebands, separated from the center band by an integral multiple of the spinning frequency. The MAS technique improves the resolution to the point that one may resolve the positions of the center bands, provided one may identify them. To have the spectra be essentially the maps of center bands (and therefore like solution-state spectra) requires one to spin the sample at speeds such that the first sideband is well outside the bandwidth of the dispersion. Since this may be practically difficult, most solid-state spectra exhibit some sidebands. With

Using Quadrupolar Coupling to Identify Materials 391

392 Part I

Chemistry

Part I

used as catalysts for reactions such as the conversion of methanol to higher alkyl structures give clues to structural changes in catalyst accompanying adsorption of the substrate [26].

Structure Determination An analytical measurement of great importance is the determination of physical structure. As seen above, the chemical connectivity can be inferred from the isotropic chemical shielding through its known correlation with chemical group identity. Physical structure is equally important in specifying the nature of materials, e.g. synthetic polymers or biological systems. The classical means to identify physical structure is through diffraction methods with X-rays or neutrons. These methods, however, require single crystals to extract maximal information. Solid-state NMR spectroscopy can be used to define physical structure to a degree, without the creation of single-crystal samples. The structural information is obtained by measurement of dipole–dipole coupling between two nuclear spins, the magnitude of which depends only on the distance between the spins. One may divide the sorts of dipolar-coupling measurements into two types: (1) homonuclear dipolar coupling experiments and (2) heteronuclear dipolar coupling experiments. Depending on how the experiment is run, these are sometimes called recoupling experiments [27]. As an example, selective enrichment of nuclei at points in amyloid fibrils allows the measurement of distances in these systems, an important piece of information in understanding the structures implicated in Alzheimer’s disease [28,29]. Combining several dipolar measurements with restrictions from chemical shielding limits the possible structures that, for example, a protein can adopt [30]. The study of structures of many systems, e.g. catalysts [31], can be addressed by these kinds of measurements. One may use the dipole-measurement techniques to determine distance constraints in technologically important systems such as polymer electrolytes [32] or to determine hydrogen-bond distances with great precision [33]. There are ways to address structure in solids that do not directly involve the dipolar coupling to determine distances. For example, knowledge of the orientation of the chemical-shielding tensor in the local molecular frame may be used to determine the orientation in partially oriented samples, as was demonstrated for uniaxial deformation of poly(tetrafluoroethylene) [34]. The orientation distribution function can be discerned for partially ordered samples from other NMR parameters such as the deuterium quadrupolar line shape [35]. By correlating chemical shielding as a function of orientation in a field, one may study biaxial orientation, as has been shown for the carbon resonances of poly(ethylene terephthalate)

[36]. Such experiments show the strength of multidimensional techniques in the NMR of solids, in this case to determine orientation quantitatively. The identification of relations between parameters such as the quadrupolar coupling constant and local structure allows one to infer structural features from measurements of these parameters. An example of this sort of relationship is the connection between the oxygen quadrupolar parameters and the Si–O–Si bond angles in silicates [37]. Another example that shows the utility of measurements of quadrupolar coupling constants involves determining the state of ionic materials, for which the electric-field gradient is a strong indicator of structure [38].

Quantification with Solid-State NMR Spectroscopy A principal concern of the analyst is determining the amounts of identifiable species in a sample, about which not everything is known. The resolution of a question very often involves measuring with great precision and accuracy tiny amounts of material in a sample. Anyone involved in analytical chemistry receives requests for such determinations and appreciates the difficulty of achieving accuracy and precision at very low levels. For example, the identification and quantification of species in samples of environmental importance is of crucial importance in determining the nature of contamination. NMR spectroscopy, like other forms of spectroscopy, can—in principle—provide a means to determine the amounts of various species in a sample by careful intensity measurements. However, there are caveats that accompany that statement. It is important to remember that the NMR experiment gives the intensities of the various unique nuclear magnetizations under the conditions of preparation in the experiment. These magnetization intensities may not be proportional to the number of spins in each environment in every experiment. A simple and obvious example of quantification with NMR spectroscopy is the measurement of relative intensities in solution-state NMR spectroscopy, a principal means of specification of chemical structure. As most NMR spectroscopists know, one must account for effects, such as incomplete relaxation or nuclear overhauser enhancements, to make such experiments relatively quantitative. Even with these precautions, the measurements provide information on the relative amounts of material not absolute amounts. Inclusion of a material of known concentration may be used to determine concentrations of unknown materials absolutely, through ratio to the known material. The situation in solid-state NMR spectroscopy is more complex than that in solution-state NMR spectroscopy.

Analytical Aspects of Solid-State NMR Spectroscopy

Uniform excitation of transitions is a problem when observing quadrupolar nuclei with large coupling constants. In such cases, the spectroscopy often yields only one transition not the full spectrum. Comparison of this spectrum to that from a material with a smaller coupling constant may skew the determination of the amount of material. The multiple-quantum methods for studying quadrupolar nuclei have been promoted in recent years. The study on how to relate signal intensity to concentrations is ongoing [44,45]. The usual NMR determination of relative numbers of spins in a material is not an absolute measurement. To determine the absolute numbers of spins in a sample, one must compare the intensities to a known amount of a standard material. In solution-state NMR, the standard can be frequently added to the solution, which allows easy comparison. For a solid-state measurement, one may add the material as a component of a physical mixture in the sample region, in analogy to the solution-state process. Comparison of intensities, taking into account the ramification discussed above, allows an absolute measurement of number of spins. However, care must be exercised to ensure uniform excitation of all parts of the sample, including the reference.

Summary NMR spectroscopy is the preeminent technique for determination of many material properties. This was obvious even in the early days of solution-state NMR, a fact that resulted in its relatively quick adoption by chemists. The application of NMR spectrocopy to solids has led to a similar utility for a wider range of solid materials. The analysis may involve identification of species, determination of structure or “sizes” of various interactions, and relative or absolute quantification of species. To analyze a material properly requires a careful consideration of all factors that affect the spectroscopy.

Acknowledgment The support of the US National Science Foundation through Grant # CHE-0411790 is acknowledged.

References 1. Melchior MT, Vaughan DEW, Jacobson AJ. J. Am. Chem. Soc. 1982;104:4859. 2. Fyfe CA, Gobbi GC, Murphy WJ, Ozubko RS, Slack DA. J. Am. Chem. Soc. 1984;106:4435–8. 3. Engelhardt G, Lohse U, Lippmaa E, Tarmak M, Maegi M. Z. Anorg. Allg. Chem. 1981;482:49. 4. Schaefer J, Stejskal EO, Buchdahl R. Macromolecules. 1977;10:384.

Part I

Many techniques used in solid-state NMR spectroscopy create magnetization by complex manipulation of spin interactions that do not necessarily affect all spins in the same manner. The quintessential example of this effect is the cross-polarization technique for creating magnetization in rare spins. Because of the different kinetics of polarization transfer, the magnetizations created for various centers may not be in the ratio of the number of nuclei at those centers. In certain cases, one may model NMR processes, from which one produces a relative quantitation of spins [39]. In other cases that are particularly important for organic materials, the magnetization development may be very complex and not fit by simple models, leading to the situation in which it is difficult to quantify rare spins directly [40]. Certain sequences that simplify or enhance spectroscopic features do so by canceling portions of the magnetization (e.g. the part in sidebands in TOSS sequence). Comparison of magnetization amplitudes in this sort of experiment can never be guaranteed to reflect the ratio of numbers of spins in various environments, without invoking some assumption about the nature of the chemical shielding at various sites [41]. Thus, it is often necessary to avoid cross-polarization and other spectroscopicenhancement techniques to ensure that the signal intensity ratios represent ratios of numbers of spins. So, for example, in measurements to detect 13 C in organic solids, it is preferred to measure intensities in spectra obtained with direct excitation, rather than with cross-polarization, to avoid these complications. However, this may be difficult or impossible because of the 13 C long relaxation times for pure crystalline solids. The measurement of relative numbers of spins in various environments may be hampered by certain interactions that make some spins “invisible” in NMR spectroscopy. Such is the case for materials like coal that contain paramagnetic centers [42]. In a complex, heterogeneous substance such as coal, if the paramagnetic centers are clustered in one phase, the NMR spectrum may not adequately represent that phase relative to another, resulting in incorrect quantitation. There has been a long tradition of using NMR spectroscopy of abundant spins to quantify materials such as polymers but often at low resolution. Because of spin diffusion, differential relaxation of these abundant spins is not necessarily a problem for this spin system. Thus, relative intensities may be used to determine the relative numbers of spins in various environments. For example, it has been shown that multiplepulse NMR experiments may be used to infer the relative numbers of spins in the amorphous regions of poly(ethylene) [43]. Even in that relatively simple case, it was found necessary to extrapolate intensity ratios on the duty factor of the experiment because that affected the relative intensities of the crystalline and amorphous phases.

References 393

394 Part I

Chemistry

Part I

5. Sherriff BL, Tisdale MA, Sayer BG, Schwarz HP, Knyf M. Archaeometry. 1995;37:95. 6. Bardet M, Foray MF, Maron S, Gonclaves P, Tran QK. Carbohydr. Polym. 2004;57:419. 7. Beckmann PA, Burbank KS, Clemo KM, Slonaker EN, Averill K, Dybowski C, Figueroa JS, Glatfelter A, Koch S, LiableSands LM, Rheingold AL. J. Chem. Phys. 2000;113:1958. 8. Padden BE, Zell MT, Dong Z, Schroeder SA, Grant DJW, Munson EJ. Anal. Chem. 1999;71:3325. 9. Grant D, Paul E. J. Am. Chem. Soc. 1964;86:2984. 10. VanderHart DL. J. Chem. Phys. 1976;64:830. 11. Duncan TM. A Compilation of Chemical Shift Anisotropies. The Farragut Press: Chicago, 1990. 12. Pines A, Gibby M, Waugh J. J. Chem. Phys. 1973;59:569. 13. Murphy PD, Gerstein BC. J. Am. Chem. Soc. 1981;103:3282. 14. For example, see Neue G, Smith ML, Hepp MA, Perry DL, Dybowski C. Sol. State Nucl. Magn. Reson. 1996;6:241. 15. Kormos D, Waugh J. Anal. Chem. 1983;55:633. 16. Shore J, Zhao P, Prasad S, Huang J, Fitzgerald J. J. Phys. Chem. B. 1999;103:10617. 17. Wang P, Ansermet J, Rudaz S, Sinfelt J, Slichter C. Science. 1986;234:35. 18. For example, see Dybowski C, Neue G. Progr. Nucl. Magn. Reson. Spectrosc. 2002;41:153. 19. Herzfeld J, Berger A. J. Chem. Phys. 1980;73:6021. 20. Maciel G, Bax A, Severenyi N. J. Magn. Reson. 1983;52:147. 21. Alderman D, McGeorge G, Hu J, Pugmire R, Grant D. Mol. Phys. 1998;95:1113. 22. Hansen MR, Madsen GKH, Jakobsen HJ, Skibsted J. J. Phys. Chem. A. 2005;109:1989. 23. Schramm S, Oldfield E. J. Chem. Soc. Chem. Commun. 1982;980. 24. Huang W, Louis LT, Yap GPA, Beer R, Francesconi LC, Polenova T. J. Am. Chem. Soc. 2004;126:11564. 25. Masierak W, Emmler T, Buntkowsky G, Gutsze A. Z. Phys. Chem. 2003;217:1613.

26. Seiler M, Wang W, Hunger M. J. Phys. Chem. B. 2001;105:8143. 27. Schnell I. Progr. Nucl. Magn. Reson. Spectrosc. 2004;45:145. 28. Benzinger TL, Gregory DM, Burkoth T, Miller-Auer H, Lynn DG, Botto RE, Meredith SC. Proc. Natl. Acad. Sci. U. S. A. 1998;95:13407. 29. Balbach J, Ishii Y, Antzutkin ON, Leapman RD, Rizzo NW, Dyda F, Reed J, Tycko R. Biochemistry. 2000;39:13748. 30. Bower PV, Oyler N, Mehta MA, Long JR, Stayton PS, Drobny GP. J. Am. Chem. Soc. 1999;121:8373. 31. Kenaston NP, Bell AT, Reimer JA. J. Phys. Chem. 1994;98:894. 32. Reichert D, Pascui O, Judeinstein P, Gullion T. Chem. Phys. Lett. 2005;402:43. 33. Goward G, Schnell I, Brown SP, Spiess H-W, Kim H-D, Ishida H. Magn. Reson. Chem. 2001;39:S5. 34. Brandolini AJ, Dybowski C. J. Polym. Sci. Polym. Lett. Ed. 1983;21:423. 35. Spiess HW. Pure Appl. Chem. 1985;57:1617. 36. Henrichs PM. Macromolecules. 1987;20:2099. 37. Clark TM, Grandinetti PJ. J. Phys. Condens. Matter 2003;15:S2387. 38. Bureau B, Silly G, Buzare JY, Boulard B, Legein C. J. Phys. Condens. Matter. 2000;12:5775. 39. Maciel GE, Sindorf DW. J. Am. Chem. Soc. 1980;102:7607. 40. Smith JM, Dybowski C, Bai S. Sol. State Nucl. Magn. Reson. 2005;27:149. 41. Duer M. Introduction to Solid-State NMR Spectroscopy. Blackwell: Oxford, 2004. 42. Wind RA, Maciel GE, Botto RE. Adv. Chem. 1993;229:3. 43. Pembleton RG, Wilson RC, Gerstein BC. J. Chem. Phys. 1977;66:5133. 44. Ding S, McDowell CA. Chem. Phys. Lett. 1999;307:215. 45. Gu J, Power WP. Sol. State Nucl. Magn. Reson. 2005;27:192.

395

NMR and Its Application John R. Jones1 and Shui-Yu Lu2

1 Chemistry,

School of Biomedical and Molecular Sciences, University of Surrey, Guildford, Surrey GU2 7XH, UK; 2 Molecular Imaging Branch, National Institute of Mental Health, National Institutes of Health, 10 Center Drive, MSC 1003, Bethesda, MD 20892-1003, USA

Introduction Radiochemistry (tritium chemistry in particular) and nuclear magnetic resonance (NMR) spectroscopy are hardly ever taught within the same undergraduate degree course. This is one of the main reasons why those who become NMR spectroscopists are so reluctant to see radioactive material being used in their instruments. The main thrust for the development of 3 H NMR spectroscopy [1] has therefore come from the radiochemistry area and from those in the pharmaceutical and life sciences who appreciate the potential benefits of working with this radionuclide.

Radiochemical Facilities and Radiation Safety Ideally, one should have access to two laboratories, one where high levels, i.e. millicurie (mCi, 1 mCi = 37 MBq) quantities, and higher amounts of tritium, can be handled, and the other, less specialized, where the work is confined to the tracer level (mCi down to µCi). The Curie (Ci, 3.7 × 1010 disintegration per second) is the “old” unit of radioactivity and represents a very large amount of radioactivity, hence the frequent use of millicurie and microcurie quantities. On the other hand, the “new” SI unit, the Becquerel (Bq, 1 disintegration per second), is an extremely small amount of radioactivity so that e.g. megabecquerels (MBq), are frequently encountered. In addition, there should be a separate counting room where the scintillation spectrometer(s) are kept. The synthesis and handling of tritiated compounds should all be done in fume cupboards of the necessary specification. Operations over spill trays ensure that any contamination is limited to a specific area while regular monitoring provides the necessary reassurance. Much preliminary labeling work can be performed using the stable deuterium isotope. Where compounds at very high specific activity e.g. 20 Ci/mmol or more are required, it is necessary to obtain a supply of T2 gas but rather than use a glass vacuum line, it is better to purchase a commercially available instrument in which the tritium is stored on a uranium bed—on warming the latter sufficient T2 Graham A. Webb (ed.), Modern Magnetic Resonance, 395–398.
C 2008 Springer.

gas can be released for the proposed experiment and on completion any unused tritium can be taken up by another uranium bed specially kept for this purpose. In this way, all the tritium can be easily accounted for. Purification of tritiated compounds relies heavily on one or more radiochromatographic methods of which radio-HPLC is the most widely used.

Tritiation Procedures For most, but not all, applications, it is necessary to introduce the tritium at specific sites and for this reason the following reactions are chosen [2,3]: (a) Catalytic hydrogenation CHT CH2T

T2 Pd/C

(b) Catalytic aromatic dehalogenation (usually debromination) H3C

T2 Br

H3 C

Pd/C

T

(c) Methylation using 3 H-methyl iodide CT3I NaH

N H

N CT3

(d) Sodium [3 H]borohydride reduction O

HO H

NaBT4

H T

Part I

3H

396 Part I

Chemistry

Part I

In this way compounds of very high specific activity can be prepared e.g. the introduction of two tritium atoms to produce ethyl benzene can give a product with a maximum specific activity of 58 Ci/mmol. In addition to the above-mentioned reactions, there are a number of others, chief among them being hydrogen isotope exchange reactions that can lead to either specific or generally labeled compounds but at a lower specific activity. The reason for this is that tritiated water is now the source of the tritium and for health and safety reasons it is usually used at the Ci/cm3 level. Some of these reactions are slow, requiring many hours to come to completion. Others such as the catalytic aromatic debromination and borohydride reduction are isotopically inefficient, leading to the production of large amounts of radioactive waste (50% in the first example and 75% in the second). Consequently, there has been much interest of late in the development of new, microwave-enhanced procedures [4–6] that proceed much more rapidly (matter of minutes), more efficiently, and with the production of much reduced levels of radioactive waste. In some cases, it is no longer necessary to use a solvent while in others an ionic liquid can be used to replace the more conventional solvent. A good appreciation of how organic compounds interact with microwave irradiation is necessary in order to maximize the benefits of these new procedures [7,8].

Tritium NMR Spectroscopy Now that a large number of tritiated compounds can be rapidly produced via microwave-enhanced procedures, there is a corresponding need for a fast and sensitive analytical method for determining the pattern of labeling. Tritium possesses ideal NMR properties (Table 1)—it has a nuclear spin of 1/2 and is the most sensitive of all NMRactive nuclei (21% better than 1 H). Unfortunately, NMR spectroscopy by comparison with other analytical methods e.g. mass spectrometry is not a very sensitive method although considerable improvements have been made in recent years through the design of higher performance magnetic fields. This is not an inexpensive exercise and it

is fortunate that the most recent improvement, through the development of cryoprobes [9], is much more cost effective although the challenge of keeping the tritiated sample at or close to room temperature while the radiofrequency coils nearby are cooled to below 35 K was a formidable one. In early studies, a 10 mCi sample of a tritiated compound gave a satisfactory 3 H NMR spectrum when using a spectrometer operating at 64 MHz. A more recent example—a spectrometer operating at 533.5 MHz with a cryoprobe accessory—gave a 3 H NMR spectrum with as little as 11 µCi (S/N ratio of 21, Figure 1A). In both cases, the accumulation time was overnight. This 1000-fold improvement in sensitivity still leaves one with much higher levels of radioactivity than are used in liquid scintillation counting. Fortunately, the natural abundance of tritium (500 mT/m), fast switching (100–200 µs) gradient systems are necessary to apply high-speed sequences at high spatial resolution. However, such gradient switching can cause heating, and so an efficient cooling system is required to prevent temperature changes being transferred to the animal, which would result in altered cardiac function. Today’s clinical MR scanners commonly have static magnetic field (B0 ) strengths of 1.5–3 T, whereas animal systems range between 4.7 and 17.6 T. The signal-tonoise ratio (SNR) of an MR experiment improves with increasing B0 . Ultra-high-field magnets are important for cardiac experiments in rodents as the gain in SNR allows

for achieving the necessary spatial and temporal resolution (the SNR decreases with increasing resolution). Additionally, dedicated radio-frequency (RF) coils that are optimized in geometry and loading for a particular animal size, are required to obtain maximal SNR. Volume coils (i.e. birdcage coils [9]) are commonly used for mice as they provide an excellent homogeneity of the RF-field. They can also be applied in “quadrature mode”, resulting in an additional increase in SNR of a factor up √ to 2 [10]. A picture of a birdcage coil used for cardiac MR in mice at 11.7 T is shown in Figure 1. In rats, a combination of body coil for transmit and surface-coil for receive are typically used [11,12]. A similar configuration would provide more efficient mouse imaging, but the implementation is very difficult owing to the small size. Next to hardware and MR methods, careful consideration must be given to maintaining stable animal physiology throughout an experiment. Dedicated animal cradles, optimized in diameter and length, are required. An example illustrating such a set-up is shown in Figure 2. They typically comprise a nose cone for delivery of anaesthetic gases, a scavenging line for anaesthetic gas recovery, a temperature control system—consisting of a heating blanket and a thermocouple—for maintaining a constant body temperature, and ECG and respiratory motion sensing capabilities for physiological gating. RF filters may be useful to eliminate contamination of the MR signal by external RF noise pick-up. Additional lines are required if the animals are to be ventilated artificially, and if drugs or MR contrast agents need to be administered. The animals have to be secured within the cradle using surgical tape. Special care should be taken not to distort or compress their abdominal or chest regions. The heart rate of the animal is influenced by the body temperature and by the depth of anaesthesia. A core body temperature of 37 ◦ C has to be maintained by supporting the temperature regulation of the animal in order to ensure physiologically normal conditions. This can be achieved using special blankets that are heated by warm air or water (air has the advantage over water that it is MR invisible and it does not interfere with the RF-coil). Furthermore, the lowest possible anaesthetic level should be applied in order to minimize depression of cardiac function by anaesthesia [13]. To achieve this, anaesthetic gases are commonly used for cardiac studies as the dose can be easily titrated for the individual animal whilst in the magnet. In particular, Isoflurane causes the least cardiac depression and represents the anesthetic of choice for this purpose [14]. Due to suppression of the eye-closure reflex during anesthesia, ointment must be applied to the eyes in order to keep them moist. The recording of the ECG and the respiratory signal is not only necessary to monitor the animal inside the magnet (where visual assessment is not possible), but also to minimize the influence of cardiac and respiratory motion on the MR experiment. It is well recognized that motion artifacts

CVMR in Rodents

Methods and Requirements 837

become more pronounced with increasing magnetic field strength [15]. Figure 3 illustrates this influence on cardiac MR imaging in mice at 11.7 T. The data shown in this figure were acquired under various gating strategies. No gating

Front-paw-mounted needle electrodes, inserted subcutaneously into the front limbs of the animal, or surface electrodes are used to derive the ECG. A good coupling between the electrodes and the animal is crucial to obtain Conventional gating

Cardiac gating

Gating with SS

a

b

c

d

a’

b’

c’

d’

Fig. 3. Motional influence on cardiac imaging in mice at ultra-high magnetic fields. Transverse gradient echo images through the heart of a normal mouse acquired at 11.7 T. Both rows are identical with the image intensity in the bottom row (a –d ) increased to a maximum level in order to reveal low-level artifacts. Each column corresponds to a different gating strategy (from left to right): (a, a ) No gating, (b, b ) cardiac gating, (c, c ) respiratory gating without and (d, d ) with steady-state maintenance (SS) during respiration [16]. Each row is scaled to the same range of image intensity. The vertical signal voids present in the images are due to saturation effects from adjacent oblique slices that were acquired in the same experiment. Note the low-level artifacts visible in panel c . They are caused by interrupting the steady state during respiration and their strength depends on the MR sequence and parameters used in an experiment. Scale bars: 2 mm.

Part I

Fig. 2. Illustration of a set-up used for CMR in rodents. The shown set-up consists of the animal cradle, which is equipped with a nose cone, needle electrodes for deriving the ECG and a conductor loop for detecting respiratory motion. The animals are placed onto a heating blanket to maintain a constant body temperature of 37 ◦ C throughout the MR experiment. (See also Plate 70 on page XLVII in the Color Plate Section.)

838 Part I

Biological Sciences

Part I

robust ECG-traces that can be triggered from. Respiration can be monitored using pressure pads or conductor loops mounted on top of the chest and the abdomen of the animal as shown in Figure 2. Adequate measures have to be provided to suppress any interference due to gradient switching with the ECG and respiratory signals. An electronic gating device serves as an interface between the animal, the user and the MR-system and allows for synchronizing the experiments to the cardiac cycle and for interrupting the data acquisition during respiration (i.e. respiratory gating). This device is offered by various manufactures or

can be home built. Care has to be taken when disrupting the steady state of the spin system during respiration, because this can be the source of another—non-motion related—image artifact (see Figure 3 c ). Cassidy et al. demonstrated a simple way of maintaining steady state of the spin system during respiration: the MR sequence was continued throughout respiration with the same timing as determined by the heart beat but without acquiring data. The decision to acquire data or to maintain steady state is made during run-time without any additional user-input being required [16].

Fig. 4. Diagram of an MRI cine sequence used in rodents. Fast, spoiled gradient echo sequences are typically used for cardiac imaging in rodents. After the detection of the R-wave in the ECG, the same k-space line is acquired repeatedly with a constant value for the phase encoding gradient. The number of frames N per cardiac cycle depends on the sequence timing and the heart rate of the animal and ranges between 15–30 frames. The illustrated scheme is repeated in the next cardiac cycle with a different value for the phase encoding gradient. Thus, the product of number of phase encoding steps times number of averages cardiac cycles are required in total to obtain a full cine data set for one slice. If respiratory gating is employed, the scheme is interrupted during respiration and the imaging time is prolonged.

CVMR in Rodents

Global Cardiac Function 839

Part I

Global Cardiac Function The application of high-resolution MRI in multiframe mode (cine imaging, cine-MRI) allows for non-invasive quantification of left-ventricular mass and volumes in mice and rats. Fast, 2D spoiled gradient echo type sequences are commonly applied continuously throughout the cardiac cycle and provide high contrast between blood and myocardium. Refocused steady-state free-precession sequences, as frequently used for cardiac MRI on human scanners, are more difficult to employ at ultra-high magnetic fields due to their sensitivity to susceptibility differences. Figure 4 shows a schematic depiction of a cine experiment in rodents. Repetition times of less than 5 ms per frame freeze cardiac motion and result in 15– 30 frames per RR interval (depending on the heart rate). The echo times are B0 -field dependent and chosen such that lipid- and water-protons have an opposite phase to enhance the contrast between different tissue types [17]. The values range between 1 and 2 ms. The flip-angle of the sequence needs to be adjusted according to the chosen repetition time and respective relaxation times in order to maximize the contrast between blood and myocardium. The in-plane image resolution (before image interpolation) in mice is typically around 100–200 µm at a slice thickness of 1 mm, and in rats about 200–300 µm at a slice thickness of 1–1.5 mm. Seven to ten slices are required to cover the entire mouse heart from base to apex. Accordingly, 10–20 slices are needed to image the entire rat heart. Studies at magnetic field strengths up to 7 T commonly use cardiac gating only, whereas additional respiratory gating is essential at higher magnetic field strengths to obtain virtually artifact-free, high-quality images, as we have demonstrated quantitatively [18]. A cine study of the entire mouse heart can currently be accomplished in well under one hour, and in rats within approximately 90 mins—depending on the required spatial resolution, and whether or not respiratory gating has to be applied or not. Figure 5 shows examples of end-diastolic and endsystolic frames in short-axis and long-axis orientations of a normal mouse heart, acquired at 11.7 T. The enddiastolic frame is characterized as the one with maximal left-ventricular volume, and the end-systolic frame the one with minimal left ventricular volume, respectively. The mainly stationary tissue (relative to the imaging slice), such as cardiac or skeletal muscle, is saturated by the repeatedly applied RF-pulses and subsequently appears dark. The blood provides high signal in these images due to the inflow effect of blood into the imaging slice (brightblood images). It has to be noted that the contrast can be inverted if the imaging sequence is combined with dedicated black-blood techniques [19,20]. However, this approach usually provides a reduced temporal resolution throughout the cardiac cycle and requires longer acquisition times

Fig. 5. Cine-images of normal mouse heart. Mid-ventricular end-diastolic images through a normal mouse heart in the (a) short-axis orientation, (b) the four-chamber and (c) the twochamber long-axis orientation; both long-axis views are orthogonal to the short-axis orientation. The primed panels correspond to the respective end-systolic frames. Abbreviations: lvc, rvc— left/right ventricular cavity; lvw, rvw—left/right ventricular wall; pm—papillary muscle; lu—lungs; la, ra—left/right atrium; ao— aorta; pa, pv—pulmonary artery/vein; mv—mitral valve. Scale bars: 2 mm.

to gain sufficient signal-to-noise. The left ventricle has a characteristic doughnut shape in the short-axis view (Figure 5a, a ), which is orientated orthogonally to both longaxis views: the four-chamber view (Figure 5b, b ) and the two-chamber view (Figure 5c, c ). The papillary muscles do not appear as connected to the ventricular muscle in this end-diastolic frame (Figure 5a) and are seen as dark spots inside the bright ventricular cavity. Manual or semi-automatic segmentation of both the end-diastolic and the end-systolic short-axis frame for every slice allows for a quantitative analysis of left ventricular mass and function in rodents. Figure 6 shows the

840 Part I

Biological Sciences

Part I

a’

a

Fig. 6. Quantitative image analysis. Segmentation of end-diastolic and end-systolic frames shown in Figure 5a, a allows for quantitative measurement of left ventricular mass (white area) and volume (grey area) and subsequently calculating cardiac functional parameters. Note that the papillary muscles are counted towards the ventricular mass and not to the cavity volume. Scale bar: 2 mm.

result of the image segmentation process: the white compartment corresponds to myocardial volume and the grey compartment to the ventricular cavity volume. Ventricular mass is obtained by multiplying the myocardial volume with the density of myocardial tissue (1.05 g/cm3 [21]). Table 1 lists all relevant parameters that can be derived from a cine experiment. The quantitative analysis of the cine images is highly reproducible with low inter- and intra-observer variability that improves with increasing field strength [18]. This technique has been applied in several studies to investigate cardiac function of normal mice [22–25] and rats [11, 26]. It has also been used to study developmental

Table 1: Relevant cardiac functional parameters

Acronym Description

Definition Unit

HR

Heart rate

Beats per minute

ESV EDV EDM SV EF

µl End-systolic volume End-diastolic volume µl End-diastolic mass mg Stroke volume EDV-ESV µl Ejection fraction 100% · % SV/EDV Cardiac output SV · HR ml/min

CO

* Taken from [18].

bpm

Normal Mouse Heart* 429 ± 24 15.7 ± 1.2 43.0 ± 3.9 57.1 ± 4.2 27.4 ± 3.4 63.5 ± 2.9 11.4 ± 1.2

changes in cardiac function and mass from neonatal to adult mice [27]. Applications in transgenic mice have, for example, been shown in a model of cardiac hypertrophy [28], mice with myocardial overexpression of tumor necrosis factor-α [29,30], and adult cardiomyocyte-specific VEGF knockout mice [31]. We have used this technique to investigate the effect of orthostasis in mice and rats. Experimental ultra-highfield MR systems are commonly equipped with a vertical bore magnet for engineering reasons. Although from a physiological point of view this is the preferred design for experiments in isolated perfused organs and on aqueous solutions, animals have to be positioned in an upright position for in vivo studies on such MR systems. It is well recognized that it is impractical to investigate larger mammals and humans in the vertical position, as the effect of orthostasis reduces venous return, LV volumes and cardiac output [32]. We demonstrated that MR systems with a vertical bore can generally be used to measure cardiac function in both, mice and rats, within approximately 1– 1.5 h [18,26]. However, longer experiments may best be done in horizontal position due to detectable changes in volumes, ejection fraction and cardiac output occurring over prolonged experimental periods [33]. Cine imaging can also be used to characterize surgical animal models of human cardiac disease non-invasively. In chronically failing hearts of rodents [34–37], the degree of failure (as indicated by the ejection fraction; hearts with an EF 4.5, p < 0.05, corrected for multiple comparisons. Activation maps are overlaid on a T2 -weighted spin echo anatomical image, with each contiguous slice being 1 mm thick and moving rostrocaudally through the brain from bottom right to top left, with approximate co-ordinates relative to Bregma [108]. (See also Plate 75 on page L in the Color Plate Section.)

also similar with both models. This suggests a significant correlation between the time course of changes in blood quinelorane concentration and the observed BOLD signal changes in activated regions and that there may be a direct link between the level of activity in these brain regions and the amount of drug present within the blood.

Behavioral Modeling Using a model based on the biphasic locomotor effects of quinelorane that takes no account of changes in the behavior of control animals (Figure 4, blue curve), produces SPM maps that are almost entirely devoid of statistically significant changes in BOLD contrast (Figure 5). However, in contrast to the results produced using either

Using Dopamine Receptor Agonists 867

Part I

BLOOD QUINELORANE CONCENTRATION (ng/ml)

Preclinical Pharmacological MRI

7

6

5

4 Blood Quinelorane Concentration

3

Derived Pharmacokinetic Model 2

1

0 0

10

20

30

40

50

60

TIME POST-INJECTION (MINUTES) Fig. 2. Mean blood quinelorane concentration following administration of 30 µg/kg quinelorane (n = 3) (solid squares) and derived pharmacokinetic model after interpolation of blood quinelorane levels (solid line).

the “off/on” or pharmacokinetic models, small bilateral increases in BOLD signal were noted in the entorhinal cortex and posterior piriform cortex and small bilateral decreases in BOLD signal were present within the nucleus accumbens. The lack of statistically significant BOLD signal changes using this model is likely to be due to the biphasic shape of the curve used to model BOLD changes against. For there to be a statistically significant relationship between the model and data, BOLD signal changes would have to first decrease and then increase post-injection. The lack of statistically significant changes observed using this model shows that, with the exception of entorhinal and posterior piriform cortices, most brain regions do not show biphasic changes in BOLD contrast i.e. decreased then increased signal matching the drug-induced changes in locomotor activity at this drug dose. Using a model derived from the locomotor effects of quinelorane after adjustment to account for changes observed in control animals (Figure 4, green curve) produces statistical parametric maps that are similar to those produced using an “off/on” model (Figure 6). Bilateral increases in BOLD contrast are observed in the anterior olfactory nuclei, nucleus accumbens, and caudateputamen, with additional unilateral increases in BOLD signal within some cortical regions. Interestingly, BOLD signal increases within the olfactory nuclei and nucleus

accumbens are more statistically significant (and those changes in the caudate-putamen less significant) when modeled against the adjusted behavioral covariate than when using an “off/on” model. The adjusted behavioral covariate therefore provides a better estimate of the temporal profile of changes in BOLD signal within these regions than does an “off/on” model. This is reflected by a highly significant Pearson correlation coefficient between the adjusted behavioral model and the observed BOLD signal change in the nucleus accumbens (Figure 7), which explains the increase in statistical significance found when using this model when compared to those changes detected using a simple “off/on” model. These contrasting approaches to analysis of phMRI data demonstrate the importance of model selection when analyzing phMRI data. With the exception of the behavioral model that was not adjusted to account for locomotor effects in control animals, all models produced similar SPM maps of significant BOLD signal changes, with small variations in the spatial extent of activations within different brain regions and statistical significance of observed changes. The lack of marked differences in observed patterns of activation when using these different models is due to each model being very similar in terms of having a rapid increase in the measured variable following drug administration that remains elevated for the duration of the experiment. These results suggest that

868 Part I

Biological Sciences

Part I Fig. 3. Group statistical parametric map showing significant changes in BOLD contrast after administration of 30 µg/kg quinelorane (n = 5), analyzed using a covariate derived from quinelorane blood pharmacokinetics (Figure 2). Color scales represent T scores of significant increases (red < yellow) and decreases (blue < green) in BOLD signal, where T > 4.5, p < 0.05, corrected for multiple comparisons. Activation maps are overlaid on a T2 -weighted spin echo anatomical image, with each contiguous slice being 1 mm thick and moving rostrocaudally through the brain from bottom right to top left, with approximate co-ordinates relative to Bregma [108]. (See also Plate 76 on page LI in the Color Plate Section.)

quinelorane activates a range of limbic and basal ganglia regions in a similarly rapid and sustained fashion, with activation occurring almost immediately after drug injection and being sustained for the remainder of the experiment. Using a series of different models in this way can yield

useful additional information, particularly where a drug has an effect on several behavioral measures at different times following administration, and also where pharmacokinetic and pharmacodynamic measures are temporally divergent. Many studies have used intravenous drug

Using Dopamine Receptor Agonists 869

Part I

QUINELRANE-INDUCED LOCOMOTION

Preclinical Pharmacological MRI

90 Quinelorane 30ug/kg locomotion Quinelorane 30ug/kg locomotion interpolated

70

Quinelorane 30ug/kg locomotion interpolated minus saline locomotion interpolated Saline locomotion Saline locomotion interpolated

50

30

10 0

10

20

30

40

50

60

-10

-30

-50

MINUTES POST-INJECTION Fig. 4. Covariates derived from locomotor activity after administration of saline (red curve), 30 µg/kg quinelorane (blue curve), and after subtraction of locomotor activity in control animals from that of treated animals (green curve) in arbitrary units. (See also Plate 77 on page LII in the Color Plate Section.)

injections when producing pharmacokinetic models, with the advantage that the downward slope of the pharmacokinetic curve produced by more rapid drug clearance helps to reduce autocorrelations between the model and other confounding signal changes (such as global effects or scanner signal drift) within a time series. Whilst increasing the statistical power of analysis by allowing multiple periods of drug “stimulation,” intravenous injections are liable to cause additional physiological confounds such as marked changes in systemic blood pressure that may lead to non-neurogenic BOLD signal changes. Repeated drug challenges may, over a short time period, also affect the response to drug as receptors become progressively desensitized, and giving drugs via intravenous rather than other routes can also make comparisons with the behavioral effects of a drug in awake animals more difficult. Nevertheless, it may be desirable to compare the effects of different routes of drug administration where this alters the time course of drug-induced behavioral and BOLD signal changes in order to isolate changes in those regions of the brain that mediate particular aspects of a drug’s behavioral effect.

Accounting for Changes in Physiological Measures As BOLD contrast is dependent on rCBF, factors unrelated to neuronal activity which may cause rCBF to

increase must be controlled as carefully as possible. In the phMRI experiments using quinelorane described here, less than 20% of treated animals showed a small, transient increase in heart rate and blood pressure between 5 and 10 min after administration of quinelorane. This lack of cardiovascular changes, combined with observed BOLD signal changes being localized and not global in nature, make it unlikely that the changes in BOLD signal result from the systemic effects of quinelorane. It is also important to consider the possible effects that confounding changes in global signal intensity, which may result in part from the systemic effects of a drug, may have on the results of a phMRI experiment. For this study, changes in global signal intensity were tested against each of the models used in the data analysis to ensure that there was no statistically significant correlation between them. Having provided this assurance, the global signal was then included within the statistical model used for analysis (ANCOVA).

Confounding Drug Effects on Cerebral Blood Vessels Dopamine is known to play a role in the regulation of cerebral blood flow, as dopaminergic axons innervate intraparenchymal microvessels within the cortex, and dopamine elicits dose-dependent reductions in the diameter of such vessels [98]. Such dopaminergic regulation

870 Part I

Biological Sciences

Part I Fig. 5. Group statistical parametric map showing significant changes in BOLD contrast after administration of 30 µg/kg quinelorane (n = 5), analyzed using a covariate derived from the locomotor effects of 30 µg/kg quinelorane (Figure 4). Color scales represent T scores of significant increases (red < yellow) and decreases (blue < green) in BOLD signal, where T > 4.5, p < 0.05, corrected for multiple comparisons. Activation maps are overlaid on a T2 -weighted spin echo anatomical image, with each contiguous slice being 1 mm thick and moving rostrocaudally through the brain from bottom right to top left, with approximate co-ordinates relative to Bregma [108]. (See also Plate 78 on page LII in the Color Plate Section.)

of local cerebral cortical blood flow raises the possibility that quinelorane may be acting directly at receptors on cerebral blood vessels to elicit BOLD signal increases without a corresponding change in underlying neuronal activity. Both apomorphine and the D1 receptor agonist

SKF-38393 dose-dependently increase the diameter of cortical arterioles in vivo, an effect blocked by the D1 receptor antagonist SCH23390, suggestive of D1 receptor mediated effects [99]. In contrast, the D2 /D3 receptor agonist quinpirole increases vessel diameter only at

Preclinical Pharmacological MRI

Using Dopamine Receptor Agonists 871

Part I

Fig. 6. Group statistical parametric map showing significant changes in BOLD contrast after administration of 30 µg/kg quinelorane (n = 5), analyzed using a covariate derived from the locomotor effects of 30 µg/kg quinelorane accounting for locomotor changes in control animals (Figure 4). Color scales represent T scores of significant increases (red < yellow) and decreases (blue < green) in BOLD signal, where T > 4.5, p < 0.05, corrected for multiple comparisons. Activation maps are overlaid on a T2 -weighted spin echo anatomical image, with each contiguous slice being 1 mm thick and moving rostrocaudally through the brain from bottom right to top left, with approximate co-ordinates relative to Bregma [108]. (See also Plate 79 on page LIII in the Color Plate Section.)

high agonist concentrations, possibly via effects at histamine H2 receptors. Dopamine-mediated vessel constriction may involve activation of α-adrenoceptors and 5-HT receptors, as the effects of dopamine-induced vasoconstriction can be blocked by adrenergic and serotonergic

antagonists like phentolamine and methysergide [99,100]. The effects of dopamine and dopaminergic agonists on cortical blood vessels therefore appear to be primarily mediated by D1 receptors and/or non-dopamine receptors, and as quinelorane has extremely low affinities for

872 Part I

Biological Sciences

Part I

140 120 100 80 Adjusted Locomotion 60 Quinelorane Pharmacokinetics 40 BOLD signal in nucleus accumbens

20 0 0

5

10

15

20

25

30

35

40

45

50

55

Fig. 7. Comparison of covariates used in phMRI analysis representing locomotor activity induced by 30 µg/kg quinelorane after accounting for changes in activity in control animals (blue), quinelorane blood pharmacokinetics (red), and observed BOLD signal change in the nucleus accumbens after grand mean scaling to 100 for comparative purposes. (See also Plate 80 on page LIV in the Color Plate Section.)

D1 and non-dopamine receptors [84], it is unlikely that BOLD signal changes elicited by the low dose of quinelorane used in the experiments presented here result from a direct drug effect on cerebral blood vessels. It might also be expected that dopaminergic innervation of the cerebral vasculature might occur throughout the brain, and that any direct activation of cerebrovascular dopamine receptors would produce more global patterns of change in blood flow rather than the localized BOLD changes observed following quinelorane administration. However, the intimate relationship between the dopaminergic system and the regulation of CBF represents a potentially serious confound for all phMRI studies using dopaminergic agonists and antagonists, and particular care is needed when both planning and interpreting the results of such experiments in order to minimize these confounds.

Explaining the Absence of Signal Changes It is also important to consider why activity in regions of the brain that might be expected to respond to a particular drug appears unchanged. For example, the lack of observed BOLD signal response to quinelorane through more widespread dopaminergic regions such as the substantia nigra and ventral tegmental area may be due to a variety of factors. Firstly, there may be no changes in neuronal signaling within these regions and so no corresponding change in BOLD signal. The lack of quineloraneinduced changes in BOLD signal within the substantia nigra may therefore reflect a lack of alteration in neuronal activity in this area. Alternatively, any changes in activity

within these nuclei may not be of sufficient magnitude to elicit a detectable BOLD response. As BOLD signal is likely to reflect increased neuronal signaling at synapses rather than increased energy utilization per se [11], if there is no net change in the overall level of signaling between neurones, BOLD signal might remain unaffected by alterations in action potentials which have no overall effect on the signaling systems controlling blood flow. If the balance of pre- and post-synaptic activity changes without any net change in synaptic activity, then the spiking output from a particular region might alter without having any effect on the observed BOLD signal. It is also possible that if the drug under investigation has a direct effect on the cerebral vasculature, causing it to either dilate or constrict, this may directly counter any changes in rCBF that might be expected to occur as a result of modulated neuronal activity and so prevent a change in BOLD signal from being observed. When interpreting the results of phMRI experiments, it is therefore important to bear in mind that the absence of a change in BOLD signal cannot necessarily be taken as proof that no change in neuronal activity within a particular brain region has occurred.

Effects of Anesthesia Of particular concern when interpreting phMRI results is the effects of the anesthetic agent used on resulting patterns of brain activity. The effects of α-chloralose on neurotransmission are not well understood, with contradictory reports suggesting that it depresses basal dopamine

Preclinical Pharmacological MRI

Appropriate use of phMRI In summary, there are several points that must be considered and incorporated into every phMRI experiment in

order to have confidence in the final results. When considering whether a particular drug is suitable for investigation using phMRI, the known (or likely) effects that the drug may have directly on cerebral blood vessels, and hence CBF and CBV, need to be taken into account. If anesthesia is used, it must be at the lowest ethically acceptable level in order to minimize the confounding effects on drug-induced neuronal activation. Consideration should also be given to the known effects of different anesthetics on neurotransmitter systems. During the phMRI experiment, it is vital that extensive physiological monitoring is undertaken to provide confidence that any drug-induced patterns of activation are neuronal in origin. To do this, changes in heart rate, blood pressure, or respiration must be shown not to correlate with observed BOLD signal changes. In a similar fashion, changes in global signal should also be tested for correlations with the model of the expected BOLD signal response being used in the analysis. If patterns of activation truly result from druginduced changes in neuronal activity, then it is likely that they will bear some resemblance to either the distribution of receptors to which that drug binds, or to the innervation of neurones within the neurotransmitter system affected by the drug. The less specific the compound being studied in terms of receptor binding profiles, then the more difficult it will be to provide certainty that patterns of activation genuinely reflect changes in neuronal activity. It may also be preferable to use smaller drug doses whenever possible in order to minimize any unwanted systemic effects and activate as selective a population of receptors within the brain as possible.

The Future of phMRI The explosion in phMRI studies over recent years have shown both the potential this technique has for revealing the mechanisms by which drugs act in the brain and also the variety of different experimental situations where it can be employed. The major advantage phMRI has over other techniques is the detailed time course information it provides. PhMRI has shown how drugs with similar mechanisms of action, like cocaine and amphetamine which act on overlapping brain regions and neurotransmitter systems, can differ greatly in their onset and duration of effect. In addition, phMRI has also demonstrated a way in which the route of administration and dose can affect the measured phMRI response and thus neuronal activity. This time course information allows an assessment to be made of the brain circuitry that underlies a given drug response. As the use of phMRI becomes more routine and techniques are increasingly refined, the key advantage that phMRI has in allowing longitudinal assessment of changes in brain functioning over time will become

Part I

levels and dopamine release in response to sensory stimulations in the caudate-putamen and substantia nigra of the cat [64], but has little or no depressive effects on central dopaminergic metabolism and neurotransmission in the rat [101,102]. α-chloralose has also been shown to enhance GABAergic function by increasing the affinity and efficacy of GABA at GABA A receptors [103]. For these reasons, it has been suggested that α-chloralose may not be a suitable anesthetic for phMRI investigations of the dopamine system. However, our results show that α-chloralose is suitable for detecting the effects of D2 /D3 receptor agonist-induced changes in brain activity using BOLD phMRI. As spatial patterns of quineloraneinduced activity closely match D2 /D3 receptor distribution patterns, α-chloralose may produce a state of low basal activity within the dopaminergic system, allowing phMRI to detect the primary site of dopamine receptor agonist action within the brain. This possibility is supported by a phMRI study showing that rCBV changes in the cortex and basal ganglia that follow the administration of either cocaine or amphetamine are preserved under halothane anesthesia, but are considerably diminished or absent under αchloralose anesthesia [104]. This study also reported that following a challenge with the D2 -like receptor antagonist clozapine, rCBV increases are observed throughout the striatum under halothane anesthesia, but are much diminished when α-chloralose is used instead [104]. These data suggest that α-chloralose decreases basal dopamine release, diminishing the actions of drugs that act either by inhibiting dopamine reuptake (e.g. cocaine and amphetamine), or are antagonists like clozapine and produce their functional effects by blocking the actions of endogenous dopamine. As quinelorane is a direct agonist at D2 /D3 receptors, its effects are not dependent on either modulating or blocking endogenous dopamine, and so the actions of quinelorane at dopamine receptors would be unaffected by any α-chloralose-induced modulation of dopamine release. α-chloralose anesthesia has been shown to lower the cerebral metabolic rate of glucose utilization (CMRgluc ) compared to the unanasthetized state, but upon activation induced by somatosensory stimulation, the final level of CMRgluc increases to the same level in both states [105]. Thus a larger incremental increase in CMRgluc occurs during activation in the anesthetized state (in order to reach the same final level of activity), and this larger increase in CMRgluc may, in turn, produce a larger BOLD signal change. In this way, α-chloralose may actually serve to enhance the observed BOLD response to certain compounds.

The Future of phMRI 873

874 Part I

Biological Sciences

Part I

increasingly important. Most clinically used compounds exert their effects only after chronic treatment, and whilst this process is difficult to investigate in detail with current, highly invasive techniques, phMRI is ideally suited to investigating the change in receptor responses that follows repeated drug administration. In addition to assessing acute drug effects in na¨ıve animals, as phMRI becomes more established it will be interesting to examine the effects of drugs in specific animal disease models, as has already begun in models of addiction and drug tolerance [40,41], cerebral ischemia [106], or assessing the effects of drug interactions, like those between dopamine agonists and antagonists [67]. There is also undoubted utility for phMRI within drug discovery in areas where the synthesis of a radioligand for either PET or autoradiography is difficult, or the allosteric action of a particular drug prevents the use of a ligand. To date, most studies have relied either on drugs producing regionally specific modulations in BOLD or rCBV in order to negate the possibility that signal changes may be purely vascular and not neuronal in origin. In future, cross-validating the results of phMRI experiments with those produced by other measures, such as electrophysiology in humans [107], local field potentials [32], microdialysis [52], or metabolic markers will become increasingly important to substantiate the findings of phMRI studies, particularly as phMRI comes to be applied in an increasingly more sophisticated fashion involving the use of compounds with an unknown mechanism of action. Such complex experimental designs will also necessitate new data analysis techniques, particularly those that require no a priori knowledge of the time course over which a drug acts, and methods of studying drug-induced modulation in functional connectivity between different brain regions.

6.

7.

8.

9.

10.

11. 12.

13.

References 14. 1. Leslie RA, James MF. Pharmacological magnetic resonance imaging: a new application for functional MRI. Trends Pharmacol. Sci. 2000;21:314–8. 2. Grasby PM, Friston KJ, Bench CJ, Cowen PJ, Frith CD, Liddle PF, Frackowiak RS, Dolan RJ. The effect of the dopamine agonist, apomorphine, on regional cerebral blood flow in normal volunteers. Psychol. Med. 1993;23:605–12. 3. Lahti AC, Holcomb HH, Medoff DR, Tamminga CA. Ketamine activates psychosis and alters limbic blood flow in schizophrenia Neuroreport. 1995;19:869–72. 4. Breiter HC, Gollub RL, Weisskoff RM, Kennedy DN, Makris N., Berke JD, Goodman JM, Kantor HL, Gastfriend DR, Riorden JP, Mathew RT, Rosen BR, Hyman SE. Acute effects of cocaine on human brain activity and emotion. Neuron. 1997;19:591–611. 5. Stein EA, Pankiewicz J, Harsch HH, Cho JK, Fuller SA, Hoffmann, RG, Hawkins M, Rao SM, Bandettini PA, Bloom

15.

16. 17. 18.

19.

AS. Nicotine-induced limbic cortical activation in the human brain: a functional MRI study. Am. J. Psychiatry. 1998;155:1009–15. Burdett NG, Menon DK, Carpenter TA, Jones JG, Hall LD. Visualisation of changes in regional cerebral blood flow (rCBF) produced by ketamine using long TE gradient-echo sequences: preliminary results. Magn. Reson. Imaging. 1995;13:549–53. Chen YI, Galpern WR, Brownell AL, Matthews RT, Bogdanov M, Isacson O, Keltner JR, Beal MF, Rosen BR, Jenkins BG. Detection of dopaminergic neurotransmitter activity using pharmacologic MRI: correlation with PET, microdialysis and behavioural data. Magn. Reson. Med. 1997;38:389–98. Hagino H, Tabuchi E, Kurachi M, Saitoh O, Sun Y, Kondoh T, Ono T, Torii K. Effects of D2 receptor agonist and antagonist on brain activity in the rat assessed by functional magnetic resonance imaging. Brain Res. 1998;813:367– 73. Chen YI, Brownell AL, Galpern W, Isacson O, Bogdanov M, Beal MF, Livni E, Rosen BR, Jenkins BG. Detection of dopaminergic cell loss and neural transplantation using pharmacological MRI, PET and behavioural assessment. Neuroreport. 1999;10:2881–6. Reese T, Bjelke B, Porszasz R, Baumann D, Bochelen D, Sauter A., Rudin M. Regional brain activation by bicuculline visualized by functional magnetic resonance imaging. Time-resolved assessment of bicuculline-induced changes in local cerebral blood volume using an intravascular contrast agent. NMR Biomed. 2000;13:43–9. Attwell D, Iadecola C. The neural basis of functional imaging signals. Trends Neurosci. 2002;25:621–5. Sokoloff L, Reivich M, Kennedy C, Des Rosiers MH, Patalk CS, Pettigrew KD, Sakurada O, Shionhara M. The [14 C]deoxyglucose technique for measurement of local cerebral glucose utilisation: theory, procedure and normal values in the conscious and anaesthetised albino rat. J. Neurochem. 1977;28:897–916. Fox PT, Raichle ME. Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects. Proc. Natl. Acad. Sci. USA. 1986;83:1140–4. Detre JA, Leigh JS, Williams DS, Koretsky AP. Perfusion imaging. Magn. Reson. Med. 1992;23:37–45. Kim SG. Quantification of relative cerebral blood flow change by flow-sensitive alternating inversion recovery (FAIR) technique: application to functional mapping. Mang. Reson. Med. 1995;34:293–301. Kim SG, Ogawa S. Insights into new techniques for high resolution functional MRI. Curr. Opin. Neurobiol. 2002;12:1–9. Duong TQ, Kim DS, Ugurbil K, Kim SG. Localised cerebral blood flow response at submillimeter columnar resolution. Proc. Natl. Acad. Sci. USA. 2001;98:10904–9. Belliveau JW, Kennedy DN Jr, McKinstry RC, Buchbinder BR, Weisskoff RM, Cohen MS, Vevea JM, Brady TJ, Rosen BR. Functional mapping of the human visual cortex by magnetic resonance imaging. Science. 1991;254:716–9. Mandeville JB, Marota JJA, Kosofsky BE, Keltner JR, Weissleder R, Rosen BR, Weisskoff RM. Dynamic functional imaging of relative cerebral blood volume during rat

Preclinical Pharmacological MRI

21. 22.

23. 24.

25. 26. 27. 28.

29. 30.

31.

32. 33. 34.

35. 36.

37.

38. 39.

40.

41. 42.

43.

44. 45.

46. 47. 48.

49.

50.

51.

synthase inhibition. J. Cereb. Blood Flow Metab. 1997;17: 1191–201. Boxermann JL, Bandettini PA, Kwong KK, Baker JR, Davis TL, Rosen BR, Weisskoff RM. The intravascular contribution to fMRI signal change: Monte Carlo modelling and diffusion-weighted studies in vivo. Magn. Reson. Med. 1995;34:4–10. Duong TQ, Yacoub E, Adriany G, Hu X, Ugurbil K, Kim, S-G. Magn. Reson. Med. 2003;49:1019–27. Bandettini PA, Wong EC, Jesmanowicz R, Scott Hinks R, Hyde JS. Spin-echo and gradient-echo EPI of human brain activation using BOLD contrast: a comparative study at 1.5T. NMR Biomed. 1994;7:12–20. Lowe AS, Williams SCR, Symms MR, Stolerman IP, Shoaib M. Functional magnetic resonance imaging of drug dependence: naloxone-precipitated morphine withdrawal. Neuroimage. 2002;17:902–10. Shoaib M, Lowe AS, Williams SCR. Imaging localised dynamic changes in the nucleus accumbens following nicotine withdrawal in rats. Neuroimage. 2004;22:847–54. Shah YB, Prior MJ, Dixon AL, Morris PG, Marsden CA. Detection of cannabinoid agonist evoked increase in BOLD contrast in rats using functional magnetic resonance imaging. Neuropharmacology. 2004;46:379–87. Turner R, Jezzard P, Wen H, Kwong KK, Le Bihan D, Zeffiro T, Balaban RS. Functional mapping of the human visual cortex at 4 and 1.5 Tesla using deoxygenation contrast EPI. Magn. Reson. Med. 1993;29:277–9. Kim D-S, Garwood M. High-field magnetic resonance techniques for brain research. Curr. Opin. Neurobiol. 2003;13: 612–9. Bandettini PA, Wong EC. A hypercapnia-based normalization method for improved spatial localization of human brain activation with fMRI. NMR Biomed. 1997;10:197– 203. Wise RG, Ide K, Poulin MJ, Tracey, I. Resting fluctuations in arterial carbon dioxide induce significant low frequency variations in BOLD signal. Neuroimage. 2004;21:1652-64. Xu H, Li SJ, Bodurka J, Zhao X, Xi ZX, Stein EA. Heroininduced neuronal activation in rat brain assessed by functional MRI. Neuroreport. 2000;11:1085–92. Zaharchuk G, Mandeville JB, Bogdanov AA, Weissleder R, Rosen BR, Marota JJ. Cerebrovascular dynamics of autoregulation and hypoperfusion. An MRI study of CBF and changes in total and microvascular cerebral blood volume during hemorrhagic hypotension. Stroke. 1999;30:2197– 204; discussion 2204–5. Kalisch R, Elbel GK, Gossl C, Czisch M, Auer DP. Blood pressure changes induced by arterial blood withdrawal influence BOLD signal in anesthetised rats at 7 Tesla: implications for pharmacological MRI. Neuroimage. 2001;14:891–8. Rao SM, Salmeron BJ, Durgerian S, Janowiak JA, Fischer M, Risinger RC, Conant LL, Stein EA. Effects of methylphenidate on functional MRI blood-oxygen-leveldependent contrast. Am. J. Psychiatry. 2000;157: 1697–9. Luo F, Wu G, Li Z, Li SJ. Characterization of effects of mean arterial blood pressure induced by cocaine and cocaine methiodide on BOLD signals in rat brain. Magn. Reson. Med. 2003;49:264–70.

Part I

20.

forepaw stimulation. Magn. Reson. Med. 1998;39:615– 24. Mandeville JB, Jenkins BG, Kosofsky BE, Moskowitz MA, Rosen BR, Marota JJ. Regional sensitivity and coupling of BOLD and CBV changes during stimulation of rat brain. Magn. Reson. Med. 2001;45:443–7. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. USA. 1990;87:9868–72. Ogawa S, Menon R, Tank DW, Kim SG, Merkle H, Ellermann JM, Ugurbil K. Functional brain mapping by blood oxygen level dependent contrast MRI. Biophys. J. 1993;64:800– 12. Weiskoff RM, Zou CS, Boxerman JL, Rosen BR. Microscopic susceptibility variation and transverse relaxation theory and experiment. Magn. Reson. Med. 1994;31:601–10. Ogawa S, Tank DW, Menon R, Ellermann JM, Kim SG, Merkle H, Ugurbil K. Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proc. Natl. Acad. Sci. USA. 1992;89:5951–5. Vanzetta I., Grinvald A. Increased cortical oxidative metabolism due to sensory stimulation: implications for functional brain imaging. Science. 1999;286:1555–8. Buxton RB. The elusive initial dip. Neuroimage. 2001;13: 953–8. Fox PT, Raichle ME, Minton MA, Dence C. Nonoxidative glucose consumption during focal physiologic neuronal activity. Science. 1988;241:462–4. Kwong KK, Belliveau JW, Chesler DA, Glodberg IE, Weisskoff RM, Poncelet BP, Kennedy DN, Hoppel BE, Cohen MS, Turner R. Dynamic magnetic resonance imaging of human brain activation during primary somatosensory stimulation. Proc. Natl. Acad. Sci. USA. 1992;89:5675–9. Buxton RB, Wong EC, Frank LR. Dynamics of blood flow and oxygenation changes during brain activation: the balloon model. Magn. Reson. Med. 1998;39:855–64. Schwartz WJ, Smith CB, Davidsen L, Savaki H, Sokoloff L, Mata M, Fink DJ, Gainer H. Metabolic mapping of functional activity in the hypothalamo-neurohypophysial system of the rat. Science. 1979;205:723–5. Sibson NR, Dhankhar A, Mason GF, Rothman DL, Behar KL, Shulman RG. Stoichiometric coupling of brain glucose metabolism and glutamatergic neuronal activity. Proc. Natl. Acad. Sci. USA. 1998;95:316–21. Logothetis NK, Pauls J, Augath M, Tirnath T, Oeltermann A. Neurophysiological basis of the fMRI signal. Nature. 2001;412:150–7. Lauritzen M, Gold L. Brain function and neurophysiological correlates of signals used in functional neuroimaging. J. Neurosci. 2003;23:3972–80. Shmuel A, Yacoub E, Pfeuffer J, Van de Moortele PF, Adriany G, Hu X, Ugurbil K. Sustained negative BOLD, blood flow and oxygen consumption response and its coupling to the positive response in the human brain. Neuron. 2002;36:1195–210. Heeger DJ, Rees D. What does fMRI tell us about neuronal activity? Nat. Rev. Neurosci. 2002;3:142–51. Cholet N, Seylaz J, Lacombe P, Bonvento G. Local uncoupling of the cerebrovascular and metabolic responses to somatosensory stimulation after neuronal nitric oxide

References 875

876 Part I

Biological Sciences

Part I

52. Schwarz AJ, Zocchi A, Reese T, Gozzi A, Garzotti M, Varnier G, Curcuruto O, Sartori I, Girlanda E, Biscaro B, Crestan V, Bertani S, Heidbreder C, Bifone A. Concurrent pharmacological MRI and in situ microdialysis of cocaine reveal a complex relationship between the central hemodynamic response and local dopamine concentration. Neuroimage. 2004;23:296–304. 53. Febo M, Segarra AC, Tenney JR, Brevard ME, Duong TQ, Ferris CF. Imaging cocaine-induced changes in the mesocorticolimbic dopaminergic system of conscious rats. J. Neurosci. Methods. 2004;139:167–76. 54. Kalisch R, Salome N, Platzer S, Wigger A, Czisch M, Sommer W, Singewald N, Heilig M, Berthele A, Holsboer F, Landgraf R, Auer DP. High trait anxiety and hyporeactivity to stress of the dorsomedial prefrontal cortex: a combined phMRI and Fos study in rats. Neuroimage. 2004;23:382– 91. 55. Hansen TD, Warner DS, Todd MM, Vust LJ. Effects of nitrous oxide and volatile anaesthetics on cerebral blood flow. Br. J. Anaesth. 1989;63:290-5. 56. Hansen TD, Warner DS, Todd MM, Vust LJ. The role of cerebral metabolism in determining the local cerebral blood flow effects of volatile anesthetics: evidence for persistent flow-metabolism coupling. J. Cereb. Blood Flow Metab. 1989;9:323–8. 57. Hyder F, Behar KL, Martin MA, Blamire AM, Shulman RG. Dynamic magnetic resonance imaging of the rat brain during forepaw stimulation. J. Cereb. Blood Flow Metab. 1994;14:649–55. 58. Ueki M, Mies G, Hossmann KA. Effects of alpha-chloralose, halothane, pentobarbital and nitrous oxide anaesthesia on metabolic coupling in somatosensory cortex of rat. Acta Anaesthesiol. Scand. 1992;36:318–22. 59. Balis GU, Monroe RR. The pharmacology of chloralose. Psychopharmacologica. 1964;6:1–30. 60. Bonvento G, Charbonne R, Correze J, Borredon J, Seylaz J, Lacombe P. Is α-chloralose plus halothane induction a suitable anaesthetic regimen for cerebrovascular research? Brain Res. 1994;665:213–21. 61. Lindauer U, Villringer A, Dirnagul U. Characterization of CBF response to somatosensory stimulation: model and influence of anesthetics. Am. J. Physiol. 1993;264:H1223-8. 62. Preece M, Mukherjee B, Huang CL, Hall LD, Leslie RA, James MF. Detection of pharmacologically mediated changes in cerebral activity by functional magnetic resonance imaging: effects of sulpiride in the brain of the anaesthetised rat. Brain Res. 2001;916:107– 14. 63. Houston GC, Papadakis NG, Carpenter TA, Hall LD, Mukherjee B, James MF, Huang CL. Mapping of brain activation in response to pharmacological agents using fMRI in the rat. Magn. Reson. Imaging. 2001;19:905–19. 64. Nieoullon A, Dusticier A. Effects of α-chloralose on the activity of the nigrostriatal dopaminergic system in the cat. Eur. J. Pharmacol. 1980;65:403–10. 65. Xi ZX, Wu G, Stein EA, Li SJ. Opiate tolerance by heroin self-administration: an fMRI study in rat. Magn. Reson. Med. 2004;52:108–14. 66. Marota JJA, Mandeville JB, Weisskoff RM, Moskowitz MA, Rosen BR, Kosofsky BE. Cocaine activation discriminates

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77. 78. 79.

80.

dopaminergic projections by temporal response: an fMRI study in rat. Neuroimage. 2000;11:13–23. Schwarz A, Gozzi A, Reese T, Bertani S, Crestan V, Hagan J, Heidbreder C, Bifone A. Selective dopamine D(3) receptor antagonist SB-277011-A potentiates phMRI response to acute amphetamine challenge in the rat brain. Synapse. 2004;54:1–10. Mueggler T, Baumann D, Rausch M, Rudin M. Bicucullineinduced brain activation in mice detected by functional magnetic resonance imaging. Magn. Reson. Med. 2001;46:292–8. Zhang Z, Andersen AH, Avison MJ, Gerhardt GA, Gash DM. Functional MRI of apomorphine activation of the basal ganglia in awake rhesus monkeys. Brain Res. 2000;852:290–6. Maekawa T, Tommasino C, Shapiro HM, Keifer-Goodman J, Kohlenberger RW. Local cerebral blood flow and glucose utilization during isoflurane anesthesia in the rat. Anesthesiology. 1986;65:144–51. Lenz C, Rebel A, Van Ackern K, Kuschinsky W, Waschke KF. Local cerebral blood flow, local cerebral glucose utilization, and flow-metabolism coupling during sevoflurane versus isoflurane anaesthesia in rats. Anaesthesiology. 1998;89:1480–8. Hendrich KS, Kochanek PM, Melick JA, Schiding JK, Statler KD, Williams DS, Marion DW, Ho C. Cerebral perfusion during anaesthesia with fentanyl, isoflurane, or pentobarbital in normal rats studied by arterial spin-labelled MRI. Magn. Reson. Med. 2001;46:202–6. Sicard K, Shen Q, Brevard ME, Sullivan R, Ferris CF, King JA, Duong TQ. Regional cerebral blood flow and BOLD responses in conscious and anaesthetised rats under basal and hypercapnic conditions: implications for functional MRI studies. J. Cereb. Blood Flow Metab. 2003;23: 472–81. Antognini JF, Buonocore MH, Disbrow EA, Carstens E. Isoflurane anaesthesia blunts cerebral responses to noxious and innocuous stimuli: a fMRI study. Life Sci. 1997;24:349–54. Zhang H, Ji T, Leslie R, Hockings PD, Templeton, D, Wyrwicz AM. Characterization of D2 antagonist sulpiride effects on cerebral hemodynamics in a conscious rabbit with fMRI. Proc. Intl. Soc. Mag. Reson. Med. 2003;11:1861. Stein EA, Fuller SA, Edgemond WS, Campbell WB. Selective effects of the endogenous cannabinoid arachidonylethanolamide (anandamide) on regional cerebral blood flow in the rat. Neuropsychopharmacology. 1998;19:481–91. Friston KJ, Ashburner J, Poline JB, Frith CD, Heather JD, Frackowiak RSJ. Spatial registration and normalization of images. Hum. Brain Mapp. 1995;2:165–89. Ashburner J, Friston KJ. Nonlinear spatial normalisation using basis functions. Hum. Brain Mapp. 1999;7:254–66. Van de Moortle PF, Pfeuffer J, Glover GH, Ugurbil K, Hu X. Respiration-induced B0 fluctuations and their spatial distribution in the human brain at 7 Tesla. Magn. Reson. Med. 2002;47:888–95. Aguirre GK, Zarahn E, D’Esposito M. The inferential impact of global signal covariates in functional neuroimaging analysis. Neuroimage. 1998;8:302–6.

Preclinical Pharmacological MRI

94. Bouthenet ML, Souil E, Martres MP, Sokoloff P, Giros B, Schwartz JC. Localisation of dopamine D3 receptor mRNA in the rat brain using in situ hybridisation histochemistry: a comparison with dopamine D2 receptor mRNA. Brain Res. 1991;564:203–19. 95. Mengod G, Villaro MT, Landwehrmeyer GB, Martinez-Mir MI, Niznik HB, Sunahara RK, Seeman P, O’Dowd BF, Probst A, Palacios, JM. Visualization of dopamine D1 , D2 and D3 receptor mRNAs in human and rat brain. Neurochem. Int. 1992;20(Suppl 33s–43s). 96. Landwehrmeyer B, Mengod G, Palacios JM. Differential visualisation of dopamine D2 and D3 receptor sites in rat brain. A comparative study using in situ hybridisation histochemistry and ligand binding autoradiography. Eur. J. Neurosci. 1993;5:145–53. 97. Diaz J, Levesque D, Lammers CH, Griffon M, Martres MP, Schwartz JC, Sokoloff P. Phenotypical characterisation of neurons expressing the dopamine D3 receptor in the rat brain. Neuroscience. 1995;65:731–45. 98. Krimer LS, Muly EC, Williams GV, Goldman-Rakic PS. Dopaminergic regulation of cerebral cortical microcirculation. Nat. Neurosci. 1998;1:286–9. 99. Edvinsson L, McCulloch J, Sharkey J. Vasomotor responses of cerebral arterioles in situ to putative dopamine receptor agonists. Br. J. Pharmacol. 1985;85:403–10. 100. Iadecola C. Neurogenic control of the cerebral microcirculation: is dopamine minding the store? Nat. Neurosci. 1998;1:263–5. 101. Massott M, Longo VG. α-Chloralose and the central dopaminergic system. J. Pharm. Pharmacol. 1978;30: 667. 102. Ford APDW, Marsden CA. Influence of anaesthetics on rat striatal dopamine metabolism in vivo. Brain Res. 1986;379:162–6. 103. Garrett KM, Gan J. Enhancement of γ-aminobutyric acidA receptor activity by α-chloralose. J. Pharmacol. Exp. Ther. 1998;285:680–6. 104. Chen YI, Mandeville JB, Marota JA, Nguyen TV, Green AR, Jenkins BG. Anaesthetic filters for eliciting specific neurotransmitter effects in pharmacological MRI. Proceedings of the International Society of Magnetic Resonance in Medicine, Glasgow, 2000. 105. Shulman RG, Rothman DL, Hyder F. Stimulated changes in localised cerebral energy consumption under anaesthesia. Proc. Natl. Acad. Sci. USA. 1999;96:3245–50. 106. Reese T, Bochelen D, Baumann D, Rausch M, Sauter A, Rudin M. Impaired functionality of reperfused brain tissue following short transient focal ischemia in rats. Magn. Reson. Imaging. 2002;20:447–54. 107. Arthurs OJ, Stephenson CM, Rice K, Lupson VC, Spiegelhalter DJ, Boniface SJ, Bullmore ET. Dopaminergic effects on electrophysiological and functional MRI measures of human cortical stimulus-response power laws. Neuroimage. 2004;21:540–6. 108. Paxions G, Watson C. The Rat Brain in Stereotaxic Coordinates. Academic Press: London, 1997.

Part I

81. Lowe AS, Barker GJ, Ireland MD, Beech JS, Williams SCR. Estimating global effects from extra-cerebral tissue: inferential utility for pharmacological fMRI (in press). 82. Bloom AS, Hoffmann RG, Fuller SA, Pankiewicz J, Harsch HH, Stein EA. Determination of drug-induced changes in functional MRI signal using a pharmacokinetic model. Hum. Brain Mapp. 1999;8:235–44. 83. Wise RG, Williams P, Tracey I. Using fMRI to quantify the time dependence of remifentanil analgesia in the human brain. Neuropsychopharmacology. 2004;29:626–35. 84. Bymaster FP, Reid LR, Nichols CL, Kornfield EC, Wong DT. Elevation of acetylcholine levels in striatum of rat brain by LY163502, trans-(−) 5,5a,6,7,8,9a,10octahydro-6-propylpyrimidioquinolin-2-aminedi-hydrochloride, a potent and stereospecific (D2 ) agonist. Life Sci. 1986;38:317–22. 85. Sokoloff P, Andrieux M, Besancon R, Pilon C, Martres MP, Giros B, Schwartz JC. Pharmacology of human dopamine D3 receptor expressed in a mammalian cell line: comparison with D2 receptor. Eur. J. Pharmacol. 1992;225:331–7. 86. Bowen WP, Coldwell MC, Hicks FR, Riley GJ, Fears R. Ropinirole, a novel dopaminergic agent for the treatment of Parkinson’s disease, with selectivity for cloned dopamine D3 receptors. Br. J. Pharmacol. 1993;110(Suppl 93P). 87. Gackenheimer SL, Schaus JM, Gehlert DR. [3 H]Quinelorane binds to D2 and D3 receptors in the rat brain. J. Pharmacol. Exp. Ther. 1995;274:1558–65. 88. Levant B. The dopamine D3 receptor: neurobiology and potential clinical relevance. Pharmacol. Rev. 1997;49:231–52. 89. Coldwell MC, Boyfield I, Brown AM, Stemp G, Middlemiss DN. Pharmacological characterisation of extracellular acidification rate responses in human D2 (long), D3 and D4 receptors in Chinese hamster ovary cells. Br. J. Pharmacol. 1999;127:1135–44. 90. Kelinschmidt A, Bruhn H, Kruger G, Merboldt KD, Stoppe G, Frahm J. Effects of sedation, stimulation and placebo on cerebral blood oxygenation: a magnetic resonance neuroimaging study of psychotropic drug action. NMR Biomed. 1999;12:286–92. 91. Foreman MM, Fuller RW, Hynes MD, Gidda JS, Nichols CL, Schaus JM, Kornfeld EC, Clemens JA. Preclinical studies on quinelorane, a potent and highly selective D2 -dopaminergic agonist. J. Pharmacol. Exp. Ther. 1989;250:227–35. 92. Manzione BM, Bernstein JR, Franklin RB. Observations on the absorption, distribution, metabolism, and excretion of the dopamine (D2 ) agonist, quinelorane, in rats, mice, dogs, and monkeys. Drug Metab. Dispos. 1991;19: 54–60. 93. Sokoloff P, Giros B, Martres MP, Bouthenet ML, Schwartz JC. Molecular cloning and characterization of a novel dopamine receptor (D3 ) as a target for neuroleptics. Nature. 1990;347:146–51.

References 877

879

Kishore Bhakoo, Catherine Chapon, Johanna Jackson, and William Jones Stem Cell Imaging Group, MRC Clinical Sciences Centre, Imperial College London, London W12 0NN, UK

Introduction Cell replacement therapy is undergoing a critical transition from being a discipline of the basic sciences to being recognized as a potential component of medical practice. For multiple tissues, the use of stem cell transplantation to replace cells lost due to traumatic injury or chronic degenerative processes is being pursued in a wide range of experimental models. Cell-based therapies [1] have received much attention as novel therapeutics for treatment of cancer [2], autoimmune [3], cardiovascular [4], inflammatory [5], and degenerative diseases [6,7]. A number of native cells, antigen-specific T-lymphocytes [8], or, more recently, stem and progenitor cells have been used for these approaches. Such treatment offers the possibility of treating a wide range of serious degenerative diseases that affect millions of people worldwide for which there are currently no cures. The recognition that cell transplantation can be used for tissue repair is associated with recognition of the considerable challenges involved in implementing this approach in the clinical arena, with one of the most significant challenges being the non-invasive analysis of transplanted cells and their progeny. While multiple approaches can be used to analyze survival, dispersion, and differentiation of transplanted cells in experimental animals, none of them can be applied to clinical analysis. Whether one considers ex vivo cell labeling with fluorescent dyes or transplantation of cells expressing reporter genes (e.g. β-galactosidase, green fluorescent protein), these are methods that involve sacrifice of the animal and removal of tissue for histological procedures [9,10]. Thus, these approaches cannot be translated to human studies. Development of methods for monitoring cell grafts non-invasively, with sufficiently high sensitivity and specificity to identify and map the fate of transplanted cells, is an important aspect of application and safety assessment of stem cell therapy. MRI methods are potentially well suited for such applications as this produces non-invasive “images” of opaque tissues or structures inside the body and more importantly can be translated for pre-clinical assessments. Due to the seamless integration into the host parenchyma, and migration over long distances, cell grafts Graham A. Webb (ed.), Modern Magnetic Resonance, 879–890.
C 2008 Springer.

cannot be detected based on their mass morphology. To monitor cell migration and positional fate after transplantation, current methods use either reporter genes or chimeric animals. These methods are cumbersome, involve sacrifice of the animal and removal of tissue for histological procedures, and cannot be translated to human studies [9,10]. Therefore, development of methods for monitoring cell grafts non-invasively, with sufficiently high sensitivity and specificity to identify and map the fate of transplanted cells, is an important aspect of application and safety assessment of stem cell therapy. MRI methods are potentially well suited for such applications as this produces non-invasive “images” of opaque tissues or structures inside the body. For transplanted cells to be visualized and tracked by MRI, they need to be tagged so that they are “MR visible”. At present there are two types of MRI contrast agents used clinically. These are gadolinium chelates (e.g. Gd3+ -DTPA) or iron oxide nanoparticles. However, these reagents were designed as blood-pool contrast agents and are impermeable to cells. Several approaches have been deployed to enhance cell labeling to allow in vivo cell tracking by conjugating MRI contrast agents to a range of ancillary molecules to enhance their uptake. With the growing array of cell labeling techniques, cells tagged with various monocrystalline MR probes have been evaluated both in vitro and in vivo [11–13]. Methods for monitoring implanted stem cells noninvasively in vivo will greatly facilitate the clinical realization and optimization of the opportunities for stem cell-based therapies. Other than tracking stem cells, there are numerous examples where similar methodologies of cell tracking can aid in clinical diagnosis, such as those from a tumor or following an inflammatory response.

Intracellular MRI Contrast Agents Recent work in the design of MRI contrast agents has opened up the possibility of combining the spatial resolution available of MRI for anatomic imaging with the ability to “tag” cells, and thus enable non-invasive detection and study of cell migration from the site of implantation. In vivo monitoring of stem cells after grafting is essential for understanding their migrational dynamics, which is an important aspect in determining the overall therapeutic

Part I

Application of MRI to Cell Tracking

880 Part I

Biological Sciences

Part I

O

O

O

-

O

-

N

N 3+

Gd N

O

NH

-

O

(a) Gadolinium chelate DO3A

(b) SPIO

(c) USPIO

Fig. 1. Schematic structures of (a) Gd[DO3A], (b) SPIO, and (c) USPIO.

index of cell therapies. Despite recent advances in both the synthesis of paramagnetic molecules and the basic cell biology, methods for achieving effective cell labeling using molecular MR tags are still in their infancy.

Properties of a Good Contrast Agent for Cell Tracking Before discussing the design and utilization of contrast agent to label cells for in vivo tracking by MRI, there are several parameters that need to be considered when synthesizing an efficient contrast agent. Firstly, there is a need to deliver sufficient amounts of MRI contrast agent into cells and achieve intracellular retention. Once the contrast agent is loaded into the cells, there is a need for efficient relaxivity to obtain a high in vivo MR signal-to-noise ratio. An additional aspect for successful design is tolerable cytotoxicity of the MRI contrast agent, which should have no long-term effects on cell viability, nor compromise cellular function, e.g. metabolism and differentiation.

MRI Contrast Agent for Cell Tracking MRI contrast agents can be classified as either paramagnetic or superparamagnetic and will be discussed in some detail below.

Paramagnetic Agents Where an element has one or more unpaired electrons, it is said to be paramagnetic, as it possesses a permanent magnetic moment. Examples of these are Fe3+ , Mn2+ , and Gd3+ . The more unpaired electrons present, the greater

the magnetic moment. The effect of the magnetic moment in solution results in a dipolar magnetic interaction between the paramagnetic ion and neighboring water molecules. A fluctuation in this magnetic interaction produces a decrease in T 1/T 2 relaxation time [14]. Paramagnetic compounds produce, predominantly, a T 1 effect, giving a hyperintense region. In order to avoid the problems associated with toxicity in vivo, heavy metal ions are chelated with organic moieties in order to make them more biocompatible (Figure 1a). An in-depth discussion of the mechanism underlying the enhancement of relaxation is examined by Merbach and Toth [14].

Superparamagnetic Agents The other class of contrast agents is superparamagnetic compounds. These consist of an iron oxide core, typically 4–10 nm in diameter, where several thousand iron atoms are present. A biocompatible polymer surrounds the core to provide steric and/or electrostatic stabilization. This is required due to the large surface area to volume of the nanoparticles. If no stabilization is present, the particles spontaneously precipitate out of solution due to colloidal instability. The polymers used to stabilize the iron oxide core are typically polysaccharide-based (e.g. dextran and starch) but others have also been used, e.g. polyethylene glycol (PEG). There are two types of superparamagnetic contrast agents, superparamagnetic iron oxide (SPIO) and ultra small superparamagnetic iron oxide (USPIO). The difference between the two is illustrated in Figure 1b and c, where SPIOs consist of several magnetic cores surrounded by a polymer matrix and USPIOs are individual

Cell Tracking

Engineering Delivery Systems for Iron Oxide Contrast Agents However, none of the iron oxide-based contrast agents available for clinic studies were designed to go across cellular membranes. Therefore, numerous efforts have been made to deliver iron oxide particles into a variety of cells. One such molecule that is emerging as a useful reagent relies on the covalent binding of CLIO particles to the HIV-1 TAT peptide to enhance cellular uptake [11,13,22–24]. TAT peptide contains a membrane-translocating signal that efficiently shuttles the particles into cells and the nuclear compartment [25]. A similar approach involves the covalent conjugation of internalizing monoclonal antibodies onto SPIO particles leading to cellular uptake via endosome-mediated mechanisms [26]. Alternatively, conjugations of SPIOs with antibodies to cell surface antigens have also been deployed with some success [27]. Nevertheless, these methods are inherently restrictive to the particular antibody–receptor interaction on the target

cell line. Other methods involve the use of transfection reagents, such as those developed for the translocation of plasmid DNA into cells [28]. Even though this approach offers a more universal way to transfer iron oxide particles intracellularly, it suffers from the variability in cellular labeling; but more importantly some of these reagents have cytotoxicity characteristics associated with highly cationic transfection agents (TAs). A more detailed summary of these methods is outlined below.

Delivery of Contrast Agent with Transfection Agents TAs have been developed commercially for the delivery of genetic cargo into cells for gene therapy applications. Such delivery systems are based on different platforms, such as liposomes, cationic dextrans, dendrimers, etc. These delivery systems are usually cationic to allow a favorable interaction between both DNA–TA and TA–cell surface (both DNA and cellular surface are anionic). The mechanism affording interactions between DNA and TA are predominantly electrostatic, although van der Waals forces will also be present. The use of TAs with clinically approved contrast agents has resulted in successful labeling of several cell lines [28–32]. Complexation between the contrast agent and the TA depends on the nature of the contrast agent itself. However, where a low surface charge is present on the contrast agent, van der Waals forces will predominate. The contrast agent–transfection agent complex enters the cell by endocytosis, shown in Figure 2. The principle advantage of using this technology is its wide availability. Commercially developed transfection kits (lipofectene, SAINT-Mix, poly-l-lysine, etc.) can be obtained from several manufacturers and have been used in conjunction with easily available contrast agents (Sinerem
r , Endorem
r , etc.). The disadvantage of such a system is that the protocol must be optimized for every cell type [33], and the TAs themselves usually pose various levels of toxicity [34]; again these are cell-type dependent. It is therefore necessary to optimize the contrast agent/transfection agent ratio, in order to maintain a fine balance between labeling efficiency and cellular toxicity. Another drawback of this system is that the interaction between the contrast agent and the TA can result in a decrease in T 1 and T 2 relaxivity [29]. Furthermore, studies have also demonstrated that the more effective TAs have greater cytotoxic effects [34].

Delivery of Contrast Agent Using Specific Targeting A distinct advantage of developing de novo contrast agents is the opportunity to incorporate specific molecular

Part I

cores surrounded by a polymer. Superparamagnetic contrast agents provide predominantly a T 2 effect, but smaller particles have shown to act as a T 1 agent [14]. The relaxation mechanism for superparamagnetic particles is discussed more extensively in a review by Roch et al. [15]. SPIOs are clinically contrast agents approved by the Food and Drug Administration (FDA) for hepatic reticuloendothelial cell imaging, and are in Phase III clinical trials as blood-pool agents for use in lymphography [16]. Iron oxide particles are also being developed for a variety of different applications, namely: (a) as magnetic navigation devices for the targeted delivery of therapeutics [17,18]; (b) hyperthermia-induced tumor therapeutics under high-frequency magnetic fields [19]; and (c) for magnetic cell sorting [20]. More importantly, SPIO particles are biodegradable and can be degraded and assimilated within the body. More recently, a new class of modified USPIO has been produced known as cross-linked iron oxide (CLIO), whereby the dextran coat of the USPIO is cross-linked in the presence of epichlorohydrin, and then aminated to produce amine-terminated nanoparticles suitable for conjugation [21]. Therefore, the concept of labeling cells with one of these classes of contrast agents is extremely attractive, as one could then visualize transplanted cells non-invasively by MRI; where the labeled cells would produce either hyperintense or hypointense depending on the class of contrast agent chosen. However, most efforts to engineer cell tagging agents has concentrated on using superparamagnetic nanoparticles as they possess higher sensitivity, especially on the higher field research scanners that are now being implemented for pre-clinical or clinical studies.

Delivery of Contrast Agent Using Specific Targeting 881

882 Part I

Biological Sciences

Part I USPIO

Transfection Agent

USPIO-TA complex Cell Membrane

Membrane Invagination

Clathrin coated pit formation

Fig. 2. Schematic of the complexation of USPIO with a transfection agent and subsequent cellular uptake via endocytosis.

targeting to increase the efficiency of cellular labeling. Many different methods have been used to deliver contrast agents and will be discussed in some detail.

Membrane Permeating Peptides Membrane permeating peptides have received considerable attention as they can delivery cargoes of different sizes including: nanoparticles [35], liposomes [36], and fluorescein with extremely high efficiency and minimal cytotoxicity [37]. Various membrane permeating peptides sequences (peptide transduction domain, PTD) are found to occur naturally including HIV-1 TAT, Antennapedia transcription factor, Herpes simplex virus, and VP22 transcription factor. However, the most extensively employed delivery system has been the conjugation of TAT peptide with USPIO to facilitate labeling of a range of cell types [11,35,38,39]. The precise mechanism for cellular entry has yet to be elucidated; however, there are several indications for particular structural requirements that allow for effective and efficient cellular uptake. One of the principal requirements is the presence of multiple arginine residues on the PTD for efficient cell entry. Many naturally occurring membrane-permeating peptides contain a large number of arginine residues [40]. Another requirement appears to be the presence of negatively charged glycosaminoglycans on the cell surface [41], whilst another report suggests that heparin sulfate is required on the cell surface [42]. It was originally thought that the uptake mechanism was energy independent, and thus would not involve endocytosis.

Other studies suggested that the non-endosomal uptake was an artifact of the technique used to study the mechanism of cell entry [43]. Neverthless, what is clear, regardless of the exact mechanism, is that TAT and other polyarginine peptide sequences allow the shuttling of cargo across the cell membrane with high efficiency. The majority of studies that have investigated the delivery of CLIO into a variety of cells using PTDs have used the TAT peptide or a variation on the theme [13,23,39,44]. Studies have also demonstrated successful cellular labeling with gadolinium chelates conjugated with poly(arginine) with minimal of cytotoxic effects. However, there is some evidence of gadolinium chelate leakage from these cells [45,46].

Use of Antibodies Conjugation of contrast agents to PTDs provides a ‘generic’ model for labeling a wide variety of cell types. However, this system lacks cellular specificity. In contrast, contrast agents conjugated to antibodies, against specific cell surface antigens, provides cellular targeting with high specificity [47]. However, targeting cell surface markers, whilst being specific, has two major drawbacks. Firstly, the presence of the nanoparticles on the surface of the cell may affect its capacity to migrate and “home” effectively. Secondly, the antibody–contrast agent complex’s interaction with the cell surface antigen is reversible; thus there is a high probability that the complex can be taken up by other cells (e.g. macrophages) and provide ambiguous information on the migrational characteristics of the

Cell Tracking

will discuss the means of conjugating PTDs and antibodies to contrast agents. In the case of USPIOs, there are two ways in which conjugation can be accomplished: (i) direct attachment to the stabilizing polymer, in this case dextran will be used as a model system, and (ii) the conjugation via a heterofunctional linker.

Other Methods for Delivering Contrast Agents Alternative means of cell labeling with iron oxide nanoparticles have also been investigated. Recently, Gupta and Curtis have used lactoferrin and ceruloplasmin coated particles to label cells to target their respective receptors. This resulted in a decrease in toxicity where lactoferrin produced the least cytotoxic effects when compared with naked iron oxide particles [49]. However, these particles were shown not to be internalized and therefore face the same problem outlined above, namely loss of cell homing abilities or detachment of iron oxide from the cell surface.

Cytotoxicity and Metabolism There are several important characteristics that need to be considered when designing a new contrast agent. These include its toxicity and eventual in vivo metabolism. It is essential that the presence of the contrast agents does not affect proliferation, differentiation, its ability to integrate into host tissues, nor its cellular function. Whilst iron oxide dextran nanoparticles have been used extensively in the clinic, these nanoparticles have remained outside the cell. Recent reports have demonstrated that the intracellular presence of dextran coated iron oxide nanoparticles induces apoptosis and affects cytoskeletal properties [50]. However, in this instance only the core of the iron oxide was investigated and was quoted to be 7.8 nm by magnetic measurements. It is likely that the overall hydrodynamic radius would play an important role in the disruption of intracellular structures. Thus, particles with a smaller hydrodynamic radius may pose fewer cytotoxic effects. The mechanism of iron oxide metabolism in vivo has been known for a number of years, as these iron oxide nanoparticles were original produced for the treatment of severe anemia. It was found that the primary mechanism was the assimilation of exogenous iron oxide into the iron cycle [51] and is incorporated into rapidly regenerating hemoglobin [52].

Conjugation Chemistry: Attaching Contrast Agent to Delivery Ligand As discussed previously, there are several methods for enhancing the uptake of contrast agents. In this section, we

Conjugating Directly to USPIO: Oxidation-Reductive Amination Whilst USPIOs are coated with hydroxyl groups present on dextran backbone, these groups are relatively unreactive in an aqueous environment. To allow direct conjugation, it is therefore necessary to introduce aldehyde moieties through selective oxidation of vicinal glycols in the presence of sodium periodate as shown in Figure 3. It is then possible to produce an imine by reacting the aldehyde in the presence of an amine-terminating moiety. The imine however is unstable and therefore selective reduction of the imine is achieved through the use of sodium cyanoborohydride (Figure 4). Whilst this approach has been used successfully for conjugating biological moieties [53,54], this conjugation method can give rise to several problems. Firstly, there may be reduced affinity of the conjugated ligand, probably due to steric hindrance from the dextran backbone [54]. Another problem that is frequently encountered is the uncontrolled oxidation of the dextran coating, resulting in particle destabilization, probably due to the degradation of the dextran [53]. Moreover, degradation was also found to occur on periodate oxidized sephadex G-25, where the dextran beads were degraded during extended exposure with sodium periodate [55]. Further problems can arise due to inconsistent reaction conditions, as it has been reported that the production of aldehydes depend on the pH of the reaction environment [56].

Introducing a Linker An alternative method of conjugating the delivery/ targeting ligand to the contrast agent is via a linker. Nevertheless, there is still the issue of lack of reactivity of the hydroxyl groups under neutral aqueous conditions. This problem is overcome by the use of CLIO particles, where the dextran is cross-linked and then incubated in the presence of ammonia to produce amine-terminated particles. These on the other hand are relatively more reactive. It is then possible to couple the nanoparticles to a plethora of heterofunctional linkers, usually by the use of a succinimide ester activated compound as illustrated below (Figure 5). The advantage of using this approach is the commercial availability of several succinimide esters available for

Part I

targeted cells. These two issues have been overcome by using the OX-26 antibody in the labeling of neural precursor cells [48], which is an internalizing antibody directed against the transferrin receptor. In this instance, uptake of the particles is through receptor-mediated endocytosis.

Conjugation Chemistry: Attaching Contrast Agent to Delivery Ligand 883

884 Part I

Biological Sciences

Part I

HO

HO

H

H OH H

HO H

H

OH

OH H

O

O

OH

H

NaIO4

H

H OH H

O

O

H

OH

OH

HO

HO H

H OH H H

n

OH

n

OH OH

OH

Fig. 3. Selective oxidation of neighboring hydroxyl groups in the presence of sodium periodate producing dextran polyaldehyde.

HO

HO

H

H

OH

H

H

O O

RNH2

H

N

O

H OH H

O

O

R

H

OH H

OH

H

H OH H

OH

HO

HO H

n

OH

H

n

OH OH

OH

HO

NaCNBH3

H

H

OH H

HN R

OH

O

H

H OH H

OH

HO H

n

OH OH

Fig. 4. Reduction of an unstable imine via reductive amination in the presence of sodium cyanoborohydride to produce a stable amine.

Cell Tracking

MRI Tracking of Stem Cells in the Heart 885

NH

O O

H3C

R

R

N

H3C

O

O

Fig. 5. Functionalization of amine-terminated nanoparticles in the presence of a succinimide ester to produce an amide, e.g. SIA, where R = CH2 I.

facile conjugation, although some can be readily prepared prior to use [57]. The disadvantage to this technique is that the N -hydroxysuccinimide ester is hydrolyzed under aqueous conditions. The most common heterofunctional linkers used to date are those that allow the formation of either a thioether bond (SIA; Figure 6) or a disulfide bond (SPDP; Figure 7). These are ideal for selective conjugation to a cysteine residue either naturally present on the ligand, or introduced during synthesis. Sulfur is conjugated in preference as the rate of reaction is sulfhydryl > imidazolyl > thioether > amine [58]. SIA is preferential where a stable thioether bond is required, as SPDP produces a labile disulfide bond. The advantage of using SPDP is that the degree of conjugation O

R NH

R I

HS

MRI Tracking of Stem Cells in the Heart Myocardial infarction is by nature an irreversible injury. The extent of the infarction depends on the duration and severity of the perfusion defect [59]. Beyond contraction and fibrosis of myocardial scar, progressive ventricular remodeling of non-ischemic myocardium can further reduce cardiac function in the weeks to months after initial event [60]. Many of the therapies available to clinicians today can significantly improve the prognosis of patients following

O R

S

R

NH

Sulfur bearing ligand

SIA

can be observed by monitoring the increase in adsorption at 343 nm.

Stable thioether

Fig. 6. Nanoparticles functionalized with SIA (where R = nanoparticle) react with sulfur bearing moieties to produce a stable thioether bond. O R N O N R S HS R NH S R S S S NH

SPDP

Sulfur bearing ligand

Labile disulfide bond

P2T (monitor at 343nm)

Fig. 7. SPDP derived nanoparticles (where R = nanoparticles) react with a sulfur bearing ligand to produce a cleavable disulfide bond.

Part I

O

NH2

886 Part I

Biological Sciences

Part I

an acute myocardial infarction [60,61]. However, no pharmacological or interventional procedure used clinically has shown efficacy in replacing myocardial scar with functioning contractile tissue. Cellular agents such as fetal cardiomyocytes [62], mesenchymal stem cells (MSCs) [63], endothelial progenitor cells [64], and skeletal myoblasts [65], or embryonic stem (ES) cells [66] have already shown some efficacy to engraft in the infarct, differentiating toward a cardiomyocyte phenotype, by expressing cardiac specific proteins, preserving left ventricular function and inhibiting myocardial fibrosis. Besides, in vitro exposure of stem cells, specifically MSCs, to specific signal molecules prior to transplantation into infarcted myocardium allows their differentiation into cardiomyocyte [67] and may facilitate a successful engraftment [68]. However, verification of the status of transplanted stem cells in animal models has been performed with histological analysis. Clinical data on stem cells transplantation is in its infancy and is very limited. But preliminary clinical data have shown that stem cell transplantation for the treatment of ischemic heart failure is feasible and promising [69]. Besides MRI was used in some clinical studies to assess the improvement of contractile function after cell transplantation [70] but without any possibility to visualize the transplanted stem cells. Therefore, the ability to label stem cells with MRI-visible contrast agents [71] should enable serial tracking and quantification of transplanted stem cells, non-invasively by MRI with high spatial resolution. The programmable nature of the imaging planes allows reproducible and volumetric coverage of the heart. Moreover, this technology scales well with subject size ranging from mouse to human. Visualization of magnetically labeled endothelial progenitor cells transplanted intra-myocardially for therapeutic neovascularization in infarcted rats has been demonstrated with ex vivo MRI at 8.5 T on T 2-weighted images [72]. Garot et al. [73] have demonstrated the feasibility of in vivo MRI tracking of skeletal muscle-derived myogenic precursor cells (MPC) pre-loaded with iron oxide nanoparticles (Endorem) injected into healthy and infarcted porcine myocardium. Iron-loaded cells in the infarcted region were detected by T 2-weighted spin-echo MRI at 1.5 T. In addition, MRI guided the catheter for the injection of the labeled cells into the ischemic myocardium using Gd-DTPA delayed enhancement of the site. Moreover, post-mortem analysis demonstrated the presence of iron-loaded MPC at the center and periphery of the infarcted tissue as predicted by MRI. MSCs derived from bone marrow can be detected and tracked by MRI for up to 3 weeks [74–76]. Allogeneic ferumoxides [74] or iron fluorescent particle [75,76] were given by intra-myocardial injection in a pig model of myocardial infarction. A minimum quantity of MSCs per injection was required to be MR-detectable on T 2*- [75] or

T 2-weighted images [74] as hypointense lesions. Indeed, Kraitchman et al. have shown that using a limited number of MSC injections per animal, only some (∼70%) of the injections performed in each animal can be visualized. These promising experiments demonstrate the need for future studies to delineate the fate of injected stem cells by incorporating non-invasive tagging methods to monitor myocardial function following cell engraftment in the myocardial infarction. Consequently, MRI may lead to a better understanding of the myocardial pathophysiology as well as assessing the proper implantation and the effects of stem cell therapy by allowing a multimodal approach to evaluating anatomy, function, perfusion, and regional contractile parameters in a single non-invasive examination.

MRI Tracking of Stem Cells in the CNS Neurodegenerative diseases where cell loss is the predominate feature of the pathology and, for which there are currently no cures, cellular replacement therapy using stem cells may provide a beneficial alternative. The efficacy of cell replacement therapy was first demonstrated using engraftment of human mesencephalic tissue into the brain of patient with Parkinson’s disease [77]. Functional recovery and l-DOPA withdrawal followed by an increase in released dopamine demonstrated the functional integration of the grafted tissue. Although this was not a true stem cell transplant, it nevertheless indicated that the adult brain provides local environmental cues to undifferentiated cells to produce neuronal cell types capable of providing functional recovery. Since this pioneering experiment, many different populations of stem cells have shown to differentiate into neural phenotypes. The most obvious choice of stem cell population would be those already derived from the neural phenotype. These include neural stem cells, found in the adult subventricular zone (SVZ) and glial-restricted precursors, found in the embryonic spinal cord. Neural stem cells have been shown to differentiate into dopaminergic neurons [78], astrocytes, and oligodendrocytes [79], and spinal cord motor neurons [80]. Surprisingly, these cells are also able to transdifferentiate into other non-neural cell types such as skeletal muscle [81]. Glial-restricted progenitors, on the other hand, are restricted to the glial lineage and produce oligodendrocytes and type-1 and type-2 astrocytes [82]. ES cells are the most pluripotent of all the stem cell populations, giving rise to many cell types in the body; thus have the greatest regenerative capacity. ES cells differentiate into a variety of neural phenotypes, including dopaminergic neurons [83], serotoninergic neurons [84], neuronal precursors [85], oligodendrocytes [86], and astrocytes [87]. There are several problems, aside from the

Cell Tracking

migrated toward the lesion in the opposite hemisphere. Strong hypointense “columns” were seen by MRI in the corpus callosum. These were later confirmed as migrating blankets of cells traveling toward the ischemic lesion. GFP-expressing cells were also seen in the ischemic penumbra, and in contrast to the study by Modo et al. [92], the majority were NeuN+ suggesting that the cells had differentiated into neurons. Astrocytes and oligodendrocytes were also seen populating the surrounding area. MRI studies have also been used to track glial progenitors labeled with contrast agent. However, in these studies MRI was used to scan post-mortem tissues, following cellular transplantation. Oligodendrocyte progenitor cells (OPCs) from the CG4 cell line have greater migratory and myelinating capacity than mature oligodendrocytes. The cells were labeled with monocrystalline iron oxide nanoparticles targeted to the transferrin receptor to aid internalization of the particles [94]. The progenitor cells were grafted into the spinal cord of a myelin-deficient rat. Cellular migration was also visualized by MRI, especially in the dorsal column. Moreover, iron oxide labeled cells, fixed with paraformaldehyde and implanted in the same way, did not migrate at all; MRI contrast was seen only at the site of injection. This also suggests that the iron oxide remains localized and is not taken up by other host cells. This is of great importance if iron oxide labeling and tracking of cells is to be used clinically. The MR images were verified by histological analysis and the lesion was found to include astrocytes, microglia, and myelin. Importantly, the Prussian blue staining correlated with that for myelin, whereas it did not overlap with the GFAP+ astrocytes or microglia present. Obviously reactive gliosis and an immune reaction had occurred, but the inflammatory cells had not taken up iron oxide; the labeled OPCs were able to infiltrate the inflamed area and produce myelin. Jendelova et al. [95] used MRI to study the differential response of MSCs and ES cells in rodent models of stroke (photochemical lesion) and spinal cord compression (balloon inflation). Prior to implantation, both MSCs and ES cells were labeled with Endorem and additionally co-labeled with either BrdU or GFP, respectively. Following the induction of the lesions, either ES cells or MSCs were grafted contralateraly to the ischemic lesion. In another set of animals, either ES cells or MSCs were administered intravenously into rats with an ischemic lesion. The animals with spinal cord compression lesion were infused intravenously with MSCs. ES cells given to rats with ischemic lesions, regardless of whether given intravenously or intracerebral implantation, migrated to the lesion site within 2 weeks, as observed by MRI and subsequently confirmed with GFP visualisation. Additionally, at the site of implantation, hyper-proliferation was seen in 10% of the animals. This suggests that tumor formation had taken place, and was detected using MRI as a very large hypointense

Part I

issues of ethics, that makes ES cell therapy difficult, including the risk of inappropriate cellular differentiation and tumor formation. Mesenchymal cells derived from bone marrow have also been shown to differentiate into neural phenotypes [88]. Additionally, rat MSCs differentiate into a mixture of neural phenotypes including astrocytes, oligodendrocytes, and neurons. Upon further differentiation, GABAergic, dopaminergic, and serotoninergic neurons may also develop [89]. Neural progenitor cells have innate migratory properties. For example, neural progenitor cells isolated from the SVZ of adult or neonatal rats, when implanted into the different regions of the neonatal brain, migrate and differentiate within regions such as the olfactory bulb, cortex, and striatum. In contrast, when grafted into the adult brain, the SVZ cells only migrated to the olfactory bulb, but not to the cortex or striatum [90]. MRI was used to longitudinally track the migration of SVZ cells after implantation into the healthy rat striatum using pre-labeling cells with BrdU and lipophilic dyecoated ferromagnetic particles [91]. Furthermore, MRI revealed that the area grafted with live cells appeared to expand, whereas the area implanted with dead cells, decreased in size. Immunohistochemical analysis showed that the SVZ cells differentiated into neurons (MAP-2+ NeuN+ ) and migrated within the striatum after being cultured with bFGF. These studies revealed only localized migration. Migration over greater distances was demonstrated using non-invasive MRI studies in a stroke lesioned animal model. It was hypothesized and later confirmed that stroke damage functions as a “chemoattractant” for neural stem cells [92]. Neural stem cells derived from the Maudsley Hippocampal Clone 36 (MHP36) cell line, were labeled with the bimodal contrast agent, GRID. This enables detection both by MRI and fluorescent histology. Following a middle cerebral artery occlusion (a rodent model of stroke), the neural stem cells were grafted unilaterally into the hemisphere contralateral to the lesion. Using a combination of GRID labeled cells and MRI, it was demonstrated that following 14 days post-transplantation; most of the cells had migrated to the ipsilateral hemisphere along the corpus callosum and populated the surrounding lesion area. Moreover, upon fluorescent immunochemistry, these cells were found to be GFAP+ astrocytes and NeuN+ neuronal precursors. Hoehn et al. [93] used a similar stroke model to demonstrate the migratory properties of implanted ES cells into the brain by MRI. The cells were pre-labeled with USPIO and encapsulated with a lipofection reagent to enhance cellular uptake. The ES cells also expressed GFP as a reporter gene for immunohistochemical procedures. The labeled cells were implanted into two regions of the unaffected hemisphere: the border between the cortex and the corpus callosum, and the striatum. The labeled cells

MRI Tracking of Stem Cells in the CNS 887

888 Part I

Biological Sciences

Part I

area, much larger than the other cellular transplants [95]. Implanted MSCs also migrated to the lesion area but gathered in the necrotic tissue surrounding the lesion. Few cells entered the actual lesion and of those very few differentiated into neurons, as seen 4 weeks after implantation. Additionally, MSCs injected intravenously also migrated to the lesion site and were visible for 7 weeks postimplantation. Similarly, MSCs injected intravenously also migrated to the lesioned spinal cord, which was also confirmed with Prussian blue staining. Thus, the use of stem cell therapy to treat neurodegenerative diseases is a realistic possibility in the near future. However, the need for non-invasive imaging techniques is a prerequisite in order to monitor these transplants to determine clinical efficacy. Examination by MRI ensures that the stem cells are not only injected to the lesion site, but it also allows the monitoring of inappropriate cellular migration, and furthermore identifies damage to surrounding tissues.

MRI Tracking of Cell-Based Tumor Therapy Cell-based therapies have received much attention as novel therapeutics for the treatment of cancer [2]. For example, tumor antigen-specific lymphocytes have been used for adoptive transfer and treatment in lymphoma, melanoma, and other malignancies. But a major obstacle to an accurate evaluation of treatment efficacy and antitumor effects has been the inability to track these cytotoxic T-lymphocytes (CTLs), in vivo at sufficiently high spatial and temporal resolution [96]. Dodd et al. [97] demonstrated the distribution of T cells labeled with CLIO-TAT peptide, by in vivo MRI, in mice following intravenous administration. These studies also showed the migration of T cells loaded with CLIO-TAT to the spleen by monitoring a decrease in signal intensity observed by MRI at 4.7 T. Similar protocols have been used to detect the localization of cells in small sites such as tumors and lymph nodes in small animals. Kircher et al. [98] were the first to assess and quantitate the recruitment of systemically administered cells over time in a model of melanoma. They used an improved superparamagnetic particle (CLIO-HD: highly derivatized CLIO nanoparticles) to optimize the lymphocytes labeling at levels that can be detected in vivo via high resolution three-dimensional T 2-weighted MRI at 8.5 T. This study showed the ability to examine both cellular recruitment and therapeutic response (tumor volume change) in three dimensions, across the entire tumor simultaneously, and in quantitative and repetitive manner in the same animal using MRI. Zhang et al. [99] have reported the in vivo targeting and infiltration by magnetically labeled neural progenitor

cells, derived from the adult SVZ, to a tumor mass in a rat model of gliosarcoma. To date, this was the first study in which adult neural stem cells have been employed to target a brain tumor. The non-invasive imaging, by MRI at 7 T, monitored the dynamic migration of superparamagnetic particle-labeled neural progenitor cells toward the brain tumor. Indeed, the spatiotemporal distribution of transplanted cells and the tumor in the host brain identified by MRI was confirmed using histochemical staining and fluorescent microscopy. This study also demonstrated, using MRI, the migration of iron oxide nanoparticles labeled MSCs toward the tumor, thus confirming previous reports of stem cell infiltration of brain tumors [100,101]. Development of MRI techniques for in vivo assessement of the interaction between grafted neural progenitor cells and tumor cells in the host brain may contribute not only to our understanding of the mechanisms involved in the treatment of brain tumors with neural progenitor cells therapy but also assess the outcome of neural progenitor cell therapy both in animals and in humans with brain tumors. Tumor vasculature has attracted much interest as a potential target for cancer therapy [102]. Since cancerous growth depends on a good blood supply for nutrition and oxygen, the ability to image tumor vasculature would help to monitor the progress of cancer therapies. Brown et al. [103] demonstrated the accumulation of Sickle red blood cells (RBCs) in tumor vasculature, using MRI at 7 T. In this study, RBCs, from patients with sickle cell anemia, were loaded with Gd-DTPA and injected intravenously in rats with 9L glioma. T 1- and T 2-weighted MR images were used to monitor the infiltration of GdDTPA-loaded RBCs into the tumor mass. Additionally, changes in hemoglobin–oxygen state after administration of RBCs, without Gd-DTPA loading, was assessed using BOLD imaging. MRI allowed the visualization of the preferential aggregation of RBCs in tumor periphery. Thus, MRI can be used as a useful technique to follow the cellular migration and recruitment to monitor the progress of cell-based therapy in tumors. The high anatomical resolution together with the noninvasively in vivo imaging methodology offered by MRI, applied to monitoring implanted cells, will greatly facilitate the clinical realization and optimization of the opportunities for cell-based therapies. There are numerous examples where similar methodologies of cell tracking can aid in clinical diagnosis or can be used to trace other cells types, such as those following an inflammatory response [104,105].

Acknowledgment This work was funded by the Medical Research Council of Great Britain.

Cell Tracking

1. Dove A. Nat. Biotechnol. 2002;20:339. 2. Valone FH, Small E, MacKenzie M, Burch P, Lacy M, et al. Cancer J. 2001;7:(Suppl 2):S53. 3. Burt RK, Traynor A. Curr. Opin. Hematol. 1998;5:472. 4. Semsarian C. Intern. Med. J. 2002;32:259. 5. Bessis N, Cottard V, Saidenberg-Kermanach N, Lemeiter D, Fournier C, et al. J. Gene. Med. 2002;4:300. 6. Petit-Zeman S. Nat. Biotechnol. 2001;19:201. 7. Johnstone B, Yoo J. Expert. Opin. Biol. Ther. 2001;1:915. 8. Wen YJ, Min R, Tricot G, Barlogie B, Yi Q. Blood. 2002;99: 3280. 9. Lagasse E, Connors H, Al-Dhalimy M, Reitsma M, Dohse M, et al. Nat. Med. 2000;6:1229. 10. Orlic D, Kajstura J, Chimenti S, Jakoniuk I, Anderson SM, et al. Nature. 2001;410:701. 11. Lewin M, Carlesso N, Tung CH, Tang XW, Cory D, et al. Nat. Biotechnol. 2000;18:410. 12. Bhorade R, Weissleder R, Nakakoshi T, Moore A, and Tung CH. Bioconjug. Chem. 2000;11:301. 13. Josephson L, Tung CH, Moore A, Weissleder R. Bioconjug. Chem. 1999;10:186. 14. Merbach AE, Toth E. The Chemistry of Contrast Agents in Medical Magnetic Resonance Imaging. John Wiley and Sons, New York, 2001. 15. Roch A, Muller RN, Gillis P. J. Chem. Phys. 1999;110:5403. 16. Mack MG, Balzer JO, Straub R, Eichler K, Vogl TJ. Radiology. 2002;222:239. 17. Ito A, Shinkai M, Honda H, Kobayashi T. Cancer Gene Ther. 2001;8:649. 18. Hafeli UO, Yu J, Farudi F, Li Y, Tapolsky G. Nucl. Med. Biol. 2003;30:761. 19. Jordan A, Scholz R, Wust P, Fahling H, Krause J, et al. Int. J. Hyperthermia. 1997;13:587. 20. Roger J, Pons JN, Massart R, Halbreich M, Bacri JC. Eur. Phys. J. 1999;5:321. 21. Wunderbaldinger P, Josephson L, Weissleder R. Acad. Radiol. 2002;9(Suppl 2):S304. 22. Vives E, Brodin P, Lebleu B. J. Biol. Chem. 1997;272: 16010. 23. Zhao M, Kircher MF, Josephson L, Weissleder R. Bioconjug. Chem. 2002;13:840. 24. Fawell S, Seery J, Daikh Y, Moore C, Chen LL, et al. Proc. Natl. Acad. Sci. U. S. A. 1994;91:664. 25. Caron NJ, Torrente Y, Camirand G, Bujold M, Chapdelaine P, et al. Mol. Ther. 2001;3:310. 26. Artemov D, Mori N, Okollie B, Bhujwalla ZM. Magn. Reson. Med. 2003;49:403. 27. Kang HW, Josephson L, Petrovsky A, Weissleder R, Bogdanov A, Jr. Bioconjug. Chem. 2002;13:122. 28. Frank JA, Miller BR, Arbab AS, Zywicke HA, Jordan EK, et al. Radiology. 2003;228:480. 29. Frank JA, Zywicke H, Jordan EK, Mitchell J, Lewis BK, et al. Acad. Radiol. 2002;9(Suppl 2):S484. 30. Kalish H, Arbab AS, Miller BR, Lewis BK, Zywicke HA, et al. Magn. Reson. Med. 2003;50:275. 31. Arbab AS, Yocum GT, Kalish H, Jordan EK, Anderson SA, et al. Blood. 2004;4:1217. 32. Arbab AS, Yocum GT, Wilson LB, Parwana A, Jordan EK, et al. Mol. Imaging. 2004;3:24.

33. Bulte JW, Arbab AS, Douglas T, Frank JA. Methods Enzymol. 2004;386:275. 34. Kim SW, Ogawa T, Tabata Y, Nishimura I. J. Biomed. Mater. Res. 2004;71A:308. 35. Becker ML, Remsen EE, Pan D, Wooley KL. Bioconjug. Chem. 2004;15:699. 36. Torchilin VP, Rammohan R, Weissig V, Levchenko TS. Proc. Natl. Acad. Sci. U. S. A. 2001;98:8786. 37. Futaki S, Suzuki T, Ohashi W, Yagami T, Tanaka S, et al. J. Biol. Chem. 2001;276:5836. 38. Sundstrom JB, Mao H, Santoianni R, Villinger F, Little DM, et al. J. Acquir. Immune. Defic. Syndr. 2004;35:9. 39. Wunderbaldinger P, Josephson L, Weissleder R. Bioconjug. Chem. 2002;13:264. 40. Sakai N, Matile S. J. Am. Chem. Soc. 2003;125:14348. 41. Console S, Marty C, Garcia-Echeverria C, Schwendener R, Ballmer-Hofer K. J. Biol. Chem. 2003;278:35109. 42. Fuchs SM, Raines RT. Biochemistry. 2004;43:2438. 43. Richard JP, Melikov K, Vives E, Ramos C, Verbeure B, et al. J. Biol. Chem. 2003;278:585. 44. Koch AM, Reynolds F, Kircher MF, Merkle HP, Weissleder R, et al. Bioconjug. Chem. 2003;14:1115. 45. Allen MJ, Meade TJ. J. Biol. Inorg. Chem. 2003;8: 746. 46. Allen MJ, MacRenaris KW, Venkatasubramanian PN, Meade TJ. Chem. Biol. 2004;11:301. 47. Suwa T, Ozawa S, Ueda M, Ando N, Kitajima M. Int. J. Cancer. 1998;75:626. 48. Bulte JW, Ben-Hur T, Miller BR, Mizrachi-Kol R, Einstein O, et al. Magn. Reson. Med. 2003;50:201. 49. Gupta AK, Curtis AS. Biomaterials. 2004;25:3029. 50. Berry CC, Wells S, Charles S, Aitchison G, Curtis AS. Biomaterials. 2004;25:5405. 51. Weissleder R, Stark DD, Engelstad BL, Bacon BR, Compton CC, et al. Am. J. Roentgenol. 1989;152:167. 52. London E. J. Pharm. Sci. 2004;93:1838. 53. Molday RS, MacKenzie D. J. Immunol. Methods. 1982;52:353. 54. Shen TT, Bogdanov A, Jr, Bogdanova A, Poss K, Brady TJ, et al. Bioconjug. Chem. 1996;7:311. 55. Wilson MB, Nakane PK. J. Immunol. Methods. 1976;12: 171. 56. Novikova EV, Tishchenko EV, Iozep AA, Passet BV. Russ. J. Appl. Chem. 2002;75:985. 57. Hartman FC, Suh B, Welch MH, Barker R. J. Biol. Chem. 1973;248:8233. 58. Hermanson GT. Bioconjugate Techniques. Elsevier, Amesterdam, 1996. 59. Reimer KA, Lowe JE, Rasmussen MM, Jennings RB. Circulation. 1977;56:786. 60. Pfeffer MA. Annu. Rev. Med. 1995;46:455. 61. Ryan TJ, Antman EM, Brooks NH, Califf RM, Hillis LD, et al. J. Am. Coll. Cardiol. 1999;34:890. 62. Etzion S, Battler A, Barbash IM, Cagnano E, Zarin P, et al. J. Mol. Cell. Cardiol. 2001;33:1321. 63. Shake JG, Gruber PJ, Baumgartner WA, Senechal G, Meyers J, et al. Ann. Thorac. Surg. 2002;73:1919. 64. Kawamoto A, Gwon HC, Iwaguro H, Yamaguchi JI, Uchida S, et al. Circulation. 2001;103:634. 65. Jain M, DerSimonian H, Brenner DA, Ngoy S, Teller P, et al. Circulation. 2001;103:1920.

Part I

References

References 889

890 Part I

Biological Sciences

Part I

66. Min JY, Yang Y, Converso KL, Liu L, Huang Q, et al. J. Appl. Physiol. 2002;92:288. 67. Makino S, Fukuda K, Miyoshi S, Konishi F, Kodama H, et al. J. Clin. Invest. 1999;103:697. 68. Tomita S, Li RK, Weisel RD, Mickle DA, Kim EJ, et al. Circulation. 1999;100:II247. 69. Hamano K, Nishida M, Hirata K, Mikamo A, Li TS, et al. Jpn. Circ. J. 2001;65:845. 70. Britten MB, Abolmaali ND, Assmus B, Lehmann R, Honold J, et al. Circulation. 2003;108:2212. 71. Bulte JW, Douglas T, Witwer B, Zhang SC, Strable E, et al. Nat. Biotechnol. 2001;19:1141. 72. Weber A, Pedrosa I, Kawamoto A, Himes N, Munasinghe J, et al. Eur. J. Cardiothorac. Surg. 2004;26: 137. 73. Garot J, Unterseeh T, Teiger E, Champagne S, Chazaud B, et al. J. Am. Coll. Cardiol. 2003;41:1841. 74. Kraitchman DL, Heldman AW, Atalar E, Amado LC, Martin BJ, et al. Circulation. 2003;107:2290. 75. Hill JM, Dick AJ, Raman VK, Thompson RB, Yu ZX, et al. Circulation. 2003;108:1009. 76. Dick AJ, Guttman MA, Raman VK, Peters DC, Pessanha BS, et al. Circulation. 2003;108:2899. 77. Piccini P, Brooks DJ, Bjorklund A, Gunn RN, Grasby PM, et al. Nat. Neurosci. 1999;2:1137. 78. Ourednik J, Ourednik V, Lynch WP, Schachner M, Snyder EY. Nat. Biotechnol. 2002;20:1103. 79. Yamamoto S, Yamamoto N, Kitamura T, Nakamura K, Nakafuku M. Exp. Neurol. 2001;172:115. 80. Shibuya S, Miyamoto O, Auer RN, Itano T, Mori S, et al. Neuroscience. 2002;114:905. 81. Galli R, Borello U, Gritti A, Minasi MG, Bjornson C, et al. Nat. Neurosci. 2000;3:986. 82. Noble M, Mayer-Proschel M. In: MS Rao (Ed). Stem Cells and CNS Development. Humana Press, New Jersey, 2001, p. 123. 83. Kawasaki H, Mizuseki K, Nishikawa S, Kaneko S, Kuwana Y, et al. Neuron. 2000;28:31. 84. Lee SH, Lumelsky N, Studer L, Auerbach JM, McKay RD. Nat. Biotechnol. 2000;18:675.

85. Reubinoff BE, Itsykson P, Turetsky T, Pera MF, Reinhartz E, et al. Nat. Biotechnol. 2001;19:1134. 86. Liu S, Qu Y, Stewart TJ, Howard MJ, Chakrabortty S, et al. Proc. Natl. Acad. Sci. U. S. A. 2000;97:6126. 87. Stavridis MP, Smith AG. Biochem. Soc. Trans. 2003;31:45. 88. Weimann JM, Charlton CA, Brazelton TR, Hackman RC, Blau HM. Proc. Natl. Acad. Sci. U. S. A. 2003;100:2088. 89. Jiang Y, Jahagirdar BN, Reinhardt RL, Schwartz RE, Keene CD, et al. Nature. 2002;418:41. 90. Herrera DG, Garcia-Verdugo JM, Alvarez-Buylla A. Ann. Neurol. 1999;46:867. 91. Zhang RL, Zhang L, Zhang ZG, Morris D, Jiang Q, et al. Neuroscience. 2003;116:373. 92. Modo M, Mellodew K, Cash D, Fraser SE, Meade TJ, et al. Neuroimage. 2004;21:311. 93. Hoehn M, Kustermann E, Blunk J, Wiedermann D, Trapp T, et al. Proc. Natl. Acad. Sci. U. S. A. 2002;99:16267. 94. Bulte JW, Zhang S, van Gelderen P, Herynek V, Jordan EK, et al. Proc. Natl. Acad. Sci. U. S. A. 1999;96:15256. 95. Jendelova P, Herynek V, Urdzikova L, Glogarova K, Kroupova J, et al. J. Neurosci. Res. 2004;76:232. 96. Yee C, Riddell SR, Greenberg PD. Curr. Opin. Immunol. 2001;13:141. 97. Dodd CH, Hsu HC, Chu WJ, Yang P, Zhang HG. J. Immunol. Methods. 2001;256:89. 98. Kircher MF, Allport JR, Graves EE, Love V, Josephson L, et al. Cancer Res. 2003;63:6838. 99. Zhang Z, Jiang Q, Jiang F, Ding G, Zhang R, et al. Neuroimage. 2004;23:281. 100. Aboody KS, Brown A, Rainov NG, Bower KA, Liu S, et al. Proc. Natl. Acad. Sci. U. S. A. 2000;97:12846. 101. Lee J, Elkahloun AG, Messina SA, Ferrari N, Xi D, et al. Cancer Res. 2003;63:8877. 102. Folkman J. Nat. Med. 1995;1:27. 103. Brown SL, Ewing JR, Nagaraja TN, Swerdlow PS, Cao Y, et al. Magn. Reson. Med. 2003;50:1209. 104. Taupitz M, Schmitz S, Hamm B. Rofo. Fortschr. Geb. Rontgenstr. Neuen. Bildgeb. Verfahr. 2003;175:752. 105. Yeh TC, Zhang W, Ildstad ST, Ho C. Magn. Reson. Med. 1995;33:200.

Part I

Part I , Section 3: Applications in Marine Science

893

Foreword to Applications in Marine Science

Today, strong focus on freshness and high product quality is an essential strategy of fisheries and aquaculture. Both consumer and marked are becoming nowadays increasingly aware of all the dietary benefits of marine foods for human health. Maintaining the quality of marine foods through the whole value chain is one of the most important challenges for this industry. Therefore it is necessary to develop basic knowledge about the product composition, degradation processes, effects of processing on product shelf life as well as effective methods of quality preservation. Much of on-going research in being carried out in this field, where both traditional and more sophisticated techniques are in use. Modern Nuclear Magnetic Resonance (NMR) technique is a unique method that opens up great possibilities to study foods non-destructively and non-invasively in many different ways. In the first part of the section the application of several low field NMR techniques for use in process and quality control is demonstrated. One of them is a newly developed time domain NMR technique allows determining the most important quality parameters of fish feed as protein, carbohydrate, fat and moisture content in one measurement. This application is implemented as an at-line method in several fish feed production plants. Furthermore, the low field NMR techniques can provide with information about the fat and water in fish tissue, and a combination of the free-induction decay and the pulsed gradient spin-echo technique allows simultaneous determination of fat and water content in fatty fish species. It is shown how the structural changes in fish muscle, water binding and distribution within the muscle can be studied by the low field NMR. The available mathematical methods for extracting the information from the NMR relaxation signals are discussed as well. A great potential of the low field NMR as a user-friendly non-destructive method for measuring of important quality attributes in marine foods is demonstrated. The second part of the section deals with the high resolution NMR. It is shown to be a unique tool in lipid research, where many metabolites and components can

be studied simultaneously in relatively short time without any extraction involved. High resolution NMR enables to quantify lipid classes, to obtain fatty acid composition, to study the positional distribution of fatty acids in the triacylglycerol molecule, as well as to study the lipolysis and lipid oxidation in the same sample. Utilization of various Electron Spin Resonance (ESR) techniques is also shown to provide an additional valuable information regarding early stages of lipid oxidation in marine material. The third section presents the use of 1 H and 13 C high resolution magic angle spinning (HR MAS) NMR for quantification of the total omega-3 fatty acids in intact muscle of salmon. Application of this method for metabolic profiling of micro algae and analysis of fatty acids and storage polysaccharide structure in algae is presented as well. Ability to perform spectroscopic NMR investigations on intact cells or muscle allows to obtain a vast amount of information in a truly non-destructive way on, for example, degradation processes or biochemical processes related to fatty acid biosynthesis. Another important practical benefit of the HR MAS NMR is that both 1 H and 13 C experiments can be done on the same sample with only one simple sample preparation. In the fourth section the use of the Magnetic Resonance Imaging (MRI) as research tool in food science is demonstrated. Due to high investment costs, the sheer size of the instruments and the infrastructure needed, MRI can not presently be introduced as a standard analytical tool in aquaculture or fish processing industry. However, as a research tool taking advantage of the unique features of the method, one can obtain basic insight into a number of issues related to anatomical studies, composition and structure of tissues, distribution maps of fat, water and salt as well as temperature profiles. In some cases theoretical transport models can be used to interpret the images, like for example in MRI studies of salting or dehydration. MRI has a great potential for fish industry as a tool to study the effects of feeding regimes during the fish on-growth phase and breeding, optimisation of unit operations in the fish processing such as salting, freezing and thawing.

895

Abbreviations

TD NMR: time domain nuclear magnetic resonance FID: free induction decay CPMG: Carr–Purcell-Meiboom-Gill

ESR: electron spin resonance EPA: eicosapentaenoic acid PCA: principal component analysis Chl a: chlorophyll

NIR: near Infrared Spectroscopy

HR MAS: high resolution magic angle spinning

PARAFAC: parallel factor analysis

FA: fatty acids

PLS: partial least-squares

MRI: magnetic resonance imaging

WHC: water holding capacity

MQF: multiple quantum filter

DSC: different scanning calorimetry

SQ: single quantum

PUFA: polyunsaturated fatty acid

DQF: double quantum filter

n-3: omega-3

ATP: Adenosine 5 -triphosphate

DHA: docosahexaenoic acid

HPLC: High Performance Liquid Chromatography

SFA: saturated fatty acid

DEPT: Distortionless Enhancement by Polarization Transfer

MUFA: mono unsaturated fatty acid EGDM: ethylene glycol dimethyl ether

HETCOR: Heteronuclear Correlation spectroscopy (C,H-Cosy)

HRGC: high resolution gas chromatography

DP: Degree of Polymerization

GC-MS: gas chromatography mass spectrometry

DB: Degree of branching

897

Emil Veliyulin1 , Karl Østerhus2 , Wolfgang Burk3 , Trond Singstad4 , and Tore Skjetne5 1 SINTEF

Fisheries and Aquaculture, 7465 Trondheim, Norway 2 EWOS Innovation, 4335 Dirdal, Norway 3 Bruker Optik GmbH, Silberstreifen, D-76287 Rheinstetten, Germany 4 St.Olavs Hospital, 7465 Trondheim, Norway 5 SINTEF Petroleum Research, 7465 Trondheim, Norway

Introduction The quality of fish feed and the effectiveness of fish feed production process are important issues for both fish feed suppliers and fish farmers. A correct combination of protein, carbohydrate, fat, and moisture contents in fish feed is crucial for achieving a desirable growth rate and other key characteristics of farmed fish. Ability of fast at-line control of the composition of fish feed would give fish feed producers an advantage of more flexible control over the production, resulting in a more effective consumption of energy and raw ingredients, increasing the production speed, and improving the quality of the final product. Ability to do rapid corrections in the ongoing production, based on the at-line analysis results would substantially reduce the amount of rework and low-grade production. At the initial stage of fish feed production, a mixture of raw ingredients is prepared to produce the semi-solid matrix of the pellet. This mixture of raw ingredients mainly contains protein, carbohydrate, moisture, and fat. During the extrusion process, this mixture is compacted under high temperature and pressure to form a pellet with a porous, semi-solid wet matrix. The pellet matrix is then dried and saturated with fish or vegetable oil. Wide variety of well-known analytical procedures currently adopted by feed producers for compositional analysis of fish feed has a number of common disadvantages. These standard chemical–physical tests are usually time consuming, demanding highly trained staff, costly, most of the standard methods are destructive for the sample, often require thorough calibration, many of them utilize dangerous toxic solvents and cannot be performed at the production line. To the contrary modern time domain nuclear magnetic resonance (TD NMR) analyzers offer a wide spectrum of quick, non-destructive, and precise applications. Thanks

Graham A. Webb (ed.), Modern Magnetic Resonance, 897–903.
C 2008 Springer.

to the high automation of modern NMR instruments, most routine tests can be performed by the ground-floor personal. A TD NMR instrument can perform a number of experiments, providing with various types of information about the studied material. This is accomplished by programming and running specific NMR pulse sequences, such as, for example, “free induction decay” (FID), “Hahn echo” [1], CPMG [2], “solid echo” [3], and others. Choice of the NMR pulse sequence depends not only on the type of the information required, but also to a large extent on the physical and chemical properties of the sample. Presence of solid and liquid phases, their mobility, rigidity, and NMR relaxation times are the most important parameters to be taken into consideration. A number of TD NMR applications have recently been developed and adopted for quality control of foods such as rapid determination of fat content in meat [4,5], characterization of fat and water states in cheese [6], prediction of the content of water, oil, and protein in rape and mustard seeds [7], solid fat content analysis [8], monitoring of textural changes in frozen cod [9], and NMR relaxation time studies of intact fish flesh [10] and meat [11].

Experimental Equipment The technique has been implemented on a BRUKER minispec mq10 TD NMR analyzer (Bruker Optik GmbH, Rheinstetten, Germany) using a built-in programming language ExpSpel [12]. The instrument operated at 10 MHz proton frequency and could accommodate sample tubes of 40 mm in diameter. Sample tubes could be filled up to about 4 cm filling height to be measured in the system. This corresponded to about 30 or more pellets depending on their size.

Part I

Comprehensive Compositional Analysis of Fish Feed by Time Domain NMR

898 Part I

Marine Science

Part I

Fish Feed Samples Fish feed pellet is a porous, semi-solid matrix saturated with fish oil. Fish oil content in a pellet typically varies within 16–40 wt.%. The matrix of the pellet consists of a number of chemical compounds depending on the recipe used in the production. The most important of them are protein and carbohydrate, that together usually constitute over 95% of the solid matrix weight. Other typical ingredients of the matrix are ashes, minerals, vitamins, and pigments. In addition the solid matrix of fish feed contains certain amount of rest moisture left after drying of the extruded pellets. Amount of the rest moisture can vary within 5–10 wt.%.

130 ms). The relaxation peaks in Figure 1 were additionally broadened by the processing. Close inspection of the measured relaxation curve indicated that there is no contribution of the moisture signal at times above 6 ms. Thus it was possible to quantify the total amount of hydrogen atoms present in the fish oil by a “Hahn echo” experiment with the echo time of 6 ms. Unlike the “Hahn echo,” the amplitude of the FID signal from fish feed is proportional to the total amount of protons in the liquid phase, i.e. both fish oil and rest moisture. Thus by combining measured amplitudes of the FID and the “Hahn echo” signals both fat and moisture contents could be quantified.

NMR Measurements of Protein and Carbohydrate NMR Measurements of Fat and Moisture In a fish feed pellet, the residual moisture is always strongly bounded to the solid matrix which makes its NMR relaxation time much shorter than that of the oil. This has been confirmed by a T2 relaxation curve measurement by a CPMG technique with the following experimental parameters: echo time (TE) = 0.1 ms, relaxation delay (RD) = 2 s, number of acquired even echoes 8000, and number of scans 32. Figure 1 shows the continuous distribution of the T2 relaxation times that was calculated by inverse Laplace transform algorithm [13] implemented in the BRUKER minispec software. From the last figure, it is seen that the average relaxation time of the residual moisture (peak centered at 3.5 ms) is much shorter than that of fish oil (peak centered at about

Fig. 1. Continuous distribution of the spin–spin relaxation times in fish feed.

Detection of NMR signal from large and immobile molecules like solid proteins has always been a challenging experimental task. These large molecules have extremely short NMR relaxation times and the majority of TD NMR techniques cannot be used for quantitative measurements of solid protein content. To obtain an NMR signal from the solid matrix of fish feed, a “solid echo” technique was used that allowed to overcome the receiver “dead time” problem and recover the full signal amplitude from the sample’s solid part. The corresponding NMR sequence consists of two 90◦ pulses with different phases allowing to observe “solid” echoes from a system with dipole–dipole coupled protons [14]. If the sample contains both solid and liquid phases, the “solid echo” pulse sequence will in addition generate a FID-like tail following the “solid echo” signal [15]. This is schematically illustrated in Figure 2. In a fish feed production up to five main raw ingredients (fish meal, wheat meal, wheat gluten, soybean meal, and maize gluten) are mixed together in various proportions. Chemical composition of each raw ingredient of fish feed is always known from the preliminary wet chemical analysis. The total amplitude of the observed “solid echo” is a measure for the total amount of all protons in both solid and liquid phases of the sample. The difference between the total “solid echo” amplitude and the FID amplitude originates only from the solid part of the sample as indicated in Figure 2. The solid matrix of fish feed consists mainly of proteins, carbohydrates, ash ( 1/2 systems, 443 Saccharomyces cerevisiae Į-factor receptor, 338 Salmon. See also atlantic Salmon (Salmo salar), low-field NMR studies salt and fat distribution, 968 salt interaction in smoked, 967 salt analysis, cod fillets, 960, 961 and fat distribution in smoked Salmon, 968 interaction in smoked Salmon, 967 salty, 967 Samia cynthia ricini, See also Bombyx mori silk, 97 structure before spinning, 101 saturated fatty acids. See SFA saturation transfer difference (STD), 319, 505 across hydrogen bonds, TROSY for obeserving, 489 scaling factors (CRAMPS), 369 scaling under pulse sequences (CRAMPS), 366 scanning transmission electron microscopy. See STEM SCULPTOR package, 666 SDC implementation by pulsed ENDOR, 647 SE and GE BOLD imaging, 858 SEC–NMR. See on-line SEC–NMR second order polarization propagator approximation, 64 secondary structure, 335 bacteriorhodopsin, 336 bovine rhodopsin, 337

human erythrocyte glycophorin, 336 lactose permease, 338 secondary structure analysis (proteins) angle measurements, 736 anisotropic nuclear magnetic interactions, 736 backbone torsion angle analysis of peptides, 739 backbone torsion angles, 735 characterization of, 735 chemical shifts dependence, 735 distance measurements, 736 J coupling dependence, 735 NMR for, 735 torsion angle measurements, 736 SEDOR. See spin-echo double resonance seeds. See plant seeds seizures generalized, 807 simple partial, 807 selective solute interactions in mixed solvent systems, 417 SEMA BB-SEMA, 704 very high RF power requirement, 703 semiconductors, organic. See organic semiconductors sensitivity, 17 O MQMAS NMR, 695 issues in NMR, 353 sensitivity enhancement by cross-polarization, 469 low-Ȗ nuclear quadrupole resonance, 446 separated local field (SLF) experiment, 703 SERCA, 313 SFA, 915 SGP condition, 109 shielding constants, ab initio calculation and, 63 shielding tensor, 63, 390 ship-in-a-bottle type synthesis, 148 short gradient pulse condition, 109 side chain of PBLG, 623 signal assignments, SSNMR, 468 signal-to-noise. See SNR silanols, hydrogen-bonded, 197 silica gels, 199 silicates, 199 silk fibroins 1 H CRAMPS NMR, 595 1 13 2D H– C HETCOR NMR, 599 2D NMR, 611 HETCOR NMR, 597 Bombyx mori, 101 structure after spinning, 101 structure before spinning, 101 conformation study of, 593 Nephila clavipes Dragline, 104

Subject Index

silk fibroins (cont.) Samia cynthia ricini, 101 structure before spinning, 101 structural analysis, 101 simple partial seizures, 807 simulation polymer blends, 638 simultaneous frequency amplitude modulation (SFAM), 722 SIn, PISEMA experiments of, 703 single photon emission computed tomography (SPECT), 855 site-directed 13 C NMR, membrane proteins (monomers), 285 assignment,13C NMR signals, 291 NMR studies on membrane proteins, 291 size exclusion chromatography (SEC), 399 SLICING, 902 slow exchange, 421. See also chemical exchange small molecule interactions small molecule-organic solvent interactions, 409 small molecule–water interactions, 416 small molecule-organic solvent interactions, 409 small molecule–water interactions, 416 small organic molecules RDC and, 657 smart contrast agents, 792 smart gels characterization, 125 SNR, 759, 760, 761 brain disease, experimental models (MRI-based), 787 MRI contrast optimization, 759 sodium hydrous silicates, layered, 200 solid polymers 13 C NMR, 567 NMR chemical shifts, 43 nuclear shielding, 43 solid state polypeptides in, 591 2 solid state H NMR polypeptides dynamics, 621 19 solid-state F-NMR analysis of oriented biomembranes, 261 experimental aspects, 261 solid-state deuterium coupling tensors observation, 249 membrane lipids atudy and, 249 NMR spectroscopy of, 246 solid-state heteronuclear NMR polymer interfaces characterization and, 575 solid-state MAS NMR crystalline polymers, 615

solid-state multiple-quantum NMR disordered polymers, conformations study of, 563 solid-state NMR, 87 13 C, 567 13 1 H and C, 467 analytical and numerical tools for experiment design, 679 aspects of, 389 tools, 679 average Hamiltonian theory (AHT), 679 cross-polarization (CP) MAS, 89 effective Hamiltonian Theory (EHT), 679 experimental aspects, 473 heterogeneous catalysts study, 205 high-resolution, 467 in host–guest chemistry, 147 isotropic shielding uses in, 389 ligands and membrane-embedded receptors interactions, 319, 321 membrane-active proteins and peptides, 287 membrane-associated peptides, 263 membrane-binding proteins structure and, 299 numerical tools, 679 phospholamban, 313 PLC-į1 PH domain and, 300 polymer blends, 631, 633 polymer interfaces characterization, 575 polymer, applications of, 575 protein structure determination, 517 quadrupolar coupling, 391 quantification with, 392 signal assignments and multi-dimensional, 472 structure determination and, 392 systematic design of, 682 tools for systematic experiment design, 679 torsion angle determination, 727 solid-state proton NMR polymer interfaces characterization and, 575 solute–solvent interactions, NOE and, 409, 410 biomolecule–water interactions, 418 cross-relaxation intermolecular, 412 intermolecular cross-relaxation, determination of, 413 micelle–water interactions, 416 selective solute interactions in mixed solvent systems, 417 small molecule-organic solvent interactions, 417

small molecule–water interactions, 416 xenon–solvent interactions, 416 solution dynamics, NMR diffusometry and, 109 solution NMR methods ligands and membrane-embedded receptors interactions (NMR study), 319 solvent systems, mixed, 417 SOPPA. See second order polarization propagator approximation SOS-DFPT. See sum-over-states-density functional perturbation theory spatial information NMR measurements and, 129 spatially encoded single-scan NMR, 347, 349 spatially-resolved degradation, ESR imaging and, 179 specific enrichment, 349 spectral analysis 17 O MQMAS NMR, 695 spectral assignments polymer microstructure and, 567 spectral densities, lipid membranes, 254 spectral-spatial ESRI, 179 spectrometer requirements (CRAMPS), 367 spherical-basis treatment, protein structure determination, 517 SPI methods, 168 spin 1/2 nuclei, structural information, 148 spin counting, 150 spin density map, 159 spin diffusion, 410, 548 spin Hamiltonian EPR principles, 439 spin relaxation amorphous polymers dynamics, 607 spin space average Hamiltonians in, 363 spin system, multiple, 719 of cod liver oil, 930 spin-echo double resonance, 150 spin–lattice relaxation dioxygen in solution, paramagnetic effects of, 467 polymer structure, 539 spin–lattice relaxation times, 607 deuterium, 249 spinning sidebands, dipolar, 595 spinning speed dependence 13 C MAS spectra, 469 1 and sensitivity, H MAS NMR, 467 spin–orbit interaction EPR principles, 439 spin–spin couplings, nuclear. See nuclear spin–spin couplings spin–spin relaxation polymer structure and, 539

Subject Index

spin–spin relaxation times, 607 spin–spin terms, higher order, 443 SPIO and gadolinium chelates, in brain perfusion images, 765 in detecting tumor necrosis (apoptosis), targeted, 766 spontaneous magnetic orientation, oriented bilayer systems and, 241 spontaneous relaxation, 409 spreading depression (SD), 806 SPRITE methods, 168 SQUIDs, 449 SSNMR. See solid-state NMR stable isotope labeling glycoproteins, 285 stable isotopic labeling RNA and DNA, 671 stable-isotope-aided methods for protein NMR spectroscopy, 215 stable-isotope-assisted NMR spectroscopy structural glycobiology, 223 static tensor correlation techniques, 727 statistical analysis phMRI, 855 STEAM, 776, 777 steel-making process NMR imaging, 163 STEM Alzheimer’s Aȕ amyloid fibrils, polymorphism, 19 stem cells tracking CNS, 886 heart, 885 stereoregularity, polymeric, 541, 542 stereoregularity, polymers methyl acrylate (A)/methyl methacrylate (B) copolymer, 558 poly(methyl methacrylate), 608 stereosequences, PP, 567 stimulated relaxation, 409 STMAS Stokes–Einstein equation, 135 structural analysis biomembranes, 261 ligands and membrane-embedded receptors interactions (NMR study), 319 membrane-associated peptides, 263 nucleosides, 668 of silk fibroins, 101 secondary structure (proteins), 731 structural biology DNA, 658 NMR in, 653 oligosaccharides and small organic molecules, 661 protein structures, 658 residual dipolar couplings and, 657

RNA, 658, 659 unfolded denatured proteins, 661 structural characterization (inorganic materials), 197 structural fitting algorithm, protein structure determination, 517 structural fitting, protein structure determination, 517 fitting to 2D spectra, 576 fitting to 3D spectra, 517 structural glycobiology stable-isotope-assisted NMR spectroscopy, 218 structural stability of calcium-binding lysozyme, 497 structural studies biological solids, 700 structure determination DNA and RNA, 671 solid-state NMR and, 389 sugar chains, in vitro labeling, 213 sum-over-states-density functional perturbation theory (SOS-DFPT) approach, 67 superparamagnetic agents for cell tracking, 880 superparamagnetic iron oxide (SPIO), 880 surface acidity heterogeneous catalysts, 205 surface coils surface structures membrane proteins, 289 surfactants, NMR diffusometry and, 109 swelling process, 187. See also hygrogels, NMR imaging of synthetic polypeptides, 591 T T1 and T2 contrast, 761 T1 relaxation behavior, 13C, 546 T1 vs. T2 contrast agents, 764 T1-weighted dynamic contrast enhanced MRI (brain disease), 790 T1-weighted imaging, 787 T1ȡ-weighted imaging, 789 T2 relaxation times polymer networks structure, 584 T*2-weighted imaging, 788 T2-weighted imaging, 788 tachykinin NK-1 receptor, 339 tacticity, molecular weight dependence of, 401 TANSEMA, 704 low RF power experiments, 704 targeted SPIO in detecting tumor necrosis (apoptosis), 766 TB MO methods, 45 configuration aspects, 46 conformation aspects, 45

crystal structure aspects, 48 telomeric DNA, 743 telomeric DNA bound to hTRF2, 745 temperature imaging, 169 temporal resolution, 759 tensor correlation techniques MAS, 727 tensor correlation techniques, static, 727 tensor, shielding, 65 tensors chemical shielding, 67 NMR chemical shielding, 66 shielding, 390 tetramethylsilane, 140 thermosetting resins, 583 thermotropic applications, liquid crystals, 120 tight-binding (TB) molecular orbital (MO) theory. See TB MO methods time dependence of NOE, 411 time domain NMR, 893 tissue signature modeling, 805 tissue-specific contrast agents, 764 TOAC, 316 TORQUE experiments, 199 torsion angle analysis backbone, 736, 739 peptides, 739 torsion angle determination distance methods, 732 MAS tensor correlation techniques, 727 simultaneous (ij,ȥ) determination, 731 solid-state NMR, 727 static tensor correlation techniques, 727 ij angle techniques, 729 ȥ angle techniques, 730 torsion angle measurements, secondary structure analysis (proteins), 738 mutual orientation of anisotropic interactions, 738 torsion angles, 277 transfection agents, 881 transformation, crystalline polymers, 615 trans–gauche isomerizations, 253 transient NOE, 299 translational diffusion, 135 translational motion, 152 transmembrane helices, 338 transmembrane protein, 338 transverse aortic constriction (TAC), 835 transverse relaxation-optimized spectroscopy, 489 transverse relaxationoptimized spectroscopy (TROSY) technique, 218 TRAPDOR, 150 traumatic brain injury (TBI), 812

Subject Index

triplet instability, 82 triplet instability of CHF, 82 tritiation procedures, 395 tritium NMR spectroscopy, 396 TRNOE measurement, 299 tropoelastin, 93 TROSY 13 1 [ C, H] correlation experiments, 491 15 1 2D [ N, H], 491 and CRINEPT, 494 and CRIPT, 494 applications to nucleic acids, 493 for large biological macromolecules study, 490 for observation of scalar couplings across hydrogen bonds, 493 for RDCs measurements, 493 for resonance assignments in large molecules, 491 for studies of dynamic processes, 493 in NOE, 493 protein backbone resonances, 491 protein side-chain resonances, 491 T-Stimulus, 190 tumor angiogenesis, gadolinium agents in evaluating, 766 tumor necrosis (apoptosis), targeted SPIO in detecting, 766 tumors biology and physiology, 823 cell-based tumor therapy, 888 hypoxia, 823 lipid metabolism in, 827 pH, 826 phospholipid metabolism in, 826 U ultra small superparamagnetic iron oxide (USPIO), 880

unfolded denatured proteins, RDC and, 657 USPIO, 880, 881 V vascular MRI, 848 velocity component, 159 velocity component mapping sequences, 160 very fast MAS paramagnetic effects under, 475 very large structures, cross-correlated relaxation-induced polarization transfer for studies of, 494 vesicle systems, magnetically oriented, 243 VFMAS. See very fast MAS viable myocardium after infarction, manganese-based contrast in determining, 768 voltage-gated potassium channel, 339 volume phase transition, 187 vulcanization, 584 W water analysis atlantic Salmon (Salmo salar), 903 muscle after heating, 902, 904 low-field NMR studies, 901 atlantic salmon fillets, 960, 961 cod fillets, 960, 961 water and solute interactions biomolecule–water, 418 micelle–water, 416 small molecule–water, 416 water diffusion inside gels, 192 water holding capacity, 913 waterLOGSY technique, 320

WAXD, 563 weighted imaging T1, 787 T1ȡ, 790 T2, 788 T2*, 788 whole body phenome, 771 wide line separation spectroscopy (WISE), 580 wide-angle X-ray diffraction, 563 X xenon gas, absorbed, 639 xenon–solvent interactions, 416. See also solute–solvent interactions, NOE and XPLOR package, 657 Z Zeeman interaction EPR principles, 439 Zeeman interaction, nuclear, 439 Zeeman splitting, 353 zeolites, 125 zero field NMR, 445 demagnetization and initiation of evolution, 446 experimental aspects, 446 extensions of, 450 NMR and NQR, 450 NMR and NQR, limitations of, 482 signal evaluation, 445 zero field signal detection, 448 zero field spin Hamiltonian, 444 zero magnetic field NMR and NQR in, 445 zero-quantum experiments, disordered polymers study, 563 zeutmatography, 179

Modern Magnetic Resonance Part 2

Part 1: Applications in Chemistry, Biological and Marine Sciences Part 2: Applications in Medical and Pharmaceutical Sciences Part 3: Applications in Materials Science and Food Science

Modern Magnetic Resonance Part 2: Applications in Medical and Pharmaceutical Sciences Graham A. Webb (Ed.) Royal Society of Chemistry, London, UK

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-1-4020-3894-5 (HB) ISBN 978-1-4020-3910-2 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

www.springer.com

Printed on acid-free paper First published in 2006 Reprinted with corrections in 2008

All rights reserved.
C 2008 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

V

Contents

List of Section Editors ........................................................................................................... Preface ............................................................................................................................... Color Plate Section................................................................................................................

VII IX XI

APPLICATIONS IN MEDICAL SCIENCES Foreword ............................................................................................................................. Abbreviations ....................................................................................................................... Glossary of Terms ..................................................................................................................

975 977 981

Acquiring Neurospectroscopy in Clinical Practice, Alexander P. Lin .............................................. Application of Magnetic Resonance for the Diagnosis of Infective Brain Lesions, Uwe Himmelreich and Rakesh K. Gupta........................................................................................................ Application of 2D Magnetic Resonance Spectroscopy to the Study of Human Biopsies, Edward J. Delikatny and June Q.Y. Watzl............................................................................... Correlation of Histopathology with Magnetic Resonance Spectroscopy of Human Biopsies, Carolyn Mountford, Ian C.P. Smith, and Roger Bourne .................................... Functional MRI, Graeme D. Jackson, Regula S. Briellmann, Anthony B. Waites, Gaby S. Pell, and David F. Abbott High Resolution Magic Angle Spinning (HRMAS) Proton MRS of Surgical Specimens, Leo L. Cheng, Melissa A. Burns, and Cynthia L. Lean................................................................................... Intraoperative MRI, Isabelle Latour and Garnette R. Sutherland ................................................... In Vivo Magnetic Resonance Spectroscopy in Breast Cancer, Uma Sharma and N.R. Jagannathan........ In Vivo Molecular MR Imaging: Potential and Limits, Mathias Hoehn and Uwe Himmelreich............... In vivo 13 C MRS, Stefan Bluml............................................................................................. Magnetic Resonance Spectroscopy and Spectroscopic Imaging of the Prostate, Breast, and Liver, Mark G. Swanson, Susan M. Noworolski, and John Kurhanewicz.................................................... MR-Mammography, W.A. Kaiser ............................................................................................ Phosphorus Magnetic Resonance Spectroscopy on Biopsy and In Vivo, Geoffrey S Payne, Nada Al-Saffar, and Martin O Leach ......................................................................................................... Radio Frequency Coils for Magnetic Resonance Spectroscopy, Boguslaw Tomanek........................... Spatially Resolved Two-Dimensional MR Spectroscopy in vivo, M. Albert Thomas, Amir Huda, Hyun-Kyung Chung, Nader Binesh, and Talaignair Venkatraman....................................................

985 1005 1015 1027 1037 1051 1065 1077 1087 1099 1113 1127 1143 1163 1171

APPLICATIONS IN PHARMACEUTICAL SCIENCE Foreword ............................................................................................................................. 1187 Key Terms ........................................................................................................................... 1189 Overview of NMR in the Pharmaceutical Sciences, Horst Joachim Schirra and David J. Craik............... 1195 Instrumentation Applications of Cryogenic NMR Probe Technology for the Identification of Low-Level Impurities in Pharmaceuticals, Gary E. Martin ...................................................................... Flow NMR Techniques in the Pharmaceutical Sciences, Paul A. Keifer........................................... Developments in NMR Hyphenation for Pharmaceutical Industry, Manfred Spraul .......................... LC-NMR in Dereplication and Structure Elucidation of Herbal Drugs, Gloria Karagianis and Peter G. Waterman...........................................................................................................

1203 1205 1213 1221 1229

VI

Contents

Techniques New Approaches to NMR Data Acquisition, Assignment and Protein Structure Determination: Potential Impact in Drug Discovery, Antonio Pineda-Lucena ..................................................... Transferred Cross-Correlated Relaxation: Application to Drug/Target Complexes, T. Carlomagno and C. Griesinger................................................................................................................... Novel Uses of Paramagnets to Solve Complex Protein Structures, R. Andrew Byrd, C. Andrew Fowler, Robert L. McFeeters, and Vadim Gaponenko............................................................................ Fast Assignments of 15 N-HSQC Spectra of Proteins by Paramagnetic Labeling, Guido Pintacuda, Thomas Huber, Max A. Keniry, Ah Young Park, Nicholas E. Dixon, and Gottfried Otting........................ Phospholipid Bicelle Membrane Systems for Studying Drug Molecules, Jianxin Guo, Xiaoyu Tian, Spiro Pavlopoulos, and Alexandros Makriyannis........................................................................ Partial Alignment for Structure Determination of Organic Molecules, Burkhard Luy and Horst Kessler Measurement of Residual Dipolar Couplings and Applications in Protein NMR, Keyang Ding and Angela M. Gronenborn...................................................................................................... Using Chemical Shift Perturbations to Validate and Refine the Docking of Novel IgE Antagonists to the High-Affinity IgE Receptor, Melissa A. Starovasnik and Wayne J. Fairbrother ....... Dual-Region Hadamard-Encoding to Improve Resolution and Save Time, Ronald Crouch .................. Nonuniform Sampling in Biomolecular NMR, Mark W. Maciejewski, Alan S. Stern, Glenn F. King, and Jeffrey C. Hoch.......................................................................................................... Applications Structural Characterization of Antimicrobial Peptides by NMR Spectroscopy, Leonard T. Nguyen, Elmar J. Prenner, and Hans J. Vogel...................................................................................... Pharmaceutical Applications of Ion Channel Blockers: Use of NMR to Determine the Structure of Scorpion Toxins, Muriel Delepierre and Lourival D. Possani ......... Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β-Lactamases, Christian Damblon and Gordon C.K. Roberts................................................................................................... NMR Spectroscopy in the Analysis of Protein–Protein Interactions, David A. Gell and Joel P. Mackay.. Identification and Characterization of Ternary Complexes Using NMR Spectroscopy, Robert E. London The Transferred NOE, Mike P Williamson................................................................................ NMR Kinetic Measurements in DNA Folding and Drug Binding, Mark S. Searle, Graham Balkwill, Huw E.L. Williams, and Evripidis Gavathiotis ........................................................................... The Use of NMR in the Studies of Highly Flexible States of Proteins: Relation to Protein Function and Stability, Søren M. Kristensen, Marina R. Kasimova, and Jens J. Led.......................... NMR-based Metabonomics Techniques and Applications, John C. Lindon, Elaine Holmes, and Jeremy K. Nicholson......................................................................................................... Protein Misfolding Disease: Overview of Liquid and Solid-State High Resolution NMR Studies, Harald Schwalbe and Julia Wirmer........................................................................................ 19 F NMR Spectroscopy for Functional and Binding High-Throughput Screening, Marina Veronesi and Claudio Dalvit................................................................................................................ Applications of Receptor-Based NMR Screening in Drug Discovery, Philip J. Hajduk....................... NMR SHAPES Screening, Christopher A. Lepre and Jonathan M. Moore ............................................ NMR-Based Screening Applied to Drug Discovery Targets, Jennifer J. Gesell, Mark A. McCoy, Mary M. Senior, Yu-Sen Wang, and Daniel F. Wyss .................................................................... NMR and Structural Genomics in the Pharmaceutical Sciences, Maša ýemažar and David J. Craik ....... Author Index Subject Index

1237 1239 1247 1255 1263 1271 1279 1287 1293 1299

1305 1313

1315 1325 1331 1339 1347 1357 1363 1369 1377 1387 1393 1401 1409 1419 1429

VII

List of Section Editors Subject

Name

Type of Editor

Graham A. Webb

Editor-In-Chief

Chemistry

Hazime Saitˆo Himeji Institute of Technology and QuLiS, Hiroshima University Japan e-mail: [email protected] Isao Ando Department of Chemistry and Materials Science Tokyo Institute of Technology, Ookayama, Meguro-ku Tokyo 152-0033 Japan e-mail: [email protected] Tetsuo Asakura Department of Biotechnology Tokyo University of Agriculture and Technology Koganei, Tokyo 184-8588 Japan e-mail: [email protected]

Section Editors

Biological Sciences

Jimmy D. Bell Molecular Imaging Group, MRC Clinical Sciences Centre Hammersmith Hospital Campus, Imperial College London London, W12 OHS UK e-mail: [email protected]

Section Editor

Marine Science

M. Aursand SINTEF Fisheries and Aquaculture Ltd N-7465 Trondheim Norway e-mail: [email protected]

Section Editor

Medical Science

Carolyn Mountford Institute for Magnetic Resonance Research, and Department of Magnetic Resonance in Medicine University of Sydney, PO Box 148, St Leopards, 1590, NSW Australia email: [email protected] Uwe Himmelreich, PhD Max-Planck-Institute for Neurological Research In vivo NMR Group, Gleueler Str 50, Cologne, D-50931 Germany email: [email protected] Deborah Edward Pageturner (Editing & Research Services)

Section Editor

Assistant Editor

Subject

Name

Type of Editor

Pharmaceutical Science

David Craik Institute for Molecular Bioscience University of Queensland Brisbane 4072, Queensland Australia e-mail: [email protected]

Section Editor

Materials Science

Marcel Utz University of Connecticut, 97 N Eagleville Rd Storrs CT 06269-3136 e-mail: [email protected]

Section Editor

Food Science

Peter Belton School of Chemical Sciences and Pharmacy University of East Anglia Norwich NR4 7TJ, UK e-mail: [email protected]

Section Editor

IX

Preface

It is a great pleasure for me to Introduce the handbook of Modern Magnetic Resonance, MMR. The various techniques which comprise MMR derive essentially from three sources, all of which were produced by physicists. Today they are widely used by scientists working in many diverse areas such as chemistry, biology, materials, food, medicine and healthcare, pharmacy and marine studies. The first source of MMR studies is nuclear magnetic resonance, NMR. This provides details on the relative positions of nuclei, i.e. atoms, in a molecule. Consequently NMR provides structural information on samples which may be in the solid, liquid or gaseous state. Nuclear relaxation data yield dynamic information on the sample and the topology of the dynamic processes if the sample is undergoing a molecular change. Thus high and low resolution NMR studies provide information on all interesting aspects of molecular science. The protean nature of NMR is reflected in its many applications in chemistry, biology and physics which explore and characterize chemical reactions, molecular conformations, biochemical pathways and solid state materials, to name a few examples. Magnetic resonance imaging, MRI, is the second source of MMR data. MRI provides a three-dimensional image of a substance, and is consequently widely employed to assess materials both in vitro and in vivo. The importance of MRI studies in many areas of science and

medicine is shown by the recent award of the Nobel Prize to Lauterbur and Mansfield. The third source of MMR results is due to electron spin resonance, ESR. This is a technique for detecting unpaired electrons and their interactions with nuclear spins in a given sample. Thus ESR data are often used to complement the results of NMR experiments. Taken together NMR, MRI and ESR comprise the field of MMR, recent years have witnessed the fecundity of these techniques in many scientific areas. The present three volumes cover applications in most of these areas. Part 1 deals with Chemical Applications, Biological and Marine Sciences. Medical and Pharmaceutical Sciences are covered in Part 2. Part 3 provides examples of recent work in the Materials Science and Food Science. I wish to express my gratitude to all of the Section Editors and their many contributors for their hard work and dedication in the creation of MMR. My thanks also go to Emma Roberts and the production staff at Springer, London, for their assistance in the realization of these volumes. Royal Society of Chemistry Burlington House Piccadilly London, W1J OBA

G.A.WEBB February 2005

Please note that authors used either British or American English spelling depending on the language of their choice for individual papers.

Graham A. Webb (ed.), Modern Magnetic Resonance, IX.
C 2008 Springer.

Part II

Color Plate Section

Part II

Plate 82. See also Figure 15 on page 1002. A. DCIS with micro invasion

CH2

CH3

B. DCIS without micro invasion

CH2

CH3

4

3

2

1

0

Plate 83. See also Figure 2 on page 1031.

Plate 84. See also Figure 3 on page 1043.

XIII

Part II Plate 85. See also Figure 6 on page 1045.

Plate 86. See also Figure 8 on page 1046.

XIV

Part II

Plate 87. See also Figure 10 on page 1048.

XV

Part II

A Cho PA Cr Citrate

PPM

c

3.0

2.5

2.0

4.0

Myo-Inositol

S-Inositol

b

d Acetone (impurity)

3.5

Lactate

GPC+PC

Cr

3.0

Glu/Gln

PA

Alanine

Cho

Taurine

Lactate Myo-Inositol

a

Citrate

2.5

2.0

1.5

Lipids

1.0

PPM

B

Plate 88. See also Figure 6 on page 1059.

Glucose Choline choline kinase

G-3-P Glycerol

PC

GPC PA

PC cytidylyl transferase

PLA1

PLD

Deg r

PL

C

lysoPL

LysoPtdcho

1,2-DAG

CDP-choline TAG

Ptdcho

PC transferase

C

Plate 90. See also Figure 3 on page 1146. Plate 89. See also Figure 3 on page 1094.

XVI

Syn t hes i s

ada t i on

DHAP

Part II

4.0

A

3.0

mI-Ch

NAA

F1 (ppm)

2.0

mI

1.0

GIu

Lac

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

F2 (ppm)

4.0

B

3.0

mI-Ch

2.0

F1 (ppm)

mI

NAA

1.0

GIu

Lac

4.5

4.0

3.5

3.0

2.5 F2 (ppm)

Plate 91. See also Figure 11 on page 1181.

XVII

2.0

1.5

1.0

Part II

a)

φ3

b)

Plate 92. See also Figure 2 on page 1223.

Plate 94. See also Figure 3 on page 1251.

b f

f

Plate 93. See also Figure 4 on page 1225. Plate 95. See also Figure 4 on page 1253.

XVIII

Part II

Plate 96. See also Figure 3 on page 1260.

Plate 97. See also Figure 4 on page 1266.

Plate 98. See also Figure 1 on page 1294.

XIX

Part II Plate 99. See also Figure 3 on page 1296.

C

C

A

B C

N

N

N

C Cys9

N

Cys8

C N C

Glu8 C N N

C

C

N

Plate 101. See also Figure 2 on page 1317.

Plate 100. See also Figure 1 on page 1317.

XX

Part II

Plate 102. See also Figure 3 on page 1318.

101

T150 G36

G148 G167

106

S57 S190 G179

D152

111

G211

G214

G209

W59

V170

V155 S41 L178

116

R91

H86

D90

I188

N118

L169 S58

N

L220 A184

D56

121 W189

E30 L31

126

C168

V207

L21 F34 V39 I84

131

K171

H149

11.1

10.6

10.1

9.6

9.1

8.6

8.1

7.6

H

Plate 103. See also Figure 1 on page 1333.

XXI

7.1

6.6

6.1

5.6

Part II Plate 104. See also Figure 1 on page 1364.

Plate 105. See also Figure 1 on page 1370.

Plate 106. See also Figure 3 on page 1372.

XXII

ph 2.7

Plate 107. See also Figure 4 on page 1373.

Plate 108. See also Figure 6 on page 1374.

XXIII

Part II

ph 7.0

Part II

A)

B)

New-site

H-bond acceptor or Hydrophobic H-bond acceptor

10.0

ATP-site

14.0

12.0

Site

0.6

0.3

0.0

1H

0.6

(ppm)

Plate 110. See also Figure 4 on page 1413.

0.0

0.3

Plate 109. See also Figure 2 on page 1403. 15

N-HSQC w/ocompounds w/compounds

Thr40 Gly90

Plate 111. See also Figure 1 on page 1420. (A)

(B)

m

m

Plate 112. See also Figure 3 on page 1425.

XXIV

Part II

(A)

Hits to TxGx motif:

Hits near NTP binding site :

(B)

(C)

Plate 113. See also Figure 4 on page 1426.

XXV

Hits near motif IV:

Part II Plate 114. See also Figure 1 on page 1432.

XXVI

Part II

Part II, Section 1: Applications in Medical Science

975

Foreword to Applications in Medical Science

Magnetic resonance (MR) remains an emerging trend in technological innovation. In support is the award of four Nobel Prizes since 1991. Magnetic resonance has provided a paradigm shift in medical diagnosis and treatment planning and now plays an ever increasing role in health and medical research. Magnetic resonance imaging (MRI) is used worldwide with magnetic resonance spectroscopy (MRS), functional MRI and intraoperative MRI increasingly used in the clinic and for medical research purposes. Each is technically difficult and requires a determined effort by those who develop, and use, the methods to obtain reproducible output at multiple sites. In the case of medical applications multidisciplinary teams are needed to develop, test and implement the method for routine clinical use in a controlled environment. In this volume the authors, researchers from around the globe, have documented their technology in a way that will allow others to implement the methods for either

research or clinical use. They share their many years of experience with the readers. We would like to thank them for taking the time to write their contributions with the end-user in mind. We hope that the field of magnetic resonance in medicine will benefit from this volume by sharing the various experimental protocols and details. This will ultimately translate into improved medical care. We would like to thank Professors Robert Lenkinski, Brian Tress, Drs Roger Bourne, Peter Malycha, Alan McLaughlin, Saadallah Ramadan, Robert Savoy and Peter Stanwell for their assistance reviewing the chapters. Our special thanks to Dr Deborah Edward for the large amount of time and effort she put into editing and collating the Chapters. Thanks also to Sinead Doran for assistance with the diagrams. Carolyn Mountford D. Phil (Oxon) Uwe Himmelreich Ph.D.

977

Abbreviations

(HR)MAS: (High Resolution) Magic Angle Spinning

EPI: echo planar imaging

(S)SB: (shifted) sine bell

EPSE: Echo-Planar Spin-Echo

1D: One-Dimensional

FDA : Federal Drug Administration

1

FID: Free Induction Decay

H: MRS Proton Magnetic Resonance Spectroscopy

2D: Two Dimensional

fMRI: functional magnetic resonance imaging

31P MRS: phosphorus magnetic resonance spectroscopy

FNAB: Fine Needle Aspiration Biopsy

3D: Three-dimensional

FOV: Field of View

Ac: Acetate

FWHM: full width at half maximum

ADC: apparent diffusion coefficient

GABA: γ -amino butyric acid

AQ: acquisition time

GAMMA: General Approach to Magnetic Resonance Mathematical Analysis

Asp: Aspartate B1: Magnetic Field of RF Coil

GB: Lorentzian-Gaussianlineshape transformation fraction

BASING: Band Selective Inversion with Gradient Dephasing

GBM : Glioblastoma Multiforme

BIR: B1-Insensitive Rotation BISTRO: B1-Insensitive Train to Obliterate Signal BPH: Benign Prostatic Hyperplasis CABINET: Coherence transfer based spin-echo spectroscopy CE-MRI: contrast enhanced magnetic resonance imaging CHESS: chemical shift selective CHESS: Chemical Shift Selective Saturation CHESS: Chemical Shift Selective Suppression COSY: correlated spectroscopy CSF: Cerebral Spinal Fluid CSI: Chemical Shift Imaging CT-COSY: Constant time-based COSY CT-PRESS: Constant time-based PRESS DCIS: Ductal Carcinoma In Situ DQF: double-quantum filtered DWI: diffusion weighted imaging

Glc: Glucose Gln : Glutamine Glu: Glutamate GM: grey matter HCA: Hierarchical Cluster Analysis HCO3-: Bicarbonate HDR: haemodynamic response function HMQC: heteronuclear multiple-quantum correlation HPLC: High Pressure Liquid Chromatography HR-MAS: High Resolution Magic Angle Spinning HSQC: heteronuclear single-quantum correlation ISI: inter-stimulus interval ISIS: Image Selected In Vivo Spectroscopy J: scalar coupling constant KD: Ketogenic Diet LASER: Localization by Adiabatic Selective Refocusing LB: Lorentzian line broadening LCModel: Linear Combination of Model Spectra

978 Abbreviations

Medical Uses

L-COSY : Localized Correlated Spectroscopy

SAR: Specific Absorption Rate

LDA: Linear Discriminant Analysis

SCS: Statistical Classification Strategy

LG: Lorentzian Gaussian

SECSY: Spin-echo Correlated Spectroscopy

MAS: magic angle spinning

SI: Spectroscopic Imaging

MM: magnitude mode

SNR: Signal-to-Noise Ratio

MR: magnetic resonance

SPM: statistical parametric mapping

MR(S): magnetic resonance (spectroscopy)

SS: Slice-Selective

MRI: magnetic resonance imaging

SSB: Spinning Side Bands

MRS: Magnetic Resonance Spectroscopy

STEAM: STimulated Echo Acquisition Mode

MRSI: magnetic resonance spectroscopic imaging

STEAMCSI: Multi-voxel based STEAM

NAFLD: Non-Alcoholic Fatty Liver Disease

STEAMSV: single-voxel based STEAM

NASH: Non-Alcoholic Steatohepatitis

STIR: Short Time Inversion Recovery

NE: number of experiments

SVS: single voxel spectroscopy

NOE: Nuclear Overhauser Effect

T1 : spin-lattice or longitudinal relaxation time

NOE: Nuclear Overhauser Enhancement

T1: Longitudinal (Spin-Lattice) Relaxation Time

NS: number of transients

T2 : spin-spin or transverse relaxation time

NVG: Noun Verb Generation

T2 : Transverse (Spin-Spin) Relaxation Time

OLR: Orthographic Lexical Retrieval

TCA: Tricarboxylic Acid

OSIRIS: outer volume suppressed ISIS

TE : echo-time

OVS: Outer Volume Suppression

TOCSY: total correlated spectroscopy

PA: Polyamines

TPPI: time-proportional phase incrementation

PCA: Principal Component Analysis

TR : repetition time

PCG: posterior cingulated gyrus

TRUS: Transrectal Ultrasound

PDE: phosphodiester

U-FLARE: Ultra-fast Low Angle Rapid Acquisition with Relaxation Enhancement

PME: phosphomonoester PPM: Parts-Per-Million PRESS: Point Resolved Spectroscopic Sequence PRESS: Point Resolved Spectroscopy PRESSCSI: Multi-voxel based PRESS PRESSSV: single-voxel based PRESS PS: phase-sensitive PTC: Phosphatidylcholine PWI: Perfusion Weighted Imaging RF: Radio Frequency

VOI: Volume Of Interest VOSY: Volume Localized Spectroscopy VSS: Very Selective Suppression WALTZ: Broadband Decoupling Scheme W-F: water-to-fat ratio WM: white matter

Metabolite Abbreviations Ala: Alanine ATP: Adenosine Triphosphate

Abbreviations 979

Cho: Choline

NTP: Nucleotide Triphosphate (e.g., ATP)

Cr: Creatine

PC: Phosphocholine

Eth: Ethanolamine

PCr: Phosphocreatine

Glx: Glutamine/Glutamate

PE: Phosphoethanolamine

GPC: Glycerophosphocholine

Pi: Inorganic Phosphate

GPE: Glycerophosphoethanolamine

SI: Scyllo-Inositol

Lac: Lactate

Tau: Taurine

MI: Myo-Inositol

UDP: Uridine Diphosphate

NAA: N-acetyl aspartate

981

Glossary of Terms

Abscess: Focal infection that consists of an abscess capsule, containing fibrous material and chronic inflammation cells, and a central cavity filled with pus.

Focal infection: Infection that is results in the formation of lesions. Can be caused by bacteria, fungi, parasites or viruses.

Absolute quantitation: Determination of concentrations of metabolites in mMoles per kg of tissue or volume.

Glioblastoma Multiforme: Highly malignant tumour with a low overall survival time for most patients. They are histologically characterised by a dense cellularity, high endothelial proliferation, and often the presence of focal tissue necrosis. These tumours are can occur in a cystic form.

Astrocytoma: The most common type of brain tumor in adults; astrocytomas, known for their marked potential for malignant progression and infiltrative nature, can be classified histopathologically into three grade of malignancy: WHO grade II astrocytoma, WHO grade III anaplastic astrocytoma, and WHO grade IV GBM. B0 : Strong magnetic field, constant in time and space, generated by the superconducting magnet. B1 : Radiofrequency (RF) magnetic field generated by radiofrequency coils.

Glioblastoma Multiforme: The most malignant grade of astrocytoma (WHO grade IV), which has an overall survival time of less than two years for most patients. These tumors are histopathologically characterized according to their dense cellularity, high proliferative indices, endothelial proliferation, and most importantly the presence of focal tissue necrosis.

Benign Prostatic Hyperplasia: Enlargement of the prostate and more specifically, overgrowth of the epithelium and fibromuscular tissue of the transition zone and periurethral area.

Hetero-nuclear J-coupling: J-coupling between different species of spins, e.g. proton and carbon.

Bifunctional contrast agent: Contrast agent that increases the intrinsic contrast for two different imaging modalities.

ISIS: Image-selected in vivo spectroscopy is based on a cycle of eight acquisitions which need to be added and subtracted in the right order to get a single volume. ISIS is considerably more susceptible to motion then STEAM or PRESS and is mostly used in heteronuclear studies, where its advantage of avoiding T2-relaxation is valuable.

calar coupling: See J-coupling Contrast agent : Chemical compound that increases the intrinsic contrast in an image. f: bandwidth (in Herz) of the receiver Dendrimers: Large and complex polymers of a consistent size and a regular and highly branched architecture. They are often used for the encapsulation of smaller molecules. Ductal Carcinoma In Situ: Noninfiltrative lesions composed of malignant epithelial cells that are confined to the mammary ducts and lobules. Endocytosis : In endocytosis, the cell engulfs some of its extracellular medium including material dissolved or suspended in it. A portion of the plasma membrane is invaginated and pinched off forming a membrane-bound vesicle called an endosome. Endosome : Vesicle formed during endocytosis. η: filling factor F: noise figure of the preamplifier

Homo-nuclear J-coupling: J-coupling between the same species of spins, e.g. proton and proton.

J-coupling (or scalar coupling): Many resonances split into multiplet components. This is the result of an internal indirect interaction of two spins via the intervening electron structure of the molecule. The coupling strength is measured in Hertz (Hz) rather than ppm because it is independent of the external B0 field strength. k: Boltzman constant K: numerical factor depending on the coil geometry Lipoma: Commonly diagnosed benign tumors of adipose tissue that can develop anywhere in the body where fat is normally found. Liposarcoma: A commonly diagnosed soft tissue sarcoma in adults which be found anywhere in the body and for which there exists several types with varying clinical outcomes.

982 Glossary of Terms

Medical Uses

Meningioma: Intracranial tumors that arise in the meninges and compress the underlying brain. In general, many of these tumors are benign, however, others are malignant with the capability to metastasize, both locally and distally. Metabolite Profiling: The study of small molecules, such as lipids, peptides, amino acids and carbohydrates, which represent steady state concentrations of intermediate products or end products of cellular processes and as a result can be thought of as the ultimate response to genetic and environmental stimuli. Mo: Magnetization µo: Permeability of free space Molecular imaging: Non-invasive, in vivo visualization of cellular and molecular events in normal and pathological processes. Necrosis: Tissue and/or cell death. Neoplasm: Literally meaning “new growth”. An abnormal growth of tissue which may be benign or malignant. NOE: Nuclear Overhauser effect, the magnetization of protons dipolar-coupled to 13 C nuclei can be used to enhance the 13 C signal. While the term NOE is mainly associated with 13 C MRS, NOE enhancement can also be observed in 31 P MRS and with other nuclei. ω0: Larmor frequency Oncogenomics: The study of genes, gene sequences, and the underlying genetic alterations that appear to be involved in oncological pathways. Polarization transfer: Many interesting nuclei like 13 C suffer from low inherent sensitivity compared with proton MR. Techniques like DEPT (distortionless enhancement by polarization transfer) and INEPT (insensitive nuclei enhanced by polarization transfer) improve the 13 C sensitivity by transferring the higher polarization of coupled protons to the carbon nuclei. Special hardware with two RF channels is needed for polarization transfer experiments. A modification of DEPT and INEPT is reverse DEPT and inverse INEPT where the polarization is transfered back to utilize the higher sensitivity of the protons for observation (inverse detection). PRESS: Point-resolved spectroscopy, utilizes three 180◦ slice selective pulses along each of the spatial directions and generates signals from the overlap in form of a spinecho. Pus: Content of an infective lesion. It consists of dead and dying polymorphous cells (leucocytes), living and dead

microorganisms, tissue debris and other components of inflammation (oedema fluid and fibrin). Q: Quality factor of the coil Reactive Nodes: Lymph nodes that have been immunologically challenged. Responsive contrast agent: Contrast agent that conditionally increases the intrinsic contrast (depending on the change of environmental conditions like enzyme activation, pH change, change in ion concentration and others). SAR: Specific absorption rate. Due to inductive and dielectric losses energy from radiofrequency pulses is absorbed by tissue and mainly transferred into rotational and translational movements of water molecules which causes an increase of tissue temperature. Limits for the human brain are established by FDA guidelines. Schwannoma: Benign tumors that arise on peripheral nerves in general and on cranial nerves in particular, especially on the vestibular portion of the eighth cranial nerve. Sensitivity: The term relating to the percentage of people with disease who test positive for the disease, i.e. the percentage of patients with cancer who have positive biopsy results. SNR: Signal to noise ratio Specificity: The term referring to the percentage of people without disease who test negative for the disease, i.e. patients without cancer with cancer negative biopsy results. STEAM: Stimulated echo acquisition mode, localization method utilizing three 90◦ slice selective pulses, along each of the spatial directions. Signal, in form of a stimulated echo, from the overlap is generated in a “single shot” experiment. In contrast to PRESS, only half of the possible signal is recovered when the same echo time is used. Stem cell: Cell with broad differentiation potential that retains the capacity for unlimited self-renewal. A totipotent stem cell has the ability to differentiate to all cell types of an organism, whereas a pluripotent stem cell produces many but not all cell types. T1-relaxation, T1-relaxation time: After the magnetization vector has been flipped into the transverse plane, new magnetization builds up along the z-axis. The time after 63% (1-1/e) of the equilibrium magnetization has built up is called the T1-relaxation time. T1 and T2 relaxation is caused by time-dependent fluctuations of local magnetic fields arising mostly from the motion of molecules with electric or magnetic dipoles at the site of the spins. For accurate absolute quantitation the relaxation times of

Glossary of Terms 983

all metabolites must be known in order to correct peak intensities appropriately. T1-saturation: The repetition times TR are usually in the range of the T1-relaxation times. As a consequence of this, not all the magnetization has recovered, for example when TR = T1 only 63% of the equilibrium magnetization can be used for each scan (with the exception of the first scan) when 90◦ flip angles are used for excitation. This effect is called T1-saturation. The extreme case of saturation occurs when several RF pulses are applied within a very short time followed by dephasing gradients. This technique is used in localized 1 H MRS to remove the dominant water signal (see water suppression). T2-relaxation, T2-relaxation time: The magnetization vector can be flipped into the transverse plane by using

an RF pulse. The so generated transverse magnetization undergoes an exponential decay. The time after the magnetization has relaxed to 37% (1/e) of its amplitude is called the T2 or transverse or spin-spin relaxation time. See also T1- relaxation . Tc: probe temperature TR (repetition time): The time between each initial excitation of the magnetization is called the repetition time. Transfection: A method by which experimental DNA (or in our case contrast agents) can be incorporated into mammalian cells after interaction of the encapsulating transfection agent and the cell membrane. Vc: volume of the coil

985

Alexander P. Lin Huntington Medical Research Institutes, Rudi Schulte Research Institute, Pasadena, CA 91105, USA

Part I: Seven Secrets to Successful Spectroscopy Neurospectroscopy or magnetic resonance spectroscopy (MRS) has moved from the realm of academic research into that of the clinical world. All major MR manufacturers have aided in the endeavor by automating neurospectroscopy so that it no longer requires an MR physicist and is a push-button technique that can be run by technologists just as a typical MR sequence. Thousands of papers have demonstrated the clinical efficacy of neurospectroscopy and there are a dozens of medical reviews of how this technique can be applied across a wide range of neurological disorders. However, few papers address the practical issue of acquiring neurospectroscopy in a clinical practice. Based on over a decade of clinical experience on our 1.5 T scanner and applications training for technologists and radiologists at our international clinical neurospectroscopy courses, this chapter was developed to demonstrate proven protocols for clinical diagnosis and outline the strategies involved in acquiring successful clinical spectra.

Introduction While the goals of the MR technologist are not very different between imaging and spectroscopy, the terminology and approach are slightly different. Imagers desire clear and detailed images without motion artifacts. Spectroscopists want spectra with well-defined peaks and few artifacts as well. The composition of a peak can be broken down to two parameters: (1) peak height or how readily discernible the metabolites are from noise; this in turn is governed by the principles of signal-to-noise ratio (SNR) and (2) line width or how narrow the peaks are, which is governed by magnetic field homogeneity. It is the goal of the spectroscopist to therefore maximize these two principles using a variety of different parameters discussed in this chapter. Using the same analogy, imagers use different sequences such as T2 w or FLAIR to investigate certain pathological properties; spectroscopists use different protocols to diagnose different disorders. These protocols can be broadly categorized into focal and global protocols, each of which will require a different approach to these parameters. Graham A. Webb (ed.), Modern Magnetic Resonance, 985–1003.
C 2008 Springer.

When we began developing and growing a clinical spectroscopy program, we soon discovered that the “secret” to a successful spectroscopy service was CONSISTENCY. Once a methodology of acquisition or guideline is established, it must be maintained. In any clinical MR unit, all technologists must follow that guideline in order to guarantee spectral reliability from patient to patient. Worldwide, the same consistency is essential to ensure “universality” of clinical diagnosis.

Signal and Homogeneity In the following sections, we will be discussing different parameters that can be changed to optimize your spectroscopy. There are two components to optimization, namely: SNR and homogeneity. With a high SNR, the metabolite peaks (signal) are very easy to distinguish from the smaller peaks that surround it (noise) as you can see in Figure 1. High SNR is critical to the interpretability of spectra by both the radiologist and the automated processing software. Maintaining high SNR throughout all your acquisitions is a prime responsibility for spectroscopy quality control. The factors that affect SNR are (in order of impact): voxel size, number of averages, echo time (TE ), localization sequence, and repetition time (TR ). Homogeneity refers to the uniformity of the magnetic field present in the voxel of interest (VOI: a voxel is the box which defines the area you wish to examine). A very homogeneous field has very little change in the magnetic field throughout the VOI (Figure 2B). A heterogeneous field has differing magnetic properties throughout the VOI (Figure 2A). A “sharp” spectrum at 1.5 T reflects both consistent and relatively thin line widths (3 cc, CSI if 10: (A) Malignant and (B) Benign. Spectra were acquired over a sweep width of 3597 Hz, 8192 data points, 256 accumulations, and a relaxation delay of 2 s. Source: Reprinted from the Br. J. Surg. 2001;88:1234–40. Mountford CE, Somorjai RL, Malycha P, Gluch L, Lean C, Russell P, Barraclough B, Gillett D, Himmelreich U, Dolenko B, Nikulin AE, Smith IC. Diagnosis and Prognosis of Breast Cancer by Magnetic Resonance Spectroscopy of Fine-Needle Aspirates Analyzed Using a Statistical Classification Strategy, with permission from John Wiley & Sons Ltd on behalf of the British Journal of Surgery Society Ltd. 1H

6

3.25 : 3.05 ppm

MRS Peak Height Ratio

5

4

3

2

1

0

Benign (n=106)

Carcinoma (n=82)

Diagnosis

Fig. 4. Breast fine-needle biopsy MR spectroscopic (MRS) findings of unequivocally benign vs. infiltrating carcinoma. Data are grouped on the basis of the final histopathologic findings in tissue specimens from the aspiration site. Source: Reprinted from Radiology, 1997;204:661– 6. Mackinnon WB, Barry B, Malycha P, Gillett D, Russell P, Lean CL, Doran S, Barraclough B, Bilous M, Mountford CE. Fine Needle Biopsy of Benign Breast Lesions Distinguished from Invasive Cancer by Proton Magnetic Resonance Spectroscopy, with permission from the Radiological Society of North America.

1034 Part II

Medical Uses

Part II A. Adenocarcinoma 50%

Acly

Lip Leu lIe Val

Lys Lip

Glu/Gln

Chol

Lip Lac Thr

Cre Cit

B. Adenocarcinoma 5%

Fig. 5. 1 H MR (8.5 T, 37◦ C) spectra of prostate biopsy specimens, 256 acquisitions, sweep width 3597 Hz, and a pulse repetition time of 2.14 s. The water peak was suppressed by selective gated irradiation. (A) adenocarcinoma (50% of the tissue made up of malignant tissue), (B) adenocarcinoma (5% of the tissue made up of malignant tissue), (C) Prostatic intraepithelial neoplasia (PIN), (D) stromal BPH (95% stromal, 5% glandular), and (E) glandular BPH (85% glandular, 15% stromal) are compared. Source: Reprinted from Radiology 2003;228:144–151. Swindle P, McCredie S, Russell P, Himmelreich U, Khadra M, Lean CL, Mountford CE. Pathologic Characterization of Human Prostate Tissue with Proton Magnetic Resonance Spectroscopy, with permission from Radiological Society of North America.

C. PIN

D. Stromal BPH

E. Glandular BPH PA Cit

4

to develop a classifier. Data are divided into training and test sets. The classifier is developed from the training set and then applied to the test set to see how well it does on data with which it has not trained. The original data are then mixed again and a different training and test set used to perform the same analysis. This is done up to 10,000 times to develop up to 10,000 different classifiers.

3

2

1

ppm

This we call bootstrapping. The 10,000 classifiers are then used in combination to produce a final classifier. If at the end of stage 2 the accuracy of the final classifier is not sufficient, stage 3 of the SCS method is employed. Here other representations of the spectra, such as first or second derivatives, are used to make separate classifiers using stages 1 and 2. All stage 2 classifiers are then

Magnetic Resonance Spectroscopy of Human Biopsies

Regression Analysis Routine histopathological reporting commonly includes a semi-quantitative description (e.g. “occasional foci,” “extensive infiltration”) of the amount of structural abnormality observed microscopically. Although imprecise, such semi-quantitative descriptions of observed disease burden are a factor in treatment planning. However, correlation of the MRS characteristics of tissue with histopathological characterization of the same tissue has largely been based on a simple two-class analysis in which tissue samples are classified according to the presence or absence of a single specific characteristic such as cancer. Diagnostic interpretation of the spectra from such mixed tissue voxels would ideally report not only the presence of disease, but also the extent or partial volume of disease within the volume of interest (VOI). To address these two problems a linear regression analysis was undertaken on spectra acquired from 82 cancercontaining prostate tissue specimens. MR spectra were correlated with serial section histopathological examination of each tissue specimen for which the volume fraction of cancer tissue was estimated in 5% increments. A classifier based on five spectral features developed from a training set of 45 randomly chosen spectra was used to predict from spectra the volume fraction of cancer in a validation set of 37 tissue samples. The overall accuracy of predicted cancer volume, measured as the average accuracy of all cancer volume estimates, was 96% [26].

Future Challenges Putting magnets into a pathology environment is a new concept. The challenge is to move this technology into a routine clinical environment. This was achieved by Liposciences for serum lipoprotein analysis of serum

(www.liposcience.com). Serum specimens are easier to transport and the test is for screening purposes. Clinical acceptance testing is required and is currently being undertaken in Sydney by a group of endocrine and urology surgeons. They need to verify that the tests provide additional information that is necessary for the management of the patient. Automation of the software to drive the system and to allow comparison of spectra with classifiers will see the spectroscopy method for analyzing human pathology move into a new age.

Acknowledgments We thank Deborah Edward, Brooke O’Donnell, Sinead Doran, and Raquel Baert for assistance in preparing this manuscript. Our thanks go to all present and past collaborators in particular Professor Peter Russell and Dr. Ray Somorjai, without whom none of this would have been possible.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Mountford CE, et al. Lancet. 1986;1(8482):651–3. Mountford C, et al. Chem. Rev. 2004;104:3677–704. Metcalf D. Recent Result Cancer Res. 1966;5:5. Mountford CE, et al. Br. J. Cancer. 1980;41(6):1000–3. Doran S, et al. Am. J. Surg. 2003;185(3):232–8. Russell P, et al. Am. J. Med. 1994;96(4):383–8. Mountford CE, et al. Magn. Reson. Med. 1990;13(2):324– 31. Delikatny EJ, et al. Radiology. 1993;188(3):791–6. Swindle P, et al. Radiology. 2003;228:144–51. Ende D, et al. Proc. Soc. Magn. Reson. Med. 1993;2:273. Mackinnon WB, et al. Int. J. Gynaecol. Cancer. 1995;5:211– 21. Kuesel AC, et al. Magn. Reson. Med. 1992;27(2):349–55. Bourne R, Dzendrowskyj T, Mountford C. NMR Biomed. 2003;16(2):96–101. Braun S, Kalinowski HO, Berger S. 2nd ed. 150 and More Basic NMR Experiments: A Practical Course. Weinheim: Wiley-VCH, 1998, pp 473–5. Baenziger JE, Jarrell HC, Smith ICP. Biochemistry. 1992;31:3377–85. Cheng LL, et al. FEBS Lett. 2001;494(1–2):112–6. Leemans CR, et al. Cancer. 1994;73:187–90. Tzika AA, et al. J. Neurosurg. 2002;96(6):1023–31. Cross KJ, et al. Biochemistry. 1984;23(25):5895–7. Mountford CE, et al. Br. J. Surg. 2001;88(9):1234–40. Hahn P, et al. Cancer Res. 1997;57(16):3398–401. Somorjai RL, et al. Magn. Reson. Med. 1995;33(2):257– 63. Lean CL,et al. In: G Webb (Ed). Annual Reports NMR Spectroscopy. Guildford, UK: Academic Press, 2002, pp 71–111. Nikulin A, et al. NMR Biomed. 1998;11:209–17. Doran S, et al. Proc. ISMRM. Kyoto, Japan, 2004, p 2495. Bourne R, et al. Proc. ISMRM. Toronto, Oral, 2003.

Part II

combined to yield a meta-classifier of increased accuracy [22]. In a study of FNAB from the breast, the data were subdivided into women with cancer with (n = 29) and without (n = 32) lymph node involvement [20]. Applying the SCS yielded classification of nodal involvement or not with sensitivity and specificity of 97 and 96%, respectively. Similar classification of patients with and without tumor vascular invasion yielded sensitivity and specificity of 84 and 100%. These classifications should be of great value in determining the therapeutic path for patients. More recent data from the same authors show great promise in calculating the tumor grade and estrogen and progesterone receptor status of breast tumors, also very important in planning therapy [25].

References 1035

1036 Part II

Medical Uses

Part II

27. Somorjai RL, et al. J. Magn. Reson. Imaging. 1996;6(3): 437–44. 28. Somorjai RL, et al. J. Med. Biochem. 1999;3:17–24. 29. Menard C, et al. Int. J. Radiat. Oncol. Biol. Phys. 2001;50(2):317–23.

30. Soper R, et al. Pathology. 2002;34:417–22. 31. Wallace JC, et al. Magn. Reson. Med. 1997;38(4):569– 76. 32. Bourne R, et al. Proc. ISMRM. Toronto, Canada, 2003, p 1302.

1037

Graeme D. Jackson, Regula S. Briellmann, Anthony B. Waites, Gaby S. Pell, and David F. Abbott Brain Research Institute, Melbourne, Victoria, 3081, Australia

Principles of fMRI In recent years, it has been demonstrated that MRI is capable of detecting changes in cerebral blood volume, flow, and oxygenation that accompany a change in neuronal activity. MRI techniques that observe these activationinduced signal changes are termed “functional MRI” (fMRI). The most common contrast mechanism in fMRI is known as blood oxygenation level dependent (BOLD) contrast, first described by Ogawa et al. This contrast mechanism relies on the relationship between local changes in oxygenation and blood flow that occur with neuronal activation. The sequence of events is as follows: 1. The neuronal cells fire. 2. The small amount of energy used is recovered via oxygen metabolism, resulting in a small increase in deoxyhemoglobin in the first second or two after cell firing. 3. A vascular response also occurs, being a local increase in blood flow (perfusion). 4. In the next few seconds, the delivery of oxygen via the vascular response is far in excess of that depleted to meet the local energy demands. Therefore, paradoxically, there is a local decrease in the concentration of deoxyhemoglobin of a much greater magnitude than the initial increase. BOLD contrast relies on the sensitivity of MRI to detect the changing concentration of deoxyhemoglobin. The initial small increase in deoxyhemoglobin has been observed by some groups as a very small “initial dip” in the time course of the BOLD MR signal; however, it is the subsequent larger decrease in deoxyhemoglobin concentration that provides the substantial and robust signal increase that makes the method most useful. The signal change arises because deoxygenated hemoglobin is paramagnetic, whereas oxygenated hemoglobin is diamagnetic. Paramagnetic molecules are especially visible in MRI since they exert a strong enhancing effect on the relaxation rate of the water in the vascular environment. The transverse relaxation times, T2 and T2∗ are reduced by this mechanism. These are the contrast dependencies of spin echo and gradient echo images, respectively. Both these basic imaging techniques can therefore be used to Graham A. Webb (ed.), Modern Magnetic Resonance, 1037–1050.
C 2008 Springer.

detect BOLD contrast. Differences between these two techniques exist with regard to the extent of sensitivity to the environmental change. The T2∗ relaxation time is uniquely sensitive to the intravascular changes of water environment and therefore BOLD contrast is magnified in gradient echo images. Currently, the most commonly applied imaging technique for BOLD fMRI is echo planar imaging (EPI). It allows very rapid imaging, allowing for whole brain coverage with standard paradigms. Another more recently developed contrast mechanism for fMRI is that of perfusion. This utilizes imaging methods that are sensitive to detecting the local change in blood flow which is the initial response to the neuronal firing. This technique acts as a more direct means of detecting the activation site. However, the perfusion imaging methods suffer from limitations in signal-to-noise ratio (SNR) in comparison with the simpler BOLD imaging techniques.

Design of fMRI Trials One of the most important parts of an fMRI experiment is the designing of the appropriate paradigm that will allow identification of brain voxels involved in a particular cognitive or sensory–motor function. In this chapter, we will discuss two commonly used approaches, the “block design” and the “event-related design.” For both approaches, it is important to keep in mind that the vascular response (which is what we can measure with BOLD) follows the neuronal activity (which is what we would like to know) with a delay of several seconds.

Block Design The simplest form of a design for fMRI experiments is called a block design. This comprises alternating cycles of periods of task and rest conditions (Figure 1). Each task and rest period should not be too short, as the BOLD signal takes around 5–10 s to reach a maximum after a neuronal event. Shorter periods will lead to a reduced contrast between task and rest states. Very long periods also have a drawback, as they may resemble the low frequency fluctuations or drift in the scanner signal. Such a design would lead to confusion between the cognitive variation, which we wish to measure, and artifactual changes, which we

Part II

Functional MRI

1038 Part II

Medical Uses

Part II

Fig. 1. Scheme of a block design paradigm with alternating block and task periods of identical length.

one cycle

task

rest paradigm

wish to minimize. The ideal design allows the BOLD signal to reach a steady state within each condition in order to maximize contrast. In practice, a 40 s to 1 min cycle time (i.e. 20–30 s task then 20–30 s rest) with about 5– 8 cycles, giving a total experiment time of 4–10 min is suitable for many experiments. With more robust tasks (such as visual stimulation, which leads to large neuronal response) and higher magnetic field strengths, such as 3 T, the total experiment time may be reduced. Longer experiment times increase the chance of motion artifacts due to subject discomfort, as well as loss of concentration on the task. Block design experiments allow maximization of SNR, but also have some disadvantages. Repeating the same task may lead to the subject anticipating the task and sometimes even the response. This may considerably confound the results. A randomized design avoids this issue since the occurrence of a particular task type is unpredictable. Further, not all cognitive studies can be adapted to a block design. For example, experiments assessing the cognitive effect of an unexpected stimulus within another task, known as an oddball paradigm, cannot be performed in a block. Similarly, studies assessing unpredictably occurring physiological events, such as myoclonic jerks, cannot be performed as block designs. Block designs also do not provide information on the shape of the blood flow change in response to the task. This vascular response to a neuronal event is referred to as the hemodynamic response function (HDR). The HDR is used in event-related designs.

Event-Related Designs Event-related designs are a more flexible approach, and involve the placement of single events within a train of “rest state” conditions. The order of the events is often randomized, preventing anticipation of the task. Events with varying periods of rest allow measurement of different parts of the hemodynamic response, allowing the HDR to be identified. This can be useful in understanding the different response from different brain regions, which may be a result of either the different vasculature of the

region, or the differences in cognitive dynamics in the different regions. The disadvantage of such designs is the low SNR. This is due to the fact that the task state is not sustained for long periods, leading to a less intense vascular response. This can be partially overcome by using “clustered” random presentation of events, where the random pattern of task events is chosen so that there are clumps of task and clumps of rest, albeit with some variability. This is a compromise, giving some increase in SNR whilst maintaining the benefits of a random design. Event-related analysis techniques can also be applied to data from a block-designed study. This can either take the form of modeling the block of task as a series of events, or performing a post hoc analysis of the data. The post hoc analysis involves identifying events within a block, such as correct responses to a stimulus, and comparing them to for example incorrect responses.

Inter-Stimulus Interval Selection for Event-Related Designs It is important to consider the inter-stimulus interval (ISI) carefully. One must first decide whether a sparse or rapid event-related design is to be used. Sparse designs have a long ISI such that the hemodynamic response returns to baseline before the next event is encountered. More common are rapid event-related designs where the minimum ISI is often as little as 2 s. These designs have the advantage that they are more efficient (many more events are captured in a given period of time), and they are therefore the preferred choice if the nature of the task allows it. One must be more careful with these designs, however, as analysis methods usually rely on the assumption that the BOLD response to multiple events adds linearly. This is not always the case (especially if the ISI is much less than 2 s), and one must be careful when interpreting results in brain areas where the linearity of the BOLD response is not well known. Other issues should also be considered when choosing the ISI. The use of an ISI equal to an integer multiple of the repetition time (TR) can lead to a missed peak in the HDR. This occurs because the stimulus and measurement

Functional MRI

Modeling the Hemodynamic Response in an Even-Related Study Analysis of an event-related study requires a specific model for the HDR. If the primary goal of the experiment is to determine the precise form of the HDR, a flexible basis set is usually chosen such as a set of sine and cosine waves (a Fourier set), or a set of Gamma functions. On the other hand, if the goal is to determine where and when an event occurred, a canonical HDR is typically used. The canonical HDR is an empirically determined shape found to fit experimental data. One can improve the canonical model by including the temporal derivative (which allows flexibility in the delay of the HDR) and a dispersion term to allow a broader than typical response. This can accommodate some regional and inter-subject variation in response. The canonical model is the most powerful at detection, since it has the fewest parameters, and thus preserves the highest number of degrees of freedom. Conversely, the basis function approaches will be more sensitive if the response differs significantly from the canonical form. A combination of approaches is possible. For example, a portion of a study can be dedicated to eliciting the regional shape of the HDR in an individual (perhaps with a stimulus known to yield a large signal change), and this can then be used as a more accurate model for detection of subtle stimuli in that individual.

Principles of Experimental Design Where to Start? There are many factors that affect detection power including the type of experimental design, the scanner, the

population, and the nature of the task being imaged. In order to identify the optimal task, a clear hypothesis will aid greatly in designing the experiment. Such a hypothesis may be based on theoretical background of a particular function, or more commonly, is an extension of an observation made outside the scanner. The development of an appropriate task for fMRI scanning is often based on published, somewhat similar paradigms. A protocol can then be further refined after a few pilot studies have been undertaken. Typically, the durations of the final protocol should be a little longer than the estimated optimum, as the subjects used for pilot studies are often highly motivated individuals who will give better results than the average study population.

Block or Event-Related Design? The block design is optimal for determining the location of activity, but cannot readily be used to determine the nature of the HDR time course. Event-related designs could be optimized to efficiently determine the nature of the HDR time course; however they generally offer much less power to prove that a response actually exists at a particular location. There are many subtleties to possible designs, including hybrid block and event-related designs. In general a block design is preferred if there are no contraindications, as it gives higher statistical power. Alternative approaches should be taken only for questions that cannot be answered using a block design.

Baseline Considerations Choosing the appropriate baseline is as important as choosing the optimal task. Commonly, the baseline consists of “rest,” during which the subject is asked to do “nothing.” This is a very poorly defined state, and some paradigms improve with a more defined, engaging baseline. Contrast in fMRI arises from a measured change in signal, and it is therefore important to know from what baseline state the signal changed. Our brains are always “on,” so one must be careful when interpreting signal change related to a task. All we can really say about significant activation is that there is greater signal whilst a subject is performing a task of interest than when they are performing the baseline task. Rest can and should be considered as just a different task than the task of interest. Similarly, if signal decreases during the task of interest, one can regard this as an increase in activity during rest. Importantly, signal decrease or “deactivation” is not the same as inhibition, since inhibition is itself an active process that can contribute to an increase in signal. If inhibitory processes are involved, the reduction of activity that is observed may have been signaled from an active

Part II

are synchronized, and measurements may repeatedly occur before or after the peak of the HDR. Making matters worse, there would be a systematic difference in the time each individual slice in a single volume is acquired relative to the event, so that some slices may always be more sensitive to the event than others. If synchronous timing is to be used, it is therefore best to choose an ISI that differs from the TR such that measurements of the HDR occur at different times for many successive events. For example, if events were timed to occur each 1.3 × TR, the HDR would be effectively sampled at nine different time points before the sampling pattern began to repeat. This would be better than 1.5 × TR, for example, where only two different time points would ever be sampled. Another approach is to use jittering. Jittering involves randomly varying the ISI slightly for each event, which again leads to measurements at different points throughout the HDR.

Principles of Experimental Design 1039

1040 Part II

Medical Uses

Part II

inhibitory process adjacent to or some distance from the site of “deactivation.”

Performance Measures Proper interpretation of fMRI results generally relies on knowledge of the tasks subjects were performing during the imaging session. One common assumption is that the subjects are doing what they are told. Ideally performance measures should be undertaken concurrently with the imaging, but this is not always practical. For example, in a commonly used language paradigm subjects may be asked to generate as many words as they can that begin with a provided letter. To avoid problems associated with subject motion, this task is normally performed covertly (i.e. without vocalization). This task performance cannot be directly verified. To minimize the possibility of incorrect interpretation, one should obtain some measure of task performance. This is often the measurement of a subject’s performance on a similar task out of the scanner. Subjects should also be asked after their scan whether or not they encountered any difficulty in performing the task in the scanner. Improved interpretation of functional imaging results can often be made when performance of aspects other than the task of interest are also measured. For example, it is known that changes in physiological measures such as cardiac and respiratory rates can affect the cerebral circulation, and breath-holding especially can result in very large changes in the BOLD response that can potentially confound interpretation. Measures of skin conductance and eye tracking are also possible, and these and other measures can be used to remove variance in the data that may be unrelated to the task of interest, or to explicitly model effects that may be related to the task of interest.

Principles of Analysis There are a number of steps involved in taking raw fMRI data and turning it into colored maps of activity that can be readily interpreted. The processing pipeline contains procedures to minimize artifact, maximize signal, and aid interpretation. We will describe here a commonly used imaging pipeline for BOLD fMRI.

Data Preparation Often the first step in an analysis pipeline is extraction of the imaging data from the scanner and conversion into a format suitable for further processing. Whilst scanner manufacturers are slowly introducing functional image analysis tools into their standard software, at the present time such software is often rudimentary and inflexible.

Pre-Processing Pre-processing steps are those performed prior to statistical image analysis. Depending upon the type of acquisition they can include: r Correction for signal drop out near magnetic susceptibility inhomogeneities. The variation in magnetic susceptibility especially near tissue–air boundaries can cause extensive areas of signal loss in typical EPI acquisitions due to the inability of scanners to shim to a uniform field in these regions. Various methods have been devised to overcome this problem, including the z-shim method that involves acquiring two or more sequential acquisitions with different imposed corrective gradients that effectively image separately the areas of drop out. A post-processing step is required to combine these multiple acquisitions. r Correction for geometric distortion, e.g. Measured B0 phase maps can be used to calculate and correct for geometric distortions that plague EPI fMRI images. r Slice timing correction. Acquisition volumes most commonly consist of a series of sequentially acquired slices. Slice timing correction methods generally resample the temporal response of a voxel, applying interpolation to estimate what the response would be at the time of commencement of each acquisition volume. Of course this correction should not be used in conjunction with imaging methods that involve acquisition of an entire volume in one imaging shot. r Gross motion rejection and/or realignment. The deleterious effects of subject motion can be reduced with these steps. Gross motion rejection can also remove data contaminated by acquisition failures such as a gradient misfire. r Global intensity normalization. This step is necessary with some scanners that are not able to provide a stable signal from one acquisition to the next. If one assumes that the BOLD signal change due to the task is small relative to the mean signal of all within-brain voxels, one can normalize the global signal to a common value (proportional scaling). Unfortunately, the BOLD signal change is not always negligible in relation to the mean global signal, so there is a risk that this processing step will do more harm than good (for example by reducing the apparent size of real activation). If the MRI system is reasonably stable, then global intensity normalization should usually be avoided. r Spatial normalization. This step involves registration of the imaging data of each subject in a study to a common template, to enable voxel-wise statistical analysis of group data and/or to allow reporting of activation locations relative to a standard atlas. In the case of a single subject analysis, the derived transformation may

Functional MRI

It is important to note that the order in which these processing steps are performed can have a substantial effect on results. The order listed above is a typical scenario, however, it may not always be the best arrangement.

Statistical Analysis Statistical analysis of fMRI data is necessary because the signals of interest are usually of a similar magnitude to the noise in the data. By collecting many images and applying statistical analysis, one can enumerate the voxels that have an acceptable probability of being associated with the task of interest. The most common method of analysis is a general linear model based approach. The signal is assumed to comprise a linear combination of effects of interest, confounding effects, and a random noise term. The effect of interest in the model usually consists of a function representing the time course of tasks performed convolved with an estimate of the brain’s HDR. Confounding effects are also often included in the model, such as the realignment parameters determined in the motion correction pre-processing stage (these are often included to model any residual motion effects such as changes in the magnetization experienced by protons that are not corrected by realignment). The statistical analysis determines, for each voxel independently, the proportion of included model terms that best fit the measured data, and provides an estimate of the probability that the coefficient of the effect of interest is non-zero (implying the effect is present). Gaussian random field theory is often used to estimate a probability corrected for the multiple comparisons made in this massively univariate data. Other statistical analyses are also possible, including data driven approaches where no assumptions are made about the tasks performed. Rather the data are interrogated

in such a way as to extract structured components that are able to describe the data in some way. Such methods include principle components analysis and independent components analysis, the latter generally being more useful in the neuroimaging context. Detailed descriptions of these approaches are beyond the scope of the current discussion. However, it is worth noting that these approaches are beneficial when the nature of the expected response is not known. Results of data driven analyses can be difficult to interpret however, and model based approaches are generally more powerful when a good estimate of the expected response is available.

Software Packages There are now a number of software packages available to assist with functional imaging analysis. Probably the package in most widespread use is SPM, a powerful and flexible package for performing “statistical parametric mapping.” A comprehensive competitor with some unique advantages is FSL. Both of these packages are available free of charge. There are many other packages that are also worth considering, including several specialized utilities that have been designed to perform particular aspects of processing, analysis, and/or visualization. Often the best analysis approach will take advantage of several packages. Selection can often be a matter of determining the packages that are already in use in your geographical area, since an experienced user is often the best source of help for inexperienced users.

Artifacts and Pitfalls Since fMRI studies usually require a relatively long series of scans in order for the activation-induced signal changes to build up to a detectable level, gross subject motion may occur. Image processing techniques exist to at least partially correct for motion, as we discussed earlier, however prevention is better than cure. For example, head immobilization via straps, vacuum-bean bags, and/or use of bitebars can help minimize head motion. Immobilizing the upper part of the extremities during sensory–motor tasks can help minimize task-related motion. A further confounding factor can be task-associated physiological effects resulting in changes in brain activation, such as subconscious breath-holding during challenging tasks. Such artifacts may lead to misinterpretation of the study results, and every effort should be undertaken to minimize their occurrence and to monitor for their presence. Another common problem with fMRI is geometric distortion. The most common rapid imaging technique for fMRI is EPI. With this technique, images of the brain can be acquired in times of 100 ms or less; however the technique is particularly prone to exhibit geometric distortion.

Part II

be applied to the statistical result instead of the raw imaging data, so as to avoid resampling artifact. If this is done, nearest-neighbor resampling is used to preserve the fidelity of the statistical result. r Smoothing. Spatial and temporal smoothing with a Gaussian kernel is often performed to improve the SNR of imaging data. This pre-processing step also ensures that the variance distribution is Gaussian, allowing the valid application of Gaussian random field theory in statistical analysis. To achieve this, the smoothing kernel should have a full width at half maximum (FWHM) of at least twice the voxel size. r Linear de-trending. This is a controversial processing step as the underlying assumption that a linear trend in a voxel time course is “noise” is unlikely to be valid. There are other ways to deal with this later when statistically analyzing the images.

Artifacts and Pitfalls 1041

1042 Part II

Medical Uses

Part II

This is especially severe in areas of high susceptibility differences such as the air–tissue boundary in the temporal lobes. Correct interpretation and detection of fMRI changes in these areas is often difficult. Even in relatively susceptibility-free regions of the brain, EPI images will display subtly different distortions in comparison with slower structural scans and this complicates the overlay of function and structure that is a necessary for interpretation of results. Even when all these problems are minimized, and an appropriate activation map has been generated, one has to be careful in the interpretation of the fMRI results. A common pitfall in group studies is an inadequate number of participants, leading to group results that do not survive correction for multiple comparisons. Whereas such results may be important pilot data, directing further experiments, one has to consider that they are at best a “trend” and provide no statistical proof of an effect greater than noise. A further common problem is the interpretation that a significant “activation” is equivalent to preserved function. There are many examples in the literature, demonstrating that the degree of activation does not linearly follow performance of a function. A very poor performer and an excellent performer may both activate relatively few areas, whereas a moderate performer may show a relatively higher degree of activation. Task-related activation in a particular area may not represent “the site” of a particular function, but rather represent part of a network of areas used to perform this task, or even an associated function, such as attention to the task or eye movements in relation to the paradigm. Furthermore, just because an area is active does not necessarily mean that it is essential for the performance of the task.

Practical Applications All the fMRI examples shown here were acquired on a GE Signa 3 Tesla MR scanner (GE Medical Systems). Higher magnetic field strength is associated with a higher SNR, which can be translated into shorter imaging times for a given resolution, or higher resolution for a given imaging time, or combination of both. At the time of writing 1.5 T scanners are most common. Compared to 1.5 T, at 3 T the same fMRI paradigm could be run for a shorter period of time, or paradigms evoking typically a very small response may give sufficient signal, whereas they may not be useful at lower field strength. However, the higher field strength has also some disadvantages. There is the potential for greater energy deposition into tissue and increased susceptibility-related artifacts. Functional images are always acquired together with other sequences. The T2∗ -weighted sequence used for fMRI has a low resolution, so T1 or T2 -weighted images are usually acquired in the same planes as the T2∗ -weighted

images. Further, angiogram/venogram images are often acquired to localize vessels in the brain, in order to avoid confusion of a signal arising from a vessel with the BOLD response. fMRI in the examples below was performed using a whole brain (22 slices) gradient EPI technique (TE = 40 ms, TR = 3600 ms, acquisition matrix = 128 × 128, flip angle = 60◦ , and 4 mm thick slices, 1 mm gap).

Language fMRI Language is a cognitive brain function that involves a network of brain areas. In the majority of healthy subjects, these language-related brain areas are predominantly in the left hemisphere. Since the first reported use of fMRI for the study of language in 1993 by McCarthy et al., fMRI has become established for the purpose of assessing language lateralization. The determination of “language dominance” is an important investigation step in patients undergoing brain surgery. Many different language fMRI tasks have been developed. Some are designed to activate particular parts of the language network. In our Institute, we have developed and validated two language paradigms, both performed as a block designs. For the orthographic lexical retrieval (OLR) paradigm a letter is presented visually, and the subject asked to retrieve silently as many words as possible beginning with that letter. After 18 s a second letter is presented, giving a 36-s task period. For the noun–verb generation (NVG) paradigm, the subject is visually presented with a common noun (e.g. fish) and is asked to generate an appropriate verb (e.g. swim). A new noun is presented every 4 s, with a task period of 36 s. For both tasks the baseline consists of visual fixation of a cross hair, this rest period also has a duration of 36 s (Figure 1). Both paradigms have four cycles with an additional rest period at the start of the paradigm; the total duration of the paradigm is 5.4 min. Determination of Language Lateralization Image processing, as described above, will result in maps of activation; these can be used to quantify the lateralization of a function. The laterality index (LI) is based on the number of activated pixels in language-associated areas. The LI is calculated using the following equation: LI = (left-hemispheric pixel count—right-hemispheric pixel count)/(left-hemispheric pixel count + right-hemispheric pixel count). Examples of fMRI Language in Controls We recently performed a study in multilingual subjects using a NVG task in four different languages. For reference to this publication, see figure legend. We hypothesized that the degree of proficiency in each language would be related to the extent of functional activity measured in a region of interest analysis. Therefore, the degree of

Functional MRI

Practical Applications 1043

Part II

3.5 3 2.5 German

2

English

1.5

French

1

Italian

0.5 0 1. trial

2. trial

3. trial

Fig. 2. This diagram shows an example of an out-of-scanner performance test in a quadrilingual subject. The subject was given a cartoon with a word-free story, and was asked to produce a discourse based on this cartoon. This was repeated three times, and then a new cartoon was given to assess the discourse in another language. Results are given for the words/second produced at all three trials. Note the overall increase in word production, particularly noticeable for languages with poorer proficiency. This subject had two “good” and two “poor” languages. Source: Adapted from Briellmann RS, Saling MM, Connell AB, Waites AB, Abbott DF, Jackson GD. A high-field functional MRI study of quadri-lingual subjects. Brain Lang. 2004;89:531–42.

proficiency in each language had to be quantified in detail. We used several tests, including a discourse test, in which the subjects were asked to produce a story based on a cartoon, which displayed a word-free story. Figure 2 shows the results in one subject, demonstrating a clear difference between languages with good performance (in this subject German and English), and languages with poor performance (Italian and French). The fMRI showed that all four languages activated overlapping brain areas, corresponding to the major language regions (Figure 3). Interestingly, the number of activated voxels correlated with proficiency, so that the activated volume increased for languages in which a subject had poorer proficiency. This observation highlights an important issue in the design and interpretation of fMRI experiments. In some cases

better performance is associated with more brain activity, whilst in others the reverse is true. Correct interpretation of a novel task therefore requires some independent measure of subject performance or response. Examples of fMRI Language in Patients Atypical language lateralization is more frequently found in patients with epilepsy and some other focal and chronic brain disorders. In a series of 30 healthy controls and 30 epilepsy patients, we confirmed the increased frequency of atypical language in epilepsy patients (Figure 4). Such atypical lateralization may represent reorganization due to early disturbances of the left hemisphere, such as a perinatal vascular event. fMRI is not only used to determine language lateralization, it has been used to assess the

Fig. 3. Functional MRI of language of a quadrilingual subject. The subject performed the NVG task in four different languages. The figure shows the activation during task in comparison to rest in all four languages. Note that all languages activate similar brain areas, however, there is a subtle increase in the amount of activation in poor compared to good languages. Source: Image taken from Briellmann RS, Saling MM, Connell AB, Waites AB, Abbott DF, Jackson GD. A high-field functional MRI study of quadri-lingual subjects. Brain Lang. 2004;89:531–42. (See also Plate 84 on page XIII in the Color Plate Section.)

1044 Part II

Medical Uses

Part II

Lateralization % 100 90 80 70 60 50

Controls Patients

40 30 20 10 0

L lateralised LI >0.2

Atypical dominant LI 50 kHz. The application of MAS in solid-state studies has involved the use of high spinning rates and strong radio-frequency pulses in an attempt to decouple these homonuclear interactions. This is not the case, however, for the MAS study of cellular metabolites in biological tissues. Here, water-soluble molecules reside in the cytoplasm wherein their motion is restricted by magnetic susceptibilities caused by various interfaces with other cell structures and by inherent viscosity. Although classical “solid-effects” exist, for instance within cell membranes, they would

Magic Angle Spinning MR Spectroscopy of Human Tissues

Experimental Sample Preparation Fresh and previously frozen samples may be measured directly with HRMAS MRS [17–21]. For the purpose of preserving the metabolite concentrations within tissues, samples should not be collected or stored in any liquid medium. In cases where the size of frozen tissue exceeds the sample size required for HRMAS MRS analysis, samples must be cut on a frozen surface to avoid multiple freezing and thawing. Such a surface can be made with a thin metal plate, covered with gauze and placed on top of dry ice. Tissue samples may be washed briefly with D2 O prior to spectroscopy if they contain visible amounts of blood. However, exposure to D2 O should be brief to minimize the possible loss of cellular metabolites [22]. We have observed complete depletion of metabolites for human prostate samples (∼10 mg blocks) submerged in ∼2 ml D2 O for approximately 10 min. As always, when working with human materials of potential biohazard, universal precautions need to be practiced at all times. In particular, tissue samples will undergo spinning and even at “slow spinning speeds” of less than 1 kHz, tissue fluid can leak from the rotor cap if the cap is not in tight-fit with the rotor (Figure 1). Thus the compatibility and seal between rotor and cap should be tested

Fig. 1. Photograph of the rotor and a typical biopsy tissue sample that will be placed in the rotor using the tweezers for HRMAS MRS analysis.

using a D2 O solution prior to tissue analysis. The test can be done by comparing the weights of the rotor with solution before and after HRMAS using the same experimental conditions (temperature and spinning speed) used for tissue analysis. To maximize spectral resolution, tissue samples should be limited to the physical boundaries of the receiving coils; for example by using Kel-F inserts to create a spherical sample when measuring samples on a Bruker spectrometer (Bruker BioSpin Corp., Billerica, MA). The use of a spherical sample is recommended for it minimizes shimming efforts as well as reduces the effect of the magnetic field inhomogeneity on the broadening of spectral lines. Spectrometer Settings Before tissue measurements, the HRMAS probe should be adjusted for its magic angle with potassium bromine (KBr) following the manufacturer’s protocol for solidstate MRS. Ideally, measurements should be made at a low temperature (e.g. 4 ◦ C) to reduce tissue degradation during acquisition. The optimal spinning rate should be decided after consideration of several factors, including tissue type, metabolites of interest, and the plan for the tissue after HRMAS MRS analysis. Generally, the higher the spinning rate, up to 10 KHz as reported in the literature, the better the spectral resolution [23]. However, since high spinning rates can function as a centrifuge that can potentially disrupt tissue structures, if subsequent histopathological evaluation of the tissue is critical to the study, spinning rates must be reduced to limit any structural damage of the tissue that may interfere with histopathology. It is important to note that different tissue types endure different levels of stress. For instance, for the same spinning rate, skin tissue may be perfectly preserved in structure, while brain tissue can be completely destroyed. On the other hand, if less mobile metabolites such as lipids are the focus of HRMAS MRS evaluation, faster spinning rates may be necessary. The preservation of tissue architectures during HRMAS MRS is often critical and, as such, a number of studies have explored HRMAS MRS tissue analysis using moderate to slow spinning conditions [24–27]. These studies aimed to suppress spinning side bands (SSB) that overlapped with metabolite spectral regions of interest, using spinning rates that were not fast enough to “push” the 1st SSB beyond these regions. Interested readers should test these reported techniques for applicability to their specific tissue systems. Optimal probe shimming is another critical factor that directly affects achievable spectral resolution. We have found that shimming on the lock or tissue water signal was not as sensitive as shimming on the splitting of the lactate doublet at 1.33 ppm, which fortunately presents in most excised biological tissues. However, in order to shim

Part II

contribute only to an almost invisible spectral background that is too broad (∼50 kHz) to be measurable in a typical HRMAS spectrum of metabolites (∼5 kHz, or 10 ppm on a 500 MHz spectrometer). Overall, the magnitude of line-broadenings due to magnetic susceptibility is approximately 103 less than that caused by “solid-effects,” and can be greatly reduced by MAS.

Methodology 1053

1054 Part II

Medical Uses

Part II

interactively on the degree of splitting, it is necessary to work with the frequency domain. This should not present a challenge to most current spectrometers. Furthermore, it should be possible to establish autoshimming protocols based on this criterion. It is now accepted that malignancy related cellular marker metabolites may not present simply as present or absent, but rather as continuous changes in intensity throughout disease development and progression. Quantification of these metabolites can be extremely important if they are to accurately diagnose and characterize stages of disease. Metabolite concentrations may be estimated from HRMAS MR spectra by using either the intensity of tissue water signals or an external standard (e.g. a small piece of silicone rubber) permanently attached to the inside of the rotor or attached to the rotor inserts. Such an external standard can be calibrated with known compounds of known concentrations. Interested readers can make such compounds by dissolving known amounts of relevant metabolites in agarose gel. HRMAS MR spectra may be acquired with or without water suppression, depending on whether water intensities are required for the estimation of metabolite concentrations. A rotor-synchronized CPMG sequence may be applied to achieve a flat spectral baseline if there are undesired broad resonances from the probe background.

Histopathology The clinical utility of high-resolution tissue metabolite profiles obtained with HRMAS MRS needs to be investigated and validated by means of accurate and detailed correlation with serial-section tissue pathologies. Such correlations are particularly important for studies of human malignancies due to the heterogeneity that may be inherent in the disease. Tissue pathology can vary greatly from region to region within a single tumor, and intertumor differences may be even more pronounced. An obvious advantage of HRMAS MRS is its preservation of tissue for subsequent histopathological analyses, allowing the establishment of correlations between metabolite profiles and quantitative pathologies [28–30]. Routine clinical histopathology data is most often inadequate. Routine histopathology, in particular 5 micron slices of the tissue biopsy, provides information about the presence or absence of certain features. These features may vary from slice to slice and therefore information from a small percentage of the tissue may not correlate with the spectral profile, which consists of the weighted sum of the profiles from all tissue slices or indeed the entire piece of tissue. Hence, for certain types of tissues, particularly neoplasm samples with known heterogeneities, histopathology needs to be evaluated quantitatively. This can be achieved laboriously by the histopathologist’s

examining serially sectioned tissue slices or possibly with the assistance of computer image analysis of these same sections. The sectioning frequency differs with tissue type and should be determined in consultation with the pathologist. For instance, the optimal sectioning frequency for human prostate was found to be between 200 ∼ 400 µm.

Data Analysis Histopathological analysis in cancer diagnosis relies on the observation of variations in colors and shapes using a light microscope. Importantly, to suspect disease the pathologist must observe widespread rather than isolated changes in color and/or shape, a process requiring keen pattern recognition. Similarly, pattern recognition methods are most often required to allow the diagnosis of malignancy from MRS tissue metabolite profiles. Development and progression of malignant disease involves the simultaneous evolution of many metabolic processes. Therefore, although some individual metabolites have been reported to correlate with disease types and stages [1,2] it is more likely that the overall metabolite profile rather than changes in single metabolites will be sensitive and specific for disease diagnosis [31,32]. Sophisticated statistical classification strategies (SCS) have been developed and applied to the analysis of conventional MRS of malignant tissues [33]. To date, however, principal component analysis (PCA) using readily available statistical programs has been sufficient for revealing accurate diagnostic information from HRMAS MRS data [34,35]. However, the application of SCS may further improve these accuracies.

HRMAS MRS of Human Surgical Specimens Over recent years, HRMAS MRS has been applied biomedically to the analysis of human surgical samples, research animal tissues, and cultured cells. The scope of the methodology presented in this section will be limited to HRMAS MRS studies of human tissues and will be presented where possible in context with preceding studies of the same tissues using conventional MRS. Although presented in this context, comparison of the sensitivities and specificities obtained using conventional and HRMAS MRS are not at this stage warranted as unlike the mature discipline of conventional MRS, HRMAS MRS is still in its infancy and has reported only studies with restricted patient numbers. Although the capability of HRMAS has been clearly demonstrated in generating high-resolution spectra from which individual metabolites can be measured and such measurements were impossible with conventional methods, studies of large patient populations, as with conventional MRS studies, that allow evaluation of the sensitivity and specificity of the method have not

Magic Angle Spinning MR Spectroscopy of Human Tissues

Brain Human brain was the first study reported using proton HRMAS MRS to determine tissue pathology. Due to the relative motion stability and homogeneity compared to other organs, brain has dominated the development of in vivo MRS, and accordingly has inspired many ex vivo studies, primarily including neurodegenerative diseases and tumors, aimed at understanding metabolism and defining brain tissue chemistry. The first HRMAS MRS human studies on brain tissues from autopsies described semi-quantitative evaluation of the pathology of a neurodegenerative disease, specifically Pick disease [36]. Through MRS measurements and traditional neurohistopathology, direct and semi-quantitative correlations were found to exist between the levels of N-acetyl aspartate (NAA) and the amount of surviving neurons in varying regions of examined brain as seen in Figure 2. The study also, for the first time, demonstrated that the spectral resolution of HRMAS proton MRS was comparable with that measurable with conventional proton MRS of tissue

Fig. 2. Comparison of HRMAS MR spectra of brain tissue from the relatively unaffected primary visual cortex region (a) and the severely Pick disease affected rostral inferior temporal gyrus region (b), showing a marked decrease in NAA concentration, 8.48 µmol/g for spectrum a and 4.96 µmol/g for spectrum b. This decrease in NAA was found to correlate with an average neuronal count decrease of 33 neurons per 0.454 mm2 . The spectra were acquired at 2 ◦ C, and were scaled according to concentration of creatine at 3.03 ppm for enhanced visualization. Figure 4 from ref. [36]

Part II

yet appeared in literature. Similarly, it is not as yet possible to provide the reader with one concise and optimal method for undertaking HRMAS MRS. Detailed methods are provided for each of the studies presented and these methods discussed in terms of the study aims. In an emerging discipline such as HRMAS MRS care must be taken to rigorously address methodological variables with respect to the type of tissue being analyzed and the specific information required. Increased spectral quality will most often be obtained with higher spinning speeds but this will be at the expense of tissue preservation allowing subsequent histopathological assessment of the tissue. Some tissues are less easily destroyed than others by high spinning speeds and thus preliminary experiments need to be undertaken to determine optimal parameters. MRS analyses on tissue extracts have been studied for many years on many diseases, and have formed a large body of literature. Results from these studies are worthy of close examination and review but will not, however, be included in this review due to the following: the pathology of the tissue samples most often remains incomplete and/or the degree of extraction cannot be certain.

HRMAS MRS of Human Surgical Specimens 1055

1056 Part II

Medical Uses

Part II

Fig. 3. Comparison of human brain proton MRS acquired with (a) HRMAS on intact tissue and (b) conventional method on extract solution. The spectral resolution was comparable while relative intensities for certain metabolites varied.

extracts (Figure 3), and superior to that obtained using conventional MRS of intact tissue (Figure 4). Another early study, also on autopsy brain tissues, correlated MRS data and stereological pathology for Alzheimer’s disease and confirmed NAA concentration to be proportional to neuronal density (Figure 5). In this study, 7 human brains were examined, 3 of which were Alzheimer diseased and 4 which presented as normal human control brains. Figure 5 demonstrates the quantitative nature of the NAA and neuronal count relationship, as the correlation intercepted at zero (–0.29 ± 1.15 µmol/g) [37]. The MRS study of brain tumor intact tissues using conventional MRS suggested the diagnostic importance of lipid and lipid metabolites. Kuesel and colleagues reported correlations between MR lipid signal intensities and the amount of necrosis in astrocytomas [38]. With 42 cases, they showed that the intensity of the mobile

fatty acyl –CH=CH- resonance at 5.3 ppm, differentiated 0, 1–5 and 10–40% of necrosis with statistical significance [39]. These results represented a great potential use for ex vivo MRS for astrocytoma diagnosis, in particular for differentiating Grade III and Grade IV tumors, which have radically different prognoses. The technique also ensured that necrotic foci often missed by clinical pathology were identified. Working with conventional MR spectra of brain tumor tissues, Rutter et al. attempted to categorize tumors according to 1D peak ratios (3.1–3.4 vs. 1.1– 1.5 ppm), T2 values of peaks at 1.3 ppm and cross-peaks on 2D COSY spectra (0.9 ppm with 1.35 ppm, representing methyl-methylene couplings; and 1.3 ppm with 2.05 ppm for couplings between methylene groups in fatty acids) [40]. With 38 samples studied from 33 subjects (including normal tissue, astrocytoma, GBM, meningiomas, and metastases), they were able to use peak ratios to

Magic Angle Spinning MR Spectroscopy of Human Tissues

differentiate GBM from astrocytomas and normal tissues. They also found that the T2 of the 1.3 ppm peak could be fitted by a double exponential function and that the long fraction of T2 values could be used to group both GBM and metastasis from normal tissue. However, tissue spectra measured with conventional method of low resolution

20

NAA

total

(µmol/g)

AD n = 3 15

10

All n = 7 r = 0.828 P = 0.021 Ctrl n = 4 r = 0.989 P = 0.011

5

0 0

0.5

1

1.5

2

2.5

3

Neuronal Counts (×104/mm3) Fig. 5. A statistically significant linear correlation was found to exist between the number of neurons and the concentration of NAAtotal , measured from 3 Alzheimer diseased and 4 normal control brains (the dotted line, r = 0.828, P = 0.021). The solid line represents the linear correlation obtained when only normal control brains were included (r = 0.989, P = 0.011). Figure 3 from ref. [37].

prevented these studies from observation of individual brain metabolites other than identification of broad lipid peaks. HRMAS MRS of brain tumors allowed for the firsttime detailed profiles of water-soluble metabolites and lipids to be obtained from the same spectra [31]. In a study of 19 brain tumors, including astrocytomas, GBM, meningiomas, Schwannomas, and normal brain, metabolite concentrations both in absolute units and relative ratios normalized to the creatine resonance at 3.03 ppm were reported and T2 values for these metabolites were measured ex vivo for the first time (Table 1). Metabolite concentrations from intact tissues and tissue extracts from the same tumors were compared to each other and to the literature values. While the concentrations for some metabolites, measured by HRMAS MRS, were similar to those in extracts, others showed much higher concentrations in tissue than in extracts. Metabolite concentrations and T2 values accurately differentiated tumor types based on clinical data, but detailed histopathology of the tissue samples was not performed due to the false belief at the time that HRMAS MRS damaged tissue such that subsequent histopathological analysis was not possible. The important fact that accurate histopathological evaluation of tissue samples after HRMAS MRS is possible and the necessity of performing quantitative pathology on the same tissues after MRS was reported by Dr. Anthony and colleagues [28]. Tissue pathologies from both HRMAS MRS analyzed tissue and adjacent tissue that had not undergone HRMAS MRS before histopathological evaluation were analyzed semi-quantitatively for each region of the brain tumor. The quantitative histopathological data obtained from the study showed that adjacent specimens from the same tumor region shared similar histopathological features. Although quantitative differences were noted, these differences were most likely due to extensive tumor microheterogeity and not the result of the HRMAS MRS procedure. Futhermore, correlations between the amount of tumor necrosis and the concentrations of mobile lipids (R 2 = 0.961, p < 0.020) and lactate (R 2 = 0.939, p < 0.032), as well as between the numbers of glioma cells and the ratio of phosphocholine (PC) to choline resonances (R 2 = 0.936, p < 0.033) were observed. The strong linear correlation between tissue necrosis (%area) and lipids (mM) indicated that the amount of tissue necrosis can be estimated using the measured concentration of lipids from HRMAS MRS, and that according to the long T2 s these lipids are relatively mobile consistent with them being products of cell membrane degradation. Additionally, the results of the correlation obtained between the number of glioma cells and the phospocholine to choline resonances suggested the importance of measuring and quantifying these two resonances separately, which is difficult with both in vivo and ex vivo

Part II

Fig. 4. Comparison of human brain proton MRS acquired with (a) conventional MRS and (b) HRMAS MRS. Figure 2 from ref. [36].

HRMAS MRS of Human Surgical Specimens 1057

1058 Part II

Medical Uses

Part II

Table 1: Matrix of selected brain metabolite concentrations measured with HRMAS MRS for differentiation between different pathological specimens NAA, in the table, includes both measured resonances of NAA at 2.01ppm and acetate at 1.92 ppm (see text for details); Numbers in parentheses represent resonance chemical shift in ppm. The resonance at 3.93 is tentatively assigned to the Cr metabolite. As an example of the use of this matrix, the Chol resonance can be used to differentiate low-grade/anaplastic astrocytomas from GBMs with a significance of p < 0.05. Similarly, the glycine resonance (Gly) can be used to distinguish GBMs from Schwannomas with a p < 0.005 Normal Normal

LG and AAa NAAb Lac

(1.33)c

LG&AA GBMs

GBMs 3.93)c

Cr (?, Gly (3.55)d Chol (3.20)b Cr (3.03)b NAAf Lac (1.33)b Chol (3.20)e

Schwannomas (3.20)b

Chol NAAb Lac (1.33)d

Meningiomas Cr (?, 3.93)b Chol (3.20)e Glu (2.35)d Ala (1.48)b

Glu (2.35)b Gly (3.55)d

Gly (3.55)b

Schwannomas Meningiomas a LG

and AA, low-grade/anaplastic astrocytomas < 0.05, two-tailed student’s t test. c p < 0.0005, two-tailed student’s t test. d p < 0.005, two-tailed student’s t test. e p < 0.05, calculated according to one-tailed student’s t test, based on the hypothesis that chol increases in tumors. f p < 0.00005, two-tailed student’s t test. bp

conventional MRS methods, but readily achievable with MAS. It is expected that HRMAS MR spectra of tissue rather than MRS of tissue extracts will provide metabolite information more closely related to that in the in vivo brain. However, cautions should be exercised when comparing in vivo MRS and HRMAS MRS data, as in vivo MRS will always be broad-line in nature. Nevertheless, a number of studies, both in adult [41] and pediatric [42] brain tumors, have concluded that there is good agreement between in vivo and ex vivo tissue MR spectra with high-resolution ex vivo results providing both insight into which metabolites reside within the broad resonances observed in vivo and a link between in vivo MRS evaluations and neuropathologies.

Prostate The search for marker metabolites of prostate cancer has been inspired by the current state of prostate cancer pathology, wherein more than 70% of newly diagnosed cases are categorized with similar Gleason scores (6 or 7), but for which individual patient outcomes within these tumors are drastically different and unpredictable. Hahn and colleagues reported the first intact prostate tissue conventional MRS study of 66 benign prostatic hyperplasia (BPH) and 21 prostate cancer samples from

50 patients [43]. They divided the proton spectral region between 0.5 and 3.55 ppm into 50 equal subregions, and applied multivariate linear-discriminant analysis to the point-reduced spectra. The study found six spectral regions including those containing citrate, glutamate, and taurine to be sensitive in differentiating BPH from cancer with an overall accuracy of 96.6%. This algorithm was tested by the same research group on another group of 140 samples from 35 patients after radiotherapy to test for the sensitivity of spectroscopy analysis in differentiating cancer positive vs. cancer negative samples [44]. After eliminating 24 samples that did not have sufficient signal-to-noise ratios, they reported, with the remaining 116 spectra, the sensitivity and specificity of tissue spectra in identifying cancer samples to be 88.9 and 92%, respectively. Van der Graaf and colleagues presented another interesting study of intact prostate tissue combining conventional MRS and high-pressure liquid chromatography (HPLC) analysis to measure the relationship between polyamines (PA) and prostate cancer [45]. Although they observed PA in the proton spectra and measured statistically significant drops of PA in cancer samples with HPLC, no correlation between PA levels measured by MRS and those determined by HPLC were presented for the same cases due to very limited number of samples analyzed. Nevertheless, the study suggested the existence

Magic Angle Spinning MR Spectroscopy of Human Tissues

HRMAS MRS of Human Surgical Specimens 1059

Part II

Cho PA Cr Citrate

PPM

c

3.0

2.5

2.0

4.0

Myo-Inositol

S-Inositol

b

d Acetone (impurity)

3.5

Lactate

GPC+PC

Cr

3.0

Glu/Gln

PA

Alanine

Cho

Taurine

Lactate Myo-Inositol

a

Citrate

2.5

2.0

1.5

Lipids

1.0

PPM

Fig. 6. Results from a presurgical 3D-MRSI of a 56-year-old prostate cancer patient were concordant with histopathologically classified malignancy of H and E stained tissue samples of excised tissue with Gleason 3 + 4 prostate cancer, d. Spectrum b, shows elevated levels of choline and low levels of citrate and polyamines relative to creatine in the 0.24 cm3 voxel of a. The postsurgical HRMAS MR spectrum (c) confirms the 3D-MRSI results with enhanced resolution that also identifies elevated levels of GPC + PC relative to choline. Figure 2 from ref. [30]. (See also Plate 88 on page XVI in the Color Plate Section.)

of a potential prostate biomarker that MRS may be able to quantify, as well as a direction for future studies. More recently, Mountford et al. reported a conventional MRS study of 71 prostate samples from 41 patients who underwent both cancer and non-cancer prostate surgeries [46]. By using peak ratios 3.2/3.0 ppm (choline-to-creatine) and 1.3/1.7 ppm (lipid-to-lysine), they were able to differentiate malignant from benign tissues with 97% sensitivity and 88% specificity. Tomlins and colleagues experimented with proton HRMAS MRS analysis on human prostate tissues with both 1D and 2D MR spectroscopy. In their qualitative report, they confirmed the usefulness of HRMAS MRS in producing high-resolution spectra from intact human prostate tissues [47]. A second HRMAS MRS study of human prostate tissues from 16 patients was the first to include quantitative pathology on the MRS specimens and to include multiple subjects [29]. The results of the study proved the validity of HRMAS MRS for the accurate determination of tissue histopathology. Both citrate and spermine were quantified from the tissue HRMAS MR spectra and shown to be linearly correlated with the amounts of prostate normal epithelium. In 2003, Swanson and colleagues reported an interesting study of HRMAS

MRS of tissues from 26 patients harvested post surgically under the guidance of 3D-MRSI from lesions that had been analyzed using in vivo MRS prior to prostatectomy. Figure 6 shows the resulting spectra from a 56-year-old prostate cancer patient [30]. By combining the MRS results with quantitative pathology of the same tissue after spectroscopy, metabolite discriminators (i.e. ratios of citrate, polyamine, and choline compounds to creatine) were found to differentiate normal prostate epithelial tissue from cancer and stromal tissue. Furthermore, a correlation between the intensity of MIB-1 immunohistochemical staining and the ratio of choline to creatine resonances was reported, supporting their findings in vivo. These results were dependent on the assumption that the creatine concentration did not alter during the disease process, which awaits verification.

Breast Human breast tissue was difficult to analyze using conventional proton MRS due to high lipid content. As pioneers of this work, Mountford and colleagues noted that to acquire a spectrum that was diagnostically meaningful,

1060 Part II

Medical Uses

Part II

fine-needle aspiration biopsy (FNAB) instead of regular tissue samples had to be utilized [48]. In a study of 218 FNAB samples from 191 patients with benign lesions, ductal carcinoma in situ (DCIS), and invasive carcinoma, they found that by using the resonance peak height ratio threshold of 1.7 (3.25 vs. 3.05 ppm; or choline-compounds vs. creatine), benign lesions could be differentiated from carcinoma with 95% sensitivity and 96% specificity. However, this single ratio identifier only worked if the signalto-noise ratio, particularly for choline peaks at 3.25 ppm, was above 10. To extend the MRS differentiation capability below this limit, and more importantly to search for a more robust diagnostic protocol of computerized spectral analysis, Mountford et al., along with Smith et al. at the NRC, Canada, used a pattern recognition method, SCS, developed to utilize the entire MR spectrum instead of only the aforementioned two peaks for the purpose of analysis [9]. From measurements conducted on 140 samples, they reported an overall accuracy of 93% in distinguishing benign from malignant tumors using SCS. Furthermore, they reported that using SCS classifiers, lymph node involvement and tumor vascular invasion could be predicted with 95 and 94% overall accuracies, respectively. The advantage of SCS in relating proton FNAB spectra with breast cancer diagnosis and possibly prognosis is evident and its potential in clinical usage is apparent. However, the more advanced the SCS, the less evident were the direct connections with individual metabolites. HRMAS MRS overcame many of the limitations of conventional MRS32. Even with such lipid-rich tissue, HRMAS MRS successfully produced high-resolution proton spectra. Results from a study at 400 MHz of 19 cases of ductal carcinomas showed that both the high fat contents and individual cellular metabolites could be measured from the same spectrum. Particularly, from these spectra, PC could be quantified separately from choline; and the ratio between PC and choline was found to correlate with tumor grades, as shown in Figure 7. A more recent detailed HRMAS MRS study at 600 MHz of 10 ductal carcinomas by Sitter et al. compared HRMAS MR spectra with those obtained using conventional MRS of perchloric acid tissue extracts [20]. The study concluded that for breast tissue, HRMAS MRS was able to achieve spectral resolutions approaching those obtained using conventional MRS of extracts. 2D J-resolved and COSY MRS was used to accurately assign metabolites.

Fig. 7. Examples of observed correlations between histopathological grades of breast ductal carcinomas and the means for metabolic intensities, particularly PC over choline (A) and lactate over choline (B) measured with HRMAS MRS. Of note, the HRMAS method provided a means to differentiate PC from choline in the such a lipid-rich tissue. Figure 4 from ref. [32].

distinguish invasive from pre-invasive epithelial malignancy [50]. More recently, Sitter and colleagues reported the use of HRMAS MRS of cervical tissue from eight cancer and eight non-cancer patients [.35]. In addition to their presentation of detailed resonance assignments for cervical metabolites, they were able to use PCA to separate cancer from non-cancer samples with the first principal component (PC1). This principal component was comprised primarily of lactate, the methyl, and methylene groups of lipids, and, to a lesser extent, the choline-containing compounds. This component represented 63% of the variations in the spectra.

Cervix Kidney Mountford and colleagues began analysis of human cervical biopsies with conventional proton MRS in the early 1990s [49]. Broad-line resonances at 0.9 ppm (CH3), 1.3 ppm (CH2), and 3.8–4.2 ppm (CH) were found to

HRMAS MRS analysis of kidney tissue has been shown to diagnose renal cell carcinoma based on an increased intensity of lipid resonances in malignant compared to

Magic Angle Spinning MR Spectroscopy of Human Tissues

1.0

(a)

.5

PC1

0.0 −0.5 −1.0 −1.5 −2.0 −2.5 −3

−2

−1

0

1

2

sample was not spun [53,54]. For this initial study, spectra of liposarcoma and lipoma were dominated by signals from lipids, interfering with the ability to observe resonances from other metabolites. More recently, Chen and colleagues have developed a double pulsed field gradient selective echo (DPFGSE) technique in combination with HRMAS MRS to selectively excite spectral regions that contain less abundant non-lipid metabolites [23]. With their new technique they were able to observe and quantify phosphatidylcholine (PTC), PC and choline, and found, using six paired samples from six patients, that the ratio of PTC/PC differentiated normal fat from lipomalike well-differentiated liposarcoma with statistical significance ( p < 0.001).

PC2

(b)

Lymph Nodes Similarity 0.0

0.5

1.0

Fig. 8. Using PCA (a) and HCA (b), tumor samples can be differentiated from non-tumor samples. Key: ( r) normal tissue; (
) renal cell carcinoma; () sample from bladder metastasis in renal collecting duct; and () lung metastasis in renal cortex. Figure 3 from ref. [.35].

control tissues [51,52]. These studies were not designed to test the utility of HRMAS MRS in the diagnosis of renal cell carcinoma, but rather to illustrate the feasibility of HRMAS MRS for assignment of resonances and identification of metabolites as well as to develop twodimensional HRMAS MRS techniques for the assessment of intact tissues. Later, a study of 22 paired control and tumor samples of human renal cortex used computer-based pattern recognition techniques to show that using PCA and hierarchical cluster analysis (HCA), tumor samples were differentiated from non-tumor samples, as shown in Figure 8 [34].

Sarcoma HRMAS MRS has been shown to be a useful technique in the analysis of human sarcoma tissue. Singer and colleagues reported a comparison of 2D TOCSY spectra with and without HRMAS MRS, which demonstrated the superior resolution of the MAS technique by revealing less intense cross peaks that were not observed when the

Accurate detection of metastatic deposits in lymph nodes is often the most important predictor of cancer patient prognosis. Histopathology is subject to sampling and observer error and thus there is a real need for a new, rapid, and cost-effective method with high accuracy for this purpose. An early study identifying malignant cells in lymph node tissue using 1D 1 H MRS was limited by high fat content of the tissue and the low spectral resolution obtained. Two-dimensional (2D) techniques successfully resolved resonances and allowed the complex spectra from lymph nodes to be assigned and correlated with detailed histopathology [10]. Choline and lactate were found to be key metabolites diagnostic of the presence of metastases in lymph nodes using the 2D method. However, these 2D MR experiments require long acquisition times (4–5 h) during which sample degradation may occur, potentially compromising the outcome of the measurement. In addition, conventional 1D and 2D techniques cannot address the problem of resonance broadening directly and thus useful diagnostic information may be missed. Alternatively, MRS of FNAB of lymph node tissues has been used to diagnose the presence of metastases in lymph nodes from melanoma patients [55]. Spectra of node tissue containing metastatic melanoma were characterized by the presence of a distinct resonance from choline-containing metabolites at 3.2 ppm. The ratio of the integrals of resonances from lipid and metabolites (1.8–2.5 ppm region) and “choline” (3.1–3.3 ppm region) distinguished benign nodes (38.8 ± 34.3) from melanoma containing nodes (7.2 ± 6.8) with p < 0.012 (separate ttest) [55]. This method assumes the chemistry of the node detectable by MRS changes immediately upon the node being infiltrated by metastatic cells. This is supported by studies in a rat lymph node metastasis model where MRS detected the presence of malignant cells prior to clusters of metastatic cells being identified by histopathology.

Part II

1.5

HRMAS MRS of Human Surgical Specimens 1061

1062 Part II

Medical Uses

Part II

Instead, single malignant cells were observed unclustered throughout the node, the presence of which was confirmed by growing these nodes in nude mice, which subsequently developed tumors. Datasets in the study of MRS on FNAB as a method of diagnosing lymph node metastases remain, however, small and until larger databases confirm these findings the possibility that the method introduces a sampling error cannot be excluded. The application of HRMAS MRS to intact tissue analysis was in fact initiated from the proton MR detection of lymph node metastases in a rat model for breast cancer metastasis. Since then, HRMAS MRS has also been successfully applied to the proton MR analyses of human lymph nodes from breast cancer patients. Malignant nodes were distinguished from reactive nodes based on the peak height ratio of the smallest to largest peaks resolved in the 0.9 ppm resonance multiplied by the peak height ratio of the choline and creatine resonances at 3.2 and 3.0 ppm, respectively (Figure 9). Three of four axillary nodes from breast cancer patients that were clinically suspicious but diagnosed cancer free on routine histopathology had an HRMAS MRS ratio consistent with reactive nodes. The fourth node in this category was, by HRMAS MRS criteria, malignant. Review of the histology of these nodes using increased magnification and immunohistochemical staining identified malignancy in the fourth node missed during routine examination.

Fig. 9. HRMAS MRS of Human Lymph Nodes: Malignant nodes were distinguished from normal (reactive) nodes based on the MRS Peak Height Ratio (ratio of the smallest to largest peaks resolved in the 0.9 ppm resonance multiplied by the ratio of the choline and creatine resonances at 3.2 and 3.0 ppm). One of four axillary nodes that were clinically suspicious but cancer free on routine histopathology had an HRMAS MRS ratio consistent with malignancy. Review of the histology for all four nodes identified malignancy in this one node missed by routine histopathology.

Future Developments and Conclusions The development of proton HRMAS MRS for tissue analysis has not only drastically simplified the procedure of obtaining high resolution cellular metabolite spectra directly from intact tissue, but more importantly allows quantitative pathology to be conducted on the same tissue after MRS measurements for correlation with individual metabolite concentrations or collective alterations in overall metabolite profiles. The high-resolution spectra generated by HRMAS MRS contain many more resonance peaks for analysis than conventional MRS. This greatly improves the likelihood that sophisticated data analysis methods can accurately distinguish between differing pathologies. However, the increased discriminatory power of the method relies on precise histopatholgical information about the tissue assessed by HRMAS MRS being available. This is only possible when detailed serial section histopathological methods are undertaken to substantially increase the precision of the information currently obtained from clinical pathology.

Glossary of Terms Astrocytoma—The most common type of brain tumor in adults; astrocytomas, known for their marked potential for malignant progression and infiltrative nature, can be classified histopathologically into three grade of malignancy: WHO Grade II astrocytoma, WHO Grade III anaplastic astrocytoma, and WHO Grade IV GBM. Benign Prostatic Hyperplasia—Enlargement of the prostate and more specifically, overgrowth of the epithelium and fibromuscular tissue of the transition zone and periurethral area. Ductal Carcinoma In Situ—Noninfiltrative lesions composed of malignant epithelial cells that are confined to the mammary ducts and lobules. Glioblastoma Multiforme—The most malignant grade of astrocytoma (WHO Grade IV), which has an overall survival time of less than 2 years for most patients. These tumors are histopathologically characterized according to their dense cellularity, high proliferative indices, endothelial proliferation, and most importantly the presence of focal tissue necrosis. Lipoma—Commonly diagnosed benign tumors of adipose tissue that can develop anywhere in the body where fat is normally found. Liposarcoma—A commonly diagnosed soft tissue sarcoma in adults which be found anywhere in the body and for which there exists several types with varying clinical outcomes. Meningioma—Intracranial tumors that arise in the meninges and compress the underlying brain. In general, many of these tumors are benign, however, others are

Magic Angle Spinning MR Spectroscopy of Human Tissues

Acknowledgments The authors thank their former and present staff, students, and collaborators involved in the research presented in this review whom have all worked diligently towards developing MRS and HRMAS MRS for determining tissue pathology. We also thank our colleagues and friends from around the world for allowing us to describe their work and for their many interesting discussions and contributions to this review. Particular thanks to Carolyn Mountford, Peter Russell, and Peter Malycha for their encouragement and support. We thank Sinead Doran and Deborah Edward for helping prepare the manuscript. LLC and MAB acknowledge grant supports from PHS/NIH CA095624 and from DOD W81XWH-04-1-0190.

References 1. Mountford C, Doran S, Lean C, Russell P. Chem. Rev. 2004;104:3677. 2. Lean C, Somorja R, Smith I, Russell P, Mountford C. In: AG Webb (Ed). Accurate Diagnosis and Prognosis of Human Cancers by Proton MRS and A Three Stage Classification Strategy. Guildford, UK, 2002.

3. Pomper MG. In VT DeVita Jr, S Hellman, SA Rosenberg (Ed). Functional and Metabolic Imaging. Philadelphia, 2001. 4. Nicholson JK, Lindon JC, Holmes E. Xenobiotica 1999;29: 1181. 5. Bettelheim R. Price K, Gelber R, Davis B, Castiglione M, Goldhirsch A, Neville A, Lancet 1990;335:1565. 6. Mountford CE, Lean CL, Mackinnon WB, Russell P. In GA Webb (Ed). The Use of Proton MR in Cancer Pathology in Annual Reports on NMR Spectroscopy. Academic Press. London 1993;27:173. 7. Russell P, Lean C, Delbridge L, May G, Dowd S, CE Mountford Am. J. Med. 1994;96:383. 8. Doran ST, Falk GL, Somorjai RL, Lean CL, Himmelreich U, Philips J, Russell P, Dolenko B, Nikulin AE, Mountford CE. Am. J. Surg. 2003;185:232. 9. Mountford CE, Somorjai RL, Malycha P, Gluch L, Lean C, Russell P, Barraclough B, Gillett D, Himmelreich U, Dolenko B, Nikulin AE, Smith IC. Br. J. Surg. 2001;88:1234. 10. Mountford C, Lean C, Hancock R, Dowd S, Mackinnon W, Tattersall M, Russell P. Invasion and Metastasis, 1993; 13:57. 11. Kuesel A, Kroft T, Saunders J, Prefontaine M, Mikhael N, Smith I. Magn. Reson. Med. 1992;27:340. 12. Andrew E, Bradbury A, Eades R. Nature 1958;182:1695. 13. Lowe I. Phy. Rev. Lett. 1959;2:285. 14. VanderHart D, Earl W, Garroway A. J. Magn. Reson. 1981;44:361. 15. Maricq M, Waugh J. J. Chem. Phys. 1979;70:3300. 16. Cheng LL, Lean CL, Bogdanova A, Wright SC Jr, Ackerman JL, Brady TJ, Garrido L. Magn Reson Med. 1996;36:653. 17. Middleton DA, Bradley DP, Connor SC, Mullins PG, Reid DG, Magn. Reson. Med. 1998;40:166. 18. Garrod S, Humpfer E, Spraul M, Connor SC, Polley S, Connelly J, Lindon JC, Nicholson JK, Holmes E. Magn. Reson. Med. 1999;41:1108. 19. Waters NJ, Garrod S, Farrant RD, Haselden JN, Connor SC, Connelly J, Lindon JC, Holmes E, Nicholson JK. Anal. Biochem. 2000;282:16. 20. Sitter B, Sonnewald U, Spraul M, Fjosne HE, Gribbestad IS. NMR Biomed. 2002;15:327. 21. Wu CL, Taylor JL, He WL, Zepeda AG, Halpern EF, Bielecki A, Gonzalez RG, Cheng LL. Magn. Reson. Med. 2003;50:1307– 1311. 22. Bourne R, Dzendrowskyj T, Mountford C. NMR Biomed. 2003;16:96. 23. Chen JH, Enloe BM, Fletcher CD, Cory DG, Singer S. J. Am. Chem. Soc. 2001;123:9200. 24. Wind RA, Hu JZ, Rommereim DN. Magn. Reson. Med. 2001;46:213. 25. Hu JZ, Rommereim DN, Wind RA. Magn. Reson. Med. 2002;47:829. 26. Hu JZ, Wind RA. J. Magn. Reson. 2003;163:149. 27. Taylor JL, Wu CL, Cory D, Gonzalez RG, Bielecki A, Cheng LL. Magn. Reson. Med. 2003;50:627. 28. Cheng LL, Anthony DC, Comite AR, Black PM, Tzika AA, Gonzalez RG, Neuro-oncol. 2000;2:87. 29. Cheng LL, Wu C, Smith MR, Gonzalez RG. FEBS Lett. 2001;494:112. 30. Swanson MG, Vigneron DB, Tabatabai ZL, Males RG, Schmitt L, Carroll PR, James JK, Hurd RE, Kurhanewicz J. Magn. Reson. Med. 2003;50:944.

Part II

malignant with the capability to metastasize, both locally and distally. Metabolite Profiling—The study of small molecules, such as lipids, peptides, amino acids, and carbohydrates, which represent steady-state concentrations of intermediate products or end products of cellular processes and as a result can be thought of as the ultimate response to genetic and environmental stimuli. Necrosis—Tissue and/or cell death. Neoplasm—Literally meaning “new growth.” An abnormal growth of tissue which may be benign or malignant. Oncogenomics—The study of genes, gene sequences, and the underlying genetic alterations that appear to be involved in oncological pathways. Reactive Nodes—Lymph nodes that have been immunologically challenged. Schwannoma—Benign tumors that arise on peripheral nerves in general and on cranial nerves in particular, especially on the vestibular portion of the eighth cranial nerve. Sensitivity—The term relating to the percentage of people with disease who test positive for the disease, i.e. the percentage of patients with cancer who have positive biopsy results. Specificity—The term referring to the percentage of people without disease who test negative for the disease, i.e. patients without cancer with cancer negative biopsy results.

References 1063

1064 Part II

Medical Uses

Part II

31. Cheng LL, Chang IW, Louis DN, Gonzalez RG. Cancer Res. 1998;58:1825. 32. Cheng LL, Chang IW, Smith BL, Gonzalez RG. J. Magn. Reson. 1998;135:194. 33. Somorjai R, Dolenko B, Nikulin A, Pizzi N, Scarth G, Zhilkin P, Halliday W, Fewer D, Hill N, Ross I, West M, Smith I, Donnelly S, Kuesel A, Briere K. J. Magn. Reson. Imaging. 1996;6:437. 34. Tate AR, Foxall PJ, Holmes E, Moka D, Spraul M, Nicholson JK, Lindon JC. NMR Biomed. 2000;13:64. 35. Sitter B, Bathen T, Hagen B, Arentz C, Skjeldestad FE, Gribbestad IS. MAGMA. 2004;16:174. 36. Cheng LL, Ma MJ, Becerra L, Ptak T, Tracey I, Lackner A, Gonzalez RG. Proc. Natl. Acad. Sci. U.S.A. 1997;94:6408. 37. Cheng LL, Newell K, Mallory AE, Hyman BT, Gonzalez RG. Magn. Reson. Imag. 2002;20:527. 38. Kuesel A, Donnelly S, Halliday W, Sutherland G, Smith I. NMR Biomed. 1994;7:172. 39. Kuesel AC, Briere KM, Halliday WC, Sutherland GR, Donnelly SM, Smith ICP. Anticancer Res. 1996;16:1485. 40. Rutter A, Hugenholtz H, Saunders J, Smith I. J. Neurochem. 1995;64:1655. 41. Barton SJ, Howe FA, Tomlins AM, Cudlip SA, Nicholson JK, Bell BA, Griffiths JR. Magma 1999;8:121. 42. Tzika AA, Cheng LL, Goumnerova L, Madsen JR, Zurakowski D, Astrakas LG, Zarifi MK, Scott RM, Anthony DC, Gonzalez RG, Black PM. J. Neurosurg. 2002;96:1023. 43. Hahn P, Smith I, Leboldus L, Littman C, Somorjai R, Bezabeh T. Cancer Res. 1997;57:3398.

44. Menard C, Smith IC, Somorjai RL, Leboldus L, Patel R, Littman C, Robertson SJ, Bezabeh T. Int. J. Radia. Oncol. Biol. Phys. 2001;50:317. 45. van der Graaf M, Schipper RG, Oosterhof GO, Schalken JA, Verhofstad AA, Heerschap A. Magma 2000;10:153. 46. Swindle P, McCredie S, Russell P, Himmelreich U, Khadra M, Lean C, Mountford C. Radiology 2003;228:144. 47. Tomlins A, Foxall P, Lindon J, Lynch M, Spraul M, Everett J, Nicholson J. Analy. Common. 1998;35:113. 48. Mackinnon WB, Barry PA, Malycha PL, Gillett DJ, Russell P, Lean CL, Doran ST, Barraclough BH, Bilous M, Mountford CE. Radiology, 1997;204:661. 49. Mountford CE, Delikatny EJ, Dyne M, Holmes KT, Mackinnon WB, Ford R, Hunter JC, Truskett ID, Russell P. Magn. Reson. Med. 1990;13:324. 50. Delikatny E, Russell P, Hunter J, Hancock R, Atkinson K, van Haaften-Day C, Mountford C. Radiology, 1993;188:791. 51. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall P, Nicholson J, Lindon J. Analy. Common. 1997;34:107. 52. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall PJ, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1998;17:125. 53. Millis KK, Maas WE, Cory DG, Singer S. Magn. Reson. Med. 1997;38:399. 54. Millis K, Weybright P, Campbell N, Fletcher JA, Fletcher CD, Cory DG, Singer S. Magn. Reson. Med. 1999;41:257. 55. Lean CL, Bourne R, Thompson JF, Scolyer RA, Stretch J, Li LX, Russell P, Mountford C. Melanoma Res. 2003;13:259.

1065

Isabelle Latour and Garnette R. Sutherland Division of Neurosurgery, Seaman Family MR Research Centre, Foothills Hospital, Calgary, AB, Canada T2N 1N4

Nothing is certain and everything changes ——Leonardo da Vinci

Contemporary neurosurgeons depend on precise lesion localization to maximize surgery and clinical outcome. Advances in neurosurgical technique and instrumentation evolved as surgical corridors became more precise. Magnetic resonance imaging (MRI) has considerably improved the diagnosis, treatment, and follow-up of patients with neurological disease. With the widespread use of MRI, it is difficult to conceive that, before the introduction of imaging technology, neurosurgery relied on precarious tools.

Historical Milestones in Neurology Cranial trephination was used thousands of years ago by Neolithic man to liberate individuals thought to be possessed by evil spirits [1,2]. It resurfaced as a treatment for infantile convulsions and was still performed as late as the mid-19th century for both trauma and epilepsy [3]. Franz Joseph Gall proposed that intellectual faculties are highly compartmentalized in the brain and that examination of the surface of the skull reflects the extent to which an aptitude is developed [4]. The use of scientific evidence to determine cortical compartmentalization of function was first initiated by Paul Broca [5]. The development of imaging techniques began with the discovery of X-rays by Wilhelm R¨ontgen in 1895. In 1918, Walter Dandy made the serendipitous discovery of intraventricular air in a patient who had suffered cranial trauma [6]. This observation led to the development of pneumoencephalography (Figure 1) as a method for localizing intracranial pathology, based on ventricular displacement by cerebral mass lesions. A decade later, Moniz introduced cerebral angiography, a method still used to visualize vascular abnormalities [7]. Ultrasound was subsequently developed as a non-invasive method of localizing central nervous system (CNS) pathology [8]. Originating from ideas of Cormack [9] and Hounsfield [10], brain imaging was revolutionized in the mid-1970s by the invention of computerized tomography (CT). This discovery significantly enhanced diagnostic accuracy and improved surgical planning and outcome. The subsequent development of MRI by Lauterbur [11] and Mansfield [12] provided enhanced tissue contrast Graham A. Webb (ed.), Modern Magnetic Resonance, 1065–1076.
C 2008 Springer.

and remains the best imaging modality for diagnosis and determining the effect of treatment on both the lesion and the brain.

Principles of Intraoperative Imaging Intraoperative imaging allows acquisition of near realtime images in the operating room (OR). As surgery often disrupts the brain environment, acquiring instant images at any time during and after completion of the surgery represents a considerable advantage. CT imaging was introduced into the OR in 1982 [13]. The first report of intraoperative MRI (iMRI), by Black et al. was in 1997 [14]. The use of intraoperative CT declined following the introduction of iMRI, in part due to lower tissue contrast and image artifact due to surgery and instrumentation. MR technology confers non-invasive, radiation-free, multiplanar, real-time, high-resolution images of the nervous system. MRI systems that include magnets of ≥1.5 T and high performance gradients provide a unique method for the evaluation of diffusion and perfusion tensor imaging, as well as a way to assess brain function. Diffusion images measure the mobility of water molecules in tissue. As pathological states can disrupt barriers that normally restrict water motion, diffusion imaging reveals information about tissue integrity [15]. Perfusion images represent the rate at which blood is delivered to the tissue, and can therefore be used to identify areas displaying an increase or decrease in blood flow [16]. Functional MRI allows in a non-invasive way the precise identification of brain areas undergoing a sharp increase in oxygen supply induced by a certain cognitive process or task [17]. Essential components of an iMRI system are the magnet, gradients, radio frequency (RF) coil designs, the operating table, and the RF shielding. Real-time, three-dimensional (3D) navigation software allows the neurosurgeon to localize surgical instruments and intracranial targets in 3D space. This improves localization of intracranial pathology from which more minimalist approaches may be directed.

Hardware and Configuration Three different types of magnets can be used for MRI: permanent, resistive, and superconducting. iMRI has utilized

Part II

Intraoperative MRI

1066 Part II

Medical Uses

Part II

Fig. 1. Coronal (A) and lateral (B) pneumoencephalography. The lateral ventricles are visible and show no evidence of displacement.

magnets of 0.12–1.5 T, with gradient strength ranging from 12 to 33 mT/m (Table 1). Gradient strength and performance is important for rapid imaging sequence, in particular echo-planar imaging (EPI), diffusion, perfusion, and fMRI. For iMRI, inductively coupled [18] or volume coils [19] can be used. The full spectrum of imaging sequences including T1 , T2 , MR angiography, diffusion, perfusion, and fluid attenuated inversion recovery (FLAIR) have been applied to the evaluation of the utility of iMRI as an adjunct to surgery. The first iMRI system, manufactured by General Electric, consisted of a 0.5 T vertical, biplanar double doughnut magnet configuration [14]. A significant advantage of this setup is that real-time images can be acquired frequently, without interrupting surgery nor having to move the patient or the magnet. Although this initial system represented a major advance in intraoperative imaging technology, it has limitations. The system requires positioning the patient and surgeon between two magnetic poles. This results in a constrained working environment, restricted patient positioning, and the requirement for MR-compatible instrumentation. Furthermore, the magnetic field strength and gradient performance are below that needed for contemporary EPI. To

partly overcome these issues, 0.2 and 0.3 T, horizontal, biplanar imaging systems were produced by Siemens and Hitachi, respectively [20,21]. For these, the surgical area is located outside of the magnetic field and the patient is moved into the magnet for image acquisition. Physically separating surgery and imaging resolved the problem of operating within a magnetic field. However, this requires interruption of surgery and moving the anesthetized, monitored patient from the surgical site to the magnet. In need of better image resolution and to maintain a patient focused environment, a movable, ceiling-mounted wide bore 1.5 T magnet was developed and manufactured by IMRIS [22–24]. With this system, conventional neurosurgical procedures are maintained and the magnet is moved to the patient for imaging (Figure 2). The higher magnetic field and improved gradient performance provides highresolution and rapid imaging including echo planar. The system includes a versatile MR-compatible OR table accommodating various patient positioning including prone, supine, and lateral. Philips and Siemens have recently developed and installed 1.5 T iMRI systems based on stationary magnets. In the Siemens configuration, surgery takes place in the fringe field and, for imaging, the OR

Table 1: iMRI systems with main advantage and disadvantage

Manufacturer Odin Siemens Hitachi General Electric Philips Siemens IMRIS

Magnet strength

Gradient performance

0.12 0.2 0.3 0.5 1.5 1.5 1.5

25 mT/m 20 mT/m 21 mT/m 12 mT/m 33 mT/m 30 mT/m 20 mT/m

Main advantage

Main disadvantage

Portable Unrestricted instrumentation Unrestricted instrumentation Real-time imaging Image resolution Image resolution Image resolution

Image resolution Image resolution Need to move patient for imaging Requires MR-compatible tools Need to move patient for imaging Need to move patient for imaging Interruption of surgery for imaging

Intraoperative MRI

Clinical Applications of iMRI 1067

table is rotated to the magnet [25]. Philips designed a 1.5 T interventional/neurosurgery suite where the patient is moved by a floating tabletop from the operative table to the imaging platform for image acquisition [26]. Using the Philips system, surgeons have performed stereotaxy by reaching into the magnet, using iMRI for accurate instrument placement. To decrease the impact of iMRI in the OR environment, a small highly portable 0.12 T iMRI system, positioned under the OR table, was developed by Odin [27]. While this represented a unique innovation, imaging quality, and patient positioning are compromised by the low magnetic field and close opposition of the two magnetic poles.

Clinical Applications of iMRI The development of iMRI has been driven by neurosurgical desire to have intraoperative imaging of CNS pathology following the induction of anesthesia and patient positioning, at various stages of the surgical dissection and as a measure of quality assurance prior to reversal of anesthesia. Despite these advantages, instrument size, weight, magnetic field, RF interference, and cost continue to pose significant challenges for widespread acceptance. Also, there has been no randomized controlled trial demonstrating superiority of iMRI-assisted neurosurgery over standard neurosurgical procedures. This in part reflects traditional education bias, the difficulty of recruiting patients for trials, crossover, and the relatively

recent introduction of iMRI technology. Nevertheless, case studies and existing databases of patients who underwent iMRI procedures represent important sources of information to assess potential efficacy. Non-randomized retrospective studies matching patients based on prognostic factors are considered valuable. Examples where the use of iMRI led to modifications of the surgical procedure are also accepted as supportive evidence of utility. To fully evaluate the benefit of iMRI, randomized clinical trials, with blinded outcome assessment and longterm follow-up, are needed. At the University of Calgary, iMRI has been used to study surgery in 530 patients, 61 of which were pediatric (4 months to 18 years), with various CNS pathologies (Table 2). For each condition, surgical planning iMRI was performed after induction of anesthesia and patient positioning; interdissection iMRI was performed at various stages of the resection and quality assurance iMRI was acquired after wound closure, but before emergence from anesthesia. On average, an imaging session required 30 min. This included reregistration of surgical navigation. A total of 1930 MR studies were performed (Table 3). In 225 patients, trajectories were established by coupling surgical planning images with frameless stereotaxy (BrainLAB, Heimstetten, Germany). Time required for imaging depended on the specific MRI sequences used. Scout images (field of view 30 cm; TR = 35 ms; TE = 15 ms; matrix size 128 × 256; slice thickness 10 mm; one average) were obtained in 15 s. T1 -weighted images were obtained with spin-echo sequences (field of view 26 cm; TR = 400 ms; TE = 12 ms;

Part II

Fig. 2. Ceiling-mounted 1.5 T magnet being moved into the surgical field, together with the OR table and local RF shielding.

1068 Part II

Medical Uses

Part II

Table 2: Patient characteristics grouped by pathological finding

Neoplasia Low-grade glioma High-grade glioma Meningioma Pituitary adenoma Metastatic tumor Schwannoma Vascular AVM Cavernoma Aneurysm ECIC bypass Epilepsy Temporal lobe Corpus callosotomy Miscellaneous Abscess Angiofibroma Central neurocytoma Chondromyxoid fibroma Choroid plexus carcinoma Colloid cyst Cortical dysplasia Craniopharyngioma Cysticercosis Dermoid cyst DNT tumor Hydrocephalus Ganglion cell tumor Germinoma Glomus jugular tumor Hamartoma ICH Lymphoma Medulloblastoma Optic nerve biopsy Osteoma Pachymeningitis Parotid adenoma Plasmacytoma Rathke’s cleft cyst Sarcoma Tubularsclerosis Other Spine Liver resection

Number, gender

Mean

SD

Age range

Total

47 F, 64 M 30 F, 47 M 33 F, 15 M 13 F, 26 M 5 F, 5 M 3 F, 2 M

36 43 57 52 49 38

17 16 16 15 14 20

3–77 6–80 24–84 15–80 26–63 14–59

111 77 48 39 10 5

9 M, 9 F 16 M, 11 F 10 F, 2 M 2 M, 1 F

35 35 55 74

14 14 13

8–56 24–69 37–76

18 27 12 3

34 M, 30 F 1 F–1 M

36 13

12

8–55 13–28

64 2

1 2M 1 F, 1 M F F 2 F, 2 M 5 F, 4 M 3 M, 3 F M 1 F,1 M 1 F,1 M 2M M M M M M 2 F–3 M 6 F, 1 M M 1 F, 1 M F 2M 2F F M

27 16 20 22 15 48 19 23 32 36 10 20 41 12 60 32 38 45 26 40 67 34 29

13 F, 8 M 5 M, 2 F

45 68

1 1

19–20

11 14 25

34–49 1–35 13–51

24 13

10–62 5–35

21

23 months–81 58–83

17 67

1 2 2 1 1 4 9 6 1 2 2 2 1 1 1 1 1 5 7 1 2 1 2 2 2 1 2 14 21 7

Intraoperative MRI

Epilepsy 1069

Part II

Table 3: Number and type of intraoperative MR imaging sequences MR technique

Surgical planning Interdissection Quality assurance Total number

Patients

MR studies

T1

T1 -Gd

T2

MRA

503 300 298

923 550 457 1930

425 278 253 956

255 126 103 484

243 146 101 490

7 3 4 14

matrix size 192 × 256; slice thickness = 4 mm; two or three averages) in ∼4 min. T2 -weighted images with fast spin-echo sequences (field of view 26 cm; TR = 4000 ms; TE = 90 ms; matrix size 256 × 256; slice thickness 4 mm; two averages) and FLAIR fast spin-echo sequences (field of view 26 cm; TR , 6000 ms; TE , 96 ms; matrix size 192 × 256; slice thickness 4 mm; two averages) were obtained in ∼4 min. The following sections summarize the impact of iMRI on various neuropathologies.

Neoplasia CNS neoplasia represented the largest group of patients (n = 337) undergoing iMRI-assisted neurosurgery. Similarly to other iMRI units, we have observed a case selection of patients with low-grade glioma, pituitary adenoma, and childhood tumors. This in part reflects the growing body of knowledge that outcome for these patients correlates to the extent of resection. iMRI application to tumor resection was beneficial for many reasons. Assessment of the lesion immediately prior to surgery may show progression or regression of the lesion, as compared to the diagnostic study. In some cases, this dramatically changed the planned procedure and occasionally led to the decision not to proceed, as in four of the cases (Figure 3). Surgical planning iMRI resulted in more accurate, smaller craniotomies, and narrower surgical corridors (Figure 4). Such precise operative procedures are thought to have resulted in improved outcome. Interdissection iMRI, in addition to correcting for intraoperative brain shift for frameless stereotaxy, provided an ideal method for resection control (Figure 5). This was advantageous when tumor borders were difficult to identify during microdissection, such as in many of the patients with low-grade glioma. Unsuspected residual neoplasm was observed in 9% high-grade glioma, 20% low-grade glioma, 33% pituitary adenoma, and 2% meningioma. Quality assurance iMRI, at the end of the procedure, confirmed the lack of acute surgical complications such as postoperative hemorrhage, obstructive hydrocephalus, or cerebral edema. This knowledge facilitated early postoperative care.

Spreading enhancement, particularly in patients with malignant neoplasms, was observed (Figure 6). While the nature of this phenomenon needs to be determined, it is likely to be multifactorial, reflecting disrupted blood– brain barrier, surgical manipulation, and extension of neoplasm outside borders visible with diagnostic or iMRI. This spreading enhancement was not observed in patients with meningioma, a benign encapsulated tumor, despite considerable gadolinium (Gd) tumor enhancement, and extensive surgical manipulation, including, in many cases, disruption of the arachnoid and pia. In selected cases, biopsy of the area of spreading enhancement confirmed the presence of tumor. Based on these observations, we now limit, in patients with suspected malignant neoplasms, the administration of Gd until late in the surgical dissection. Randomized clinical trials, where patients are randomly assigned to undergo either iMRI-assisted or traditional surgery for neoplasia, are essential to assess the efficacy of iMRI over standard procedures. Patients with low-grade glioma would be ideal for such a study, as surgical resection is often incomplete and outcome may well correlate with gross resection. While the primary outcome for such a study would be the interval before tumor recurrence, other secondary outcome measures could include: accuracy of craniotomy placement, extent of tumor resection, operative complications, and length of hospital stay. A positive study would establish iMRI as the standard of care. Apart from surgery, iMRI can be useful to track biopsy needle and to monitor probe placement at the tumor site for brachytherapy [28], delivering viral vectors, chemotherapeutics, or heat (thermal ablation therapy) [29].

Epilepsy The surgical management of epilepsy is based on a number of variables including the type of epilepsy, location of the epileptogenic focus, patient wishes, and expertise of the surgeon. The immediate goal of surgery is maximal safe resection of the epileptogenic tissue or

1070 Part II

Medical Uses

Part II Fig. 3. T1 -Gd coronal 1.5 T MR images obtained at diagnosis (upper), surgical planning iMRI (middle) and 3-month follow-up (lower). At the time of surgical planning iMRI, the lesion had virtually disappeared and the procedure was aborted.

Intraoperative MRI

Epilepsy 1071

Part II

Fig. 4. Surgical planning T1 -Gd iMRI showing a large intraventricular lesion, together with registration for surgical navigation (BrainLAB, Heimstetten).

anatomical and functional disconnection, to eliminate or reduce the number of clinically significant seizures without causing major deficit. Of the 530 patients, 75 presented with intractable epilepsy. iMRI provided precise identification of abnormal tissue such as cortical dysplasia or medial temporal lobe sclerosis. When coupled to surgical navigation technology, targeting epileptogenic structures, in particular the medial temporal lobe, was

optimized. As the magnet can be moved from the operative site, iMRI did not interfere with the use of surgical adjuncts such as electrocorticography, cortical mapping, and Cavitron Ultra-Sonic Aspirator (CUSA, Valleylab, Boulder, CO, USA). Selective amygdalohippocampectomy was performed in 36 of our 64 patients with temporal lobe epilepsy. We and others have found this to be as effective as temporal lobectomy for seizure control

1072 Part II

Medical Uses

Part II Fig. 5. Surgical planning T1 -Gd iMRI obtained from a 5-year-old boy with pilocytic astrocytoma show the large tumor with associated brain shift (upper). Interdissection images (lower) revealed unsuspected residual tumor (arrows).

[30]. Despite the fact that this procedure is routinely performed, anatomical targets can become unclear during microdissection. When approaching through the mid temporal gyrus, the amygdala may be hard to define, while with trans-sylvian dissections, it is often difficult to determine the extent of hippocampal removal. In 19% of patients with temporal lobe epilepsy, interdissection iMRI, showed unsuspected residual tissue, which was then removed (Figure 7). Notwithstanding these encouraging results, a randomized trial is necessary to both confirm the observations and determine the impact of iMRI on clinical outcome. The assessment of cortical dysplasia requires the signal-to-noise ratio of 1.5 T or higher-field MR systems. In 4 patients with cortical dysplasia, extent of the abnormality was well assessed with the 1.5 T iMRI system, enhancing craniotomy placement and resection control.

Vascular disorders Surgical planning iMRI significantly improved craniotomy placement for patients with arteriovenous malformation (AVM) and cavernous angioma, making lesion exposure relatively straightforward. Among the 60 patients with cerebral vascular pathology, 18 had AVM and 27 had cavernous angioma. We have, as of yet, not used

diffusion/perfusion MR imaging to evaluate the effect of resecting AVM on the adjacent brain. The literature suggests that AVM resection may result in hyperperfusion of the adjacent brain. It is this consequence that is thought to contribute to the relatively high frequency (10%) of postoperative hemorrhage [31]. There were 12 patients with aneurysm and 3 had extracranial carotid occlusion. For all vascular procedures, interdissection iMRI and quality assurance iMRI confirmed complete resection of the target or treatment of the abnormality. For patients with aneurysm, intraoperative MR angiography accurately revealed vascular anatomy and clip placement. However, clip-induced susceptibility and/or eddy current artifact interfered with evaluation of the aneurysm neck. Intraoperative assessment of the aneurysm neck will require the development of MRinvisible clips. EPI was used to identify diffusion/ perfusion abnormality in the distribution of distal vascular branches, including perforators. The presence of such abnormalities would allow clip repositioning prior to irreversible brain injury (Figure 8). We have not encountered such changes and none of the aneurysm patients have suffered deficit. Patients with carotid occlusion were symptomatic with recurrent transient ischemic attacks, and diagnostic MRI showed diffusion/perfusion mismatch. Extracranial to intracranial bypass resulted in immediate correction of the

Intraoperative MRI

Future Directions 1073

Part II

Fig. 6. Surgical planning (upper), interdissection (middle), and day 2 postoperative (lower) T1 -Gd iMRI of a patient with recurrent medulloblastoma. Diffusion of contrast well beyond the suspected tumor margin can be observed on interdissection but not on follow-up images.

perfusion deficit as shown with iMRI. Diffusion-weighted EPI quality requires a homogenous magnetic field, excellent eddy current compensation, strong gradients with fast switching times, and excellent shimming [32].

Spine iMRI was used as an adjunct to spine surgery in 21 patients. The procedures included anterior cervical discectomy, cervical corpectomy with instrumentation, and the transoral resection of various C1/2 pathologies. iMRI with surgical navigation provided precise targeting of spinal pathology particularly lesions involving the C1/2 region, that are difficult to visualize through surgical corridors designed to limit trauma associated with extensive exposure. Interdissection images accurately determined the extent of decompression and showed the relationship of the surgical

dissection to neural vascular structures (Figure 9). Extending iMRI capability to the thoracic and lumbar spine will be increasingly important as spinal surgery becomes reliant on microsurgical technique including restricted surgical corridors. The modified phased array spine RF coils are used to maximize images of the spine. The volume coil is used to transmit the RF power while the phased array coils work in the receive only mode.

Future Directions Technology While imaging techniques have greatly enhanced diagnostic accuracy and improved surgical outcomes, technical limitations necessitate further iMRI improvement. Increasing the magnetic field would further enhance image

1074 Part II

Medical Uses

Part II

Fig. 7. Surgical planning T1 -weighted iMRI from a patient with intractable temporal lobe epilepsy, showing the targeted amygdala and hippocampus (upper). Interdissection images (middle) indicate unsuspected residual amygdala (arrow). Quality assurance iMRI reveals complete dissection of the target (bottom).

Fig. 8. Surgical planning (upper) and interdissection (lower) T1 -weighted, MR angiography, diffusion, and perfusion iMRI studies obtained from a patient with a giant left PICA aneurysm. Interdissection studies obtained 5–10 min following clip ligation of PICA show flow change within the aneurysm, parent vessel occlusion, aneurysm obliteration, and no diffusion or perfusion abnormality.

Intraoperative MRI

Bioinformatics 1075

resolution by improved signal-to-noise ratio, although this might also lead to an amplification of susceptibility artifacts in certain areas. Improvements in RF coil design need to take advantage of improved technology while not limiting surgery. Larger RF coils, not restricting access to the operative field, would be an asset. Additional reduction in total acquisition time could be achieved by the application of phased array technology that allows the acquisition of parallel images. Enlarging the magnet bore facilitates patient positioning and the introduction of surgical technologies. To take full advantage of an iMRI system and maximize its use, the magnet should ideally be movable and sited for use in two or more rooms.

Bioinformatics Software development has made possible 3D reconstruction of iMRI including lesion segmentation. 3D reconstruction augments lesion localization, diagnosis, and surgery planning. MRI and spectroscopic characteristics may correlate with outcome [33,34]. Multimodal fusion of MR images provides information as to the relationship of eloquent brain and vasculature to the planned surgical corridor and target.

Robotics The introduction of iMRI has provided an imaging adjunct to surgery. Despite this and other technologies,

such as microscopy, surgeons operate much like they did 100 years ago, depending to a large extent on hand–eye coordination. Robotic advances, such as tremor filtration and motion scaling permit maximal use of magnification and enable precise, tremor-free tool manipulation during surgery. Integrating robotic technology with iMRI would allow surgeons to manipulate images in real time and reduce limitations imposed by iMRI, particularly those related to a restricted working environment (GE double doughnut configuration) and disruptions in surgery related to image acquisition (1.5 T IMRIS, Siemens or Philips systems). At the University of Calgary, we are developing an MRI-compatible, image-guided, ambidextrous robot called neuroArm [35]. Our industrial collaborator, MD Robotics (Brampton, Ontario, Canada), has successfully developed safety-critical robotic systems for the space shuttle and International Space Station. NeuroArm consists of a neurosurgical robot, main controller, and a workstation. The robot has two MR-compatible articulated arms with dextrous mechanical manipulators that grasp and move surgical tools. Each arm has 8 degrees of freedom (DOF) and are small enough to work within the 68 cm working diameter of the 1.5 T IMRIS magnet and capable of 30 µm precision (Figure 10). A 3-DOF strain gauge sensor system was integrated to provide haptic feedback to the surgeon. The workstation incorporates a computer processor, two hand controllers to manipulate the robot arms, a controller for positioning the microscope and lights, and three types of display and data recorders.

Part II

Fig. 9. T1 - and T2 -weighted iMRI from patient with retro-odontoid synovial cyst. Surgical planning iMRI (upper) show significant spinal cord compression. Interdissection iMRI show unsuspected residual spinal cord compression at the C2 level (middle). Quality assurance iMRI show optimal decompression (lower).

1076 Part II

Medical Uses

Part II

7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Fig. 10. NeuroArm in position for stereotaxy. The base is equipped with an extension platform (not shown), allowing the manipulator arms to reach inside the magnet bore.

19. 20.

The interface maximizes ergonomic comfort and minimizes surgeon fatigue. The robotic platform and imaging environment include surgical simulation capability. This will be of increasing importance as advances in surgical technologies and associated techniques increase case complexity and produce a tyranny of choice. Risk-free rehearsal and development of patient specific interventional recipes will allow surgeons to safely take full advantage of iMRI and robotic technologies.

Acknowledgments The authors would like to thank Boguslaw Tomanek and John Saunders for their kind suggestions. The iMRI program has been supported by a grant from the Canadian Foundation for Innovation. The iMRI system is being marketed by IMRIS ([email protected]).

References 1. Clower WT, Finger S. Neurosurgery. 2001;49:1417. 2. Piek J, Lidke G, Terberger T, von Smekal U, Gaab MR. Neurosurgery. 1999;45:147. 3. Barker FG II. J. Neurosurg. 1993;79:948. 4. Gall FJ. Sur les fonctions du cerveau et sur celles de chacune de ses parties avec des observations sur la possibilit´e de reconnaitre les instincts, les penchans, les talens, ou les dispositions morales et intellectuelles des hommes et des animaux, par la configuration de leur cerveau et de leur tˆete, Vol 1. J.B. Bailli`ere: Paris, 1822. 5. Broca P. Bull. Soc. Anat. Paris. 1861;6:398. 6. Dandy WE. Ann. Surg. 1919;70:397.

21. 22. 23. 24.

25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35.

Moniz E. Rev. Neural. 1927;2:72. Fry WJ, Fry FJ. IRE Trans. Med. Electron. 1960;ME-7:166. Cormack AM. Phys. Med. Biol. 1973;18:195. Hounsfield GN. Br. J. Radiol. 1973;46:1016. Lauterbur PC. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 1980;289:483. Mansfield P, Maudsley AA. Br. J. Radiol. 1977;50:188. Lunsford LD. Appl. Neurophysiol. 1982;45:374. Black PM, Moriarty T, Alexander E III, Stieg P, Woodard EJ, Gleason PL, Martin CH, Kikinis R, Schwartz RB, Jolesz FA. Neurosurgery. 1997;41:831. Le Bihan D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N, Chabriat H. J. Magn. Reson. Imaging. 2001; 13:534. Keston P, Murray AD, Jackson A. Clin. Radiol. 2003;58:505. Le Bihan D. J. Neuroradiol. 1996;23:1. Hoult DI, Saunders JK, Sutherland GR, Sharp J, Gervin M, Kolansky HG, Kripiakevich DL, Procca A, Sebastian RA, Dombay A, Rayner DL, Roberts FA, Tomanek B. J. Magn. Reson. Imaging. 2001;13:78. Steidle G, Graf H, Schick F. Magn. Reson. Imaging. 2004;22:171. Steinmeier R, Fahlbusch R, Ganslandt O, Nimsky C, Buchfelder M, Kaus M, Heigl T, Lenz G, Kuth R, Huk W. Neurosurgery. 1998;43:739. McPherson CM, Bohinski RJ, Dagnew E, Warnick RE, Tew JM. Acta Neurochir. Suppl. 2003;85:39. Sutherland GR, Kaibara T, Louw DF, Hoult DI, Tomanek B, Saunders J. J. Neurosurg. 1999;91:804. Kaibara T, Saunders JK, Sutherland GR. Neurosurgery. 2000;47(1):131. Sutherland GR, Kaibara T, Wallace C, Tomanek B, Richter M. Intraoperative assessment of aneurysm clipping using magnetic resonance angiography and diffusion-weighted imaging: technical case report. Neurosurgery. 2002;50(4): 893. Nimsky C, Ganslandt O, von Keller B, Fahlbusch R. Acta Neurochir. Suppl. 2003;88:21. Hall WA, Liu H, Martin AJ, Truwit CL. Top. Magn. Reson. Imaging. 2000;11:203. Schulder M, Jacobs A, Carmel PW. Stereotact. Funct. Neurosurg. 1998;76:151. Susil RC, Camphausen K, Choyke P, McVeigh ER, Gustafson GS, Ning H, Miller RW, Atalar E, Coleman CN, Menard C. Magn. Reson. Med. 2004;52:683. Dick EA, Joarder R, de Jode M, Taylor-Robinson SD, Thomas HC, Foster GR, Gedroyc WM. Clin. Radiol. 2003; 58:112. Olivier A. Can. J. Neurol. Sci. 2000;27(Suppl 1):S68. Schaller C, Urbach H, Schramm J, Meyer B. Neurosurgery. 2002;51(4):921. Fischer H, Ladebeck R. Echo-Planar Imaging: Theory, Technique and Application. Springer: New York, 1998, pp 179– 200. Ekinci G, Akpinar IN, Baltacioglu F, Erzen C, Kilic T, Elmaci I, Pamir N. Eur. J. Radiol. 2003;45(2):99. Wu WC, Chen CY, Chung HW, Juan CJ, Hsueh CJ, Gao HW. Am. J. Neuroradiol. 2002;23(10):1775. Louw DF, Fielding T, McBeth PB, Gregoris D, Newhook P, Sutherland GR. Neurosurgery. 2004;54(3):525.

1077

Uma Sharma and N.R. Jagannathan Department of NMR, All India Institute of Medical Sciences, New Delhi 110029, India

Abbreviations: MR, magnetic resonance; MRI, magnetic resonance imaging; MRS, magnetic resonance spectroscopy; 1 H MRS, proton magnetic resonance spectroscopy; 31 P MRS, phosphorus magnetic resonance spectroscopy; CE-MRI, contrast-enhanced magnetic resonance imaging; VOI, volume of interest; CSI, chemical shift imaging; MRSI, magnetic resonance spectroscopic imaging; STEAM, stimulated echo acquisition mode; PRESS, point resolved spectroscopy; CHESS, chemical shift selective; W–F, water-to-fat ratio; PME, phosphomonoester; PDE, phosphodiester; PCr, phosphocreatine; SV, single-voxel.

Introduction Breast cancer is the most common cancer affecting women and is a significant health care problem, worldwide [1–3]. Breast cancer is classified as either ductal or lobular, depending on its morphological origin and similarities to the normal cellular components. Considerable heterogeneity among breast tumors demands highly accurate diagnostic techniques. Methods like X-ray mammography, ultrasound, and physical examination are often limited in sensitivity and specificity, especially in young women [4–6]. Magnetic resonance (MR) mammography and contrast-enhanced magnetic resonance imaging (CEMRI) methodologies have greatly improved the ability to differentiate malignant from benign breast tumors [7–11] (see chapter by Kaiser). However, the reported specificity of the above studies is widely variable [8] and does not provide any metabolic information. Recently, there has been an increasing interest in augmenting breast MRI with magnetic resonance spectroscopy (MRS) investigations by using the biochemical information to increase the specificity in differentiating benign from malignant tumors. Results of in vivo proton (1 H) [12–25] and phosphorus (31 P) [26–28] MRS, ex vivo and in vitro [29–32], and cell line [33,34] studies reported to date suggest that malignant breast tumors exhibit elevated levels of choline-containing compounds (total choline, TCho). In the case of MRS applied to fine needle aspiration biopsies (FNAB), the creatine to choline ratio can provide the pathological diagnosis with accuraGraham A. Webb (ed.), Modern Magnetic Resonance, 1077–1086.
C 2008 Springer.

cies in the late 90th percentile [32,35]. This chapter reviews the potential of in vivo MRS (both 1 H and 31 P) in breast cancer especially in relation to the various methodological issues involved, technical limitations, diagnostic specificity, clinical applications, and future directions.

In vivo Localization in MRS The basic requirement of in vivo localized MRS is to acquire the signal from a particular volume of interest (VOI) with optimal sensitivity. In the last 20 years, the availability of gradients has led to the development of localization techniques to obtain spectra from specific VOI. Magnetic field gradients in either the B0 field as in imaging or the B1 (radio frequency) field are used for localization of a particular volume. In vivo MRS began with the analysis of isolated tissues and surface regions. Most 31 P studies on breast cancer use surface coils, which are one of the earliest methods of localization. Surface coils provide rough localization. However, spatial selectivity can be achieved by varying the radio frequency pulse length. The disadvantages include extremely inhomogeneous transverse magnetic field, difficulty in assessing VOI that is below the surface, and contamination of signals from extraneous tissues. For these reasons, surface coils are used with other techniques like depth resolved spectroscopy, DRESS [36], which uses B0 magnetic field gradients and a frequency selective pulse for localization. Depth resolved spectroscopy is also possible by using B1 changes [37]. Currently employed image guided localization methods use proton images acquired in three orthogonal planes (transverse, sagittal, and coronal) to guide the placement of VOI. The magnetic field gradients spatially encode the resonance frequencies and the frequency selective pulse excites the spin distribution within the sample. Position, size of VOI, and the actual shape depend on the slice profile of the selective pulses. Localization methods acquire spectra either as single-voxel (SV) or as multivoxel [chemical shift imaging (CSI) or magnetic resonance spectroscopic imaging (MRSI)]. MRSI offers many advantages over SV techniques, such as localized spectra from many locations and small VOIs can be acquired simultaneously which can be used to obtain the metabolite

Part II

In Vivo Magnetic Resonance Spectroscopy in Breast Cancer

1078 Part II

Medical Uses

jor problem. Chemical shift selective (CHESS) radio frequency pulses that excite a limited narrow band (∼60 Hz) of frequencies corresponding to the water signal are used to suppress the water signal [44]. Sequences that simultaneously suppress both water and lipid resonances are also being developed (vide infra). 31

P MR Spectroscopy

31

P MR spectra provide information on cellular metabolism by measuring various phosphorus metabolites and have been reviewed recently [26,27]. Most studies were carried out at 1.5 T with the patient positioned either supine or prone with a dual 1 H/31 P surface coil covering the area of the tumor. Shimming is carried out at 1 H frequency. The measurement methodology adopted by different groups varies in relation to the use of repetition time, size of coil used, and other experimental parameters. Figure 1A shows the 31 P MR spectrum obtained from a normal volunteer, and reflects the level of various

B

PDE 5

0

-5

Chemical shift (ppm)

β - NTP

α - NTP

γ - NTP

β - NTP

γ - NTP

Pi 10

α - NTP

PDE

PME Pi

PCr

PCr

A

PME

Part II

images. The pixel intensity in these images is proportional to the relative concentrations of the metabolites which is useful in visually assessing the spatial variation in the metabolite concentrations. However, the technique suffers from some disadvantages, such as signal from a voxel is prone to contamination from outside the VOI. This contamination occurs due to discrete and finite sampling, since the number of phase-encoding steps is typically more limited than normal imaging acquisitions. In addition, shimming the magnetic field over large volume especially in breast due to heterogeneity of tissues and long acquisition times are other disadvantages. In recent times, echo-planar imaging-based MRSI sequences which offer less image acquisition times have also become available. The most commonly used pulse sequences in breast cancer MRS are image selected in vivo spectroscopy (ISIS) [38], stimulated echo acquisition mode (STEAM) [39,40], point resolved spectroscopy (PRESS) [41], and CSI or MRSI [42,43]. In breast 1 H MRS, detection of resonances from metabolites with lower concentration in the presence of a large water and lipid signals is a ma-

-10

-15

10

5

0

-5

-10

-15

Chemical shift (ppm)

Fig. 1. (A) 31 P MR spectrum from normal breast tissue of a volunteer. Peak assignment corresponds to adenosine triphosphate (ATP), phosphodiesters (PDE), phosphomonoesters (PME), phosphocreatine (PCr), and inorganic phosphate (Pi). (B) 31 P MR spectrum from a patient suffering from infiltrating duct carcinoma of the breast. Spectra were obtained using a 10 cm surface coil with a single RF pulse.

Breast MRS

MR Spectroscopy of Breast 1079

has been shown in recent years in evaluating the role of in vivo 1 H MRS in breast cancer. 1

H MR Spectroscopy of Breast

Presently seven groups are involved in breast in vivo 1 H MRS [12–25]. Katz-Brull et al. [54] have reviewed these reports and carried out pooled analysis of the existing 1 H MRS data. However, several methodological issues need to be addressed and these are outlined below.

Methodological Issues Related to Proton in vivo Breast MRS Most in vivo 1 H MRS of human breast cancer studies reported to date were performed at 1.5 T except for the work of two groups [24,55,56]. The patient is usually positioned prone and this has some technical advantages, in that a tumor may be relatively distant from other structures that provide a substantial signal, such as chest wall muscle. However, positioning a surface coil is difficult in prone position and thus, a relatively large volume coil is often employed which has the disadvantage of potentially reducing the signal-to-noise ratio. In the supine position, a surface coil can be used, but contamination from chest wall and motion artifacts may increase. An additional disadvantage is the loss of signal from lesions that are not close to the surface coil of the breast. A dedicated single or bilateral surface receive coil (custom/commercial make) is used for optimal signal reception with the body coil as transmitter. Automatic tuning and electronic decoupling is used for optimal signal reception. Some groups have used a transmit–receive breast coil (usually custom made) for their study [15,23,24]. Patients are positioned prone with the breast fitting into the cup of the coil. Followed by localizer images, T1 weighted images in sagittal plane are obtained using a spin-echo sequence. To identify precisely the full extent of the irregular, speculated border of malignant tumors, lipid saturated proton MR images in the transverse, and coronal planes are acquired. Some centers use CE-MRI to localize the tumor region. Kvistad et al. used non-contrast MRI, when the tumor was easy to localize, otherwise CEMRI was used for positioning the voxel [16]. Caution should be exercised when MRS is performed after using contrast agents to identify the tumor, since gadolinium has been reported to influence the in vivo detection of TCho [57]. Depending on the tumor size, voxels of appropriate dimension are chosen and positioned well within the tumor area. Magnetic field shimming must be carried out both globally and over the voxel region to optimize the field homogeneity and for good water-suppression. In our experience, the line width after voxel shimming typically

Part II

phosphorus containing metabolites. In general, considerable variations in the 31 P metabolite profile of the normal breast were observed during the menstrual cycle [45,46]. Pre-menopausal women showed a rise in phosphomonoester (PME) signal at the late follicular stage of the menstrual cycle and less signal at the early luteal phase [46]. Phosphodiester (PDE) was maximal at the early follicular phase, and minimal at the late follicular stage. Twelves et al. [45] have reported lower PME and a high PDE/PME ratios during the second week of menstrual cycle. It is also reported that PME is increased in lactating breast compared with non-lactating post-menopausal women [45,47]. In general, 31 P MR signals are lower in post-menopausal women compared with pre-menopausal women [45,46]. Figure 1B shows the 31 P MR spectrum of a patient suffering from infiltrating duct carcinoma with elevated levels of PME and PDE, similar to that reported by other researchers [26,28,48–51]. In general, breast tumors have higher total phosphate content compared to the normal breast tissues of non-menstruating women [26,48]. The potential of 31 P MRS in distinguishing benign lesions from malignant has also been reported on the basis of elevated PME [49,50]. Biopsy results demonstrate that carcinomas contain three times more adenosione triphosphate and PME than benign breast lesions [29]. Recently, the utility of in vivo 31 P MRS for patients with various malignant and benign diseases has been evaluated [28]. Significant differences in PME and the ratio of PDE to phosphocreatine in malignant tumors were observed compared to normal breast tissue. However, no difference was observed between malignant and benign tumors suggesting that 31 P MRS is not helpful in differential diagnosis between benign and malignant lesions [28]. The use of 31 P MRS to monitor the response of breast tumors to chemotherapy has also been reviewed [26]. Serial measurements of localized 31 P MRS of breast cancer patients using the image guided in vivo spectroscopy sequence [26] was carried out both before and after chemotherapy showing decreased PME that was found to be associated with the response of the disease, while an increase in PME was associated with the disease progression [12,45,51–53]. Even though 31 P MRS studies were promising, they gave diverse results [28]. Moreover, its use in characterizing the tumors in vivo is hampered by lower MR sensitivity for detecting 31 P signals (sensitivity 1/10 of 1 H MRS). To achieve a similar signal-to-noise ratio of metabolites detected by 1 H MRS, a 31 P MRS study requires a voxel that is about 10 times larger than that used in 1 H MRS and hence can be performed only on large sized tumors. In addition, 31 P MRS requires special hardware that may not be available with all clinical scanners. By contrast, 1 H MRS examination can easily be integrated into a routine MRI examination. In view of this, considerable interest

1H

1080 Part II

Medical Uses

Part II

should correspond to 10–25 Hz for the lipid peak in case of normal breast tissue and 5–20 Hz for the water peak in patients with breast tumors. If this is not achieved, the quality of water-suppression will be poor leading to broad or no detectable choline signal. Solvent suppression pulses can be optimized to have narrow bandwidth excitation to minimize spectral perturbations outside the desired area. Water-suppressed and un-suppressed localized spec-

tra with an adequate number of scans are acquired using an appropriate echo time (TE = 30/135/270/350 ms). Chemical shifts are reported using water as the internal standard at 4.7 ppm. The total study time per patient ranges between 60 and 75 min including MRI and MRS. Table 1 summarizes details of the pulse sequence, acquisition parameters, etc. used in various studies reported in the literature.

Table 1: Summary of experimental details used in various 1 H MRS studies

Study and reference

Patients, number of cases studied, and other details

Sequence used and SV/MRSI

Water or lipid suppression sequence used

Field, coil used, methods, acquisition parameters, sensitivity, and specificity 1.5 T; single breast multicoil receive; CE-MRI; TR = 2 s; TE = 31, 270 ms; voxel size = 0.7–9.8 ml; sensitivity = 70%; specificity = 86% 1.5 T; bilateral surface receive coil; non-CE-MRI; TR = 3 s; TE = 135 ms; voxel size = 1–8 ml

Roebuck et al. [13]

Malignant, n = 10 (pre-therapy); benign, n = 7

STEAM; SV

Water; CHESS

Jagannathan et al. [14,17]

Malignant, n = 17 (pre-therapy); n = 34 (post-therapy); benign, n = 2; controls, n = 14 Malignant , n = 11 (pre-therapy); n=1 (post-therapy); benign, n = 11; control, n = 11; lactating, n = 7 Malignant, n = 46 (pre-therapy); n = 35 (post-therapy); benign, n = 14; control, n = 16; lactating, n = 1 Malignant, n = 23 (pre-therapy); benign, n = 15

STEAM; SV

Water; CHESS

PRESS; SV

Water frequency selective inversion

1.5 T; single breast receive coil; non-CE and CE-MRI; TR = 2 s; TE = 135, 350, 450 ms; voxel size = 1–25.2 ml; sensitivity = 82%; specificity = 82%

STEAM; SV

Water; CHESS

1.5 T; bilateral surface receive coil; non-CE-MRI; TR = 3 s; TE = 135 ms; voxel size = 1–27 ml; sensitivity = 78%; specificity = 86%

STEAM; SV

Water; CHESS

Yeung et al. [19]

Malignant, n = 24 (pre-therapy); benign, n = 6

PRESS; SV

Gribbestad et al. [15]

Malignant, n = 12 (pre-therapy); control, n = 10 Malignant, n = 6 (pre-therapy) Malignant, n = 21 (pre-therapy); control, n = 43; lactating , n = 3

PRESS; SV

Water frequency selective inversion recovery Water frequency selective inversion Water and lipid using MEGA VSS and CHESS

1.5 T; phased array multicoil receive; non-CE and CE-MRI; TR = 2 s; TE = 31, 270 ms; voxel size = 1–3.4 ml; sensitivity = 83%; specificity = 87% 1.5 T; double breast receive coil; CE-MRI; TR = 2 s; TE = 38, 135, 270 ms; voxel size = 1–95 ml; sensitivity = 92%; specificity = 83% 1.5 T; double breast transmit-receive coil; non-CE-MRI; TR = 2 s; TE = 136 ms; voxel size = 8–27 ml 1.5 T; standard coil; CE-MRI; TR = 1.6 s; TE = 270 ms; voxel size = 3.4 ml 1.5 T; single breast transmit–receive coil; non-CE-MRI; TR = 2 s; TE = 135, 350 ms; voxel size = 3.4 ml

Kvistad et al. [16]

Jagannathan et al. [22]

Cecil et al. [18]

Gribbestad et al. [61] Stanwell et al. [23]

PRESS; SV PRESS; SV

Breast MRS

1H

MR Spectroscopy of Breast 1081

Study and reference Bolan et al. [24]

Thomas et al. [66] Jacobs et al. [25]

Meisamy et al. [56]

Patients, number of cases studied, and other details Malignant, n = 86 (pre-therapy); n = 14 (post-therapy); control, n = 5 Malignant, n = 8; control, n = 9 Malignant, n = 8 (pre-therapy); benign, n = 7 Malignant, n = 16; MRS both pre- and post-therapy

Sequence used and SV/MRSI

Water or lipid suppression sequence used

Field, coil used, methods, acquisition parameters, sensitivity, and specificity

LASER; SV

VAPOR

4 T; single breast transmit–receive coil; CE-MRI; TR= 3 s; TE = 45–196 ms; voxel size = 0.4–6 ml

2D, SV CABINET

No suppression

PRESS; MRSI

Water and lipid; CHESS and STIR VAPOR

1.5 T; phased array coil; TR = 2 s; TE = 30 ms; voxel size = 1 ml 1.5 T; phased array coil; TR = 2 s; TE = 272 ms; voxel size = 1 ml

SV, localization with adiabatic selective refocusing

4 T; single breast quadrature transmit–receive coil; TR = 3 s; TE = 45–196 ms; voxel size = variable dimensions according to tumor size

Note: TR = time for repetition; TE = time for echo; LASER = localization by adiabatic selective refocusing; STIR = short inversion time recovery; VAPOR = variable pulse power and optimized relaxation; CABINET = coherence transfer based spin-echo spectroscopy; MEGA = MEscher GArwood (named after the authors); VSS = very selective suppression.

In our work, the observation of TCho was based on strict experimental criteria [22], namely: (i) the line width of the un-suppressed water peak to be around 5–20 Hz and (ii) the ratio of the water-suppression factor ≥20. If these criteria were not met, the data were discarded [22]. It is our experience that these criteria are essential since the observation of choline may be affected by poor shimming, the relative position of the voxel in relation to the surface coil, coil loading, and partial volume effects [22,24,54]. Moreover, SV-MRS of breast is sensitive to the size and placement of the voxel because of the heterogeneous distribution of TCho in the breast [24]. In addition, quantification of metabolite levels in breast is more difficult because of the heterogeneous distribution of the glandular and adipose tissues. Absolute concentration of TCho in breast lesions has also been reported using a phantom containing 1 mmol/l solution of choline placed in the multicoil array [13]. Recently, Bolan et al. used water as an internal reference peak for quantification of TCho [24]. This approach compensates for the partial volume of adipose tissue in the voxel and leads to a molal (mol/kg) concentration of water-soluble metabolites. Use of an appropriate echo time in breast in vivo spectroscopy involves a trade-off between signal intensity (high with short echo time) and signal contrast (the ability to resolve the composite choline signal from the lipid signal, which is higher with long echo time). The use of long echo times (≥135 ms) is preferred, despite the loss of signal intensity, for improved visibility of TCho signal because of a decreased over-

lap with the lipids [13,19,22]. Recently, Stanwell et al. have demonstrated the importance of referencing and optimized post-acquisitional data processing to improve the spectral resolution of MR spectra [23]. This allows the resolution of the composite choline resonance into its constituent components and helps in improving the specificity of in vivo MRS method [23]. In addition, water-tofat (W–F) (lipid) ratio can also be calculated using the respective peak areas (using area under the water and the major lipid peak) from the un-suppressed MR spectra [12,14,17]. Interpretation of 1 H and 31 P spectra should include the age and the menopausal status of the patient. Postmenopausal women have more adipose tissue compared to glandular tissue. In pre-menopausal women, glandular tissue can give an adequate signal, and this needs to be considered both in the measurement strategies as well as in interpreting the spectra obtained. In a study by Stanwell et al. composite choline signal was observed in three non-lactating normal volunteers [23]. The chemical shift of this resonance was found to differ from that recorded from patients with malignant disease. When 1 H MRS on two of these volunteers was repeated once a week for a period of 8 weeks [23], in one volunteer the choline resonance was absent at different stages throughout the menstrual cycle but re-appeared in the next cycle. In the other volunteer, the resonance was always detectable but appeared to fluctuate during the 8-week period. However, long-term follow-up of these individuals has revealed no disease.

Part II

Table 1: (Continued)

B

8

6

–CH2–

––CH– –CH– H2O

Part II

A

–(CH2)n–

Medical Uses

–CH3

1082 Part II

4

2

0

Chemical shift (ppm) Fig. 2. (A) T1 -weighted axial image from a volunteer using spin-echo sequence with the following parameters: TR = 600 ms, TE = 15 ms, and slice thickness = 4 mm. (B) 1 H MR spectrum acquired from an 8 ml voxel localized in the normal breast tissue at TE = 135 ms using PRESS sequence.

Comparison of 1 H MRS of Normal and Malignant Breast Tissues Water-to-Fat Ratio A typical 1 H MR spectrum obtained from our laboratory using STEAM sequence (without water-suppression) from the normal breast tissue of a volunteer is shown in Figure 2B while the corresponding voxel location in Figure 2A. Lipid resonance dominates and the assignment of the various lipid resonance peaks are as shown. A 1 H spectrum without water-suppression from the tumor of a patient suffering from infiltrating duct carcinoma is shown in Figure 3B and the voxel location in Figure 3A. In tumor spectrum, the water peak dominates with low contributions from protons of the fatty acid chains in comparison to the normal breast tissue of controls and from the unaffected contralateral breast of the patient. Elevated W–F ratio characterizes the breast tumor. Similar findings have been documented by several others [12,14,17]. Observation of elevated W–F ratio in breast cancer patients is in agreement with the generally established trend that tumors have considerably higher water content [12,58,59]. Mackinnon et al. [60] reported changes in lipid content

with tumor development and progression. In patients receiving chemotherapy, resulting in the reduction of primary tumor size, the W–F ratio showed a statistically significant ( p < 0.01) decrease compared with the pretherapy value, thus providing a non-invasive indicator of favorable clinical outcome of chemotherapy [14,17]. The method thus provides the potential for non-invasively monitoring and assessing the response of breast cancer to chemotherapy. However, due to overlap of W–F values between benign and malignant lesions it has limited utility [59] and hence a search for other biochemical/s, which can be used as marker for malignancy, was undertaken. Choline Signal Water-suppressed 1 H MR spectra of breast tumor showed, in addition to the residual water and lipid peak, a peak at 3.2 ppm due to TCho as shown in Figure 3C. In general, the observation of TCho from breast tumors is hampered by the presence of huge water and lipid signals. However, if the strict experimental criteria outlined earlier are followed, then the observation of TCho is feasible. Evaluation of a pulse sequence to suppress simultaneously both the water and lipid signals to improve the detection

Breast MRS

1H

MR Spectroscopy of Breast 1083

Part II

B

2.5

–(CH2)n–

>

×10 3

H2O

A

C

2.0

80

1.5

60

0

0.0 6

4 2 Chemical shift (ppm)

–CH3

20

–CH2–

–(CH2)n–

0.5

Cho

40

1.0

0

4

3 2 1 Chemical shift (ppm)

0

Fig. 3. (A) T2 -weighted lipid-suppressed sagittal MR image of a patient suffering from infiltrating duct carcinoma of the right breast, the tumor is seen as hyper-intense area. 1 H MR spectrum from an 8 ml voxel positioned in the tumor of the same patient obtained at TE = 135 ms using PRESS sequence: (B) without water-suppression and (C) with water-suppression.

of choline resonance is currently underway in our laboratory (see Figure 4). Lipid/water signal reduction is obtained using the spectral saturation method, by which transverse magnetization is dephased selectively before and after the second slice selective 180˚ spin-echo pulse. This dephasing is defined to affect both the water and lipid signals from 0.7 to 2 ppm, following the procedure of Mescher et al. [61]. Recently, Gribbestad et al. have also used a similar sequence to suppress the signals from water and lipid [62]. Of the various TCho (Cho, glycerophosphocholine, and phosphocholine) that contribute to the in vivo peak at 3.2 ppm, an increase in phosphocholine has been documented [23,33,63]. Elevation of TCho levels in malignant breast tumors compared to benign cases from in vitro MRS of FNAB and tissue has also been reported [30,31,32]. Results of 1 H in vivo breast MRS investigations to date have

shown that the TCho resonance is specific to malignant tissue and can be used to differentiate cancerous from benign tissue. Combined analysis of the published reports reveal that the overall sensitivity and specificity of MRS is 83 and 85%, respectively [54]. The details are as presented in Table 1. In younger women, differentiation of benign from malignant lesion is important since the incidence of benign breast disease in young women is high [64,65] and mammography examination has lower sensitivity. Results from Cecil et al. [18], Yeung et al. [19], and Roebuck et al. [13] show that in younger patients (≤40 years of age), 1 H MRS has a sensitivity of 100% and a specificity of 89–100% in detecting malignancy. Several factors that limit the sensitivity of 1 H MRS were discussed earlier and the false negative results reported in many studies are ascribed mainly to technical limitations. For example, if the tumor size is small,

Medical Uses

Part II

A

Cho

1084 Part II

B 0.8

0.6

0.4

0.2 0.0

-0.2 4

3

2 1 Chemical shift (ppm)

0

Fig. 4. (A) T2 -weighted lipid-suppressed sagittal MR image of a patient suffering from infiltrating duct carcinoma. (B) 1 H MR spectrum from an 8 ml voxel positioned in the tumor of the same patient obtained at TE = 270 ms using PRESS sequence with both the water and the lipid peak simultaneously suppressed.

relatively long acquisition time is required resulting in the patient becoming restless leading to movement artifacts. This also could lead to incorrect sampling of the tumor, thereby leading to contributions from surrounding fatty tissue. Moreover, movement of a patient would also lead to incorrect sampling and hence to false diagnosis. Incorrect sampling may also play a role in the detection of the choline signal. Few benign cases also showed TCho signal [16,18,19,22]. Detection of choline from the normal breast tissue of lactating women has also been reported [16,22,23]. This observation implies that the interpretation of spectra from such patients needs to be evaluated with caution and also points out the limitation of using TCho as a marker of malignancy. Stanwell et al. have documented that the main contribution to the TCho resonance observed in lactating breast is from the glycerophosphocholine [23]. They have documented that optimized post-processing of the spectra resolved a resonance at 3.22 ppm, consistent with phosphocholine, in patients with cancer. In contrast, the spectra recorded for three false positive volunteers and the three lactating mothers had a resonance centered at 3.28 ppm (possibly taurine, myo-inositol, or glycerophosphocholine). Recently Jacobs et al. [25] have reported on a proton MRSI of breast cancer patients; their results show that TCho from malignant tissue was significantly elevated compared to benign breast tissue. The potential advantages of MRSI over SV-MRS include the ability to assess multiple lesions and tissues with normal appearance, as

well as to distinguish lesion borders and infiltration into the surrounding tissue. Another major concern of in vivo breast MRS is the spectral overlap, in particular the dominant peaks due to lipids. To address this issue, Thomas et al. have carried out localized two-dimensional correlation spectroscopy experiments [66] and have shown that the W–F ratio can be calculated using two-dimensional spectral peak volumes. In addition, peaks due to unsaturated fatty acids can also be differentiated from saturated fatty acids. The TCho signal can also be used to monitor tumor response to different therapeutic regimens [22,56]. In a study carried out in our laboratory on 44 infiltrating duct carcinoma patients, the TCho peak was either reduced or absent after the third and/or sixth cycle of neoadjuvant chemotherapy [22].

Future Directions and Conclusions Results obtained to date indicate that the sensitivity of MRS may be limited by various technical factors rather than by intrinsic properties of the tumors. Therefore, any improvement in signal-to-noise ratio that will effectively enhance the detection of TCho resonance may increase the sensitivity and improve the diagnostic potential especially in relation to 1 H MRS. Moreover, the detection of choline signal can be increased by performing MRS at higher fields [24,55]. Use of special radio frequency pulse sequences, like simultaneous suppression of water and

Breast MRS

Acknowledgments The financial grant (SP/S0/B27/95 and SP/S0/B21/2001) to NRJ from the Department of Science and Technology, Government of India, is gratefully acknowledged. We wish to thank Prof. G. Govil for critical evaluation of the manuscript and for many useful suggestions. The authors thank Drs. Mahesh Kumar, V. Seenu, O. Coshic, S.N. Dwivedi, Profs. A. Srivastava, P. K. Julka, G.K. Rath, Dr. Stefan Roell, and Mr. Virendra Kumar for their support and help.

References 1. Greenlee RT, Hill-Harmon MB, Murray T, Thun M. CA Cancer J. Clin. 2001;51:15. 2. National Cancer for Health Statistics. SEER Cancer Statistics Review (1973–1995). Bethesda, MD: US National Cancer Institute, 1998. 3. National Cancer Registry Programme (NCRP): Consolidated Report (1990–1996). An Incidence and Distribution of Cancer. Indian Council of Medical Research, New Delhi, 2001. 4. Stavros AT, Thickman D, Rapp CL, Dennis MA, Parker SH, Sisney GA.Radiology. 1995;196:123. 5. Fenlon HM, Phelan N, Tierney S, Gorey T, Ennis JT. Clin. Radiol. 1998;53:17. 6. Birdwell R, Ikeda D, O’Shaughnessy KF, Sickles EA. Radiology. 2001;219:192. 7. Schnall MD, Orel SD. Magnetic Resonance Imaging Clinics of North America: Breast MR Imaging, Vol. 9. WB Saunders Co.: Philadelphia, 2001.

8. Orel SG, Schnall MD. Radiology. 2001;220:13. 9. Sinha S, Sinha U. Ann. N. Y. Acad. Sci. 2002;980:95. 10. Kneeshaw PJ, Turnbull LW, Drew PJ. Br. J. Cancer. 2003;13:4. 11. Schnall MD. Radiol. Clin. N. Am. 2003;41:43. 12. Sijens PE, Wijrdeman HK, Moerland MA, Bakker CGJ, Vermeulen JW, Luyten PR. Radiology. 1988;169:615. 13. Roebuck JR, Cecil KM, Schnall MD, Lenkinski RE. Radiology. 1998;209:269. 14. Jagannathan NR, Singh M, Govindaraju V, Raghunathan P, Coshic O, Julka PK, Rath GK. NMR Biomed. 1998;11:414. 15. Gribbestad IS, Singstad TE, Nilsen G, Fjosne HE, Engan T, Haugen OA, Rinck PA. J. Magn. Reson. Imaging. 1998;8:1191. 16. Kvistad KA, Bakken IJ, Gribbestad IS, Ehrnholm RJB, Steiner L, Fjosne HE, Haraldseth O. J. Magn. Reson. Imaging. 1999;10:159. 17. Jagannathan NR, Kumar M, Raghunathan P, Coshic O, Julka PK, Rath GK. Curr. Sci. 1999;76:777. 18. Cecil KM, Schnall MD, Siegelman ES, Lenkinski RE. Breast Cancer Res. Treat. 2001;68:45. 19. Yeung DKW, Cheung HS, Tse GMK. Radiology. 2001;220:40. 20. Lee JK, Tsai SC, Ho YJ, Chanclai SP, Kao CH. Anticancer Res. 2001;21:1481. 21. Bakken IJ, Gribbestad IS, Singstad TE, Kvistad KA. Magn. Reson. Med. 2001;46:189. 22. Jagannathan NR, Seenu V, Coshic O, Dwivedi SN, Julka PK, Srivastava A, Rath GK. Br. J. Cancer. 2001;84:1016. 23. Stanwell P, Gluch L, Clark D, Tomanek B, Baker L, Giuffre B, Lean C, Malycha P, Mountford CE. Eur. Radiol. 2005;15:1037. 24. Bolan PJ, Meisamy S, Baker EH, Lin J, Emory T, Nelson M, Everson LI, Yee D, Garwood M. Magn. Reson. Med. 2003;50:1134. 25. Jacobs MA, Barker PB, Bottomley PA, Bhujwalla Z, Bluemke DA. J. Magn. Reson. Imaging. 2004;19:68. 26. Leach MO, Varrill M, Glaholm J, Smith TAD, Collins DJ, Payne GS, Sharp JC, Ronen SM, McCready VR, Powles TJ, Smith IE. NMR Biomed. 1998;11:314. 27. Ronen SM, Leach MO. Breast Cancer Res. 2001;3:36. 28. Park JM, Park JH. Korean J. Radiol. 2001;2:80. 29. Degani H, Horowitz R, Itzchak Y. Radiology. 1986;161: 53. 30. Gribbestad IS, Fjosne HE, Haugen OA, Nilsen G, Krane J, Petersen SB, Kvinnsland S. Anticancer Res. 1993;13: 1973. 31. Gribbestad IS, Petersen SB, Fjosne HE, Kvinnsland S, Krane J. NMR Biomed. 1994;7:181. 32. Mackinnon WB, Barry PA, Malycha PL, Gillett DJ, Russell P, Lean CL, Doran ST, Barraclough BH, Bilous M, Mountford CE. Radiology. 1997;204:661. 33. Glunde K, Jie C, Bhujwalla ZM. Cancer Res. 2004;15:4270. 34. Ting YT, Sherr D, Degani H. Anticancer Res. 1996;16:1381. 35. Mountford CE, Somorjai RL, Malycha P, Gluch L, Lean C, Russell P, Barraclough B, Gillett D, Himmelreich U, Dolenko B, Nikulin AE, Smith ICP. Br. J. Surg. 2001;88:1234. 36. Bottomley PA. U.S. Patent 4. 1984;480:228. 37. Bendall MR, Gordon RE. J. Magn. Reson. 1983;53:365. 38. Ordidge RJ, Connelly A, Lohman JAB. J. Magn. Reson. 1986;66:283. 39. Frahm J, Merboldt KD, Hanicke W. J. Magn. Reson. 1987;72:502.

Part II

lipid signals that are specifically optimized for detection of the signal at 3.2 ppm, would also improve the sensitivity of detection of TCho. Use of respiratory-gated radio frequency sequences would improve motion related artifacts. In addition, advances in the design of MR coils should improve detection of TCho. Use of metabolic imaging will also allow exploration of tumor heterogeneity and characterization. Currently, breast MRS is not routinely performed as part of a breast MRI examination in many centers, in part because it is technically challenging and time consuming. More research focusing on early diagnosis with improved sensitivity and specificity and evaluation of response to chemotherapeutic agents in breast cancer patients would be clinically useful. Presently, MRS and MRI are complementary tools to histology, mammography, and other accepted techniques. Increasing use of these methods is expected for basic research, clinical investigations, and ultimately for patient diagnosis. The sensitivity and specificity of in vivo MR particularly for small lesions needs to be improved before MRS can be incorporated into clinical practice.

References 1085

1086 Part II

Medical Uses

Part II

40. Frahm J, Bruhn H, Gyngell ML, Merboldt KD, Hanicke W, Sauter R. Magn. Reson. Med. 1989;9:79. 41. Orididge RJ, Bendall MR, Gordon RE, Connelly A. Volume selection for in vivo biological spectroscopy. In: G Govil, CL Khetrapal, A Saran, AS Tata (Eds). Magnetic Resonance in Biology and Medicine. McGraw Hill: New Delhi, 1985, p 387. 42. Brown TR, Kincaid BM, Ugurbil K. Proc. Natl. Sci. U.S.A. 1982;79:3523. 43. Maudsley AA, Hilal SK, Perman WH, Simon HE. J. Magn. Reson. 1983;51:147. 44. Haase A, Frahm J, Hanicke W, Matthei D. Phys. Med. Biol. 1985;30:341. 45. Twelves CJ, Porter DA, Lowry M, Dobbs NA, Graves PE, Smith MA, Rubens RD, Richards MA. Br. J. Cancer. 1994;69:1151. 46. Payne GS, Dowsett M, Leach MO. Breast. 1994;3:20. 47. Twelves CJ, Lowry M, Porter DA, Dobbs NA, Graves PE, Smith MA, Richards MA. Br. J. Radiol. 1994;67:36. 48. Ng TC, Grundfest S, Vijaykumar S, Baldwin NJ, Majors AW, Karalis I, Meaney TF, Shin KH, Thomas FJ, Tubbs R. Magn. Reson. Med. 1989;10:125. 49. Merchant TE, Gierke LW, Meneses P, Glonek T. Cancer Res. 1988;48:5112. 50. Merchant TE, Meneses P, Gierke LW, Otter WD, Glonek T. Br. J. Cancer. 1991;63:693. 51. Kalra R, Wade KE, Hands L, Styles P, Camplejohn R, Greenall M, Adams GE, Harris AL, Radda GK. Br. J. Cancer. 1993;67:1145. 52. Glaholm J, Leach MO, Collins DJ, Mansi J, Sharp JC, Madden A, Smith IE, Mc Cready VR. Lancet. 1989;1:1326.

53. Redmond OM, Stack JP, O’Connor NG, Carney DN, Dervan PA, Hurson BJ, Ennis JT. Magn. Reson. Med. 1992;25: 30. 54. Katz-Brull R, Lavin PT, Lenkinski RE. J. Natl. Cancer Inst. 2002;94:1197. 55. Roebuck JR, Lenkinski RE, Bolinger L, Schnall MD. Proc. Int. Soc. Magn. Reson. Med. 1996;4:1246. 56. Meisamy S, Bolan PJ, Baker EH, Bliss RL, Gulbahce E, Everson LI, Nelson MT, Emory TH, Tuttle TM, Yee D, Garwood M. Radiology. 2004;233:424. 57. Sijens PE, vander Bent MJ, Nowak PJ, van Dijk P, Oudkerk M. Magn. Reson. Med. 1997;37:222. 58. Bakker CJG, Vriend J. Phys. Med. Biol. 1983;28:331. 59. Jagannathan NR, Seenu V, Kumar M. Radiology. 2002; 223:281. 60. Mackinnon WB, Huschtscha L, Dent K, Hancock R, Paraskeva C, Mountford CE. Int. J. Cancer. 1994;59: 248. 61. Mescher M, Tannus A, Johnson MO’N, Garwood M. J. Magn. Reson. 1996;A123:226. 62. Gribbestad IS, Singstad TE, Kristofferson A, Kvistad KA, Johnson IT, Lundgren S, Roell S. Proc. Int. Soc. Magn. Reson. Med. 2004;11:2044. 63. Katz-Brull R, Margalit R, Bendel P, Degani H. MAGMA 1998;6:44. 64. Johnstone PA, Moore EM, Carrillo R, Goepfert CJ. Cancer. 2001;91:1075. 65. Wang J, Shih TT, Hsu JC, Li YW. Clin. Imaging. 2000;24: 96. 66. Thomas MA, Binesh N, Yue K, DeBruhl N. J. Magn. Reson. Imaging. 2001;14:181.

1087

Mathias Hoehn and Uwe Himmelreich In-vivo-NMR-Laboratory, Max-Planck-Institute for Neurological Research, Cologne, Germany

Introduction The recent years have seen dramatic progress in molecular imaging across the board of various techniques. Molecular imaging, in the context of the authors’ goals in this chapter, is termed as the non-invasive, in vivo visualization of cellular and molecular events in normal and pathological processes. In particular, optical and in vivo bioluminescence imaging (BLI) techniques [1,2] but also positron emission tomography (PET) [3,4] have set the pace. All these high-resolution imaging modalities are based on the same general idea that the presence of specific contrast agents, incorporated or intracellularly induced by cellular activities, will provide the necessary contrast of the cells or compartments of interest against the background of the host tissue [1,5]. For the purpose of specific contrast generation, either tracers or contrast agents are incorporated in cells or attached via selective antigen–antibody bindings. Alternatively, cells are, in close cooperation with molecular biologists, transfected with the goal of expressing selected markers. These markers are considered “reporter genes,” as the genes responsible for the marker expression are coupled to other predefined genetic expression patterns. Thus, the reporter genes can be chosen to be expressed constitutively (i.e. permanently) or conditionally (i.e. dependent on specific gene activities, linked to particular cellular dynamics or functional activity). Magnetic resonance imaging (MRI) has joined this group of imaging modalities with cellular/molecular focus only very recently. The reason for this “late appearance” as a player in the field of molecular imaging, despite its high spatial three-dimensional (3D) resolution, is based on the fact that labeling strategies of cells for this imaging technique are not as intuitively available and not directly accessible. For example, optical techniques exploit very successfully the existence of naturally occurring chromophores, genetic code of which is transfected in a selected group of cells. The most successful single chromophore to date is probably the green fluorescent protein “GFP,” originally found in certain jellyfish. For MRI applications of microscopic resolution and for the study of cellular dynamical processes, cells can be effectively loaded with iron oxide nanoparticles. This cell loading generates a pronounced contrast in T2∗ -weighted Graham A. Webb (ed.), Modern Magnetic Resonance, 1087–1098.
C 2008 Springer.

images. This approach is by now well established and has already led to a variety of applications in diverse medical but also in developmental biological areas. In the latter field, individual cells of an embryo have been labeled [6,7] and, consequently, the fate of the labeled cells during further cell division and selective cell migration was followed in vivo. In the medical arena, there are two major application fields: one dealing with the monitoring of macrophage activity during inflammatory processes in various organs including liver [8,9] as well as brain [10–12] and another field dealing with stem cell implantation in the hope of successful tissue regeneration [13–16] (Figure 1). This chapter is not aiming to review the broad application field. For this purpose, the reader is referred to a couple of excellent recent reviews [17,18]. The basis for the success of these applications in animal models rests in all cases on fundamental considerations concerning the reliable detectability of the cells of interest and the monitoring of their fate in space and time. The present chapter will review the methodological and technical aspects of the ultrahigh-resolution MRI technique, and, most of all, the key tool of this strategy: cell labeling. The goal of the present chapter is to provide the reader with an insight into the methodological requirements for successful Molecular MRI, and to prepare an overview of the various strategies from which the reader is then free to determine an optimal strategy for particular applications.

Detectability Intrinsic contrast is in most cases insufficient to monitor cells in vivo. Cell-specific contrast must be generated in order to distinguish between targeted cells and other tissue or cells. Generation of sufficient contrast depends on (a) the availability of high-affinity contrast agents, (b) the ability of these contrast agents to overcome biological delivery barriers (e.g. cell membranes; blood–brain barrier), (c) minimal disturbance of the cellular microenvironment (low toxicity), (d) amplification strategies that result in appropriate concentrations, and (e) the availability of sensitive, high-resolution imaging techniques. Cells can be labeled in vivo or by pre-labeling, ex vivo. Paramagnetic MRI contrast agents modulate the spin– lattice relaxation time (T1 ) of rapidly exchanging water

Part II

In Vivo Molecular MR Imaging: Potential and Limits

1088 Part II

Medical Uses

Part II Fig. 1. Coronal sections through a rat brain at 6 and 8 days following implantation of embryonic stem cells into the left hemisphere, contralateral to the induced transient 60 min focal cerebral ischemia. 3D data sets were recorded with isotropic spatial resolution of 78 µm with a scan time of approximately 70 min. The primary implantation sites in the cortex and the striatum are indicated by white arrows. Note at 6 days, the discrete dark line (dark arrow) along the corpus callosum between the cortical implantation site and the ventricular wall showing cells migrating toward the lesioned hemisphere. At 8 days postimplantation, a dark region becomes visible in the dorsal part of the lesioned territory reflecting the first arrival of USPIO-labeled cells.

molecules and alter the local homogeneity of the magnetic field (T2 and T2∗ relaxation times). Cells labeled with contrast agents generate hypointense contrast in T2 -weighted (T2 w) images (labeled e.g. with iron oxide particles) or hyperintense contrast in T1 -weighted (T1 w) images (labeled e.g. with Mn or lanthanide chelates). These contrast agents most commonly consist of nanoparticles (iron oxide, T2 w), inorganic chelates (Gd, Eu, and Mn chelates; T1 w), or particles contained in a polymer sphere (iron oxide microspheres). A precondition for a contrast agent is that the relaxivity of exchanging/surrounding protons is sufficiently modified. The large magnetic susceptibility of small paramagnetic particles can affect much larger regions of the image than is suggested by the actual size of the particle [19]. The limits of cell detectability are defined by the spatial resolution, the signal-to-noise ratio (SNR), and the contrast-to-noise ratio (CNR).

Contrast Agents for Molecular Imaging The suitability of magnetic markers for molecular MR imaging is influenced by the change of the MR signal caused by alterations in the magnetic susceptibility and expressed by the relaxivity (R). Contrast agents for molecular imaging must provide a high relaxivity in vivo without compromising the cellular microenvironment. This sets certain limitations for the concentrations of contrast

agents. The choice of contrast agents also depends on the intrinsic background signal of the respective application. Dextran coated superparamagnetic iron oxide particles (SPIOs) [20] and ultrasmall SPIOs (USPIOs) have magnetic moments more than three orders of magnitude larger than those of clinically applied contrast agents [21]. Iron oxide particles can be used for T1 - [22], T2 - [12], and T2∗ [23] weighted imaging. Due to the large magnetic moment of (U)SPIOs, the relaxivities R2 and R2∗ are several thousand times larger than R1 [24]. (U)SPIOs are well suited for cell tracking because of the long-range susceptibility effects. Relaxivity (R2 /R2∗ ) of iron particles can easily be tested in vitro by comparison of signal quench in the presence of respective contrast agents. R2∗ relaxivity increases with increasing concentrations of the iron oxide particles but will be saturated (complete quench of signal) [24–26]. Saturation of magnetization occurs once the mass of iron per cell exceeds a certain threshold [24]. When normalized to iron content, larger particles greatly decrease T2∗ and enhance the susceptibility effect when compared to smaller particles [27]. The T2∗ effect of iron oxide based contrast agents is also larger than T1 effects of Mn- and lanthanide-chelates because water molecules must be in close proximity of the paramagnetic center to generate T1 contrast [28]. Crich et al. estimated the minimum detectable number of Gd-HPDO3A labeled cells in a phantom at approximately 103 cells [29]. Visualization is attainable when cellular concentrations of Gd-contrast agents are in the order of 10−4 –10−5 molar [30,31]. Magnetic properties that are most commonly modified for the design of contrast agents that predominantly increase R1 relaxivity (hyperintensity in T1 w MR images) include: (a) hydration number (q)—water bound to agent (b) exchange rate of water molecules (τ m ) (c) rotational correlation time of complex (τ r ). In order to achieve sufficient contrast with T1 w contrast agents in vivo, accumulation of contrast agent to millimolar concentrations is usually required [32]. A system that entraps several units of a Gd chelate is apoferritin, resulting in an increase of relaxivity by almost 20 times compared to the free contrast agent [33,34]. Other approaches to concentrate Gd chelates at the site of interest include self-assembling aggregates [30,35,36]. It was suggested in recent studies that even picomolar binding is sufficient to achieve detectable CNR if high loads of particles were delivered to the cell [37]. The rotational correlation time of a contrast agent can be altered by changing the viscosity of the environment or by interaction with large molecules [38]. The latter has resulted in the design of contrast agents that specifically bind to certain proteins [39,40].

Molecular MR Imaging

Detectability 1089

Part II

Resolution Low image resolution can attenuate the relaxation effects through partial volume effects and thereby reduce detectability of labeled cells. The spatial resolution is limited by the strength of the gradients and the matrix size. The acquisition of MR images with a resolution in the order of tens of microns is currently possible within a reasonable time frame. The trade-off of high resolution is a reduced SNR. A precondition for detection of single labeled cells by MRI is that the number of voxels is higher than the number of individual cells in the respective volume. Dodd et al. have demonstrated a good correlation of individual cell distribution between MRI and microscopic methods, proving the detectability of single cells by MRI [41]. Larger susceptibility effects, like for micron-size iron oxide particles, result in higher contrast and permit detection at lower resolution [27,42,43]. These microspheres can be considered a point source that creates a localized field larger than the field created by the evenly distributed, smaller (U)SPIOs, resulting in detectability of single cells with a spatial resolution as low as 200 µm [42].

Signal- and Contrast-to-Noise Ratio The dilemma of cell tracking in animal models is that acquisition time is limited and resolution has to be as high as possible in order to meet the requirements set by working with anaesthetized animals and being able to detect small numbers of cells in vivo. This limits the options to maximize SNR and CNR. Apart from longer acquisition times, the SNR can be increased by stronger magnetic fields, smaller RF coils, and cryoprobes [44,45]. Pulse sequences, repetition time (TR ), echo time (TE ), and pulse length have to be chosen carefully and adjusted for each animal model. Optimization of MR parameters has been reviewed extensively [46,47]. The CNR is influenced by the efficiency of the labeling strategy. Visually apparent contrast is defined as CNR ≥ 5, with CNR = |(IA − IB )N −1 | (IA , IB —signal intensities of the two adjacent regions, N —noise level). Using the relaxivity of the respective contrast agents and T1 /T2 —values of the tissue, acquisition parameters (TR , TE ) can be optimized to maximize CNR [31,37]. It was demonstrated that detection of single cells in vitro is feasible [41,42,48].

Quantification Cells can be quantified in vitro by the construction of simple gelatin phantoms [25,49] (Figure 2). If a sufficient spatial resolution is provided in 3D MR images and

cm 2

1

0

Fig. 2. Agarose gel phantom containing embryonic stem cells transfected with iron oxide particles. The holes A–E and G contain approximately 104 cells and the hole F contains 5 × 102 cells, which were suspended in 50 µl agarose. Cells were incubated with different concentrations of iron oxide nanoparticles prior to suspension (amount of iron oxide per 5 × 105 cells: A, 40 mg; B, 18 mg; C and F, 5 mg; D, 1 mg; E, 0.2 mg; and G, no iron oxide particles). The 3D gradient-echo MR image (FLASH) was acquired with the following parameters: field strength = 7 T, matrix = 256K × 128K × 64K data points; TR = 250 ms, TE = 20 ms, FOV = 4 cm × 4 cm).

homogeneous phantoms are utilized individual cells labeled with iron oxide particles can be counted in high dilutions [49]. Bowen et al. have addressed the quantification of iron oxide concentrations in cell suspensions based on relaxivities [24]. Relaxation rates of compartmentalized (U)SPIOs correlated with predicted values using the static dephasing theory [50]. Best correlations were achieved using (R2∗ –R2 ) under the dilute pertuber assumption, indicating the suitability for quantification for cases where volume fractions of cells containing iron oxide are relatively small, as is the case, e.g. for implanted cells in vivo [24]. This model has only been tested in vitro. Experimentally, detection of cell clusters of tens to hundreds of (U)SPIO-labeled cells was demonstrated in animal models [14,51]. It was suggested that the utilization of micronsize iron oxide particles are suitable to detect single cells in vivo [52]. An additional advantage of these latter particles is that single particles or cells can be detected in vitro even with relatively low resolution (due to the large diameter and magnetic moment), and that the label is not diluted

1090 Part II

Medical Uses

Part II

beyond the minimum observable threshold by cell division but remains within the cell for up to 1 year [27,43,52]. If the relaxivity of lanthanide contrast agents is known model-based predictions of local concentrations of T1 contrast agents can be reliably performed by measurement of T1 values [31,37]. If particle loading of cells and relaxivity of the contrast agent remain constant after implantation this could result in quantitative MRI in in vivo models. Contrary to these results, Crich et al. have found that cell pellets with variable Gd-chelate loading show a decrease of R1 relaxivity if concentrations exceed 0.1 µmol/mg protein [29]. Possible explanations include competing T2∗ effects, resulting in signal broadening.

Cell Labeling Route of Uptake Conventional MR cell labeling relies on surface attachment of contrast agents to the cell [53–56]. Contrast agents are often chemically modified so that they bind unspecifically or to specific cell types. A common strategy involves linking a contrast agent with a monoclonal antibody that is recognized by receptors on the cell surface [55]. For review see Ref. [30]. These strategies are generally not suitable for in vivo applications, where interactions with other cells have to be avoided, as for stem cell tracking. Extracellular labeling may result in detachment of the label or transfer to other cells. We will therefore focus on strategies to internalize contrast agents. Probably the most effective route to internalize compounds into cells is by endocytosis. Hereby, intracellular vesicles (endosomes) are formed after introflection of portions of the cell membrane and fusion of its extremities. The endosomes contain compounds that are either bound to the membrane or present in the extracellular medium. Endocytosis is most effective in phagocytic cells (also called phagocytosis). Paramagnetic iron oxide particles but also Gd chelates have been successfully internalized into macrophages, monocytes, mononuclear T-cells, and oligodendrocyte progenitors [13,26,29,57–61]. In order to internalize high quantities of a contrast agent, receptormediated endocytosis is sometimes the method of choice. Hereby, the substrate in the extracellular medium interacts with receptor molecules on the membrane (like contrast agents linked with a monoclonal antibody) and are internalized in high concentrations [26,62]. Hinds and co-workers [27,43] have demonstrated that even micron-scale particles (0.9–2.8 µm) of an aggregate of iron oxide magnetite that is encapsulated in a polymer, are incorporated into perinuclear endosomes by stem and progenitor cells [48]. This labeling strategy has been highly efficient (almost 100%). It was also found that concentrating (U)SPIOs into cells increases the relative R2∗

sensitivity, compared to uniformly distributed particles [26,63]. In order to increase the uptake of contrast agents that do not spontaneously accumulate in endosomes, magnetoliposomes [64], and lectins [65] have been used in several experiments. A method developed in molecular biology for transfection of oligonucleotides, lipofection has been successfully adopted for labeling of cells with contrast agents [14]. Commercially available polycationic transfection agents encapsulate (U)SPIOs through electrostatic interaction [66]. The transfection agents shuttle the (U)SPIOs into the cell and form endosomes that finally merge into liposomes. This procedure has successfully been applied for labeling T-cells, stem cells, and other mammalian cells [14,66–69]. A similar approach is the labeling of cells with magnetodendrimers, which consist of (U)SPIOs coated with carboxylated dendrimers [51]. These compounds are also used as transfection agents and result in the formation of endosomes [70]. A broad range of cells has been labeled with dendrimers [15,51,71,72]. Others have produced transgenic cells expressing, for example, the transferrin receptor and are thus susceptible to the uptake of conjugates of transferrin and monocrystalline iron oxide nanoparticles [73,74]. Many larger probes are membrane impermeable and fail to attain intracellular concentrations required for MRI. Peptides or proteins mediating membrane translocation signals (MTS) have been coupled with contrast agents for internalization. MTS from the HIV tat peptide have been linked to SPIOs and lanthanide chelates for transport into several cell types [75–78]. Hereby, the contrast agents most likely accumulate in the cytoplasm [77]. Allen and Meade have suggested a relatively simple system for internalization that consists of a polyarginine-modified Gd DOTA system [79]. Microinjection of contrast agents was used in embryology for cell tracking [6,7]. Increased contrast was detectable in the developed embryo compared to controls, indicating sufficient contrast with cell division. However, this method has only applications limited to injection of large single cells in developing embryos. It is not practical for tracking populations of cells due to its labor-intensive preparation and the dilution of the label with cell division [7]. Zhang et al. adopted another method from molecular biology, where so-called biolistic “gene-guns” are used to introduce oligonucleotides into cells [80,81]. Biolistic labeling was performed with ferromagnetic particles and resulted in a labeling efficiency of 95% of surviving cells. Similarly to microinjection, cell labeling remained stable. Electroporation is a process that permeabilizes cell membranes by exposure to an external electric field. Rapid resealing of the membrane pores traps internalized

Molecular MR Imaging

In vivo and ex vivo Labeling In vivo labeling of cells is possible if phagocytes take up contrast agents that were administered either intravenously or after injection in the targeted organ. This labeling strategy is based on naturally occurring or targeted, mediated endocytosis (mainly macrophages). Hereby, macrophages accumulate the iron oxide particles from the blood, whereas the remaining iron oxide in the blood (or other organs) is washed out. It was first introduced for hepatic imaging [84]. Subsequent development of clinically approved USPIOs have resulted in labeling strategies to distinguish between normal tissue and tumor metastases in lymph nodes [85] and bone marrow [86] based on the uptake of these particles by macrophages in those organs. The uptake of iron oxide by macrophages has also been utilized in some disease models to highlight increased macrophage activity in areas of inflammation. Such applications include visualization of atherosclerotic plaques [87], visualization of increased macrophage activity in the kidney after injury and organ rejection [88,89], and macrophage infiltration in brain lesions following ischemia [10,90]. Direct injection of micron-size iron oxide particles into various regions of the sub-ventricular zone of the rat brain resulted in in vivo labeling of neuronal stem cells by endocytosis, allowing their migration to be tracked along the rostral migratory stream to the olfactory bulb [52]. Coupling of MTS peptides with nanoparticles and antibody or receptor binding ligands has been suggested as a possible targeted in vivo labeling strategy [91]. Paturneau-Jouas et al. have used electroporation after intraperitoneal injection of Gd chelates for internalization into muscle cells in vivo [83]. So far, targeted in vivo labeling was mainly demonstrated for phagocytes. In all other cases, pre-labeling of cells according to the procedures listed above is necessary. Hereby, in vitro pre-labeled cells are transferred to the targeted organ or tissue by injection.

Responsive Contrast Agents Contrast agents have been developed with the aim of reporting on the physiological status and metabolic activity

of cells [30,92]. Most of these contrast agents are Gd based chelates with one or more potential coordination sites blocked in the inactive state, preventing free access of exchanging water molecules to the paramagnetic ion. Lowe et al. synthesized an MR contrast agent that is responsive to changes in pH, generating no contrast at high pH but high contrast at low pH [93]. The Gd DO3A derived agent contains a sulfonamide nitrogen, which is protonated at low pH. Therefore, it cannot coordinate at low pH and provides free access to exchanging water molecules. Meade and co-workers developed a probe that is responsive to intracellular changes in calcium concentrations [94,95]. Similarly, two carboxyl groups of the chelate coordinate either to the Gd3+ ion (inactive) or to the Ca2+ ion (active). Probes responsive to enzyme activity (e.g. β-galactosidase) have been developed [96] and applied for cell visualization in vivo [6]. Contrast agents responsive to pO2 were developed on the basis of the BOLD effect [97] and coupling an increase in τ r with and redox switch [98]. Other contrast agents utilize the so-called chemical exchange saturation transfer (CEST) agents to be responsive to variable concentrations of targeted molecules [99,100]. Aime et al. have suggested an insoluble contrast agent that is activated by intracellular enzymatic solubilization [30]. However, the effective incorporation of the contrast agent in its insoluble state is still an unresolved issue.

In vivo MRI Experiments Cell tracking in vivo requires the acquisition of isotropic 3D MR images of living animals at near-cellular resolution in order to minimize partial volume effects. Temporal constraints are set by working with live animals so that spatial resolution is typically in the range of tens up to a hundred of microns. The principle challenge is to obtain sufficient SNR. Hardware improvements often target the reduction of thermal noise voltages from the sample by using small surface RF coils [101]. Design of the receiver coil aims for low coil resistance. This can be achieved by cryo-cooled probes [102], which have been applied to micro-imaging in animal models [44,45]. RF coils made of high temperature superconducting material have also been demonstrated to reduce coil loss substantially in vitro and in animal models [103,104]. Animal models require usually anaesthetized animals. Apart from the obvious associated disadvantages, it has the advantage of decreasing movement artifacts in MR images as the animal can be firmly fixed in an animal holder. MRI of animal models is usually performed at a high field strength (>4 T). While this is an advantage for improving the SNR it has the disadvantage of an increase in the T1 /T2 ratio, resulting in an increase in the acquisition

Part II

compounds inside the cell. Widely used to introduce foreign DNA into cells, it has also been used for drug delivery [82]. Paturneau-Jouas et al. have used electroporation for the uptake of low-molecular weight Gd chelates in vivo [83]. It remains to be seen how efficient this labeling method is and how stressful it proves for the respective cells.

In vivo MRI Experiments 1091

1092 Part II

Medical Uses

Part II

time and in a loss of intrinsic contrast in the brain. The increase in the magnetic field strength, compared to clinical systems, also results in more pronounced magnetic susceptibility effects. While this can lead to unwanted artifacts in the MR images, it has also the advantage that magnetically labeled cells can be more easily visualized. Gradient-echo sequences tend to be more susceptible to those artifacts than spin-echo experiments. Data acquisition methods utilized for cell tracking are similar to those employed for clinical MRI (spin-echo, gradient-echo, diffusion-weighted imaging, inversion recovery, and magnetization transfer sequences). Rapid acquisition methods are commonly used for animal models, namely gradient-echo fast-low-angle-shot (FLASH) [105] and echo planar imaging (EPI) [106]. The utilization of MR microscopy for animal models was reviewed by Ahrens et al. [107]. Pulse sequences that emphasize signal loss due to field inhomogeneities are beneficial for (U)SPIOs. Cell tracking in animals is usually performed using 3D gradientecho sequences (FLASH) for T2 /T2∗ effects [14,51,62] and two-dimensional (2D) multi-slice or 3D spin-echo sequences for T1 effects [6,29,57]. Unless very strong T1 weighting is used with TR s of approximately 100 ms, spinecho sequences are not fast enough to permit the acquisition of a high-resolution 3D image within the time frame required for anaesthetized animals. 2D multi-slice images have the disadvantage of strong partial volume effects. A resolution of 100 to 200 µm is usually sufficient for the visualization of small cell clusters. The detection of single cells is not realistic as naturally occurring movements like breathing and muscle twitching are in the order of hundreds of microns, resulting in blurring effects. Additional blurring effects are caused by diffusion [41]. The acquisition of 3D data sets is, at any rate, a requirement for the co-registration of data sets in longitudinal studies and for the co-registration with independently recorded, non-MR data sets. The effects of gradient strength, TR and TE to optimize SNR and CNR in MRI have been studied extensively [41,46,47]. For conditions of large magnetic moments due to (U)SPIOs loaded cells in animal models, the greatest sensitivity was observed with R2∗ , rather than R2 or R1 [24]. Low gradient strengths through the use of low acquisition bandwidths will enhance susceptibility effects. This is also beneficial in terms of SNR, due to reduced diffusion loss. An intrinsic problem for tracking iron oxide labeled cells in animal models are other areas of susceptibility induced field inhomogeneities (signal loss in T2 /T2∗ w images), namely from airspaces, but more importantly, from blood clots/hemorrhages and tissue interfaces. Seppenwoolde et al. have developed a technique that exploits the natural dipolar field surrounding paramagnetic markers by introducing a dephasing gradient on the slice-select

axis to spoil signal across all the sample except in regions where the local gradients surrounding the paramagnetic marker are of the right amplitude and orientation to combine constructively with the added dephasing gradient to result in refocusing of the lost signal [108]. Coristine et al. have applied this “white-marker” technique to distinguish iron oxide labeled cells from other sources of field inhomogeneities [109]. The optimal acquisition parameters need to be determined for each combination of cell type, animal model, and contrast agent.

Biological Aspects of Cell Labeling Reproducible and efficient visualization of cells with MRI does not only require optimized MR properties of contrast agents but also biological properties that result in stable labels without affecting other cellular processes.

Toxicity Many contrast agents are administered in coated (polymer coating) or chelated (in particular the lanthanides) form, in order to reduce toxicity. Commercially available and clinically tested contrast agents are often used for molecular imaging, for which short and medium term toxicity is minimal. However, one has to consider that contrast agents for molecular imaging are often applied for longitudinal labeling, and in concentrations much higher than clinically applied. Immediate toxic effects can easily be tested by growth and viability studies in cell suspensions. It is more difficult to assess long-term stability (and toxicity) of coated contrast agents and the effect of contrast agents on cellular processes like differentiation. It has been reported that very high concentrations of (U)SPIOs for cell labeling (up to 2 mg iron/ml) can cause decreased proliferation and even cell death [110]. This is in particular true after the formation of reactive oxygen species [111]. Some dextran coatings for iron oxide particles are biodegradable [112]. Released iron enters into the cellular iron metabolic pathways. Iron in high concentrations is known to interact with a number of other metabolic pathways [113,114]. Overloading can result in increased oxidative stress. Although iron oxide can be released by biodegradation, one has to consider that this will occur gradually resulting in non-toxic free iron concentrations. Stroh et al. have found that decomposition of the coating of iron oxide particles does not affect cell growth but causes oxidative stress [113]. These authors have suggested the application of antioxidants or iron chelators to prevent toxic effects of leaking iron. T-cells labeled with SPIOs by endocytosis retain their function [58,115]. Sapiro et al. have shown in longitudinal studies that even large iron oxide loaded microspheres do

Molecular MR Imaging

Assessment and Stability of Label It is crucial for in vivo cell tracking to estimate the efficiency of cell labeling, the internalization of the label, and for how long these cells are detectable (i.e. dilution of the label; Figure 3). Various approaches have been taken to assess in vitro labeled cells by using agar or gelatin based phantoms (Figure 2). It was found that uptake of (U)SPIOs as well as Gd chelates by endocytosis is linear with dose [24,29]. Linearity was also shown for (U)SPIO uptake with incubation

time [24,118]. Interestingly, uptake is increased with particle size [118], which also explains the excellent uptake of micron-size microspheres [27,48]. The labeling efficiency can be assessed by NMR relaxometry [29,30,7,78] or by other techniques like atomic absorption spectroscopy to determine the iron content of USPIO-labeled cells [14]. Internalization of contrast agents (Figure 3) can also be verified by staining techniques in combination with (electron) microscopy. The visibility of cells with time depends primarily on the dilution of the contrast agent due to cell proliferation and on degradation processes of the coating of contrast agents. Stem cells labeled with iron oxide particles have been monitored in vivo for up to 7 weeks in rat brains [14] and for up to 3 months in a spinal cord implantation model [72]. Embryonic stem cells were detectable, with contrast condition comparable to the originally labeled culture, after 20 cycles of cell division, which is an increase of cell numbers by a factor of 106 [14,25,49]. In another study, the complete disappearance of iron particles has been observed due to dilution after proliferation within 2–3 weeks [69]. An advantage of large micron-size particle is that a single particle per cell may result in sufficient contrast. Hill et al. were able to detect mesenchymal stem cells for up to 3 months [43]. Shapiro and co-workers monitored microsphere labeled cells for 1 year [52]. Cell labeling with lanthanide chelates is more affected by dilution after cell division due to its lower sensitivity. Modo and co-workers were able to visualize stem cells for up to 7 days after labeling with a Gd chelate [13]. Histological evaluation at the endpoint of longitudinal MR studies is necessary to evaluate MR tracking of cells in vivo. Other causes for loss of label are exocytosis (Figure 3) and immune reaction with a subsequent transfer of contrast agents to macrophages. Co-location can be estimated by co-registration of MR images with stained histological sections. Due to the invasiveness of histological methods, co-registration with MR images will always be problematic due to swelling, distortion, etc. Possible approaches include corroboration of MRI detection of labeled cells by using Prussian blue staining against iron [62], validation with X-gal staining [51], or electron-dense inclusions for visualization by electron microscopy [61]. In addition, dual staining detects co-localization of contrast agent with certain cell types. An example is the Prussian blue and ED1+ staining to discriminate USPIO-labeled macrophages [10,119]. The use of bifunctional contrast agents allows coregistration of histology and MRI without recourse to other methods. Dual labeling strategies can be classified in methods that use dual contrast agents and methods that utilize cells that can be visualized by specific modalities (e.g. cells that express certain proteins). Modo labeled stem cells with a Gd-Rhodamine dextran (GRID) that is

Part II

not influence cell proliferation and differentiation at low concentrations [48,52]. Hill et al. found a dose-dependent alteration in cell proliferation but not in differentiation for mesenchymal stem cells [43]. Also, magnetodendrimers and (U)SPIOs did not cause differences between labeled and unlabeled cells in terms of proliferation, viability, and differentiation of neural stem cells in short term in vitro experiments [14,51,71]. Frank and co-workers have studied the effect of superparamagnetic iron oxide nanoparticles and transfection agents extensively [67–69,116]. No long-term effect on cell viability, rate of apoptosis, phenotype, proliferation, activation, and differentiation was found. The toxicity of lanthanide contrast agents has also been studied extensively [38]. As lanthanides are known to be highly toxic, stable chelates are required for protection. Gd-HPDO3A is one of the best tolerated contrast agents with an LD50 > 10 mmol/kg in rats [38]. Although well tolerated in clinical applications, high dose toxicity for molecular imaging application requires further investigations. Several studies have confirmed that Gd-dextrans remain intact in cells [29] and are tolerated by cells in culture and in vivo after implantation [13,29,57]. In addition to metal ions, the interactions of chelates, transfection agents, or antibodies have to be considered for the assessment of toxicity. Although concern has been raised about potential release of reactive iron species from tat-labeled cells [67], functional tests with and without HIV tat peptide based contrast agents have shown no difference between labeled and unlabeled cells [77]. When used alone, some transfection agents are reported to be toxic [116]. However, iron oxide complexed with transfection agents does not affect viability, differentiation, or proliferation [68,69,116]. It was found in one study that the differentiation of stem cells can be influenced by transfecting agents [117]. This highlights the need to study possible biological side effects for each labeling system and cell type. In addition to toxicological aspects of the contrast agent, one has to consider that labeling procedures like electroporation, microinjection, or biolistic “gene-guns” can cause mechanical damage to cells.

Biological Aspects of Cell Labeling 1093

1094 Part II

Medical Uses

Part II Fig. 3. Investigation into the persistence of iron oxide label after incorporation into HeLa cells or mesenchymal stem cells (MSC). On the left side, Prussian blue staining of cell cultures is depicted at various times after cell labeling. It is clear that the blue stain reflecting the iron content is rapidly lost in the heavily proliferating HeLa cells, while dark blue stains are still detected in the (very little proliferating) MSCs even after 44 days. The bar diagrams on the right show the iron content as determined in the cell culture medium (red) and in the cells (yellow), with the situation for the HeLa cells above, and the MSCs below. Again, the iron content remains quite stable for the MSCs, both in the cytosol and in the media. The HeLa cells have a maximum of iron directly following labeling, which is rapidly lost again (by leakage). This is reflected in the medium where the iron content transiently increases during the phase when the cells lose their iron content. Modified with permission from Ref. [69]. (See also Plate 89 on page XVI in the Color Plate Section.)

both magnetically and fluorescently detectable [13,57]. Hereby, MRI results can be assessed by subsequent fluorescence microscopy on histological sections. In addition, fluorescence microscopy allows additional staining to detect co-localization with markers for cell differentiation [13].

Similarly, labeling with Mn2+ as contrast agent allows validation by an independent method. Mn2+ has been used as an in vivo, trans-synaptic, MRI-detectable neuronal tract tracer that enters selectively neurons via Ca2+ channels [120–122]. The fate of Mn2+ can be verified by independent fluorescent dye staining.

Molecular MR Imaging

Summary In summary, the successful design of a molecular imaging study of cells by MRI depends on a series of methodological evaluations and optimizations. As several of these optimization steps will prove to be dependent on the particular circumstances of the investigation, a checklist of aspects to be assessed before in vivo experiments are started can be compiled and is given here. 1. Select contrast agent dependent on intrinsic background (e.g. signal loss vs. signal gain), 2. optimize internalization of contrast agent, 3. check residence duration of contrast agent in cells/dilution of label, 4. check toxicity of label/labeling procedure, 5. consider independent/dual (e.g. in combination with optical) labeling options, generated by cells or colocalized on contrast agent, 6. relaxivity of labeled cells in vitro, 7. MRI sensitivity, and 8. hardware optimization for scan-time minimization. As some of these considerations and tests will vary very much with cell type, disease model, etc. it is always advisable to perform the respective assessments anew when new cell lines or animal models or contrast agent is considered for use.

contributions of in vivo molecular MRI in the future. Expertise in in vivo MRI, including the understanding of animal physiology and pathophysiology, is a central prerequisite. The second partner will contribute expertise in contrast agent chemistry to design and synthesize the new generation of so-called “smart contrast agents.” These could be agents that are activated or deactivated under external control or under control of the cell status. Finally, the third discipline is molecular biology to design transgenic cells, so-called “smart cells.” These cells will be able to activate or deactivate added smart contrast agents according to the selected reporter gene. Alternatively, such transgenic cells may even produce their own intracellular contrast agent, as has recently been shown in an impressive investigation by the laboratory of Neeman [123], where cells under TET-on/TET-off switches could be shifted between modes where they either produced a strong T2 contrast or not. In the activated mode, these cells upregulated the intracellular iron depot ferritin, resulting in (non-toxic) massive accumulation of iron in the cell. The use of MR reporter genes that are genetically expressed and generate contrast coupled to the regulation of genes or by the presence/absence of exogenous substrates opens new doors for the non-invasive analysis of cell functions in various animal models under true in vivo conditions. In conclusion, with the synergy of the above-described three complementary disciplines, molecular MR imaging is only beginning to explore its own potential for future biomedical studies on cellular and molecular levels.

Acknowledgment Financial support of the Hertie-Foundation and of the German-Israeli Cooperation project by the German Ministry of Education and Research (BMBF) is gratefully acknowledged. UH gratefully acknowledges an international reintegration grant (#013080) by the European Commission.

Outlook

Glossary of Terms

With a more widespread application of cell labeling, the above discussed procedures will become more routine and established for a broad range of cell lines and applications. An integrative view on sensitivity improvement of MRI systems will most likely lead to voxel sizes reflecting a handful of cells, under true in vivo conditions, compatible with scan times that result in anesthesia periods tolerable for weakened animals of various disease models, in longitudinal investigations. A close interaction between specialists of three different disciplines is seen as the basis for the exciting

Contrast agent—Chemical compound that increases the intrinsic contrast in an image. Bifunctional contrast agent—Contrast agent that increases the intrinsic contrast for two different imaging modalities. Responsive contrast agent—Contrast agent that conditionally increases the intrinsic contrast (depending on the change of environmental conditions like enzyme activation, pH change, change in ion concentration, and others). Dendrimers—Large and complex polymers of a consistent size and a regular and highly branched architecture.

Part II

Similar approaches, combining iron oxide particles with fluorophores were taken to track stem cells in vivo [27,43]. In contrast to dual contrast agents, co-localization can also be achieved by labeling stem cells that express fluorescence properties with USPIO [14]. These genetically modified stem cells express the GFP that can be detected on histological sections. In contrast to the other approaches, only those cells show fluorescence that is still viable at the time of termination. Hereby, dead cells and cells whose label had been transferred to macrophages can be distinguished from the targeted cells.

Glossary of Terms 1095

1096 Part II

Medical Uses

Part II

They are often used for the encapsulation of smaller molecules. Endocytosis—In endocytosis, the cell engulfs some of its extracellular medium including material dissolved or suspended in it. A portion of the plasma membrane is invaginated and pinched off forming a membrane-bound vesicle called an endosome. Endosome—Vesicle formed during endocytosis. Molecular imaging—Non-invasive, in vivo visualization of cellular and molecular events in normal and pathological processes. Stem cell—Cell with broad differentiation potential that retains the capacity for unlimited self-renewal. A totipotent stem cell has the ability to differentiate to all cell types of an organism, whereas a pluripotent stem cell produces many but not all cell types. Transfection—A method by which experimental DNA (or in our case contrast agents) can be incorporated into mammalian cells after interaction of the encapsulating transfection agent and the cell membrane.

References 1. Weissleder R, Mahmood U. Radiology. 2001;219:316. 2. Edinger M, Cao YA, Hornig YS, Jenkins DE, Verneris MR, Bachmann MH, Negrin RS, Contag CH. Eur. J. Cancer. 2002;16:2128. 3. Blasberg R, Tjuvajev JG. Mol. Imaging. 2002;1:160. 4. Gambhir SS. Nat. Rev. Cancer. 2002;9:683. 5. Weissleder R, Ntziachristos V. Nat. Med. 2003;1:123. 6. Louie AY, H¨uber MM, Ahrens ET, Rothb¨acher U, Moats R, Jacobs RE, Fraser SE, Meade RJ. Nat. Biotechnol. 2000;18:321. 7. Jacobs RE, Fraser SE. Science. 1994;263:681. 8. Weissleder R. Radiology. 1994;193:593. 9. Stark DD, Weissleder R, Elizondo G, Hahn PF, Saini S, Todd LE, Wittenberg J, Ferrucci JT. Radiology. 1988;168:297. 10. Saleh A, Wiedermann D, Schroeter M, Jonkmanns C, Jander S, Hoehn M. NMR Biomed. 2004;17:163. 11. Saleh A, Schroeter M, Jonkmanns C, Hartung HP, M¨odder U, Jander S. Brain. 2004;127:1670. 12. Dousset V, Delalande C, Ballarino L, Quesson B, Seilhan D, Coussemacq M, Thiaudiere E, Brochet B, Canioni P, Caille JM. Magn. Reson. Med. 1999;41:329. 13. Modo M, Cash D, Mellodew K, Williams SCR, Fraser SE, Meade TJ, Price J, Hodges H. Neuroimage. 2002;17: 803. 14. Hoehn M, K¨ustermann E, Blunk J, Wiedermann D, Trapp T, Wecker S, F¨ocking M, Arnold H, Heschler J, Fleischmann BK, Schwindt C, B¨uhrle C. Proc. Natl. Acad. Sci. U.S.A. 2002;99:16267. 15. Bulte JWM, Ben-Hur T, Miller BR, Mizrachi-Kol R, Einstein O, Reinhartz E, Zywicke HA, Douglas TD, Frank JA. Magn. Reson. Med. 2003;50:201. 16. Veizovic T, Beech JS, Stroemer RP, Watson WP, Hodges H. Stroke. 2001;32:1012. 17. Bulte JWM, Modo M, Hoehn M. Mol. Imaging. 2005;4:143.

18. Bulte JWM, Duncan ID, Frank JA, Cereb. J. Blood Flow Metab. 2002;22:899. 19. Lauterbur PCM, Bernardo ML Jr, Mendonca Dias MH, Hedges LK. Proc. Int. Soc. Magn. Reson. Med. 1986;5:229. 20. Fahlvik AK, Klaveness J, Stark DD. J. Magn. Reson. Imaging. 1993;3:187. 21. Bulte JWM, Brooks RA, Moskowitz BM, Bryant LH Jr, Frank JA. Magn. Reson. Med. 1999;42:379. 22. Moore A, Marecos E, Bogdanov A Jr, Weissleder R. Radiology. 2000;214:568. 23. Furman-Haran E, Margalit R, Grobgeld D, Degani H. J. Magn. Reson. Imaging. 1998;8:634. 24. Bowen CV, Zhang X, Saab G, Gareau PJ, Rutt BK. Magn. Reson. Med. 2002;48:52. 25. K¨ustermann E, Wiedermann D, B¨uhrle C, Arnold H, F¨ocking M, Trapp T, Hoehn M. Proc. Int. Soc. Magn. Reson. Med. 2002;10:1255. 26. Weissleder R, Cheng HC, Bogdanowa A, Bogdanow A Jr. J. Magn. Reson. Imaging. 1997;7:258. 27. Hinds KA, Hill JM, Shapiro EM, Laukkanen MO, Silva AC, Combs CA, Varney TR, Balaban RS, Koretsky AP, Dunbar CE. Blood. 2003;102:867. 28. Lauffer RB. Chem. Rev. 1987;87:901. 29. Crich SG, Biancone L, Cantaluppi V, Duo D, Esposito G, Russo S, Camussi G, Aime S. Magn. Reson. Med. 2004;51:938. 30. Aime S, Cabella C, Colombatto S, Crich SG, Gianolio E, Maggioni F. J. Magn. Reson. Imaging. 2002;16:394. 31. Ahrens ET, Rothb¨acher U, Jacobs RE, Fraser SE. Proc. Natl. Acad. Sci. U.S.A. 1998;95:8443. 32. Wickline SA, Lanza GM. J. Cell Biochem. 2002;39:90. 33. Harrison PM, Arosio P. Biochim. Biophys. Acta. 1996;1275:161. 34. Aime S, Crich SG, Frullano L. Angew. Chem. Int. Ed. Engl. 2002;114:1059. 35. Aime S, Botta M, Fedeli F, Gianolio E, Terreno E, Anelli PL. Chem. Eur. J. 2001;7:5261. 36. Bogdanov A, Matuszewski L, Bremer C, Petrovsky A, Weissleder R. Mol. Imaging. 2002;1:16. 37. Morawski AM, Winter PM, Crowder KC, Caruthers SD, Fuhrhop RW, Scott MJ, Robertson JD, Abendschein DR, Lanza GM, Wickline SA. Magn. Reson. Med. 2004;51: 480. 38. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB. Chem. Rev. 1999;99:2293. 39. Nivorozhkin AL, Kolodziej AF, Caravan P, Greenfield MT, Lauffer RB, McMurry TJ. Angew. Chem. Int. Ed. Engl. 2001;40:2903. 40. De Leon-Rodriguez LM, Ortiz A, Weiner AL, Zhang S, Kovacs Z, Kodadek T, Sherry T. J. Am. Chem. Soc. 2002;124:3514. 41. Dodd SJ, Williams M, Suhan JP, Williams DS, Koretsky AP, Ho C. Biophys. J. 1999;76:103. 42. Sapiro EM, Skrtic S, Hill JM, Dumbar CE, Koretsky AP. Proc. Int. Soc. Magn. Reson. Med. 2003;11:229. 43. Hill JM, Dick AJ, Raman VK, Thompson RB, Yu ZX, Hinds KA, Pessanha BSS, Guttman MA, Varney TR, Martin BJ, Dunbar CE, McVeigh ER, Lederman RJ. Circulation. 2003;108:1009. 44. Wecker S, K¨ustermann E, Radermacher B, Hoehn M. Proc. Int. Soc. Magn. Reson. Med. 2002;10:878.

Molecular MR Imaging

73. Moore A, Josephson L, Bhorade RM, Basilion JP, Weissleder R. Radiology. 2001;221:244. 74. Weissleder R, Moore A, Mahmood U, Bhorade R, Benveniste H, Chiocca EA, Basilion JP. Nat. Med. 2000;6:351. 75. Bhorade R, Weissleder R, Nakakoshi T, Moore A, Tung CH. Bioconjug. Chem. 2000;11:301. 76. Lewin M, Carlesso N, Tung CH, Tang XW, Cory D, Scadden DT, Weissleder R. Nat. Biotechnol. 2000;18:410. 77. Dodd CH, Hsu HC, Chu WJ, Yang P, Zhang HG, Mountz JD Jr, Zinn K, Forder J, Josephson L, Weissleder R, Mountz JM, Mountz JD. J. Immunol. Methods. 2001;256:89. 78. Josephson L, Tung CH, Moore A, Weissleder R. Bioconjug. Chem. 1999;10:186. 79. Allen MJ, Meade TJ. J. Biol. Inorg. Chem. 2003;8:746. 80. Zhang RL, Zhang L, Zhang ZG, Morris D, Jiang Q, Wang L, Zhang LJ, Chopp M. Neuroscience. 2003;116:373. 81. Zhang ZG, Jiang Q, Zhang R, Zhang L, Wang L, Zhang L, Arniego P, Ho KL, Chopp M. Ann. Neurol. 2003; 53:259. 82. Rols MP, Delteil C, Golzio M, Dumond P, Cros S, Teissie J. Nat. Biotechnol. 1998;16:168. 83. Paturneau-Jouas M, Parzy E, Vidal G, Carlier PG, Wary C, Vilquin JT, de Kerviler E, Schwartz K, Leroy-Willig A. Radiology. 2003;228:768. 84. Stark DD, Weissleder R, Elizondo G, Hahn PF, Saini S, Todd LE, Wittenberg J, Ferrucci JT. Radiology. 1988; 168:297. 85. Weissleder R, Heautot JF, Schaffer BK, Nossiff N, Papisov MI, Bogdanov A Jr, Brady TJ. Radiology. 1994;191:225. 86. Seneterre E, Weissleder P, Jaramillo D, Reimer P, Lee AS, Brady TJ, Wittenberg J. Radiology. 1991;179:529. 87. Ruehm SG, Corot C, Vogt P, Kolb S, Debatin JF. Circulation. 2001;103:415. 88. Beckmann N, Cannet C, Fringeli-Tanner M, Baumann D, Pally C, Bruns C, Zerwes HG, Andriambeloson E, Bigaud M. Magn. Reson. Med. 2003;49:459. 89. Zhang Y, Dodd SJ, Hendrich KS, Williams M, Ho C. Kidney Int. 2000;58:1300. 90. Rausch M, Baumann D, Neubacher U, Rudin M. NMR Biomed. 2002;15:278. 91. Wunderbaldinger P, Josephson L, Weissleder R. Bioconjug. Chem. 2002;13:264. 92. Meade TJ, Taylor AK, Bull S. Curr. Opin. Neurobiol. 2003;13:597. 93. Lowe MP, Parker D, Reany O, Aime S, Botta M, Castellano G, Gianolio E, Pagliarin R. J. Am. Chem. Soc. 2001;123:7601. 94. Li WH, Parigi G, Fragai M, Luchinat C, Meade TJ. Inorg. Chem. 2002;41:4018. 95. Li WH, Fraser SE, Meade TJ. J. Am. Chem. Soc. 1999;121:1413. 96. Moats RA, Fraser SE, Meade TJ. Angew. Chem. Int. Ed. Engl. 1997;36:726. 97. Burai L, Scopelliti R, Toth E. Chem. Commun. 2002;20:2366. 98. Aime S, Botta M, Gianolio E, Terreno E. Angew. Chem. Int. Ed. Engl. 2002;39:747. 99. Aime S, Castelli DD, Terreno E. Angew. Chem. Int. Ed. Engl. 2002;41:4334. 100. Aime S, Botta M, Mainero V, Terreno E. Magn. Reson. Med. 2001;47:10.

Part II

45. Wright AC, Song HK, Wehrli FW. Magn. Reson. Med. 2000;43:163–169. 46. Vinitski S, Griffey R, Fuka M, Matwiyoff N, Prost R. Magn. Reson. Med. 1987;5:278. 47. Haacke EM. Magn. Reson. Med. 1987;4:407. 48. Shapiro EM, Koretsky AP. Proc. Int. Soc. Magn. Reson. Med. 2004;12:1734. 49. Wiedermann D, K¨ustermann E, Blunk J, Wecker S, B¨uhrlez C, Schwindt W, Trapp T, F¨ocking M, Hescheler J, Hoehn M. Proc. Int. Soc. Magn. Reson. Med. 2002;10:1257. 50. Yablonskiy DA, Haacke EM. Magn. Reson. Med. 1994;32:749. 51. Bulte JWM, Douglas T, Witwer B, Zhang SC, Strable E, Lewis BK, Zywicke H, Miller B, van Gelderen P, Moskowitz BM, Duncan ID, Frank JA. Nat. Biotechnol. 2001;19: 1141. 52. Shapiro EM, Skrtic S, Koretsky AP. Proc. Int. Soc. Magn. Reson. Med. 2004;12:166. 53. Safarik I, Safarikova M. J. Chromatogr. 1999;B722:33. 54. H¨ogemann D, Basilion JP. Eur. J. Mol. Med. 2002;29:400. 55. Sipkins DA, Cheresh DR, Kazemi MR, Nevin LM, Bednarski MD, Li KCP. Nat. Med. 1998;4:623. 56. Lemieux GA, Yarema KJ, Jacobs CL, Bertozzi CR. J. Am. Chem. Soc. 1999;121:4278. 57. Modo M, Mellodew K, Cash D, Fraser SE, Meade TJ, Price J, Williams SCR. Neuroimage. 2004;21:311. 58. Yeh TC, Zhang W, Ildstad ST, Ho C. Magn. Reson. Med. 1993;30:617. 59. Sipe JC, Filippi M, Martino G, Furlan R, Rocca MA, Rovaris M, Bergami A, Zyrozz J, Scotti G, Comi G. Magn. Reson. Imaging. 1999;17:1521. 60. Moore A, Weissleder R, Bogdanov A Jr. J. Magn. Reson. Imaging. 1997;7:1140. 61. Franklin RJM, Blaschuk KL, Mearchell MC, Prestoz LLC, Setzu A, Brindle KM, ffrench-Constant C. Neuroreport. 1999;10:3961. 62. Bulte JWM, Zhang SC, van Gelderen P, Heryneki V, Jordan EK, Duncan ID, Frank JA. Proc. Natl. Acad. Sic. U.S.A. 1999;96:15256. 63. Majumdar S, Zoghbi SS, Gore JC. Magn. Reson. Med. 1989;10:289. 64. Bulte JWM, Ma LD, Magin RL, Kamman RL, Hulstaert CE, Go KG, The TH, de Leij L. Magn. Reson. Med. 1993;29:32. 65. Bulte JWM, Laughlin PG, Jordan EK, Tran VA, Vymazal J, Frank JA. Acad. Radiol. 1996;3:S301. 66. Kalish H, Arbab AS, Miller BR, Lewis BK, Zywicke HA, Bulte JWM, Bryant LH, Frank JA. Magn. Reson. Med. 2003;50:275. 67. Frank JA, Miller BR, Arbab AS, Zywicke HA, Jordan EK, Lewis BK, Bryant LH Jr, Bulte JWM. Radiology. 2003;228:480. 68. Arbab AS, Yocum GT, Kalish H, Jordan EK, Anderson SA, Khakoo AY, Read EJ, Frank JA. Blood. 2004. 69. Arbab AS, Bashaw LA, Miller BR, Jordan EK, Lewis BK, Kalish H, Frank JA. Radiology. 2003;229:838. 70. Zhang ZY, Smith BD. Bioconjug. Chem. 2000;11:805. 71. Walter GA, Cahill KS, Huard J, Feng H, Douglas T, Sweeney HL, Bulte JWM. Magn. Reson. Med. 2004;51:273. 72. Lee IH, Bulte JWM, Schweinhardt P, Douglas T, Trifunovski A, Hofstetter C, Olson L, Spengera C. Exp. Neurol. 2004;187:509.

References 1097

1098 Part II

Medical Uses

Part II

101. Hoult DI, Lauterbur PC. J. Magn. Reson. 1979;34:425. 102. Styles P, Soffe NF, Scott CA, Cragg DA, Row F, White DJ, White PCJ. J. Magn. Reson. 1984;60:397. 103. Black RD, Early TA, Roemer PB, Mueller OM, Mogro-Campero A, Turner LG, Johnson GA. Science. 1993;259:793–795. 104. Miller JR, Hurlston SE, Ma QY, Face DW, Kountz DJ, MacFall JR, Hedlund LW, Johnson GA. Magn. Reson. Med. 1999;41:72. 105. Haase A, Frahm J, Matthaei D, Hnicke W, Merboldt KD. J. Magn. Reson. 1986;67:258. 106. Stehling MK, Turner R, Mansfield P. Science. 1991;254:43. 107. Ahrens ET, Narasimhan PT, Nakada T, Jacobs RE. Prog. Nucl. Magn. Reson. Spectrosc. 2002;40:275. 108. Seppenwoolde JH, Viergever MA, Bakker CJG. Magn. Reson. Med. 2003;50:784. 109. Coristine AJ, Foster P, Deoni SC, Heyn C, Rutt BK. Proc. Int. Soc. Magn. Reson. Med. 2004;12:163. 110. van den Bos EJ, Wagner A, Mahrholdt H, Thompson RB, Morimoto Y, Sutton BS, Judd RM, Taylor DA. Cell Transpl. 2003;12:743. 111. Emerit J, Beaumont C, Trivin F. Biomed. Pharmacother. 2001;55:333. 112. Weissleder R, Stark DD, Engelstad BL. Am. J. Roentenol. 1989;152:167.

113. Stroh A, Zimmer C, Gutzeit C, Jakstadt M, Marschinke F, Jung T, Pilgrimm H, Grune T. Free Radiat. Biol. Med. 2004;36:976. 114. Pouliquen D, le Jeune JJ, Perdrisot R, Ermias A, Jallet P. Magn. Reson. Imaging. 1991;9:275. 115. Yeh TC, Zhang W, Ildstad ST, Ho C. Magn. Reson. Med. 1995;33:200. 116. Arbab AS, Yocum GT, Wilson LB, Parwana A, Jordan EK, Kalish H, Frank JA. Mol. Imaging. 2004;324. 117. Kostura L, Mackay D, Pittenger MF, Kraitchman DL, Bulte JWM. Proc. Int. Soc. Magn. Reson. Med. 2004;12:167. 118. Pratten MK, Lloyd JB. Biochim. Biophys. Acta. 1986;881:307. 119. Schroeder M, Saleh A, Wiedermann D, Hoehn M, Jander S. Magn. Reson. Med. 2005;52:403. 120. Pautler RG, Silva AC, Koretsky AP. Magn. Reson. Med. 1998;40:740. 121. Saleem KS, Pauls JM, Augath M, Trinath T, Prause BA, Hashikawa T, Logothetis NK. Neuron. 2002:34: 685. 122. van der Linden A, Verhoye M, van Meir V, Tindemans I, Eens M, Absil P, Balthazart J. Neuroscience. 2002; 112:467. 123. Cohen B, Dafni H, Meir G, Neeman M. Proc. Int. Soc. Magn. Reson. Med. 2004;11:1707.

1099

Stefan Bluml Childrens Hospital Los Angeles, Department of Radiology, USC Keck School of Medicine Los Angeles, CA

Abbreviations: MRS = Magnetic Resonance Spectroscopy; MRI = Magnetic Resonance Imaging; SNR = Signal-to-noise ratio; RF = Radiofrequency; FDA = federal drug administration; SAR = specific absorption rate; B1 = magnetic field of RF coil; NOE = nuclear Overhauser effect; VOI = volume of interest; LCModel = linear combination of model spectra; TCA = tricarboxylic acid; Glc = glucose; Glu = glutamate; Gln = glutamine; Asp = aspartate; NAA = N-acetylaspartate; GABA = γ-amino butyric acid; Lac = lactate; Ala = alanine; HCO− 3 = bicarbonate, Ac = acetate; KD = ketogenic diet.

Introduction The application of 13 C Magnetic Resonance Spectroscopy (MRS) for basic research/medical applications is challenging and, despite being available for more than 25 years, only a few groups have attempted 13 C MRS in vivo. However, the few studies undertaken illustrate the great promise of in vivo 13 C MRS. Interest in 13 C MRS has grown considerably in recent years. The potential of 13 C arises from its biggest handicap: Low natural abundance [and the compromised sensitivity (≈1/50 of 1 H)] renders 13 C spectroscopy in vivo very difficulty due to the inherently very low signal-to-noise ratio (SNR). This low natural abundance, on the other hand, is the key to new, exciting applications of in vivo MRS: Investigation of metabolic pathways and the measurement of flux rates in vivo in animal and humans after 13 C enriched substrate infusion. The interested reader is referred to a recent issue of NMR in Biomedicine [1] exclusively dedicated to the application of 13 C MRS to study biological systems for a more detailed update on progress in 13 C MRS than given below. Instead, in this chapter, basic experimental procedures of in vivo 13 C MRS, with emphasis on experimental work in humans, are illustrated and discussed for the researcher interested in conducting similar experimental work in the future.

Graham A. Webb (ed.), Modern Magnetic Resonance, 1099–1112.
C 2008 Springer.

Methods Equipment A list of additional equipment needed beyond that required for MR imaging or 1 H MRS is provided in Table 1. Radiofrequency (RF) Coils Even for “niche” manufacturers, 13 C RF coils are too “exotic” and most groups conducting in vivo 13 C research have built their own coils. Volume coils are more flexible and have a special advantage when quantitation of cerebral metabolites is the goal. Nevertheless, surface coils have exclusively been used for 13 C MRS since they furnish the much-needed extra signal required for in vivo 13 C MRS. Because all 13 C methods either apply direct detection with proton-decoupling, polarization transfer, or heteronuclear editing, dual-tuned coils (1 H and 13 C frequencies) are required. Two concentric circular surface coils, a small 13 C coil and a larger 1 H coil, have been used by several investigators. However, the need to block the 1 H flux of the inner 13 C coil resulted in poor performance of the 1 H coil in the sensitive volume of the 13 C coil. A much improved assembly of two semi-orthogonal surface 1 H coils, operated in quadrature mode, and a single linear 13 C coil (Figure 1) was first introduced by Adriany and Gruetter [2] and has since been adopted by other groups [37]. The quadrature 1 H coil assembly offers twofold improved efficiency over linear-polarized coils in the power requirement. This is important to avoid excessive power RF absorption by tissue during proton-decoupling (see below). Broadband Excitation/Detection Even if dual-tuned RF coils are available, not all magnetic resonance imaging (MRI) scanners can progress beyond 1 H excitation and detection. Broadband amplifiers covering the frequency of 13 C are essential. Similarly, RF receivers tuned to the 13 C resonance frequency must be provided along with several other hardware features.

Part II

In vivo 13C MRS

1100 Part II

Medical Uses

Part II

Table 1: Equipment needed for 13 C MRS beyond that required for standard MR imaging Hardware

Purpose

Specifications/comments

RF Coil

13 C

and 1 H excitation, receive, editing, decoupling

Dual-tuned 13 C-1H coil, 13 C surface coil/surface coil assembly, quadrature 1 H coil for efficient decoupling

Broadband receiver

To digitalize, filter, and store RF signal at 13 C frequency

Low noise figure

Transmit/receive switch for 13 C with pre-amp

Protects 13 C receiver from high power RF during excitation, boosts the gain of the 13 C signal during receive

Low noise figure (50 W of CW RF power. The output of the amplifier should be high-pass filtered to suppress 13 C frequencies.

RF source and modulation capabilities

Generation of a decoupling sequence at the 1 H frequency

Typical operating requirements: e.g. WALTZ-4 sequence with a 90 degree pulse range from 0–9.9 ms (in steps of 0.1 ms), bilevel (NOE/decoupling) operation, and manual control of the WALTZ-4 pulse width, attenuation control of output RF power over a 50 dB range in 1 dB steps.

Power meter

Power monitoring

Able to measure forward/reflected RF power. Values (in Watts or dB) should be displayed in real-time, e.g. on a computer screen. In normal operations, hard power limits should force the automatic shutdown of the amplifier if the output exceeds the limits. In addition, it should be possible to establish soft limits that generate warning signals or messages.

System/sequence performance evaluation

A spherical phantom containing one or more 13 C compounds suitable to test system performance/demonstrate decoupling.

Broadband capabilities

2nd RF channel or “Stand-alone” decoupler 1H

exciter, 1 H amplifierb

Miscellaneous Phantom a The

manufacturers all have specifications for exciter output levels, linearity, phase stability, etc.— the specs for broadband are always a little lower than those for the narrow band imaging exciter. b With amplifiers capable of high CW output, it is advisable that some safety interlock is available to insure that a power limit, e.g. 100 W, cannot be exceeded in normal operation. Typical amplifier specifications: 1000 W (peak)/100 W (CW) into a 50 Ohm load. The RF power amplifier should be electrically connected to the MR scanner so that all automatic, emergency electrical power shutdown procedures apply.

Proton-Decoupler and Power Monitoring Along with the capability to perform heteronuclear (other than1 H) MRS comes the need to enhance sensitivity and specificity of chemical analyses by proton-decoupling and heteronuclear polarization transfer and editing. This involves excitation at different frequencies, and accordingly

requires a second RF channel and amplifier. Since it appears that there is a renewed interest in advanced multinuclear spectroscopy, the major manufacturers of clinical MR scanners are all offering broadband capabilities and second RF channels. There are two basic options for decoupling; a fully integrated, second RF exciter channel

In vivo 13 C MRS

Methods 1101

B

1 1

H Coil

H Coil

13

C Coil

Fig. 1. Dual-tuned RF coil for in vivo 13 C MRS of human brain. (A) Schematic of the “half-volume” coil assembly as originally proposed by Adriany and Gruetter [2] and since then adopted by other groups. The two orthogonal 1 H coils can be used for MRI, 1 H MRS, as well as for proton-decoupling during 13 C signal reception. A single 13 C surface coil is mounted immediately adjacent to the head rest. The overlap of the 1 H coils and the position of the individual coils needed to be adjusted to minimize interaction between the different coils. It is also important to minimize the distance between 13 C coil and the region of interest to optimize SNR. (B) Picture of the coil assembly used at Huntington Medical Research Institutes, Pasadena, California. A hybrid coupler is necessary to operate the two 1 H coils 90◦ out of phase for quadrature excitation and detection. “Tune and match” is accomplished for each coil individually with three separate tuning boxes. Coils, cables, and connectors are “color-coded” to avoid mistakes when connecting the coil assembly with the MR scanner.

that is part of the MR scanner, or an independent or “Stand-alone” sequence generator with simple timing signals provided by the MR scanner. The fully integrated channel requires pulse sequence programming but gives the user direct access to the decoupling sequence and parameters. Standalone decouplers exist in several flavours, ranging from simple pre-programmed decoupling sequences with hardware control of all parameters to more complex signal generating devices under computer control. Excessive RF power deposition during decoupling can result in potentially damaging heating of tissue. Federal drug administration (FDA) guidelines limit the deposited energy below levels which are considered to be of any risk [8,9]. Manufacturers have developed decouplers with built-in power monitoring and automatic shutdown should a certain threshold for the average power be exceeded. It should be noted that since for decoupled 13 C MRS 1 H surface coils are being used, the local specific absorption rate (SAR) may vary with the magnetic field profile of the proton coils (B1 ). It is possible that the average SAR over the whole sensitive volume is below the FDA approved limit, whereas the local SAR of tissue adjacent to the 1 H decoupling coils may be above that threshold [2,10]. Proton-Decoupling and NOE Proton-decoupling and the nuclear Overhauser effect (NOE) are essential tools for 13 C MRS. It is not advised

to attempt 13 C MRS without decoupling/NOE. Even the simplest application, such as measuring the lipid profile of a human leg [11], would be quite demanding without the improved SNR and resolution facilitated by decoupling and NOE. Proton-Decoupling Heteronuclear scalar couplings between 1 H and 13 C nuclei result in a splitting of 13 C resonance lines in to multiplets. The goal of proton-decoupling is to suppress these couplings and to collapse multiplets into singlets, simplifying the spectrum and improving the signal intensity. This is achieved by irradiation of decoupling proton RF pulses during the readout of 13 C signal over the whole range of the spectrum (= broadband decoupling). The most commonly used proton-decoupling RF sequence for in vivo 13 C MRS is WALTZ [12–15]. Several WALTZ decoupling schemes have been developed named WALTZ-4, WALTZ-8, and WALTZ-16 and they all appear to work well for in vivo applications. NOE—The NOE is based on changes in the polarization, which occur in coupled systems when the population of one of the systems is manipulated. In contrast to scalar coupling, which is facilitated indirectly by the electron bond, it is the dipolar coupling between 13 C and 1 H nuclei, which generates the NOE. In fluids, dipolar coupling does not result in a line splitting due to “Motional Narrowing,”

Part II

Headrest A

1102 Part II

Medical Uses

Part II

however, it is still an effective relaxation mechanism resulting in changes of signal intensities. While it was necessary to irradiate at the proton frequency during the readout of 13 C signal for proton decoupling, NOE enhancement is generated by irradiation at the proton frequency before the 13 C signal is detected. Excitation of protons coupled to 13 C has an impact on the polarization of 13 C. In particular, if the protons are saturated an increase of the polarization of 13 C can be observed. Since the maximum NOE enhancement for two nuclei A and B in fluids is given by, NOE = 1 + 1/2 γ B /γ A , a threefold signal enhancement can be achieved for 13 C MRS (γ 1H ≈ 4γ 13C ). To have maximum NOE enhancement for all resonances it is necessary to irradiate RF over the full chemical shift range of dipolar coupled protons. As for proton decoupling, a WALTZ-n cycle can be employed, however, it is usually sufficient to irradiate at a much lower power levels (≈1/10 of decoupling power). The optimum choice for the center frequency and the bandwidth for NOE may be different from what works best for decoupling.

evolution period, to detect the signal at the 13 C frequency. Because the heteronuclear J-coupling constants between 1 H and 13 C nuclei are very similar (≈120 Hz for methyl groups, ≈160 Hz for aromatic groups), sequence parameters, in particular the delay for coupling evolution, can be adjusted to the optimum detection of all signals. There are two distinct features of this method. (i) Proton excitation can be combined with localization such as imageselected in vivo spectroscopy (ISIS) [21] greatly reducing the chemical shift displacement error [4,22]. (ii) 13 C detection retains the large dispersion and thus spectral resolution of 13 C (Figure 3). This method also allows recovering the maximum fourfold sensitivity gain due to the higher magnetization of protons. During detection at the 13 C frequency, proton decoupling is applied to generate a simplified spectrum for better interpretation. Polarization transfer becomes less advantageous in situations where the resonance of interest has a very short T2 (e.g. glycogen). If T2 is short relative to the time delay necessary to facilitate polarization transfer, a substantial loss of signal can occur [17].

Pulse Sequences for in vivo 13 C MRS More detailed discussions of pulse sequences and data acquisition strategies than provided below can be found elsewhere [16,17] and in references therein. 13

C Excitation and 13 C Detection

Direct detection 13 C MRS is the least sensitive method and spectra with sufficient quality can only be acquired from very large volumes [18] or without any localization [3]. Gradient localized direct 13 C MRS is further compromised by considerable chemical shift displacement errors due to the large chemical shift dispersion of 13 C spectra. On the other hand, spectra acquired with this method offer high spectral resolution and an impressive number of 13 C resonances, all measured simultaneously under identical conditions (Figure 2). In particular when a simple “pulse and acquire” sequence is used, the impact of acquisition parameters on the spectral pattern is small and sequence parameter adjustment is quick. These features may be important in a clinical setting where experimental simplicity is important [19,20]. Direct 13 C excitation and detection requires proton-decoupling and NOE to improve the spectral quality.

Polarization Transfer A better SNR can be achieved with heteronuclear polarization transfer. The basic concept of polarization transfer is to excite protons coupled to 13 C and, as the polarization is transferred to the adjacent 13 C nucleus during the

Indirect Detection Indirect detection can improve greatly the sensitivity by detecting protons attached to 13 C utilizing heteronuclear scalar couplings. In the simplest case, protons coupled to a 13 C nucleus will form two resonances symmetrically placed around the center of the singlet generated by a proton attached to a 12 C nucleus. In practice however, the 1 H spectra of chemicals such as glutamate and glutamine are complex because of home- and heteronuclear couplings and resonances overlap due to the small chemical shift dispersion of 1 H MRS. This renders the detection of 13 C coupled protons difficult. It is therefore necessary to apply editing pulses at the 13 C frequency, which will result in a modulation of a spectrum, e.g. an inversion of the signal from 13 C bound protons relative to 12 C bound protons. When unedited and edited spectra are analyzed, the difference spectra will represent protons attached to 13 C while the added spectrum represents protons bound to 12 C. This approach is referred to as proton-observed, carbon-edited (POCE) [23]. Indirect detection offers the highest SNR among the methods discussed here and has the advantage of localization at the proton frequency, reducing chemical shift displacement artifacts. Another advantage of indirect detection is that only hardware for the transmission of RF at the 13 C frequency is required and that the proton channel of MR systems, used for signal detection, is usually well engineered, guaranteeing good performance (Figure 4). A disadvantage is that the high spectral resolution of 13 C is sacrificed. While this is acceptable for simple spectra with selective isotopic enrichment, it may

In vivo 13 C MRS

Glc1 Glc1

Glu2

Ketogenic Diet

Glu4 Gln 4 Gln2 Asp2

Glu3

F

E

Asp3

Lac3

65-90 min .

HCO3

Part II

D

Glu5 Gln5

Checking System Performance 1103

Ala 3

30-50 min .

Glu1 Gln1 10-20 min .

0-10 min . 180

170

NAA1 Gln1

160

100

150

CrC=N

40

A

30

6

B

mI

PCr C=N Cr/Cho

HCO 3

180

50

Glycerol

C

Glu1 NAA 4,C=O Gln5 Glu5

90

170

160

Gln 2 Glu 2

Glc 3,5

150

?

NAA 2

9

NAA3 Cr3 Tau 2

Tau 3

3 C=O

1

70

2

150

100

60

50

80

4

8 7 10

12

40

60

40

20

Fig. 2. In vivo natural abundance and 13 C MR spectra after substrate infusion of the human brain acquired with “pulse-and-acquire” and direct 13 C detection. (A–C) Shown is a natural abundance 13 C MR spectrum of a patient with Canavan disease. B + C are expansions of A to allow a more detailed inspection. Lipid signal originating from the skull and subcutaneous fat dominates the spectrum due to non-localized acquisition. (D) 13 C spectrum (upper trace) obtained from ketogenic diet (KD) patients 30–80 min after infusion [1-13 C] acetate. The lower trace is the difference spectrum computed by subtraction of the baseline spectrum, acquired before infusion start, (not shown) to directly illustrates 13 C label accumulation. 13 C label incorporation into bicarbonate (HCO3 ), Glu C5 and Gln C5 and, Glu C1 and Gln C1 can be observed. (E) The time course of glucose uptake and metabolism to glutamate and other products in the brain in a normal adult subject after i.v. [1-13 C] glucose infusion. Difference spectra were calculated to subtract prominent lipid peaks. 13 C enriched peaks of Glu C1,2,3,4 , Gln C1,2,3,4 , Asp C2,3 , NAA C2,3 can be observed [36]. (F) Although lactate (Lac) and alanine (Ala) concentrations are too low to be detected with natural abundance 13 C MRS in normal brain, 13 C enriched resonances are readily observed after [1-13 C] glucose infusion. All spectra were acquired with the dual-tuned half-volume head coil shown in Figure 1 on a GE, 1.5 T clinical scanner.

be more problematic when high resolution of complex spectra with many components is the goal.

Checking System Performance Once the required hardware is installed and an MR sequence developed, the next essential step is the thorough evaluation of the system performance.

Low SNR Low overall SNR of 13 C MRS may be caused by several reasons. A poorly designed or tuned 13 C coil will not provide adequate SNR. Also, poor SNR may be caused by site specific factors. RF noise at 13 C frequency may be generated by other MR equipment used for MRI and not filtered appropriately. Also, equipment not related with MR may emit RF at the frequency of 13 C and adversely

1104 Part II

Medical Uses

Part II Fig. 3. Localized 13 C MR spectrum of the human occipital lobe The spectrum was acquired from 72-ml volume 60–110 min after start of [1-13 C] glucose infusion on a 4 T system using polarization transfer, proton excited carbon detected, for improved SNR and localization (Figure provided by Rolf Gruetter, Ph.D. and reproduced with permission from Dev Neurosci 1998 S. Karger AG).

interfere with the quality of a spectrum. There are many tests that can be performed to isolate the cause for insufficient SNR, however, the questions the new investigator faces first are: “How does a good spectrum look like?” “Is the performance of my system the best possible?” For tCr

NAA

tCr tCho

αH1-GlC 12C 13C 13C

×4

Glx-H2 Gln-H4 tCr GABA-H4

Glu-H4

Glx-H3 GABA-H3 Lac-H3

×8

PPM 5.0

4.0

3.0

2.0

1.0

0.0

Fig. 4. Indirect detection of 13 C label accumulation The spectrum was acquired from rat brain at 9.4 T (180 ul volume, TR = 4000 ms, TE = 8.5 ms, 512 averages). It was acquired between 110 and 140 min following the intravenous infusion of [U-13 C6]glucose (i.e. uniformily labelled glucose). The upper trace is the regular 1 H spectrum, while the bottom spectrum is the 1 H-[13 C] edited fraction (i.e. it only shows resonances from protons attached to 13 C). The sequence is described in de Graaf et al. [37] (figure provided by Robin de Graaf, Ph.D. and reproduced with permission from Wiley-Liss. Inc., 2004).

that reason, it is recommended that the investigator should team-up with a reference site with experience with 13 C MRS. The coil, a phantom, and specific instruction on how to perform a test should be sent to a collaborating site with the request to acquire data for comparison. Even better, a new investigator may want to visit an established site for 13 C MRS to witness a quality scan and to ensure that scans at the reference site and at the local site are performed under identical conditions. Once a problem has been identified or narrowed down, the manufacturer of the MR system or the RF coil can be approached with a specific request to fix a problem. Crosstalk Crosstalk between 1 H and 13 C coils during decoupling is caused by the transmission of RF at the proton frequency with relatively high power while the small 13 C signal is received. It can cause two unwanted effects. (i) Although, the 1 H signal is suppressed by subsequent filtering, the 13 C pre-amplifier usually placed immediately after the 13 C coil, sees both 13 C and 1 H signal. Both signals may be amplified. Since the 1 H signal is much larger than the 13 C signal it may saturate the pre-amplifier. This results in an overall scaling (down) of the decoupled 13 C spectrum. It may not affect the SNR but is nevertheless unwanted since an absolute comparison of scans becomes very difficult. (ii) Unacceptable artifacts are caused when decoupling causes spikes at the 13 C frequency. Detecting of these two problems is relatively straightforward. If the maximum amplitude of decoupled spectra becomes smaller after further increasing decoupling power, saturation may have occurred. Spikes are readily detectable in the spectrum because of the dramatic adverse impact on the SNR. The problems caused by crosstalk of coils can be minimized by additional filtering of the 13 C signal and by using coil assemblies with efficient built-in proton traps. Reducing the decoupling power is useful only when it does not result in incompletely decoupled spectra. Incomplete Decoupling Insufficient decoupling power results in a “hybrid” spectrum, partially non-decoupled and partially decoupled. These spectra are not quantifiable and should not be interpreted. Instabilities Scanner and coil instabilities are often another concern. Electronic components may heat-up during a pro-longed acquisitions and change the tuning of a coil, an amplifier may not deliver the same power over an extend period of time, etc. Usually these variations are small, however, in particular when difference spectroscopy is attempted or the time course of a signal is observed over a long time, the quality of the results and the accuracy of the interpretation

In vivo 13 C MRS

Reproducibility The nature of a study defines when insufficient reproducibility (= different results from the same subject on different occasions) is a problem. Hardware imperfections can cause insufficient reproducibility. However, often the positioning of a subject, selection of region of interest, adjustment of acquisition parameters, shimming, etc. is not consistent thus limiting the reproducibility. These problems can be minimized/avoided by generating macros (if possible), dedicated protocols, and clear written instruction on how to conduct a scan. Macros, protocols, and instructions should be tested by scanning different phantoms. Sequence parameters should not be “guessed” when starting an acquisition. Ideally, good and reproducible spectra should be obtained without the need to “manually” adjust scan parameters in each cause. If the lack of reproducibility is small compared with the postulated differences between two groups of subjects, it only complicates the study by the need to examine more individuals in order to reach significance. It is always a good strategy to conduct several studies in the same subject to investigate this matter. There are of course limitations to that when a study is very complex such as a lengthy examination with substrate infusion.

Data Processing High quality spectra are of no use if the information they contain is not or incorrectly extracted. Therefore appropriate data processing is as important as the acquisition of the spectra. In general, data processing should be as user independent as possible to prevent unavoidable operator bias. Since the signal generated by a specific nucleus is proportional to the number of equivalent nuclei in the volume of interest (VOI), in vivo concentrations of a metabolite M (in mmol/kg tissue or mmol/l) can be determined by [M] = CMRS AM , where CMRS comprises all correction factors (T1 saturation, T2 relaxation, VOI, number of equivalent nuclei, receiver gains, etc.) and AM is the total area under a curve. Resonances in decoupled 13 C spectra are usually well resolved and line fitting of individual resonances is relatively straightforward. However, when quantifying spectra of low SNR (13 C spectra are compromised by low SNR), the investigator may want to keep in mind that not all parameters of a curve are measured with the same accuracy. In particular the determination of the position of a resonance on the chemical shift axis and the amplitude can be measured with a much higher accuracy than the width (or area). Therefore, most of the time, it is advantageous to measure peak amplitudes instead of

areas. On the other hand, the amplitude is affected by the homogeneity whereas the area is not. Nevertheless, in a well-designed study, spectra are acquired from comparable regions of interest and the variation of the homogeneity across subjects should be very small and there should be no correction necessary. In cases where small systematic variations could be expected (e.g. when a comparison of spectra from different brain region is attempted), it is still better to perform an adjustable lineshape correction (see below) and to measure amplitudes rather than areas. If due to the nature of a study, large systematic differences of the field homogeneity is expected then areas rather than amplitudes should be evaluated.

Lineshape Transformation For spectra of low SNR and partially overlapping peaks it is generally advisable to perform a lineshape transformation. A common lineshape transformation is the Lorentzian-to-Gaussian transformation. The first step is a negative Lorentzian linebroadening to create a spectrum with a standardized linewidth. E.g. a spectrum with resonances with 4 Hz linewidth will be corrected by −3 Hz whereas a spectrum with resonances with 4.5 Hz linewidth will be corrected by −3.5 Hz to result in standardized spectra with a 1 Hz linewidth. The second step is a positive Gaussian linebroadening to improve SNR. Although this lineshape transformation is not linear (= the relative areas of peaks with different linewidths are not the same after transformation), in practice, this imperfection is more than compensated for by the fact that Gaussian lines can be fitted much easier than Lorentzian lines because of the reduced overlap of adjacent resonances.

LCModel for 1 H MRS 13

C spectra are usually well resolved and fitting of the individual resonances can be achieved relatively straightforward. Should the user decide to use inverse detection, proton spectra need to be processed. Proton spectra suffer from the narrow chemical shift range, complex pattern due to homonuclear couplings, and a broad baseline due to macromolecules. On the other hand, good processing software for 1 H MRS is commercially available. LCModel (Linear combination of model spectra, Stephen Provencher Inc., Oakville, Ontario, Canada) is a welldeveloped, robust, and at the same time flexible software package [24]. LCModel fits the in vivo data by finding the best linear combination of spectra obtained from phantoms with known concentrations. LCModel is most suited for standard 1 H MRS as carried out in a clinical environment (e.g. 1.5 T, 2.0 T, 3.0 T, PRESS, STEAM) with pre-measured basis spectra available from the vendor. For

Part II

may be compromised. Scanner and coil instabilities can readily be checked in vitro by scanning phantoms.

Data Processing 1105

1106 Part II

Medical Uses

Part II

applications with complex acquisition sequences and unusual sequence parameters the user may need to prepare his/her own phantoms to acquire the appropriate model spectra the specific experimental conditions. Although this can be tedious when many metabolites are being analyzed, it is time well spent.

Modeling and Determination of Flux Rates Infusion of a 13 C labeled substrate allows the metabolism of the substrate to be followed by sequential 13 C MRS. A mathematical model, basically a set of differential equations, of the metabolism is then needed to determine the in vivo flux rates (Figure 5). It needs to be pointed out that the formulation of the differential equations for a bi-

ological model is the easy part—finding the right model is the tricky part. Assumptions and simplifications are necessary, even if the correct model is known [e.g. for the tricarboxylic acid (TCA) cycle] because only a limited number of metabolites can be measured with sufficient SNR. E.g. the measurement of the TCA-cycle rate after [1-13 C] labeled glucose infusion is based to a large extent on the time course of the glutamate C4 enrichment. However, glutamate is NOT a metabolite of the TCA-cycle. Only because the pool sizes of TCA-cycle intermediates are small and the exchange rate between α-ketoglutarate and glutamate is fast (there are some controversies in the literature on how fast “fast” is), it is reasonable to assume that glutamate C4 acts as a trap for 13 C label entering

Fig. 5. (A) Dynamic 13 C MRS involves the sequential acquisition of 13 C spectra after substrate infusion to 13 C label accumulation. (B) The next step is the processing and quantitation of spectra to generate time courses of 13 C concentration (or 13 C enrichment). (C + D) A set of differential equations is derived from the metabolic model applied for an experiment. (E) By iteratively varying flux parameters (and pool concentrations) of the mathematical model to optimize the fit with experimental data, flux rates (and pool concentrations) can be determined.

In vivo 13 C MRS

A simple two Compartment Model—Mathematical Description V1

V2

S0 −→ S1 −→ S2 In this model S0 is the initial substrate and S1 , S2 are intermediates of S0 metabolism. [13 S0 ], [13 S1 ], [13 S2 ] are the 13 C enriched pools; e.g. before infusion [13 S1 ] = 1.1% [S1 ]. V1 and V2 are the fluxes from S0 to S1 and from S1 to S2 . Assuming that the total concentration of S1 remains constant (V1 = V2 ), the 13 C flux into and out of pool S1 is described by: d 13 [13 S0 ] [13 S1 ] [ S1 ] = V1 − V2 . dt [S0 ] [S1 ]

(1)

In the special case that the fractional enrichment of substrate S0 is constant ([13 S0 ]/[S0 ] = E 0 = constant), the 13 C enrichment of S1 will exponentially approach that of the precursor:
  [13 S1 ](t) = E 0 [S1 ] × 1 − exp V1 t [S1 ] (2) Since the time constant for the pool to turn over is the ratio of pool size and flux rate ([S1 ]/V1 ), large metabolic pools take longer to turn over whereas small pools will quickly have the same enrichment as the precursor. For this hypothetical model, the concentration of 13 C enriched S1 , would be measured by sequential MRS. Once the total pool size is known (e.g. from literature, quantitative natural abundance 13 C, or 1 H MRS), the flux rate V1 can be obtained by fitting an exponential function to the measured time course for [13 S1 ]. Real biological systems, although analytical differential equations for each metabolic pool can be written readily, are too complex for analytical solutions and numerical solutions need to be found by a step-by-step integration. Software (CWave, Graeme Mason, Ph.D., Yale University) for the design of models and analysis of 13 C-labeling

studies is available after signing a license agreement [28]. To obtain a copy contact Graeme Mason, Ph.D., at the Magnetic Resonance Center, Yale University, School of Medicine.

Miscellaneous Even when all technical obstacles have been resolved, the investigator still needs to plan any study very carefully. In vivo 13 C studies, in particular of human requiring large amounts of 13 C labeled substrates, are expensive, lengthy, and the subsequent data analysis requires a considerable effort. To avoid frustration and jeopardizing the success of a 13 C experiment, all “peripheral” steps of a study need to be planned with equal due diligence than the acquisition of the 13 C spectra. Depending on the biological question asked, it needs to be decided what substrate should be infused for how long and in what fashion. For some applications oral administration may be appropriate [29,30,26] which would simplify the procedure considerably because one inter-venous (i.v.) infusion line could be eliminated. In particular for studies of humans, i.v. administered substrates should be prepared with care, kept refrigerated, and used within a couple of days. It is usually necessary to determine fractional 13 C enrichment of the substrate in plasma. Therefore the drawing, storage, and analysis of blood samples need to be planned. A new investigator is advised to carefully read the “Methods” sections of previous publications [22,25,28] and references therein.

Applications Although there are undoubtedly applications for in vivo natural abundance 13 C MRS, to harvest the full potential of 13 C MRS, enrichment of metabolites with 13 C via intravenous or oral administration of 13 C labeled substrates is necessary. Glucose is the principal substrate for energy metabolism for both neurons and glia cells in the brain and also facilitates the de novo synthesis of many neurochemicals. In normal human adults, i.v. infused [1-13 C] labeled glucose (Glc) passes the blood-brain barrier and is readily metabolized. 13 C enrichment of individual carbon atoms of glutamate (Glu), glutamine (Gln), aspartate (Asp), Nacetylaspartate (NAA), γ-amino butyric acid (GABA), lactate (Lac), alanine (Ala), and bicarbonate (HCO− 3 ) follows [5,19,20,22,25,31]. From repeated 13 C MR spectra acquired during studies over 2–3 h, the in vivo rates of several of the principal bioenergetic pathways of normal adult brain have been determined [22,25,4]. The role of 13 C glucose MRS in diseased brain remains a matter of speculation as there are only a few studies

Part II

the TCA-cycle at acetyl-Coa [25]. It is beyond the scope of this chapter to get involved in the discussion of current complex models of glucose and acetate metabolism in animal and human brain and it is recommended that the interested reader studies recently published articles [1]. Sometimes it is possible to simplify the mathematical model considerably to answer very specific questions. Examples are the determination of the N-acetyl-asparate synthesis rate in humans in vivo, the glutamine synthesis rate in human hepatic encephalopathy, or the measurement of the astro-glial TCA-cycle rate after [1-13 C] acetate infusion [20,26,27].

Applications 1107

1108 Part II

Medical Uses

Part II Fig. 6. 13 C MRS of human hepatic encephalopathy (HE) Hepatic encephalopathy is a metabolic brain disorder caused by liver dysfunction and incomplete removal of ammonia and other neurotoxins from the blood. (A+B) Spectra acquired over a period of 60–120 min after a 20 min infusion of [1-13 C] glucose of an adult control (A) and a patient with severe chronic HE (grade III–IV) are shown. 13 C enrichment of Glu C2 , Asp C2 , Asp C3 is reduced while 13 C accumulation in Gln C2 is comparable to the control. Note the absence of natural abundance myo-inositol (mI) signal in HE. (C+D) Difference spectra, to enable inspection of 13 C accumulation in Glu C4 , Gln C4 , Glu C4 , and Gln C4 , of a control and a chronic HE patient (grade I-II) acquired 80–100 min after infusion start, are shown. Spectra were scaled to represent absolute concentrations. Note the apparent reduction of 13 C incorporation into Glu C4 , C2 , Asp C2 ,3 in HE. Even though affected by the subtraction analysis to a larger extent in HE than in control, the Gln C4 resonance appears strikingly prominent in spectra of the patient. Label accumulation in NAA C2 ,3 observed in the control was not detected in HE. The pattern of label accumulation in HE is consistent with an overall reduction of glucose oxidation and altered glutamate/glutamine cycling.

conducted. Striking abnormalities in glucose metabolism and glutamate and glutamine label accumulation were observed in patients with chronic hepatic encephalopathy (Figure 6). Glucose oxidation was also reduced in a juvenile with hypoxic injury and in a premature infant. Abnormalities were also detected in pediatric patients with leukodystrophies and in children with mitochondrial disorders [19,20]. The information obtained from following the fate of 13 C labeled glucose goes beyond that of providing a rate for energy production. A tight coupling between cerebral glucose metabolism and glutamate neurotransmitter flux in humans has been proposed by Magistretti et al. [32]. Aspartate, (a neurotransmitter?), can be studied in vivo in humans by its 13 C label accumulation. The role of NAA in mammalian brain, a neuronal/axonal marker which is central for its diagnostic power in 1 H MRS, is incompletely understood. NAA synthesis can be measured

with 13 C MRS after glucose infusion in a clinical setting [26] (Figure 7). The application of 13 C MRS after substrate infusion is by no means limited to glucose. Glucose is convenient because of its rapid oxidation and the fast appearance of 13 C label in its metabolites and its non-toxicity even at extremely high concentrations. However, the use of other substrates may further enhance the potential of 13 C MRS as a research and diagnostic tool in human brain disease. The candidate next best to glucose appears to be acetate [7,27]. Acetate (Ac) is metabolized to acetyl-CoA only in the glial compartment [33]. Using the same MR technique as for [1−13 C] glucose, glial acetate metabolism can be investigated in normal and diseased brain. 13 C label accumulation in HCO3 , in Glu C5 and Gln C5 (first turn of the TCA-cycle), and in Glu C1 and Gln C1 (second turn) can be observed after [1−13 C] Ac in fusion. The rate of Ac

Fig. 7. Determination of N-acetyl-aspartate synthesis rate, VNAA , in humans in vivo (A) 13 C labeling scheme used to predict the transfer of 13 C label from Glu4 to Asp2,3 . Only relevant TCA-cycle intermediates are shown (B). N-acetyl-aspartate (NAA) synthesis by condensation of aspartate and Acetyl CoA. (C) Human brain natural abundance (baseline) and 13 C enriched spectra after [1-13 C] glucose infusion. The time intervals indicate the time of acquisition relative to the commencement of [1-13 C] glucose infusion. Peak intensities of Asp2 and NAA2 (as well as Glu C2 and Gln C2 ) increased progressively as 13 C label accumulates. (D) Fractional 13 C enrichment of NAA (E NAA2 ) versus time was lower in patients with aspartoacylase deficiency (Canavan Disease) than in control subjects. Linear regression lines are superimposed on data points t > 50 min in each group. The calculated mean rate of in vivo NAA synthesis in normal subjects was 9.2 ± 3.9 nmol/min/g. In patients with Canavan Disease statistically significant reductions of [Asp] and [Glu] were observed. NAA pool size was increased and VNAA was significantly lower (3.6 ± 0.1 nmol/min/g, p < 0.001) than in controls. VTCA = TCA-cycle rate, Vx = α-ketoglutarate/glutamate exchange rate. Vgln = glutamine synthesis rate, Vxa = oxaloacetate/aspartate exchange rate, Cit = citrate, Suc = succinate, Pyr = pyruvate, AcCoA = Acetyl CoA.

1110 Part II

Medical Uses

Part II

oxidation in human brain was estimated to be ≈20% of the total neuronal/glial TCA-cycle rate in fasted human brain [7,27]. Recently it was demonstrated that Glu C5 and Gln C5 , which are primarily derived from acetate in the glial compartment, accumulated more in patients on ketogenic diet (KD) than in controls, whilst accumulation of bicarbonate was similar [34] (Figure 8). These results are consistent with altered glutamate-glutamine neutrotransmitter cycling and adaptation to ketogenic diet with up-regulation of acetate oxidation relative to glucose oxidation. KD is a treatment option for epileptic patients, which is particularly effective in children. The biochemical mechanisms why KD improves seizure activity are incompletely understood. In vivo 13 C MRS may elucidate a possibly biochemical basis for reduction in seizures during KD therapy. These studies indicate that 13 C MRS is an appropriate tool to investigate diseases which are believed to originate in glial cells. Together with data from glucose infusion experiments this could result in a more complete understanding of cerebral metabolism in normal and diseased human brain.

Hyperpolarized 13 C Compounds Recently, dynamic nuclear polarization was used to obtain highly polarized 13 C nuclei in endogenous substances such as isotopically enriched 13 C urea [35]. Methods have been developed to perform the polarization process outside the body and a nuclear polarization of 50% for 13 C appears to be possible [35]. This polarization corresponds with an enhancement of 500,000 (!) compared with the thermal equilibrium polarization at 1.5 T for 13 C. Hyperpolarization vanishes due to longitudinal relaxation (T1 ) and spectroscopy studies with the goal to investigate metabolic pathways face two major obstacles. (i) The hyperpolarized substrate needs to reach the organ of interest quickly and (ii) turnover rates need to be sufficiently fast to allow the observation of metabolites of a hyperpolarized substrate. Therefore, the application of hyperpolarized substrates to study normal and abnormal metabolism depends mainly on the capability to prepare metabolically active, non-toxic substrates with long T1 -times. Once that is achieved, the promise of 13 C MRS with endogenous substances is enormous. Already, Golman et al. [35] report a T1 -time of 20 s for 13 C urea in vivo. The polarization of such a substrate would still be ≈10 times higher than the thermal equilibrium polarization 4 min after injection. A hyperpolarized compound with a T1 of 40 s would have a ≈1000 (!) higher polarization and a hypothetical 12 day acquisition time could be shortened to 1 s without compromising SNR! As these compounds become available, detailed studies of chemicals of low concentrations in tissue, first in animals and then in humans, may be feasible.

In particular, the study of neurotransmitters, until now out of reach for spectroscopist may become a reality.

Acknowledgments The author likes to thank Thomas Raidy, Ph.D. and Mark Albers, BS for discussions and for using their expertise with MR hardware. The author also likes to thank Rolf Gruetter, Ph.D. and Robin A. de Graaf, Ph.D. for providing figures.

Glossary Absolute quantitation—Determination of concentrations of metabolites in mMoles per kg of tissue or volume. B0 —Strong magnetic field, constant in time and space, generated by the superconducting magnet. B1 —Radiofrequency (RF) magnetic field generated by radiofrequency coils. Hetero-nuclear J-coupling—J-coupling between different species of spins, e.g. proton and carbon. Homo-nuclear J-coupling—J-coupling between the same species of spins, e.g. proton and proton. ISIS—Image-selected in vivo spectroscopy is based on a cycle of eight acquisitions which need to be added and subtracted in the right order to get a single volume. ISIS is considerably more susceptible to motion then STEAM or PRESS and is mostly used in heteronuclear studies, where its advantage of avoiding T2 -relaxation is valuable. J-coupling (or scalar coupling)—Many resonances split into multiplet components. This is the result of an internal indirect interaction of two spins via the intervening electron structure of the molecule. The coupling strength is measured in Hertz (Hz) rather than ppm because it is independent of the external B0 field strength. NOE—Nuclear Overhauser effect, the magnetization of protons dipolar-coupled to 13 C nuclei can be used to enhance the 13 C signal. While the term NOE is mainly associated with 13 C MRS, NOE enhancement can also be observed in 31 P MRS and with other nuclei. Polarization transfer—Many interesting nuclei like 13 C suffer from low inherent sensitivity compared with proton MR. Techniques like DEPT (distortionless enhancement by polarization transfer) and INEPT (insensitive nuclei enhanced by polarization transfer) improve the 13 C sensitivity by transferring the higher polarization of coupled protons to the carbon nuclei. Special hardware with two RF channels is needed for polarization transfer experiments. A modification of DEPT and INEPT is reverse DEPT and inverse INEPT where the polarization is transfered back to utilize the higher sensitivity of the protons for observation (inverse detection). PRESS—Point-resolved spectroscopy, utilizes three 180◦ slice selective pulses along each of the spatial directions

In vivo 13 C MRS

Applications 1111

Part II

Fig. 8. Impact of ketogenic diet (KD) on astroglial acetate oxidation (A) 13 C difference spectra obtained from ketogenic diet patients (average of three) and (B) controls after infusion of [1-13 C] acetate. 13 C label incorporation into Glu C5 and Gln C5 was more pronounced in patients. In addition, Glu C1 and Gln C1 enrichment was more prominent, whereas equivalent production of HCO3 was observed.

and generates signals from the overlap in form of a spinecho. SAR—Specific absorption rate. Due to inductive and dielectric losses energy from radiofrequency pulses is absorbed by tissue and mainly transferred into rotational and translational movements of water molecules which causes an increase of tissue temperature. Limits for the human brain are established by FDA guidelines. Scalar coupling—See J-coupling STEAM—Stimulated echo acquisition mode, localization method utilizing three 90◦ slice selective pulses, along each of the spatial directions. Signal, in form of a stimulated echo, from the overlap is generated in a “single shot” experiment. In contrast to PRESS, only half of the possible signal is recovered when the same echo time is used. T 1 -relaxation, T 1 -relaxation time—After the magnetization vector has been flipped into the transverse plane, new magnetization builds up along the z-axis. The time after 63% (1-1/e) of the equilibrium magnetization has built

up is called the T1 -relaxation time. T1 and T2 relaxation is caused by time-dependent fluctuations of local magnetic fields arising mostly from the motion of molecules with electric or magnetic dipoles at the site of the spins. For accurate absolute quantitation the relaxation times of all metabolites must be known in order to correct peak intensities appropriately. T 1 -saturation—The repetition times TR are usually in the range of the T1 -relaxation times. As a consequence of this, not all the magnetization has recovered, for example when TR = T1 only 63% of the equilibrium magnetization can be used for each scan (with the exception of the first scan) when 90◦ flip angles are used for excitation. This effect is called T1 -saturation. The extreme case of saturation occurs when several RF pulses are applied within a very short time followed by dephasing gradients. This technique is used in localized 1 H MRS to remove the dominant water signal (see water suppression). T 2 -relaxation, T 2 -relaxation time—The magnetization vector can be flipped into the transverse plane by using

1112 Part II

Medical Uses

Part II

an RF pulse. The so generated transverse magnetization undergoes an exponential decay. The time after the magnetization has relaxed to 37% (1/e) of its amplitude is called the T2 or transverse or spin-spin relaxation time. See also T1 - relaxation. T R (repetition time)—The time between each initial excitation of the magnetization is called the repetition time.

References 1. NMR in Biomed, 2003;16(6–7):301–457. Special issue: 13 C NMR studies of cerebral metabolism. 2. Adriany G, Gruetter R. J. Magn. Reson. 1997;125(1):178. 3. Bluml S. J. Magn. Reson. 1999;136(2):219. 4. Shen J, Petersen KF, Behar KL, Brown P, Nixon TW, Mason GF, Petroff OA, Shulman GI, Shulman RG, Rothman DL. Proceedings of the National Academy of Sciences of the United States of America. 1999;96(14), 8235. 5. Pan JW, Stein DT, Telang F, Lee JH, Shen J, Brown P, Cline G, Mason GF, Shulman GI, Rothman DL, Hetherington HP. Magn. Reson. Med. 2000;44(5):673. 6. Chhina N, Kuestermann E, Halliday J, Simpson LJ, Macdonald IA, Bachelard HS, Morris PG. J. Neurosci. Res. 2001;66(5):737. 7. Lebon V, Petersen KF, Cline GW, Shen J, Mason GF, Dufour S, Behar KL, Shulman GI, Rothman DL. J. Neurosci. 2002;22(5):1523. 8. FDA. Federal Register, 1988;53:7575. 9. FDA. US Department of Health and Human Services, Food and Drug Administration, http://www.fda.gov/cdrh/ode/ mri340.pdf(http://www.fda.gov/cdrh/ode/mri340.pdf) (1998). 10. van den Bergh AJ, van den Boogert HJ, Heerschap A. Magn. Reson. Med. 1998;39(4):642. 11. Hwang JH, Bluml S, Leaf A, Ross BD. NMR Biomed. 2003; 16(3):160. 12. Levitt MH, Freeman R. J. Magn. Reson. 1981;43:502. 13. Levitt MH. J. Magn. Reson. 1982;48:234. 14. Shaka AJ KJ, Frenkiel AJ, Freeman R. J. Magn. Reson. 1983;52:335. 15. Shaka AJ KJ, Freeman R. J Magn Reson, 1983;52:313. 16. de Graaf RA, Mason G, Pantel AB, Behar KL, Rothman DL. NMR Biomed. 2003;16:339.

17. Gruetter R, Adriany G, Choi IY, Henry P-G, Lei H, Oz G. NMR Biomed. 2003;16:313. 18. Gruetter R, Rothman DL, Novotny EJ, Shulman RG. Magn. Reson. Med. 1992;25(1):204. 19. Bluml S, Moreno A, Hwang JH, Ross BD. NMR in Biomed. 2001;14(1):19. 20. Bluml S, Moreno-Torres A, Ross BD. Magn. Reson. Med. 2001;45(6):981. 21. Ordidge RJ, Connelly A, Lohman JAB. J. Magn. Reson. 1986;66:283. 22. Gruetter R, Seaquist ER, Kim S, Ugurbil K. Dev. Neurosci. 1998;20(4–5):380. 23. Rothman DL, Behar KL, Hetherington HP, den Hollander JA, Bendall MR, Petroff OA, Shulman RG. Proc. Natl. Acad. Sci. U.S.A. 1985;82:1633. 24. Provencher SW. Magn. Reson. Med. 1993;30(6):672. 25. Mason GF, Gruetter R, Rothman DL, Behar KL, Shulman RG, Novotny EJ. J. Cereb. Blood. Flow. Metab. 1995;15(1):12. 26. Moreno A, Ross BD, Bluml S. J. Neurochem. 2001;77(1): 347. 27. Bluml S, Moreno-Torres A. Shic F, Nguy CH, Ross BD. NMR Biomed. 2002;15(1):1. 28. Mason GF, Falk Petersen K, de Graaf RA, Kanamatsu T, Otsuki T, Shulman GI, Rothman DL. Brain Res. Brain Res. Protoc. 2003;10(3):181. 29. Watanabe H, Umeda M, Ishihara Y, Okamoto K, Oshio K, Kanamatsu T, Tsukada Y. Magn. Reson. Med. 2000;43(4): 525. 30. Moreno A, Bluml S, Hwang JH, Ross BD. Magn. Reson. Med. 2001;46(1):39. 31. Beckmann N, Turkalj I, Seelig J, Keller U. Biochemistry. 1991;30(26):6362. 32. Magistretti PJ, Pellerin L, Rothman DL. Shulman RG. Science. 1999;283(5401):496. 33. Muir D, Berl S, Clarke DD. Brain Res. 1986;380(2):336. 34. Bluml S, Shic F, Lai L, Yahya K, Lin A, Ross BD. In Proceedings, 10th International Society of Magnetic Resonance in Medicine, Honolulu, 417 (2002). 35. Golman K, Ardenkjaer-Larsen JH, Petersson JS, Mansson S, Leunbach I. Proc. Natl. Acad. Sci. U.S.A. 2003;100(18): 10435. 36. Bluml S, Hwang JH, Moreno A, Ross BD. J. Magn. Reson. 2000;143(2):292. 37. de Graaf RA, Brown PB, Mason GF, Rothman DL, Behar KL. Magn. Reson. Med. 2003;49(1):37.

1113

Mark G. Swanson, Susan M. Noworolski, and John Kurhanewicz Center for Molecular and Functional Imaging, Department of Radiology, University of California, San Francisco, CA 94107, USA

Metabolite Abbreviations: Ala, Alanine; ATP, Adenosine triphosphate; Cho, Choline; Cr, Creatine; Eth, Ethanolamine; Glx, Glutamine/Glutamate; GPC, Glycerophosphocholine; GPE, Glycerophosphoethanolamine; Lac, Lactate; MI, myo-Inositol; NAA, N-acetyl aspartate; NTP, Nucleotide triphosphate (e.g. ATP); PC, Phosphocholine; PCr, Phosphocreatine; PE, Phosphoethanolamine; Pi, Inorganic phosphate; SI, scylloInositol; Tau, Taurine; UDP, Uridine diphosphate.

Introduction In this chapter, we describe the current and potential role of magnetic resonance spectroscopy (MRS) and magnetic resonance spectroscopic imaging (MRSI) in organs of the body with an emphasis on the technical aspects of applications in prostate and breast cancer, and diseases of the liver. In contrast to anatomical magnetic resonance imaging (MRI), which detects changes in the relaxivity or density of bulk tissue water, spectroscopy detects small molecular weight metabolites within the cytosol of cells or within extracellular spaces such as glands or ducts. The addition of spectroscopy has been shown to improve the ability (i.e. sensitivity, specificity, and accuracy [1]) of conventional MRI to detect and stage prostate and breast cancer [2–4] and has shown promise in the evaluation of primary and metastatic liver tumors [5] and other liver diseases [6,7]. However, the potential of spectroscopy is even greater because multiple metabolic markers may be combined to provide an independent assessment of disease state, cancer aggressiveness, and therapeutic response [8–10]. The clinical use of spectroscopy as an adjunct to MRI has expanded dramatically over the past several years. This has been due to both the need to answer clinically relevant questions and recent technical advances in hardware and software that have provided improvements in the spatial and time resolution of the spectral data and have resulted in the incorporation of this technology on commercial MR scanners. These breakthroughs have allowed the routine addition of spectroscopy sequences to clinical MRI exams, and have led

Graham A. Webb (ed.), Modern Magnetic Resonance, 1113–1125.
C 2008 Springer.

to spectroscopy being factored into the clinical decision process. Historically, 31 P and 1 H have been the nuclei of choice for in vivo MRS in the body and each has distinct advantages and disadvantages. The major advantage of 1 H spectroscopy is its high sensitivity, which is necessary to achieve high spatial resolution (< 1 cm3 ) in a clinically reasonable amount of time. Because the sensitivity of 31 P is only 6.6% that of 1 H, much larger voxel sizes (typically >8 cm3 ) must be used to achieve the same sensitivity in the same amount of time. 31 P MRS also suffers from long T1 and short T2 relaxation times relative to 1 H. The inherently low sensitivity of 31 P MRS can be improved by broadband 1 H decoupling (e.g. WALTZ) during the acquisition, which sharpens signals by collapsing multiplets and produces a large nuclear overhauser enhancement (NOE) [11], and through the use of higher magnetic field clinical MR scanners. The major advantages of 31 P MRS are that no water or lipid suppression is needed and there is less spectral overlap because a relatively small number of metabolites are dispersed over a large spectral window (∼25 ppm1 ). However, both 1 H and 31 P spectroscopy require additional hardware, software, and postprocessing and display tools, much of which can now be purchased in the form of spectroscopy packages from the major MR manufacturers General Electric, Siemens, and Philips. Several metabolites are present in high enough concentrations (>1 mM) to be detected by 1 H or 31 P MRS, although many have not yet been detected or fully exploited in vivo in the body. The 1 H MR spectrum spans a frequency range of about 10 ppm and is centered around water at 4.8 ppm. To date, the resonances

1

Parts-per-million (ppm) is a dimensionless unit used to describe differences (usually from zero) in chemical shift (δ) independent of the applied frequency (B0 ). Frequencies in Hertz (Hz) are converted to ppm by dividing the value by the carrier frequency in MHz and multiplying by 106 . For example, a magnetic field strength of 1.5 T corresponds to a 1 H carrier frequency of ∼63 MHz, such that 1 ppm ≈ 63 Hz, whereas at 3.0 T, 1 ppm ≈ 126 Hz.

Part II

Magnetic Resonance Spectroscopy and Spectroscopic Imaging of the Prostate, Breast, and Liver

1114 Part II

Medical Uses

Part II

upfield of water have received the most attention, including lactate (Lac, δ = 1.33 and 4.12 ppm), alanine (Ala, δ = 1.48 and 3.78 ppm), glutamine/glutamate (Glx, δ = 2.04, 2.11, 2.35, and 3.76 ppm), taurine (Tau, δ = 3.26 and 3.43 ppm), myo-inositol (mI, δ = 3.28, 3.54, 3.60, and 4.05 ppm), scyllo-inositol (sI, δ = 3.35 ppm), creatine/phosphocreatine (Cr/PCr, δ = 3.04 and 3.93 ppm), the choline containing compounds choline (Cho, δ = 3.21, 3.55, and 4.07 ppm), phosphocholine (PC, δ = 3.23, 3.62, and 4.18 ppm) and glycerophosphocholine (GPC, δ = 3.24, 3.68, and 4.34 ppm), and the ethanolamine containing compounds ethanolamine (Eth, δ = 3.15 and 3.80 ppm), phosphoethanolamine (PE, δ = 3.22 and 3.99 ppm), and glycerophosphoethanolamine (GPE, δ = 3.30 and 4.12 ppm). Healthy prostate tissue is unique in that citrate (δ = 2.55 and 2.71 ppm) [12] and polyamines (predominantly spermine, δ = 3.11, 2.09, and 1.78 ppm) [13] are also present in very high concentrations and can be readily observed by 1 H MRS. There is also much interest in observing glucose (3.43, 3.80, and 5.23 ppm) and uridine diphosphate (UDP) sugars (5.5–6.1 ppm), which resonate very close to water, because of the role of increased glycolysis in cancer [14]. It has also been demonstrated that amide proton transfer from the downfield exchangeable amide protons of proteins and peptides (∼7.8−8.8 ppm) to water can improve sensitivity by several orders of magnitude and provide a novel imaging mechanism [15]. The major contributors to the 31 P MR spectrum include inorganic phosphate (Pi, δ = 2.26 ppm), phosphocreatine (PCr, δ = −2.89 ppm), the phosphomonoesters (PC and PE, δ = 3.76), the phosphodiesters (GPC and GPE, δ = 0.11 and 0.74 ppm), diphosphodiesters (e.g. UDP sugars, δ = −11.07 and −12.76 ppm), and nucleotide phosphates [e.g. adenosine triphosphate (ATP), δ = −5.24, −10.37, and −19.02 ppm] [16]. 31 P chemical shifts are typically referenced to either in vivo PCr or external 85% H3 PO4 (as listed here). It should be noted that when 31 P chemical shifts are reported with PCr set to 0.00 ppm, the values are 2.89 ppm greater than the corresponding values referenced with H3 PO4 set to 0.00 ppm. It is also widely believed that 13 C spectroscopy, which to date has been primarily limited to the brain [17,18], will play a major role in the future of body MRS as higher field human scanners, 13 C labeled substrates (e.g. glucose, acetate, and pyruvate), and commercial 13 C hyperpolarizers [19] become more widely available and approved for human studies. However, because 13 C MRS is described elsewhere in this book, it will not be further discussed in this chapter. Previous MR studies on prostate, breast, and liver tumors have identified elevated levels of phosphomonoesters and phosphodiesters (detected by 31 P MRS) [5,16,20–24] and elevated levels of the composite Cho res-

onance (detected by 1 H MRS) relative to normal healthy tissues [4,12,20,25–29]. Although the in vivo 1 H signal that is attributed to Cho contains contributions from Cho, PC, GPC, Eth, PE, GPE, Tau, mI, and sI, the Cho head group contains nine equivalent protons; consequently, a small increase in concentration results in a large increase in signal intensity. In cancer, the observed increases in Cho and Eth containing compounds have been primarily associated with increased cell membrane synthesis and degradation [24]. However, changes in cell density and altered phospholipid metabolism with cancer evolution and progression also contribute to the observed changes in phospholipid metabolites [27,30]. In spectroscopy it is also very important to identify markers for healthy or normal tissue that can be used for quantitation purposes. In 31 P MRS, metabolite ratios are often calculated relative to Pi, PCr, or NTP. In studies of cancer, before and after therapy, the phosphomonoester to phosphodiester ratio is particularly useful because it compares markers of proliferation (PC and PE) to markers of cellular breakdown or apoptosis (GPC and GPE). 1 H MRS has been highly successful in the prostate and brain because, in addition to increased Cho, there are unique markers for healthy tissue that decrease in cancer [31]. Consequently, the ratios of Cho to citrate in the prostate and Cho to N-acetyl aspartate (NAA) in the brain are significantly greater in regions of cancer compared to surrounding healthy tissues [2,32]. However, because citrate and NAA are not observed in any other tissues of the body, novel markers for healthy tissue, which may or may not change with malignancy or disease, are needed.

Techniques for Spectroscopy and Spectroscopic Imaging of the Body Spatial Localization The most common localization schemes for single voxel spectroscopy are image selected in vivo spectroscopy (ISIS) [33], stimulated echo acquisition mode (STEAM) [34], and point resolved spectroscopy (PRESS) [35]. ISIS consists of a series of selective inversion (i.e. 180◦ ) pulses, which are turned on and off according to an eight-step encoding scheme, in the presence of magnetic field gradients. Because the magnetization remains along the “z” axis prior to the read pulse, ISIS is relatively insensitive to T2 relaxation, and therefore has historically been popular for 31 P MRS. However, ISIS is particularly sensitive to motion because the eight transients must be added and subtracted to achieve spatial localization. Consequently, STEAM and PRESS, which are capable of three-dimensional (3D) localization in a

Magnetic Resonance Spectroscopy and Spectroscopic Imaging

Improved Volume Selection and Outer Volume Suppression Spectroscopy studies in the body are critically dependent on accurate volume selection, since the region of interest is often adjacent to regions of lipid or air-tissue interfaces, which can significantly impair spectral quality. A recent technical advance for 1 H MRS has been the substitution of optimally shaped rf pulses, e.g. Shinnar-Le Roux pulses [38], in place of conventional sinc-shaped pulses for improved volume selection in PRESS acquisitions. Although low tip angle pulses can produce reasonably good slice profiles, optimized pulses are essential for 90◦ and especially 180◦ excitations [38]. Water saturation performance can also be improved using shaped pulses; however, due to the imperfect excitation profiles of the PRESS spin-echo pulses, even with Shinnar-Le Roux pulses, sig-

nificant contamination from lipids outside the PRESS selected region can still occur. Several groups have used outer volume suppression (OVS) sequences to better conform the volume of interest [39–43]. These sequences utilized optimized pulses or special excitation schemes to shape the excitation volume to the region of interest. However, due to the non-rectangular suppression profiles of these pulses residual unsuppressed water and lipid signals at the band edges often rendered large portions of the spectral array unusable. Quadratic phase pulse designs, e.g. very selective suppression (VSS) pulses [44], can provide excellent spatial selectivity, high effective bandwidths, and improved B1 and T1 insensitivity compared to conventional OVS pulses. VSS pulses can be inserted just before the PRESS excitation pulses and are used to better define the edges of the PRESS box. Additional VSS pulses can also be graphically prescribed in order to shape the selected volume to the region of interest to exclude regions of lipid or air tissue interfaces. Because of the imperfect PRESS excitation profile, the effects of chemical shift misregistration and the rounded edges of the PRESS selected volume can be dramatically reduced by overprescribing the PRESS selection by ∼20–30% and applying the VSS pulses to define the desired dimensions of the box as shown in Figure 1. The use of graphical VSS pulses to exclude periprostatic lipids in the prostate is illustrated in Figure 2A and B.

Water and Lipid Suppression In order to detect the resonances of biological interest, the 110 molar water resonance must be suppressed by approximately 1000–10,000 fold and spectral contamination from lipids outside the volume of interest must be minimized as much as possible. Good shimming is absolutely essential for water and lipid suppression. Although automated shimming routines are often adequate and should be used as a starting point, it is well worth the extra time to manually shim and visually assess the shape of the water resonance and the FID. The ability to obtain a narrow water line width (1 cm from the edges of the liver, avoiding vessels, and spectra were individually phased and frequency shifted before averaging.

body. The increased sensitivity provided by higher field MR scanners, improved pulse sequences and technology, and the ability to include additional metabolic markers will also have a major impact on the clinical potential of body spectroscopy in the future.

Acknowledgments The authors wish to acknowledge and thank Drs. Michael Garwood and Patrick J. Bolan, Center for Magnetic Resonance Research, University of Minnesota School of Medicine, Minneapolis, for providing the breast figure presented in this chapter, and Dr. Aliya Qayyum, Department of Radiology, University of California, San Francisco, for providing the liver figures presented in this chapter.

Glossary of Terms BASING BIR BISTRO CHESS CSI EPSE FID HR-MAS ISIS LASER MRI MRS

Band selective inversion with gradient dephasing B1 -insensitive rotation B1 -insensitive train to obliterate signal Chemical shift selective saturation Chemical shift imaging Echo-planar spin-echo Free induction decay High-resolution magic angle spinning Image selected in vivo spectroscopy Localization by adiabatic selective refocusing Magnetic resonance imaging Magnetic resonance spectroscopy

Magnetic Resonance Spectroscopy and Spectroscopic Imaging

NAFLD NASH NOE OVS PPM PRESS STEAM STIR T1 T2 TRUS VSS WALTZ

Magnetic resonance spectroscopic imaging Non-alcoholic fatty liver disease Non-alcoholic steatohepatitis Nuclear overhauser enhancement Outer volume suppression Parts-per-million Point resolved spectroscopy Stimulated echo acquisition mode Short time inversion recovery Longitudinal (spin–lattice) relaxation time Transverse (spin–spin) relaxation time Transrectal ultrasound Very selective suppression Broadband decoupling scheme

References 1. Obuchowski NA. Radiology. 2003;229:3. 2. Scheidler J, Hricak H, Vigneron DB, Yu KK, Sokolov DL, Huang LR, Zaloudek CJ, Nelson SJ, Carroll PR, Kurhanewicz J. Radiology. 1999;213:473. 3. Yu KK, Scheidler J, Hricak H, Vigneron DB, Zaloudek CJ, Males RG, Nelson SJ, Carroll PR, Kurhanewicz J. Radiology. 1999;213:481. 4. Katz-Brull R, Lavin PT, Lenkinski RE. J. Natl. Cancer. Inst. 2002;94:1197. 5. Bell JD, Bhakoo KK. NMR Biomed. 1998;11:354. 6. Cho SG, Kim MY, Kim HJ, Kim YS, Choi W, Shin SH, Hong KC, Kim YB, Lee JH, Suh CH. Radiology. 2001;221: 740. 7. Heiken JP, Lee JK, Dixon WT. Radiology. 1985;157:707. 8. Mueller-Lisse UG, Swanson MG, Vigneron DB, Hricak H, Bessette A, Males RG, Wood PJ, Noworolski S, Nelson SJ, Barken I, Carroll PR, Kurhanewicz J. Magn. Reson. Med. 2001;46:49. 9. Mueller-Lisse UG, Vigneron DB, Hricak H, Swanson MG, Carroll PR, Bessette A, Scheidler J, Srivastava A, Males RG, Cha I, Kurhanewicz J. Radiology. 2001;221:380. 10. Arias-Mendoza F, Smith MR, Brown TR. Acad. Radiol. 2004;11:368. 11. van Sluis R, Payne GS, Leach MO. Magn. Reson. Med. 1995;34:893. 12. Kurhanewicz J, Vigneron DB, Hricak H, Narayan P, Carroll P, Nelson SJ. Radiology. 1996;198:795. 13. Cheng LL, Wu CL, Smith MR, Gonzalez RG. FEBS Lett. 2001;494:112. 14. Stubbs M, Bashford CL, Griffiths JR. Curr. Mol. Med. 2003; 3:49. 15. Zhou J, Lal B, Wilson DA, Laterra J, van Zijl PC. Magn. Reson. Med. 2003;50:1120. 16. Kurhanewicz J, Dahiya R, Macdonald JM, Jajodia P, Chang LH, James TL, Narayan P. NMR Biomed. 1992;5:185. 17. Ross B, Lin A, Harris K, Bhattacharya P, Schweinsburg B. NMR Biomed. 2003;16:358. 18. Morris P, Bachelard H. NMR Biomed. 2003;16:303.

19. Ardenkjaer-Larsen JH, Fridlund B, Gram A, Hansson G, Hansson L, Lerche MH, Servin R, Thaning M, Golman K. Proc. Natl. Acad. Sci. U.S.A. 2003;100:10158. 20. Cornel EB, Smits GA, Oosterhof GO, Karthaus HF, Deburyne FM, Schalken JA, Heerschap A. J. Urol. 1993;150:2019. 21. Kalra R, Wade KE, Hands L, Styles P, Camplejohn R, Greenall M, Adams GE, Harris AL, Radda GK. Br. J. Cancer. 1993;67:1145. 22. Kurhanewicz J, Thomas A, Jajodia P, Weiner MW, James TL, Vigneron DB, Narayan P. Magn. Reson. Med. 1991;22:404. 23. Ronen SM, Leach MO. Breast Cancer Res. 2001;3:36. 24. Daly PF, Lyon RC, Faustino PJ, Cohen JS. J. Biol. Chem. 1987;262:14875. 25. Cecil KM, Schnall MD, Siegelman ES, Lenkinski RE. Breast Cancer Res. Treat. 2001;68:45. 26. Kurhanewicz J, Dahiya R, Macdonald JM, Chang LH, James TL, Narayan P. Magn. Reson. Med. 1993;29:149. 27. Aboagye EO, Bhujwalla ZM. Cancer Res. 1999;59:80. 28. Kuo YT, Li CW, Chen CY, Jao J, Wu DK, Liu GC. J. Magn. Reson. Imaging. 2004;19:598. 29. Soper R, Himmelreich U, Painter D, Somorjai RL, Lean CL, Dolenko B, Mountford CE, Russell P. Pathology. 2002;34: 417. 30. Schiebler ML, Tomaszewski JE, Bezzi M, Pollack HM, Kressel HY, Cohen EK, Altman HG, Gefter WB, Wein AJ, Axel L. Radiology. 1989;172:131. 31. Kurhanewicz J, Vigneron DB, Nelson SJ. Neoplasia. 2000; 2:166. 32. Vigneron D, Bollen A, McDermott M, Wald L, Day M, Moyher-Noworolski S, Henry R, Chang S, Berger M, Dillon W, Nelson S. Magn. Reson. Imaging. 2001;19:89. 33. Ordidge RJ, Connelly A, Lohman JAB. J. Magn. Reson. 1986;66:283. 34. Frahm J, Merboldt KD, Hanicke W. J. Magn. Reson. 1987;72:502. 35. Bottomley PA. Ann. N.Y. Acad. Sci. 1987;508:333. 36. Moonen CT, von KM, van ZP, Cohen J, Gillen J, Daly P, Wolf G. NMR Biomed. 1989;2:201. 37. Geppert C, Dreher W, Leibfritz D. MAGMA. 2003;16:144. 38. Pauly J, Le Roux P, Nishimura D, Macovski A. IEEE Trans. Med. Imaging. 1991;10:53. 39. Shungu DC, Glickson JD. Magn. Reson. Med. 1993;30: 661. 40. Posse S, Tedeschi G, Risinger R, Ogg R, Le Bihan D. Magn. Reson. Med. 1995;33:34. 41. St Lawrence K, Lee TY, Henkelman M. Magn. Reson. Med. 1998;40:944. 42. Le Roux P, Gilles RJ, McKinnon GC, Carlier PG. J. Magn. Reson. Imaging. 1998;8:1022. 43. Luo Y, de Graaf RA, DelaBarre L, Tannus A, Garwood M. Magn. Reson. Med. 2001;45:1095. 44. Tran TK, Vigneron DB, Sailasuta N, Tropp J, Le Roux P, Kurhanewicz J, Nelson S, Hurd R. Magn. Reson. Med. 2000;43:23. 45. Haase A, Frahm J, Hanicke W, Matthaei D. Phys. Med. Biol. 1985;30:341. 46. Bydder GM, Young IR. J. Comput. Assist. Tomogr. 1985; 9:659. 47. Star-Lack J, Nelson SJ, Kurhanewicz J, Huang LR, Vigneron DB. Magn. Reson. Med. 1997;38:311.

Part II

MRSI

References 1123

1124 Part II

Medical Uses

Part II

48. Schricker AA, Pauly JM, Kurhanewicz J, Swanson MG, Vigneron DB. Magn. Reson. Med. 2001;46:1079. 49. Brown TR, Kincaid BM, Ugurbil K. Proc. Natl. Acad. Sci. U.S.A. 1982;79:3523. 50. Maudsley AA, Hilal SK, Simon HE, Wittekoek S. Radiology. 1984;153:745. 51. Moyher SE, Vigneron DB, Nelson SJ. J. Magn. Reson. Imaging. 1995;5:139. 52. Liney GP, Turnbull LW, Knowles AJ. J. Magn. Reson. Imaging. 1998;8:994. 53. DeGraaf RA, Nicolay K. Concepts Magn. Reson. 1997;9:247. 54. Garwood M, DelaBarre L. J. Magn. Reson. 2001;153: 155. 55. de Graaf RA, Luo Y, Garwood M, Nicolay K. J. Magn. Reson. B. 1996;113:35. 56. Schwarz AJ, Leach MO. Phys. Med. Biol. 2000;45:2105. 57. Katz-Brull R, Rofsky NM, Lenkinski RE. Magn. Reson. Med. 2003;50:461. 58. Tyszka JM, Silverman JM. Magn. Reson. Med. 1998;39:1. 59. Nelson SJ. Magn. Reson. Med. 2001;46:228. 60. Spielman D, Webb P, Macovski A. Magn. Reson. Med. 1989; 12:38. 61. Nelson SJ, Brown TR. J. Magn. Reson. 1987;75:229. 62. Nelson S, Brown T. J. Magn. Reson. 1989;84:95. 63. Derby K, Hawryszko C, Tropp J. Magn. Reson. Med. 1989; 12:235. 64. Provencher SW. NMR Biomed. 2001;14:260. 65. Tropp JS, Sugiura S, Derby KA, Suzuki Y, Hawryszko C, Yamagata H, Klein JE, Ortendahl DA, Kaufman L, Acosta GF. Radiology. 1988;169:207. 66. Twieg DB, Meyerhoff DJ, Hubesch B, Roth K, Sappey MD, Boska MD, Gober JR, Schaefer S, Weiner MW. Magn. Reson. Med. 1989;12:291. 67. Lenkinski RE, Holland GA, Allman T, Vogele K, Kressel HY, Grossman RI, Charles HC, Engeseth HR, Flamig D, MacFall JR. Radiology. 1988;169:201. 68. Vigneron DB, Nelson SJ, Murphy BJ, Kelley DA, Kessler HB, Brown TR, Taylor JS. Radiology. 1990;177:643. 69. Bracewell RN. The Fourier Transform and Its Applications. McGraw-Hill: New York, 1978, p 104. 70. Marcei TH, Booker HR. J. Magn. Reson. 1984;57:157. 71. Swanson MG, Vigneron DB, Tabatabai ZL, Males RG, Schmitt L, Carroll PR, James JK, Hurd RE, Kurhanewicz J. Magn. Reson. Med. 2003;50:944. 72. Bolan PJ, DelaBarre L, Baker EH, Merkle H, Everson LI, Yee D, Garwood M. Magn. Reson. Med. 2002;48:215. 73. Bolan PJ, Meisamy S, Baker EH, Lin J, Emory T, Nelson M, Everson LI, Yee D, Garwood M. Magn. Reson. Med. 2003; 50:1134. 74. Li CW, Negendank WG, Murphy-Boesch J, Padavic-Shaller K, Brown TR. NMR Biomed. 1996;9:141. 75. Gonen O, Murphy-Boesch J, Li CW, Padavic-Shaller K, Negendank WG, Brown TR. Magn. Reson. Med. 1997;37: 164. 76. Dixon RM, Frahm J. Magn. Reson. Med. 1994;31:482. 77. Kim DY, Kim KB, Kim OD, Kim JK. J. Korean Med. Sci. 1998;13:49. 78. Kasimos JN, Merchant TE, Gierke LW, Glonek T. Cancer Res. 1990;50:527.

79. Bottomley PA, Hardy CJ, Roemer PB. Magn. Reson. Med. 1990;14:425. 80. Deslauriers R, Kupriyanov VV. Biochem. Cell. Biol. 1998; 76:510. 81. Neubauer S, Horn M, Hahn D, Kochsiek K. Mol. Cell. Biochem. 1998;184:439. 82. Kreis R, Boesch C. J. Magn. Reson. B. 1996;113:103. 83. Boska MD, Nelson JA, Sripathi N, Pipinos II, Shepard AD, Welch KM. Magn. Reson. Med. 1999;41:1145. 84. Meyerspeer M, Krssak M, Moser E. Magn. Reson. Med. 2003;49:620. 85. Li CW, Kuesel AC, Padavic-Shaller KA, Murphy-Boesch J, Eisenberg BL, Schmidt RG, von Roemeling RW, Patchefsky AS, Brown TR, Negendank WG. Cancer Res. 1996;56: 2964. 86. Sijens PE. NMR Biomed. 1998;11:341. 87. Negendank WG, Padavic-Shaller KA, Li CW, Murphy-Boesch J, Stoyanova R, Krigel RL, Schilder RJ, Smith MR, Brown TR. Cancer Res. 1995;55:3286. 88. Griffiths JR, Tate AR, Howe FA, Stubbs M. Eur. J. Cancer. 2002;38:2085. 89. Narayan P, Vigneron DB, Jajodia P, Anderson CM, Hedgcock MW, Tanagho EA, James TL. Magn. Reson. Med. 1989;11:209. 90. Narayan P, Jajodia P, Kurhanewicz J, Thomas A, MacDonald J, Hubesch B, Hedgcock M, Anderson CM, James TL, Tanagho EA et al. J. Urol. 1991;146:66. 91. Coakley FV, Kurhanewicz J, Lu Y, Jones KD, Swanson MG, Chang SD, Carroll PR, Hricak H. Radiology. 2002;223:91. 92. Kurhanewicz J, Swanson MG, Nelson SJ, Vigneron DB. J. Magn. Reson. Imaging. 2002;16:451. 93. Pickett B, Ten Haken RK, Kurhanewicz J, Qayyum A, Shinohara K, Fein B, Roach M. Int. J. Radiat. Oncol. Biol. Phys. 2004;59:665. 94. Coakley FV, Qayyum A, Swanson MG, Lu Y, Roach M, Pickett B, Shinohara K, Vigneron DB, Kurhanewicz J. Radiology. 2004;233:441. 95. Costello LC, Franklin RB. Prostate. 1991;18:25. 96. Costello LC, Franklin RB. Prostate. 1991;19:181. 97. Heby O. Differentiation. 1981;19:1. 98. Heston WD. Cancer Surv. 1991;11:217. 99. Shukla-Dave A, Hricak H, Eberhardt SC, Olgac S, Muruganandham M, Scardino PT, Reuter VE, Koutcher JA, Zakian KL. Radiology. 2004;231(3):717. 100. Kaji Y, Kurhanewicz J, Hricak H, Sokolov DL, Huang LR, Nelson SJ, Vigneron DB. Radiology. 1998;206(3):785. 101. Orel SG, Schnall MD. Radiology. 2001;220:13. 102. Buchberger W, Niehoff A, Obrist P, DeKoekkoek-Doll P, Dunser M. Semin. Ultrasound CT MR. 2000;21:325. 103. Degani H, Gusis V, Weinstein D, Fields S, Strano S. Nat. Med. 1997;3:780. 104. Esserman L, Hylton N, George T, Weidner N. Breast J. 1999;5:13. 105. Leach MO, Verrill M, Glaholm J, Smith TA, Collins DJ, Payne GS, Sharp JC, Ronen SM, McCready VR, Powles TJ, Smith IE. NMR Biomed. 1998;11:314. 106. Hurd RE, Gurr D, Sailasuta N. Magn. Reson. Med. 1998;40: 343. 107. Gribbestad IS, Petersen SB, Fjosne HE, Kvinnsland S, Krane J. NMR Biomed. 1994;7:181.

Magnetic Resonance Spectroscopy and Spectroscopic Imaging

112. Ryysy L, Hakkinen AM, Goto T, Vehkavaara S, Westerbacka J, Halavaara J, Yki-Jarvinen H. Diabetes. 2000;49:749. 113. Sutinen J, Hakkinen AM, Westerbacka J, Seppala-Lindroos A, Vehkavaara S, Halavaara J, Jarvinen A, Ristola M, YkiJarvinen H. AIDS. 2002;16:2183. 114. Rosen BR, Carter EA, Pykett IL, Buchbinder BR, Brady TJ. Radiology. 1985;154:469. 115. Szczepaniak LS, Babcock EE, Schick F, Dobbins RL, Garg A, Burns DK, McGarry JD, Stein DT. Am. J. Physiol. 1999;276:E977.

Part II

108. Stanwell P, Gluch L, Clark D, Tomanek B, Baker L, Giuffre B, Lean C, Malycha P, Mountford C. Eur. Radiol. 2005;15:1037. 109. Star-Lack J, Vigneron DB, Pauly J, Kurhanewicz J, Nelson SJ. J. Magn. Reson. Imaging. 1997;7:745. 110. Pickren J, Tsukada Y, Lane WW. In: L Weiss, HA Gilbert (Eds). Analysis of Autopsy Data. GK Hall and Company: Boston, 1982, p 2. 111. Thomsen C, Becker U, Winkler K, Christoffersen P, Jensen M, Henriksen O. Magn. Reson. Imaging. 1994;12:487.

References 1125

1127

W. A. Kaiser Institute for Diagnostic and Interventional Radiology, Friedrich Schiller University, Jena, Germany

Introduction MR-Imaging (MRI) of the breast, or MR-Mammography (MRM), has probably been one of the most controversially discussed indications for MRI in the last 20 years. Whereas MRI of the brain and spine or of the musculoskeletal system was rather quickly accepted as a valuable and helpful tool to detect diseases, descriptions of MRM initially ranged from “probably not useful” to “unclear” to “wait and see”. However, in the last 5 years or so a striking wave of publications has underlined the significant additional information provided by MRM. In this chapter a brief overview about the history of MRM is followed by a description of the pathophysiological basics and the present status and indication. An overview of the different techniques follows and the description of future challenges will finalize this report.

History of MRM The development of MRM included several phases: MR-Tomography in general had first to be developed, then special breast coils (single breast coils, followed by double breast coils later combined imaging/interventional coils) built and optimized. Contrast media were used at first in spin–echo sequences, later in sophisticated the so-called “dynamic” techniques using gradient echo sequences for a better differential diagnosis. 1. By 1971 different relaxation times in tumor tissue in comparison to normal tissue were measured using in vitro experiments [1]. However, MRI of the human body was possible only after the application of local gradient fields [2]. First, tissue samples of the breast were published by Mansfield et al. [3]. These tissue samples were taken about 90 min after operation. First, in vivo images of the breasts from 65 patients (128 breasts) including seven cancers was undertaken using whole-body MR-machines in 1982 [4]. In these reports a whole-body scanner with a field strength of 0.045 T was used and patients imaged in a supine position providing two proton-density-weighted images and six T1 -relaxation-time measurements per patient. In 1983, El Yousef et al. [5] published results from Graham A. Webb (ed.), Modern Magnetic Resonance, 1127–1141.
C 2008 Springer.

2 patients using an experimental surface coil in a field strength of 0.03 T and reported a reduced signal intensity of both cancers. In a following publication of the same group [6] in 1984, images of 10 volunteers and 45 patients were reported describing a reduced signal intensity of breast lesions in spin echo and inversionrecovery sequences. 2. From 1983, onwards special breast coils were developed which were used only for breast imaging and later sold commercially. Until 1986 only single breast coils were available [7,8]. After first-disappointing trials in a supine position the examinations were performed in a prone position to minimize movement and artefacts caused by breathing. However, at that time the available spin-echo- and inversion-recovery sequences did not allow a definite detection and differentiation of small lesions in all cases, yet the advantage of imaging in thin slices in any orientation in a variable soft tissue contrast without X-rays was already clear [9–12]. Further progress was made with the development and introduction of the MR-contrast-medium Gadolinium-DTPA [13]. Experiments were possible in analogy to computer tomographic results using ionized X-ray contrast medium and radioactive iodine uptake from the mid seventies [14,15]. The first group to acquire and use contrast medium was Heywang et al. [16]. Initially spin-echo sequences in relatively long examination times and high contrast dosages were applied and reported; however, the uptake of contrast medium of cancers, normal tissue, and proliferative changes could not be differentiated sufficiently. After the introduction of fast echo gradient sequences [17] the first dynamic examinations using repetitive measurements of the same slices before and, in short time intervals, after contrast injection were established [18,19]. However, how to achieve better differentiation between benign and malignant lesions was a major issue amongst scientists [20]. The following development of a double breast coil [21] allowed routine measurements of both breasts of a patient in a single examination in a good signalto-noise (SNR) ratio. These initial steps were followed by a phase of evaluation of dynamic techniques using different measurement sequences and dosages, which was mainly performed in Europe. The result of the enormous variety of measurement techniques was, at

Part II

MR-Mammography

1128 Part II

Medical Uses

Part II

first, Babylonian confusion and a broad range of opinions about the usefulness of MRM [22–72], especially before 1995. 3. Using MRM small lesions of a few millimeters could be detected, which were often not seen by X-ray or ultrasound. Therefore, a need for the development of interventional MR procedures was rising, at first for positioning of markers, later for biopsy and therapeutic removal [73–80]. These techniques currently require an additional MR image to be taken later, i.e. a second measurement and a repeated use of the MR-device and contrast injection. At present single breast biopsy coils in the so-called “closed” as well as in “open” MR-devices are tested. A bilateral coil for combined imaging and simultaneous intervention has been used in clinical tests since 1997 [81].

Pathophysiological Background of MRM The secret of the very high sensitivity of MRM for the detection of breast lesions is the detection of tumor-based angiogenesis [82–84]. It is well-established that any tumor of more than 2 mm needs an increased perfusion to selectively direct nutrients towards the malignant tumor. Fifty years ago it was found that the implantation of breast cancer cells induces new micro-vessels after 3 days post implantation. Meanwhile numerous publications especially from Judah Folkman, have clarified that human breast cancers, like other malignant tumors, possess the angiogenentic power as well [85–91]. Over 20 years ago [92] it was established that, in a “far” distance of 3 cm from the cancer, an increased concentration of tumor angiogenetic factors can be measured. Probably induced by

Fig. 1. Normal enhancement, Inflow-phenomenon (40-year-old patient, dense breast) (A) Original dynamic scans showing the uptake of contrast over 7 min (upper left = precontrast, upper right = 1-min-postcontrast, lower left = 2-min-postcontrast, lower right = 7 min postcontrast), slow progressive enhancement especially in the outer portions of the parenchyma, bilateral, symmetric enhancement. The contrast usually arrives via lateral thoracic artery in the outer upper quadrant at first (inflow phenomenon) (arrows). (The left breast is displayed on the right side of the image, the right breast is displayed on the left side of the images.) (B) The inflow phenomenon is best delineated in the corresponding subtraction images (arrows) (upper left = subtraction 1-min-postcontrast minus precontrast, upper right = subtraction 2-min-postcontrast minus precontrast, lower left = subtraction 3-min-postcontrast minus precontrast, lower left = subtraction 7-min-postcontrast minus precontrast). (C) So-called “mosaic-image” (upper left = precontrast, upper right = 1-min-postcontrast, lower left = subtraction 1-min-postcontrast minus precontrast, lower left = T2 w-TSE image; all images in identical sclice positions).

MR-Mammography

MR-Mammography 1129

Part II

Fig. 1. (Continued)

1130 Part II

Medical Uses

Part II

hypoxia, acidosis, and many tumor angiogenetic factors, and eventually combined by an inactivation of the p53 suppressor gene, the so-called “tumor angiogenetic switch” is induced and new vessels arise through the sprouting of capillaries from pre-existing micro-vessels. Host capillaries dilate and become hyperpermeable, fibrin exudes through the leaky capillaries, proteinases, and collagenases break basal membranes, endothelial cells migrate across ruptures, form loops and canalize and finally a “functional neovasculature” called “tumor angiogenesis” is built.

Unique properties of tumor vessels are increased tumor blood volume, increased arterio-venous shunt formation, altered capillary bed transit time, increased interstitial pressure, and increased capillary permeability. Among more than fifty so-called “tumor-angiogenetic factors” the vascular permeability factor (VPF), vascular endothelial growth factor (VEGF), and basic fibroblastic growth factor (bFGF) are most important. Various pharmacokinetic models revealed that, in particular, differences in tumor vessel permeability are the

Fig. 2. Invasive lobular cancer and Lobular cancer in situ (LCIS, 78-year-old patient). (A) Original dynamic scans showing the uptake of contrast over 7 min (upper left = precontrast, upper right = 1-min-postcontrast, lower left = 2-min-postcontrast, lower right = 7 min postcontrast), strikingly enhancing area in a diameter of 1.3 cm behind the nipple in the right breast (arrow) with a maximum signal rise within the first minute after the injection of contrast followed by a decrease of signal intensity (“washoutphenomenon”) which is very typical for a cancer. In addition to that, nearly the complete right parenchyma shows a reticular, network-like enhancement compared to the other non-affected left breast (ellipse). (B) This reticular enhancement (ellipse) can clearly be detected in the corresponding subtraction images besides the focal mass enhancement of the invasive cancer (arrow). Most of this network-loke enhancement representing LCIS is steadily continuously enhancing with confluent irregular lesions in late scans (upper left = subtraction 1-min-postcontrast minus precontrast, upper right = subtraction 2-min-postcontrast minus precontrast, lower left = subtraction 3-min-postcontrast minus precontrast, lower left = subtraction 7-min-postcontrast minus precontrast). (C) The “mosaic-image” (upper left = precontrast, upper right = 1-min-postcontrast, lower left = subtraction 1-min-postcontrast minus precontrast, lower left = T2 w-TSE image in identical sclice positions) shows the invasive cancer dark in T2 w-scans (arrow), whereas the LCIS is unspecific in T2 w-signal intensity (ellipse). The main hallmark for invasive cancer is a striking “washin” in a typical range for cancers followed by a “washout-phenomenon;” the main hallmark for non-invasive cancer is the asymmetric enhancement in relation to the opposite breast (both the invasive and the non-invasive cancer had not been described by X-ray mammography and ultrasound in this case).

MR-Mammography

MR-Mammography 1131

Part II

Fig. 2. (Continued)

1132 Part II

Medical Uses

Part II

defining feature of tumor angiogenetically based uptake of contrast medium in breast cancer [93–96]. As a consequence of this pathophysiological basis MRM is looking for the type, the amount and the morphology of enhancement within the neighborhood of, and in, a lesion [97–102]. As mentioned above, malignant lesions are characterized by a tumor angiogenetic network of tumor-specific micro-vessels; thus, they show a rapid initial increase of signal intensity in T1 -weighted images (T1 w-images) after the injection of contrast medium within the first 1–2 min after injection, followed by a plateau, or the so-called “washout” effect, i.e. a decrease of signal intensity. This “washout” effect is probably caused by the arterio-venous-shunts within the angio-

genetic network. Most benign tumors show a significantly different type of enhancement, either a slow progressive uptake or a very rapid increase with a further signal rise after the initial enhancement. However, about 20% of benign lesions (especially myxoid fibro-adenomas) might also show a strong initial enhancement followed by a plateau or a washout effect. It is vital that the kinetic analysis is drawn only from the non-necrotic part of the tumor and that the maximum signal increase is used; a tumor contains all types of cells and compartments including haemorrhage, necrosis, and fibrosis besides vital malignant tumor cells. The so-called “wash-in”-phenomenon (i.e. the initial enhancement before the bending of the curve) is the most sensitive

Fig. 3. Fibrous fibroadenoma (41-year-old patient with a lump in the lateral part of the right breast. The left breast is displayed on the right side of the image, the right breast is displayed on the left side of the image). (A) The original dynamic scans show a slow steady area of enhancement in the lateral portion of the right breast (arrow) without any cancer-like washin or washoutphenomenon. (upper left = precontrast, upper right = 1-min-postcontrast, lower left = 2-min-postcontrast, lower right = 7 min postcontrast). (B). The corresponding subtraction images (upper left = subtraction 1-min-postcontrast minus precontrast, upper right = subtraction 2-min-postcontrast minus precontrast, lower left = subtraction 3-min-postcontrast minus precontrast, lower left = subtraction 7-min-postcontrast minus precontrast) display this slow enhancement (arrow) better than the original images. In addition a slight motion artefact in the left breast. (C) The “mosaic-image” (upper left = precontrast, upper right = 1-min-postcontrast, lower left = subtraction 1-min-postcontrast minus precontrast, lower left = T2 w-TSE image in identical sclice positions). The T2 w-scan show the lesion in a dark signal intensity (arrow). Typical signal change of a fibrous fibroadenoma, i.e. a slow, steady enhancement without cancer signs, dark in precontrast and dark in T2 w-scans.

MR-Mammography

MR-Mammography 1133

Part II

Fig. 3. (Continued)

1134 Part II

Medical Uses

Part II

parameter; the “washout” effect (i.e. the late signal decrease following the initial signal increase at the “cancercorner”) is the most specific parameter to diagnose a malignant tumor. In addition to kinetic data, morphological signs have to be included for the distinction between benign and malignant lesions. As malignant tumors often show a “rimenhancement” (i.e. the most enhancing vital tumor areas are in the outer rim of the lesions, because the centre consists mostly of necroses). In contrast to this, benign lesions usually enhance within the centre as well as, and as early as, in the border areas. Malignant tumors often show irregular borders and spiculation, whereas most be-

nign lesions have sharply delineated, often polylobulated margins. Benign lesions might show clear, non-enhancing septations. The vast majority of cancers (98%) show low signal-intensities in the T2 -weighted-Turbo-Spin-Echoimages (T2 w-TSE-images), whereas the strongly enhancing myxoid fibroadenomas are more or less brighter than parenchyma in T2 w-TSE-images. Malignant lesions often show the so-called “bloomingsign”, i.e. an increasing enhancement in the neighborhood outside the lesion [100]. Meanwhile more than 20 different signs in the combined morphological and kinetic analysis enable a highly reliable discrimination between a benign and malignant

Fig. 4. Myxoid or “hyaline” fibroadenoma (22-year-old patient with a lump in the lateral part of the right breast. (The left breast is displayed on the right side of the image, the right breast is displayed on the left side of the images). (A) The original dynamic scans show a 3 × 4 cm multicompartmental area in the right breast; some compartments enhance suddenly and more intensively than cancers, some compartments enhance slower, some compartments show a plateau-phenomenon or even a washout-effect (upper left = precontrast, upper right = 1-min-postcontrast, lower left = 2-min-postcontrast, lower right = 7 min postcontrast). (B) In the corresponding subtraction images (upper left = subtraction 1-min-postcontrast minus precontrast, upper right = subtraction 2-min-postcontrast minus precontrast, lower left = subtraction 3-min-postcontrast minus precontrast, lower left = subtraction 7-minpostcontrast minus precontrast) the partially very strong signal increase after contrast injection is delineated. (C) In the “mosaic-image” (upper left = precontrast, upper right = 1-min-postcontrast, lower left = subtraction 1-min-postcontrast minus precontrast, lower left = T2 w-TSE image in identical sclice positions) some components of this multicompartmenatal lesion are brighter (arrow) than parenchyma in the T2 w-scan, some are dark. Note the typical behavior of myxoid fibroadenomas: Sharp borders and intact septations throughout the dynamic scans, partially very striking enhancement, often much stronger than the range of cancer-like enhancement, no “blooming-effect” and partial brightness in T2 w-images.

MR-Mammography

MR-Mammography 1135

Part II

Fig. 4. (Continued)

1136 Part II

Medical Uses

Part II

lesion. After more than 20 years of clinical research a significant number of publications describe a very high sensitivity for MRM for detecting malignant lesions in the range of more than 95–100%. Even indications which had previously been described as “difficult”, like the identification of Ductal carcinoma in situ (DCIS), are now accepted indications for MRM [99].

Technique High temporal and spatial resolution is important for the detection and distinction between benign and malignant lesions. Today high-resolution dynamic techniques, i.e. the repeated imaging of the same slices before and in short time intervals after injection of contrast are used. The patient is placed in a prone position with the breasts hanging into a bilateral phased-array double breast coil after an iv-cannula had been placed in the cubital vein. The series of sequences in a dynamic MRM examination at 1.5 T includes (further technical details about the MRM examination are described elsewhere [101]: 1. a scout sequence (to clarify the position of the patient and breasts) 2. a coronal T1 -weighted Gradient-Echo-Sequence (T1 wGRE-sequence) in a large field-of-view (FOV) including the axillary region and the neck region (to detect enlarged lymph nodes) 3. a transversal T2 w-TSE sequence. 4. a transversal T1 w-GRE sequence (i.e. the precontrast scan) 5. an iv injection of contrast (0.1 mmol/kg Gd-DTPA) 6. a sevenfold repetition of the precontrast scan (=nr. 4, in identical imaging parameters) every minute for the next 7 min 7. an automatic image subtraction 8. a coronal post-contrast scan (identical scan as nr. 2, so that any lesion in the breasts or the lymph nodes can be evaluated in two pre- and post-contrast orientations) 9. in the case of a breast implant, a silicone-selective sequence is added (focusing on the resonance frequency of silicone, which is 100 Hz lower than the resonance frequency of fat at 1.5 T. The important T1 -weighted dynamic gradient-echosequences before and in short time intervals after the injection of contrast are performed as two dimensional(2D) gradient-echo-sequences (TR/TE/flip-angle = 100 ms/4.7 ms/80◦ ) in our institution. The FOV for scan nr. 4 is 350 mm, the slice thickness 3 mm for 36 slices in a matrix of 512 × 512 in a parallel imaging

data aquisition. The aquisition time for sequence nr. 4 is 60 s. Alternatively a dynamic three dimensional-(3D) gradient-echo-sequence can be performed which allows a better spatial resolution with thinner slices. However, the homogeneity of 3D-sequences in some machines is limited and artefacts caused by inhomogeneity deteriorate the kinetic data evaluation. Another alternative is a dynamic fat-saturationsequence (fat-sat-sequence) which usually also requires a following subtraction. After the measurements, the kinetic data evaluation (signal-intensity-time-curves) in addition to the morphological analysis is made. Modern MR-machines are able to acquire images in 512 × 512 matrices in time distances of each minute. Today parallel imaging techniques in fast gradient-echo-sequences (FLASH, FFE, GRASS) fulfill these prerequisites. Some teams prefer the so-called “fat-sat”-techniques by applying saturation pulses towards fat signal; these techniques suffered from signal inhomogeneities, longer measurement times and often single-breast examinations in the past and therefore could not be successfully applied in Europe [101].

Indications for MRM The scientific results gained so far recommend MRM for the following indications: 1. The definite identification and/or exclusion of breast cancer in cases of unclear mammographic or sonographic breasts. 2. Clarification of number, size, and location of breast lesions, i.e. clarification of multifocality and/or multicentricity; thus enabling more precise planning for treatment. 3. Differentiation between scars and/or recurrent/residual tumors after previous operation/radiation. 4. Delineation of implants, implant-ruptures, and/or lesions around implants like inflammations or benign or malignant tumors. 5. Detection of primary tumors in the case of an unknown primary tumor following a positive lymph node. 6. Delineation of diseases in the contralateral breast before operation/radiation. 7. Detection of the early effect of chemotherapy after the first doses of anticancer drugs. 8. Examination of high-risk patients (family history, genetic risk like BRCA1 and BRCA2).

MR-Mammography

Despite striking results for the sensitivity the value of MRM is still in question because of variations in specificity, the high cost of the machine and examination, its inability to detect micro-calcifications, its role in detecting DCIS, questions about different dosages of contrast etc. As with other MR-examinations, patients with ferromagnetic particles, pacemakers, and operations that have taken place within the weeks prior to the examination, and patients undertaking hormone-replacementtherapy should not be included. The major pitfalls consist of the following: 1. Signal-inhomogeneities in the image due to the rfprofile of the double-breast coil. 2. Signal-inhomogeneities and saturation artefacts in fat-sat images 3. Application of an inappropriate dose and type of contrast medium. 4. Measurement in an inappropriate (“opposed”) echotime where fat and water protons are oriented in different directions. 5. Inappropriate positioning of the breast using high compression of the breast and/or bent arms and lack of control of contrast uptake. 6. Incomplete injection of contrast medium and/or no detection of incomplete injection. 7. Movement of patient during the examination due to insufficient information. 8. Lack of kinetic analysis. 9. No inclusion of T2 -weighted-scans (T2 w-scans) 10. False data evaluation 11. Susceptibility artefacts due to small bleedings because of previous punctures and/or biopsies. 12. Omittance of normal enhancements and “inflow” effects. Since the time intensity curve is (in addition to morphological evaluations) very important for the diagnosis, the signal intensity in the first post-contrast images, in particular, is absolutely crucial for diagnosis. This critical first contrast signal intensity image depends on exact measurement parameters. Injection time, injection site, sodium chloride bolus, length of plastic tube, arm position, tuning parameters are some of the major factors affecting this critical image signal intensity. In our experience an increase of more than 90% in the first 90 s after the beginning of contrast injection (“90– 90 rule”) is a typical threshold for a malignant enhancement at 1.5T-2D-dynamic MRM. To reach this critical threshold measurement using a region of interest procedure (ROI) has to be performed carefully: Only the

“vital” tumor areas show a critical initial enhancement followed by a plateau phenomenon or a washout effect. The inclusion of necrotic areas of the tumor or surrounding normal glandular or fatty tissue will deteriorate and falsify the effect of kinetic data evaluation. It is necessary to look for the fastest enhancing portion of the lesion. It is also important not to include any enhancing focus and to describe it as malignant. Vessels are cut in different orientations in the slice and are therefore displayed in a round, oval or comet shaped manner. The detection of inflow phenomenon, the delineation of further vessels in other slices or other orientations or, if you have still any doubts, an additional MR angiography sequence clarifies the tiny enhancing area as a vessel or not. Fat saturation techniques are in relatively widespread use in the United States. They allow a high spatial resolution, but have a limited value for kinetic data evaluation. Since a pre-pulse in a low bandwidth is necessary, the adjustment parameters in pre- and post-contrast scans are not identical. Current fat-sat techniques usually have a relatively high signal inhomogeneity, so that a quantitative evaluation is difficult and subtractions are difficult, though often necessary. The measurement time is in most cases longer than in non-fat saturation techniques and the “diagnostic window” for the detection of differences between benign and malignant enhancement criteria is restricted. Fat suppression methods may be by temporal subtraction, which is ideal for fat suppression, since only enhancing structures are delineated as high signal intensity lesions. An alternative is fat suppression by pre-pulses with the above-mentioned limitations. The injection of contrast medium is made in a bolus (3 ml/s) either manually or by an automatic injector followed by a post-contrast bolus of physiological sodium chloride of 20 ml (cubital vein) or 30 ml (lower arm/hand). The procedure should be explained to the patient before positioning in the machine. No movement of, or discussion with, the patient is recommended during injection. The dosage of contrast medium should not be higher than 0.1 mmol/kg. A higher dosage shortens the already short “diagnostic window”, i.e. the time difference of only 1– 2 min where the contrast uptake between malignant and benign lesions is sufficiently different in order to enable a differential diagnosis. The evaluation begins by examining fatty tissue on both breasts to assess if the signal intensity of fat is constant. If this signal intensity of fatty tissue as an “internal standard” varies, consider changing receiver adjustment or field inhomogeneity effects or artefacts. In these cases a quantitative evaluation is difficult or impossible, since the technical problems overlap any morphological information.

Part II

Discrepancies and Pitfalls

Discrepancies and Pitfalls 1137

1138 Part II

Medical Uses

Part II

Dynamic MRM should be performed in transversal scans, because in this orientation the identification of vessels, nipple, posterior borders of implants, fatty layer between parenchyma and muscle, signal homogeneity of the coil, and signal correlation with the aorta is possible. However, if the lateral border of the parenchyma towards the axillary region and/or lymph nodes is especially important, a dynamic MRM in coronal scans is advantageous. A sagittal orientation should be avoided because of doubled measurement time and the lack of correlation with the other breast, and the difficulty of detecting inflow phenomenon and hormone effects. This image orientation might be acceptable if sagittal images can be acquired simultaneously from both breasts. If only the axillary region is important, the rotation of the phase encoding direction towards a dorso-ventral orientation is recommended to keep the lateral axillary region free of phase encoding artefacts. It is of crucial importance to use echo times where fat and water protons are in “in-phase” conditions, because Gd has a signal-increasing effect on the water protons in the voxel only in these echo times. Particularly tiny reticular structures like DCIS or lobular cancers can be imaged only in these “in-phase” images. “In-phase” conditions for 1.5 T are even numbers of 2.4 ms, i. e. 4.8, 9.6 ms etc. Odd numbers of 2.4 ms, for example 7.2 ms are “forbidden” because of the “opposed effect” of water and fat protons. This effect is field-strength dependent and should be adjusted according to the field strength which is used. During the dynamic examination it is essential not to change tuning parameters between scans and to guarantee identical pulses in pre- and post-contrast scans. No change of pulses should be allowed between pre- and postcontrast scans. Only under these conditions is a correct signal intensity/time evaluation possible. The pre-contrast signal intensity of parenchyma should be in a typical range for this sequence. If it is too low, this may be caused by an inappropriate receiver adjustment or inappropriate coil design; try to repeat the receiver adjustment. If this pre-contrast signal intensity is too high, it could also be an inappropriate receiver adjustment or a medical/biological reason (bleeding after puncture/biopsy, hormone effect, pregnancy, previous operation, and/or radiation). Diagnosis of breast lesions in MRM is always made by a combined morphological and kinetic analysis. It is the main feature of MRM that tumor typical “tumorangiogenesis” can be detected by signal-intensity-timecurves. A typical malignant lesion shows the striking initial increase (wash-in phenomenon) within the first 1–2 min followed by either a constant signal (plateau phenomenon) or—more specifically—a decrease in signal intensity (washout effect). The sudden increase and the following constancy or washout effect make the “cancer-

corner” of the signal-time-curve. Benign lesions normally show a slower initial increase which is continuously rising over the complete dynamic examination. The distribution of a lesion is described as regional or patchy or diffuse or symmetric; the lesion itself is morphologically described as a focal mass with sharp or non-sharp margins, linear, linear-branched or segmental aspects. The initial and late enhancement is described as being homogeneous, heterogeneous, rim enhancement, bright or dark septations after contrast and centripetal or centrifugal filling effect over time. A centrifugal filling (from inside to outside) is directed mainly towards a benign lesion, a centripetal filling (from outside to inside) towards a malignant lesion. Another major mistake is the inclusion of patients under hormone-replacement-therapy for MRM, because hormone effects cause bilateral, patchy, continuously enhancing spots which are confluent on late scans [103] and thus might be difficult to separate from bilateral DCIS or lobular cancers. However, in most patients DCIS and lobular cancers show unilateral enhancements and, in two-thirds, also a cancer-like wash-in and washout effect.

Future Challenges At present MRM is classified as an additional diagnostic tool in all cases of unclear mammographic and sonographic situations. The medical community needs teaching, training, increased experience, and consensus. MRM will allow a precise detection of cancers in a size as small as 3 mm. For this purpose special techniques providing high spatial and temporal resolution are needed. New specific contrast agents might enable a selective display of specific cancers and differentiate them from other lesions in the breast and in the whole body as well. The combination of MRI and MR-spectroscopy will probably increase the overall accuracy of MRM. Today modern “high-field MR-devices” (high-field for imaging means above 1.0 T) do not allow direct access to the breast during data acquisition. The patient has to be in the isocenter of the magnet field during imaging and must be moved outside the machine for a following intervention. The position of a wire marker, a core biopsy or a laser fibre has to be checked by a following MRimage, thus increasing measurement time, artefacts, and pitfalls. A MR-compatible robotic system for simultaneous imaging and immediate biopsy/intervention at high field strength is in clinical testing. Combining MRI and therapy in one session—at least for lesions smaller than 2 cm—might enable a “onestop-shop” procedure of detection and simultaneous removal in one single examination [104–105]. Preliminary

MR-Mammography

References 1. Damadian R. Science 1971;171:1151–1153. 2. Lauterbur PC. Nature 1973;242:190–191. 3. Mansfield P, Morris PG, Ordidge R. et al. Brit. J. Radiol. 1979;52:242–243. 4. Ross RJ, Thompson JS, Kim K, Bailey RA. Radiology 1982;143:195–205. 5. El Yousef SJ, Alfidi RJ, Duchesneau RH. J. Comput. Assist. Tomogr. 1983;7:215–218. 6. El Yousef SJ, Duchesneau RH, Alfidi RJ, Haaga JR, Bryan PJ, LiPuma JP. Radiology 1984;150:761–766. 7. Kaiser W. Arch. Int. de Physiol. et de Biochim. 1985;93:67–76. 8. Fritschy P, M¨uller E, Sauter R, Kaiser W. Radiology 1984;153(P):243–244. 9. Stelling CB, Wang PC, Lieber A, Mattingly SS, Griffen WO, Powell DE. Radiology 1985;154:457–462. 10. Kaiser W, Zeitler E, Fortschr. R¨ontgenstr. 1986;144.5:459– 465. 11. Kaiser W, Zeitler E, Fortschr. R¨ontgenstr. 1986;144.5:572– 579. ¨ 1985;143:207– 12. Heywang SH, Fenzl G, Edmaier M, ROFO 212. 13. Weinmann HJ, Laniado M, M¨utzel W. Physiol. Chem. Phys. Med. NRM. 1984;16:167–172. 14. Eskin BA, Parker JA, Bassett JG, George DL. Obstet. Gynecol. 1974;44:398–492. 15. Chang CHJ, Sibala JL, Lin F, Fritz SL, Gallagher JH, Dwyerill SJ, Templeton AW. Am. J. Roentgenol. 1978;131:459– 464. 16. Heywang SH, Hahn D, Schmidt H, Krischke I, Eiermann W, Bassermann RJ, Lissner J. J. Comput. Assist. Tomogr. 1986;10:199–204. 17. Haase A, Frahm J, D Matthaei et al. J. Magn. Res. 1986;67:258–268. 18. Kaiser WA, Oppelt A, KST der Mamma mit Schnellbildverfahren. MR’87, 2. Internationales Kernspintomographie Symposium, 29.1.–1.2.87, Garmisch-Partenkirchen, Referateband, 303–310, Schnetztor-Verlag, Konstanz (1987). 19. Kaiser WA, Zeitler E. Radiology 1989;170:681–686. 20. Heywang SH, Wolf A, Pruss E, Hilbertz T, Eiermann W. Radiology 1989;171:95–103. 21. Kaiser WA, Kess H. Fortschr. R¨ontgenstr. 1989;151:103– 105. 22. Kaiser WA. Eur. Front. Radiol. 1990;7:39–68. 23. Ercolani P, Giovagnoni A, Giuseppetti G, Radiol. Med. Torino. 1991;82:422–426. 24. Hachiya J, Seki T, Okada M, Radiat. Med. 1991;9:232–240. 25. Hess T, Knopp MV, Brix G, Zentralbl. Radiol. 147, 969 (1993). 26. Stack JP, Redmond OM, Codd MB. Radiology, 1990;174:491–494.

27. Boetes C, Strijk SP, Holland R, Barentsz JO, Van Der Sluis RF, Ruijs JH. Eur. Radiol. 1997;7:1231–1234. 28. Bone B, Aspelin P, Bronge L, Isberg B, Perbeck L, Veress B, Acta. Radiol. 1996;37(2):208–213. 29. Buadu LD, Murakami J, Murayama S et al. Radiology 1996;200:639–649. 30. Daldrup HE, Roberts TP, Muhler A, Gossmann A, Roberts HC, Wendland M, Rosenau W, Brasch RC, Radiologe 1997;37(9):733–740. 31. DeAngelis GA, de Lange EE, Miller LR, Morgan RF. Radiographics 1994;14:783–94. 32. el Kwae EA, Fishman JE, Bianchi MJ, Pattany PM, Kabuka MR. J. Digit. Imaging. 1998;11(2):83–93. 33. Farria DM, Gorczyca DP, Barsky SH, Sinha S, Bassett LW. Am. Roentgenol. J. 1996;167(1):187–189. 34. Fischer U, von-Heyden D, Vosshenrich R, Vieweg I, Grabbe E. Rofo. 1993;158(4):287–292. 35. Fobben ES, Rubin CZ, Kalisher L, Dembner AG, Seltzer MH, Santoro EJ, Radiology 1995;196(1):143–152. 36. Frankel SD, Sickles EA, Radiology 1997;202(3):633– 634. 37. Gilles R, Guinebretiere JM, Lucidarme O et al. [published erratum appears in Radiology 1994 Oct;193(1):285], Radiology 1994;191:625–631. 38. Gilles R, Guinebretiere JM, Shapeero LG et al. Radiology 1993;188:473–478. 39. Gilles R, Zafrani B, Guinebretiere JM et al. Radiology 1995;196:415–419. 40. Gorczyca DP, DeBruhl ND, Ahn CY, Hoyt A, Sayre JW, Nudell P, McCombs M, Shaw WW, Bassett LW. Radiology 1994;190(1):227–232. 41. Gorczyca DP. Radiology 1993;186(3):906–907. 42. Graham RA, Homer MJ, Sigler CJ et al. Am. J. Roentgenol. 1994;162:33–36. 43. Gribbestad IS, Nilsen G, Fjosne HE, Kvinnsland S, Haugen OA, Rinck PA. J. Magn. Reson. Imaging 1994;4(3), 477– 480. 44. Harms SE, Flamig DP. Magn. Reson. Imaging. Clin. Am. N. 1994;2(4):573–584. 45. Hochman MG, Orel SG, Powell CM, Schnall MD, Reynolds CA, White LN. Radiology 1997;204:123–129. 46. Hoffmann U, Brix G, Knopp MV, Hess T, Lorenz WJ, Magn. Reson. Med. 1995;33(4):506–514. 47. Hylton NM, Frankel SD, Magn. Reson. Imaging. Clin. Am. N. 1994;2(4):511–525. 48. Knopp MV, Brix G, Junkermann HJ, Sinn HP, Magn. Reson. Imaging. Clin. Am. N. 1994;2(4):633–658. 49. Kuhl CK, Kreft BP, Hauswirth A, Elevelt A, Steudel A, Reiser M, Schild HH. Rofo. 162(5):381–389 (1995). 50. Kuhl CK, Seibert C, Sommer T, Kreft B, Gieseke J, Schild HH. Rofo. 1995;163(3):219–224. 51. Kuwabara M, Nippon Igaku Hoshasen Gakkai Zasshi, 1991;51(11):1366–1374. 52. Morris EA, Schwartz LH, Dershaw DD, van Zee KJ, Abramson AF, Liberman L. Radiology 1997;205(2):437–440. 53. Muller RD, Barkhausen J, Sauerwein W, Langer R, J. Comput. Assist. Tomogr. 1998;22(3):408–412. 54. Muller-Schimpfle M, Ohmenhauser K, Stoll P, Dietz K, Claussen CD. Radiology 1997;203(1):145–149. 55. Mumtaz H, Hall-Craggs MA, Davidson T et al. Am. Roentgenol. J. 1997;169:417–424.

Part II

tests applying laser therapy, radio frequency therapy, or cryotherapy have been promising so far [106–107] as has receptor-specific near-infrared-imaging in vitro [108].

References 1139

1140 Part II

Medical Uses

Part II

56. Nunes LW, Schnall MD, Siegelman ES et al. Am. J. Roentgenol. 1997;169:409–415. 57. Orel SG, Hochman MG, Schnall MD, Reynolds C, Sullivan DC, Radiographics 1996;16:1385–1401. 58. Orel SG. Semin Ultrasound MCT R. 1996;17(5):476– 493. 59. Pedevilla M. Radiology 1997;205(2):580–581. 60. Peller M, Stehling MK, Sittek H, Kessler M, Reiser M, MAGMA 1996;4(2):105–113. 61. Piccoli CW. Magn. Reson. Imaging. Clin. Am. N. 1994;2(4):557–571. 62. Pierce WB, Harms SE, Flamig DP, Griffey RH, Evans WP, Hagans JE. Radiology 1991;181(3):757–763. 63. Rieber A, Zeitler H, Rosenthal H, Gorich J, Kreienberg R, Brambs HJ, Tomczak R. Br. Radiol. J. 1997;70(833):452– 458. 64. Rodenko GN, Harms SE, Pruneda JM et al. Am. Roentgenol. J. 1996;167:1415–1419. 65. Seki T, Hachiya J, Nitatori T, Yokoyama K, Fukushima H, Uchigasaki S, Nippon Geka Gakkai Zasshi 1996;97(5): 347–56. 66. Soderstrom CE, Harms SE, Copit DS, Evans WP, Savino DA, Krakos PA, Farrell RS Jr, Flamig DP. Radiology 1996; 201(2):427–432. 67. Soo MS, Kornguth PJ, Walsh R et al. J. Magn. Reson. Imaging. 1997;7:724–730. 68. Stelling CB. Radiol. Clin. North. Am. 1995;33(6):1187– 11204. 69. Van Goethem M, Van Breusegem L, Ceulemans R, De Schepper A, de Moor J. J. Belge. Radiol. 1995;78(1):6– 10. 70. Weinreb JC, Newstead G, Controversies in Breast MRI 1994;10(2):67–83. 71. Weinreb JC, Newstead G. Radiology 1995;196(3):593– 610. 72. Kaiser WA, Fortschr. R¨ontgenstr. 1996;165(5):425–427 (Editorial). 73. Hussman K, Renslo R, Phillips JJ, Fischer HJ, Khalkhali I, Braslau DL, Sinow RM. Work in Progress, Radiology. 1993;189:915–917. 74. Kuhl CK, Elevelt A, Leutner CC, Gieseke J, Pakos E, Schild HH. Radiology. 1997;204:667–675. 75. Thiele J, Schneider JP, Franke P, Lieberenz S, Schmidt F. Fortschr R¨ontgenstr. 1998;168:374–379. 76. Doler W, Fischer U, Metzger I, Harder D, Grabbe E. Radiology 1996;200:863–864. 77. Mahfouz AE, Rahmouni A, Zylbersztejn C, Mathieu D, Am. Roentgenol. J. 1996;167:167–169. 78. Wurdinger S, Noras H, Straube K, Michaelsen S, Kaiser WA, ECR‘97, Eur. Congr. Radiol. 2–7. March 1997, European Radiology 1997;7. S243. 79. de Souza NN, Kormos DW, Krausz T, Coutts GA, Hall AS, Burl M, Schwieso JE, Puni R, Vernon C. J. Magn. Reson. Imaging. 1995;5:525–528. 80. Silverman SG, Collick BD, Figueira MR, Khorasani R, Adams DF, Newman RW, Topulos GP. Radiology. 1995;197:175– 182. 81. Pfleiderer SO, Reichenbach JR, Azhari T, Marx C, Wurdinger S, Kaiser WA, Invest. Radiol. 2003;Jan 38(1):1–8. 82. Knopp MV, Weiss E, Sinn HP, Mattern J, Junkermann H, Radeleff J, Magener A, Brix G, Delorme S, Zuna I, G van

83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101.

102.

103. 104.

105.

Kaick. J. Magn. Reson. Imaging. 1999;Sep 10(3):260– 266. Brix G, Schreiber W, Hoffmann U, Guckel F, Hawighorst H, Knopp MV. Radiologe 1997;Jun 37(6):470–480. Buckley DL, Drew PJ, Mussurakis S, Monson JR, Horsman A. J. Magn. Reson. Imaging. 1997;May to Jun 7(3):461– 664. Algire GH. J. Int. Chir. 1953;Jul to Aug 13(4):381–384 (includes translations). Gimbrone MA Jr, Leapman SB, Cotran RS, Folkman J. J. Exp. Med. 1972;Aug 1, 136(2):261–276. Folkmann J, Long DM Jr, Becker FF. 1963;Apr 16:453– 467. Ottinetti A, Sapino A. Breast Cancer Res. Treat. 1988;Jul 11(3):241–248. Samejima N, Yamazaki KA. Jpn. J. Surg. 1988;May 18(3):235–242. Folkman J. Adv. Cancer. Res. 1985;43:175–203. Weidner N, Semple JP, Welch WR, Folkman J, Engl N. Med J. 1991;Jan 3, 324(1):1–8. Jensen HM, Chen I, DeVault MR, Lewis AE. Science 1982;Oct 15, 218(4569):293–295. Port RE, Knopp MV, Brix G. Magn. Reson. Med. 2001;Jun, 45(6):1030–1038. Lucht RE, Knopp MV, Brix G. Magn. Reson. Imaging 2001;Jan 19(1):51–57. Knopp MV, Himmelhan N, Radeleff J, Junkermann H, Hess T, Sinn HP, Brix G. Radiologe 2002 Apr ;42(4):280– 290. Buckley DL, Drew PJ, Mussurakis S, Monson JR, Horsman A. J. Magn. Reson. Imaging 1997;May to Jun 7(3), 461– 464. Kaiser WA. Magnetic-Resonance-Mammography (MRM), Springer-Verlag, 1993. Kaiser WA. In: PL Davis (Ed). MRI-Clinics of North America, Vol 2.4, Saunders, Philadelphia, 1994, pp 539– 555. H Neubauer, Li M, Kuehne-Heid R, Schneider A, Kaiser WA. Br. Radiol. J. 2003 Jan;76(901):3–12. Fischer DR, Baltzer P, Malich A, Wurdinger S, Fresmeyer MG, Marx C, Kaiser WA. Eur. Radiol. 2004 Mar; 14(3):394– 401, Epub 2003 Sep 27 (2004). Kaiser WA. Breast, Unit 21. In: EM Haake, W Lin, NYC Cheng, CP Ho, WA Kaiser, JS Lewin, ZP Liang, SK Mukhery, RC Semelka, KR Thulborn, PK Woodard (Eds). Current Protocols in Magnetic Resonance Imaging. Wiley & Sons: New York, 2004. Ikeda DM, Hylton NM, Kinkel K, Hochman MG, Kuhl CK, Kaiser WA, Weinreb JC, Smazal SF, Degani H, Viehweg P, Barclay J, Schnall MD. J. Magn. Reson. Imaging 2001;Jun 13(6):889–895. Pfleiderer SO, Sachse S, Sauner D, Marx C, Malich A, Wurdinger S, Kaiser WA. Breast Cancer Res. 6(3):R232– 238, Epub, Mar 16 (2004). Kaiser WA, Selig M, Jung R, Vagner J, S Sch¨onherr: Patentanmeldung: Manipulator f¨ur einen geschlossenen Magnetresonanztomographen (MRT). Patent Application Nr. 19941019.4; 28.08.1999, German Patent Office. Pfleiderer SO, Reichenbach JR, Azhari T, Marx C, Malich A, Schneider A, Vagner J, Fischer H, Kaiser WA. 2003 Apr;17(4):493–498.

MR-Mammography

Internet Resources http://www.mediteach.de: List of detailed courses in MRM. http://www.mrisafety.com: List of safe or unsafe materials for a quick overview.

Part II

106. Pfleiderer SO, Freesmeyer MG, Marx C, Kuhne-Heid R, Schneider A, Kaiser WA. Eur. Radiol. Dec 12(12):3009– 3014, Epub 2002 Jun 21. 107. Jeffrey SS, Birdwell RL, Ikeda DM, Daniel BL, Nowels KW, Dirbas FM, SM Griffey. Arch. Surg. 1999 Oct.;134(10): 1064–1068. 108. Hilger I, Leistner Y, Berndt A, Fritsche C, Haas KM, Kosmehl H, Kaiser WA. Eur. Radiol. 2004 Jun 14(6); 1124– 1129, Epub 2004, Apr 30.

Internet Resources 1141

1143

Geoffrey S Payne, Nada Al-Saffar, and Martin O Leach Cancer Research UK Clinical Magnetic Resonance Research Group, Royal Marsden Hospital and the Institute of Cancer Research SUTTON, Surrey, SM2 5PT, UK

Features of 31 P MRS in Tissues General Properties of 31 P 31

P is one of the most sensitive of the NMR nuclei. It has spin 1/2, a gyromagnetic ratio of 1.083 × 108 rad/s/T, and a natural abundance of 100%, giving a sensitivity of 6.63% relative to 1 H. Since many metabolites of biochemical interest are phosphorylated, 31 P MRS is a useful probe of tissue biochemistry. Spectra in vivo are relatively uncrowded as the number of 31 P-containing metabolites of sufficient concentration are relatively few. Properties of the individual metabolites of major interest are given below, while the abbreviations used for the various metabolites referred to in this chapter are given in Table 1. Note that there are nucleoside triphosphates (NTP) in addition to adenosine triphosphate (ATP), so in general we have followed the convention of referring to the corresponding peaks as NTP rather than as ATP. Spectral Dispersion. In general 31 P gives rise to spectral peaks covering a wide range of frequencies. However in tissues most peaks of interest lie within a 30 ppm range (see Figure 1). Chemical shift is often referenced relative to phosphocreatine (PCr) at 0 ppm in vivo, or to H3 PO4 in tissue extracts in vitro. Dependence of Chemical Shift on pH. Many 31 P metabolites have a pKa in the range that makes the chemical shift dependent on tissue pH. The classic example of this is the equilibrium between the two forms of inorganic phosphate (Pi) in tissue [1]; + H2 PO4 − ⇔ HPO2− 4 +H

pKa ∼ = 6.7

Since this is a dynamic equilibrium between two forms of phosphate with different chemical shifts, the net shift of the observed Pi peak depends on the relative concentrations of the two species and hence on the hydrogen ion concentration (pH). Measurement of the chemical shift of the Pi peak relative to a peak that does not shift significantly with pH (usually PCr is used) Graham A. Webb (ed.), Modern Magnetic Resonance, 1143–1161.
C 2008 Springer.

provides a non-invasive way to measure intracellular pH, as most phosphate is intracellular. Some other 31 P metabolites also demonstrate sensitivity to pH, in particular phosphocholine (PC) and phosphoethanolamine (PE) (see Figure 2). Other Factors Affecting Chemical Shift. Chemical shift is also affected by factors such as ionic strength and metal ion binding. In particular the β and γ peaks of ATP are sensitive to the magnesium concentration in the tissue [2,3]. Homonuclear J-Coupling. The main example of homonuclear coupling is found in NTP. In ATP this leads to αand γ-ATP being observed as doublets, while β-ATP is a triplet, with a coupling constant of about 16 Hz. Homonuclear decoupling is not generally required. However, allowance for coupling is necessary when measuring the T2 of coupled metabolites, such as ATP [4]. Heteronuclear J Coupling. The phosphomonoesters (PE, PC) have weak (6–7 Hz) 3-bond coupling to 1 H nuclei, forming a triplet, while the phosphodiesters couple to two-proton pairs, producing quintets (Table 1). Nearly all preclinical systems and a few of the more advanced clinical scanners have the facility to perform 1 H-decoupling to collapse these multiplets back to singlets (see below). A few research systems can also perform more complex double resonance measurements for further sensitivity enhancement.

Metabolites Observed in Tissues Most of the metabolites detectable with 31 P MRS in tissues can be seen in Figure 1, while the structures are shown in Table 1. Metabolites are involved in either energy production (high energy phosphates—ATP, PCr, sugar phosphates; NADH) or in membrane metabolism (phosphomonoesters—PC, PE; phosphodiesters—GPC, GPE; membrane phospholipids). The reactions involved are summarized in Figure 3. For more details a biochemistry textbook should be consulted [5,6].

Part II

Phosphorus Magnetic Resonance Spectroscopy on Biopsy and In Vivo

1144 Part II

Medical Uses

Part II

Table 1: Summary of main metabolites detected by 31 P MRS in vivo

Metabolite

Abbreviation

Phosphoethanolamine

PE

O

5–7

7.19

H + N

O

4.2–6.5

6.25

3.5–5.5

Singlet

H

OH O P OH O

Inorganic phosphate

N

O P

O PC

J couplinga (Hz)

Structure

O

Phosphocholine

Chemical shift relative to PCr (ppm)

OH P

Pi

O

O H2COGlycerophosphoethanolamine

GPE

O-

O-

CH

P

H2CO

3.6

b

2.96

6.03

OCH2CH2NH2

O H2COGlycerophospocholine

GPC

O-

CH O-

H2CO P

OCH2CH2N+(CH3)3

O O- H Phosphocreatine

PCr

CH3

O = P ∼ N - C - N - CH2 - C

O ATP

Adenosine

O P

O

0

Singlet

O

NH

OAdenosine triphosphate

O-

O P

-O O

OH P

-OO

-O

α-7.5 ppm β-16–19c γ -2–7.5c

Notes: a The coupling quoted is the main 3 JPH coupling. Additional JHH couplings are often present. [138]; b This was not measured. It will be similar to that of GPC. c Depends on magnesium concentration.

Additional Notes on Specific Metabolites Phosphomonoesters—phosphocholine and phosphoethanolamine. These metabolites are involved in membrane synthesis and degradation (Figure 3B [7]). PC is also involved in cell signalling [8,9]. The chemical shifts of both are pH sensitive (Figure 2). Phosphodiesters—glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE). These are produced by the action of phospholipases A1 and A2 , or phospholi-

pase B on the corresponding phospholipids (Figure 3C). The chemical shift does not depend on pH. When interpreting a “PDE” peak one must be careful to distinguish phosphodiesters from the peak of the broader phospholipid component (see below). Broad Resonances from Phospholipids and Bone. In unlocalized spectra of brain a large broad (∼750 Hz

31 P

MRS on Biopsy and In Vivo

Features of 31 P MRS in Tissues 1145

PME 1500

1000

NTP

γ

α

β

500 Pi

PDE

0

15

10

5

−5 −10 −15 0 Chemical Shift (ppm)

−20

width) symmetric resonance is observed due to immobile phosphates in bone [10]. In localized spectra of brain, liver, and kidney a broad asymmetric resonance is observed that decreases in amplitude with increasing magnetic field strength, and therefore suggesting a strong influence of chemical shift anisotropy. It has been demonstrated that the peak is produced by membrane lipid

−25

bilayers and that it may be eliminated by off-resonance presaturation [11]. Nucleoside Triphosphates. In most systems the dominant contribution is from ATP. As mentioned above, the chemical shifts of the β and γ peaks are sensitive to magnesium concentration.

Fig. 2. Chemical shift of several metabolites relative to PCr as a function of pH (calculated from .[137]).

Part II

Fig. 1. Example of a 31 P MR spectrum from a non-Hodgkins lymphoma in the groin of a patient, measured at 1.5 T. Note the absence of PCr at 0 ppm.

1146 Part II

Medical Uses

Metabolite Concentrations and Relaxation Times Some published measurements of concentrations and relaxation times of various metabolites in different tissues are summarized in Tables 2 and 3. 31

P MRS of Tissue Biopsy Samples

Methods

B

Glucose Choline choline kinase

G-3-P Glycerol

PC

GPC PA

PC cytidylyl transferase

PLA1

PLD

PL

C

lysoPL

LysoPtdcho

1,2-DAG

CDP-choline TAG

Ptdcho

Syn t hes i s

ada t i on

DHAP

Deg r

Part II

structures leading to rapid transverse relaxation such that the peaks become invisible by NMR.

A

Extraction Methods The type of extraction method to be selected depends greatly on the nature of the investigation and the analytical methods involved. To analyze cellular lipids and/or their water-soluble metabolites, it is necessary to use extraction procedures that quantitatively recover the metabolites of interest. Currently available methods include the following. Perchloric Acid (PCA). This is widely used for extraction of water-soluble cell components [12]. Modified Folch’s Method. The extraction of cellular lipids using organic solvents, usually chloroform and methanol [12].

PC transferase

C

A Combined Extraction Technique. The above two methods performed on the same cell batch [13]. The Dual Phase Extraction Method (DPE). The recovery of both the lipid and water-soluble fractions from the same biological sample by a single extraction method [12].

Fig. 3. Simplified biochemical pathways of main compounds visualized by 31 P MRS in tissues. (A) Metabolism of high energy phosphates. (B) Synthesis and degradation pathways of main membrane phospholipid. (C) Sites of action of the phospholipases on phosphatidylcholine. Phospholipase B can also bring about the successive removal of both fatty acids. (See also Plate 90 on page XVI in the Color Plate Section.)

Adenosine Diphosphate (ADP). The concentration measured by NMR is much lower than that measured biochemically after freeze-clamping of the tissue. This suggests that much of the ADP is tightly bound to larger

Chemical Shift and Concentration Referencing for 31 P MRS The separation of resonance frequencies from an arbitrary chosen reference frequency is termed the “chemical shift”, and is expressed in terms of the dimensionless unit parts per million (ppm). Depending on the phase of the extracts and the type of nucleus detected, different solvent systems and references (internal or external) are used. Aqueous Fraction. An example of stock solution used is D2 O with 10 mM ethylenediaminetetraacetate (EDTA) adjusted to pH 8.2 with tris (hydroxymethyl)methylamine (Tris) (0.5 M, pH 10.4), and methylenediphosphonic acid (MDPA) (δ = 16.8 ppm) calibrated against GPC (δ = 0.49 ppm @ pH 7.2) for referencing.

Liver (alcoholic hepatitis) Liver (alcoholic cirrhosis) Liver (viral hepatitis) Heart Heart 1.1 ± 0.3 2.1 ± 0.5

0.7 ± 0.5 1.9 ± 0.6

NMR

NMR

2.2 2.0 0.9 3.3 0.6 ± 0.3

.63 ± .20 1.05 ± .42 .60 ± .23 1.09 ± .36

.69 ±. 18 .42 ± .16 .68 ±. 24 .34 ±. 18

10 2.7 2.6 3.6 2.9 4.7 3.8 2.9 2.1 1.9 1.4 ± 0.4

.83 ± .22 1.14 ± .40

Bioch Bioch Bioch NMR NMR NMR NMR NMR NMR NMR NMR NMR NMR

NMR NMR

GPC

.74 ±. 30 1.15 ±. 43

GPE

.94 ± .16 .46 ±. 17

2.5

1.0 ± 0.2 1.3 ± 0.3 1.5 ± 0.5 1.2 ± 0.2 0.9 ± 0.1

3.1 ± 0.4 4.3 ± 0.8 3.6 ± 0.8 2.3 ± 0.4 0.9 ± 0.3

NMR NMR NMR NMR NMR

.81 ± .21 .46 ± .14

0.66 ± 0.25 1.6 1.0 ± 0.2 1.0 ± 0.2

2.32 ± 0.50 2.6 3.2 ± 0.3 4.0 ± 0.6

NMR NMR NMR

Kidney Kidney Brain cerebrum Brain— cerebellum Brain cortical GM Brain cortical WM Brain Brain Brain tumor Brain (left parietooccipital cortex) Brain (right parietoocciptal cortex) Left thalamus Right thalamus Muscle Muscle Muscle Muscle Muscle Muscle Muscle Muscle Muscle Muscle Liver

PC

Pi

PE

PME

Meth

Tissue

5.4 ± 2.0

3.3 ± 1.8

3.8 5.4 8.4 2.9 ± 1.1

4.2

10.1 ± 2.5 14.2 ± 2.6 10.2 ± 1.9 7.0 ± 0.8 2.3 ± 0.4

2.14 ± 0.91 4.9 10.8 ± 2.0 9.4 ± 3.4

PDE

0.93 ± .56 .74 ±. 48

1.26 ± .78

1.54 ± .95

MP

Table 2: Some measured concentrations of metabolites detected by 31 P MRS in different human tissues (units of mM)

2.8

1.65

5.5 5.4 5.5 5.5 (5.5) (5.5) 8.2 6.3 4.9 5.7 2.2 2.9 1.5

2.9 ± 0.2 2.9 ± 0.3 2.1 ± 0.5 2.3 ± 0.2 1.3 ± 0.2

1.6 ± 0.26 2.0 2.9 ± 0.2 2.6 ± 0.2

ATP (NTP)

8.82 ± 1.30 5.69 ± 1.02 9.0 ± 1.2 5.3 ± 1.2

17.4 17.3 17.4 25.6 23.9 24.7 17.7 32.0 24.5 22.0

3.1 ± 0.3 2.9 ± 0.4 3.8 ± 0.7 3.1 ± 0.5 1.2 ± 0.1

None None 2.9 ± 0.3 3.9 ± 0.4

PCr

[119] [173]

[151]

[151]

[141] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [67] [69] [67] [151]

[141]

[71] [71] [139] [140] [140] [141]

[95] [67] [71] [71]

Reference

Muscle Rhabdo Myosarcoma Muscle Heart Liver Liver Liver Liver

Brain (infants) Brain (infants) Muscle Muscle Muscle Muscle Muscle Muscle Muscle

Brain Brain Brain (neonate)

Tissue

3.6 ± 0.4

3.3 ± 0.2 3.5 ± 0.4 4.2 ± 0.5 4.0 ± 0.5 5.4 ± 1.7 4.6 ± 0.5 4.0 ± 0.4 4.7 ± 0.3 4.1 ± 0.5 5.3 ± 0.5 4.0 ± 0.8 3.9 ± 1.9 2.9 ± 1.3

3.20.5

3.7 ± 1.6 5.2 ± 3.9

1.6 ± 0.4 0.84 ± 0.26 0.74 ± 0.1

1.5 1.5 1.5 1.5 2.0 2.0 1.5 1.5 1.5 1.5 1.5 1.5 1.5

3T 1.5T 1.5 1.5 2.0 1.9 0.41 ± 0.1 0.8 ± 0.2 0.97 ± 0.25 0.44 ± 0.06

5.2 ± 1.0

2.0 ± 0.6

2.2 ± 0.6

4.9 ± 0.6

2.35

5.0 ± 1.0 1.4 ± 0.13 1.9 ± 0.4 1.36 ± 0.37

3.1 ± 1.9 2.0 ± 0.3

2.1 ± 0.7 1.6 ± 0.2

3.4 ± 1.1 1.8 ± 0.5

6.4 ± 1.0 2.9 ± 0.6

2.35 2.35

2.0 1.7 ± 0.4

PDE

2.5 1.8 ± 0.6

Pi

4.0 4.9 ± 1.0

PME

1.9 T 2.0T 2.35

Field strength

6.4 ± 0.2 6.1 ± 0.5 none none none

4.6 ± 0.2 5.0 ± 0.6 6.1 ± 0.3 5.6 ± 0.5 6.0 ± 0.5 6.5 ± 1.1 6.5 ± 0.1 6.5 ± 0.7 6.6 ± 0.5 6.9 ± 0.6 5.5 ± 0.2 5.8 ± 0.7 1.3 ± 0.3

3.7 ± 1.1

4.5 ± 0.3 5.4 ± 0.5 0.43 ± 0.18 0.5 ± 0.2 0.35 ± 0.06 0.33 ± 0.04

2.9 ± 0.4 4.1 ± 0.4 4.6 ± 0.3 4.5 ± 0.4 3.5 ± 1.0 4.8 ± 0.6 4.3 ± 0.5 3.9 ± 1.3 5.0 ± 0.7 5.0 ± 0.5 4.8 ± 0.5 3.4 ± 0.1 1.3 ± 0.4

2.2 ± 0.6

2.2 ± 0.6 1.5 ± 0.3

1.2 ± 0.2 0.7 1.3 ± 0.3

4.0 ± 0.4 3.0 4.4 ± 1.2 3.9 ± 1.0 3.2 ± 0.3

γ -NTP

PCr

2.6 ± 0.9 5.0 ± 0.5 0.68 ± 0.09 0.9 ± 0.3 0.46 ± 0.09

2.2 ± 0.3 2.9 ± 0.5 3.2 ± 0.5 3.4 ± 0.6 3.9 ± 0.8 3.5 ± 0.6 3.2 ± 0.2 4.2 ± 0.6 3.8 ± 0.5 3.2 ± 0.5 3.6 ± 0.5 2.8 ± 0.3 2.1 ± 0.4

2.3 ± 0.8

2.7 ± 1.0 1.7 ± 0.3

0.7 1.5 ± 0.4

α-NTP

3.5 ± 1.1 5.8 ± 1.0 0.40 ± 0.13 0.5 ± 0.1 0.35 ± 0.05

2.7 ± 0.3 3.6 ± 0.3 3.7 ± 0.6 3.8 ± 0.8 3.9 ± 0.8 3.6 ± 0.8 4.1 ± 0.3 4.1 ± 1.3 3.0 ± 0.5 4.1 ± 0.4 4.3 ± 0.6 2.8 ± 0.4 1.6 ± 1.1

2.5 ± 0.7

1.7 ± 0.6 1.3 ± 0.4

1.0 1.6 ± 0.5

β-NTP

[162] [163,164] [158] [67] [69] [165]

[67] [154] [155] [155] [156] [156] [157] [158] [150] [159] [160] [161] [161]

[153]

[153] [153]

[152] [56] [153]

Reference

Part II

Table 3A: Published values of T1 relaxation times in different human tissues

1148 Part II Medical Uses

31 P

MRS on Biopsy and In Vivo

31 P

MRS OF Tissue Biopsy Samples 1149

Tissue

Field strength

Muscle Muscle Muscle Muscle Muscle Brain Brain

1.5 1.5 1.5 1.5 3T 1.5 2.0

PME

Pi

PDE

240 ± 48 205 ± 14 148 ± 17 70

80

20

PCr

γ−NTP

α−NTP

β−NTP

425 ± 21 61a 95b 424 ± 21 334 ± 30

93 ± 3 66a 74b (16 ± 5)c 78 ± 13 89 ± 9 30c

74 ± 1 69b 75b (22 ± 6)c 55 ± 7 84 ± 6 30c

75 ± 2

150

(8 ± 2)c 55 ± 10 62 ± 3 20c

NTP

Reference [4] [166] [4] [167] [162] [168] [56]

Notes: a Spin echo with TE = 1/J; b Selective echo 90-TE /2-2(6)6(2)-Te/2-acquire; c These measurements have not allowed for modulation owing to homonuclear J coupling.

Lipid Fraction. A 2:1 mixture of CDCl3 and 40 mM methanolic EDTA (200 mM EDTA in water adjusted to pH 6.0 with CsCl, and further diluted fivefold with absolute methanol) [14]. Examples of internal references are: GPC (δ = −0.13 ppm) which is soluble in chloroform solutions, although it can be extracted with water and triphenylphosphine oxide (δ = 33.5 ppm when PtdCho δ = 0.0 ppm) or phosphocholine (δ = 3.5 ppm when PtdCho δ = 0.0 ppm) which is used as an external reference. In all cases, metabolic contents are quantified by integration, normalized relative to the reference peak, and corrected for the number of live cells extracted and cell size in cases when treatment is expected to modify cellular size. Acquisition Hardware and Condition Measurements are usually performed in vertical superconducting magnets, with field strengths in the range 7–14 T. Manufacturers provide a wide range of RF probes, that are usually equipped with a deuterium lock to ensure the field does not drift over the time of the measurement, and with a separate 1 H coil in order to perform 1 H decoupling. Most measurements will use 5 mm NMR tubes, although 2.5 mm and even 1 mm tubes are becoming available. Acquisition conditions will depend to some extent on the system and concentration being studied. The following conditions are an example of those used for cellular extracts of 0.5 − 2.0 × 108 cells measured on a 500 MHz spectrometer [15,16]. Some example spectra are given in Figure 4. Spectra are acquired at room temperature with proton broad-band decoupling to eliminate 1 H-31 P NMR multiplets (decoupler power “on” during acquisition and “off” during the interpulse delay in order to prevent the building of NOE enhancement), spectral width of 100 ppm (aqueous) and 20 ppm (lipid), 32K data points and 4K transients accumulation. The sample spinning rate is 20 Hz. Exponential multiplication to give a line-broadening of 1–2 Hz is applied before Fourier transformation.

Spectral Analysis Most spectrometers include software for processing and analysis of data. For off-line processing several packages are available, such as MestRe-C software (available free from Javier Sardina Research Group, University of Santiago de Compostela, Spain (http://qobrue.usc.es/jsgroup/ MestRe-C/MestRe-C.html)). In addition to the measurement of peak areas directly, when large numbers of samples are available it is possible to look for patterns characteristic of different sample groups using automated pattern recognition techniques [17]. MRS of Intact Tissue Samples While virtually all MRS measurements of biopsy samples are performed on tissue extracts, it has recently become possible to perform high resolution MRS directly on samples of intact tissue using the technique of magic angle spinning [18]. While most studies have used 1 H MRS, high resolution 31 P MR spectra may also be obtained [19]. Unlike extract studies the enzymes are still active, so it is necessary to perform measurements at low temperature (usually about 4◦ C). The high-energy phosphates are usually not visualized, but the metabolites involved in membrane metabolism (phosphomonoesters and phosphodiesters) are found to be reasonably stable.

Applications using 31 P MRS of Tissue Extracts In practice, 31 P MRS of biopsy samples ex vivo is not as widely used as 31 P MRS in vivo, owing to the wide variety of alternative analytical techniques available for evaluating tissue samples. However the high resolution permits the identification of many more metabolites than is possible in vivo, for example, 14 different phospholipids were identified in a study of malignant breast tumors, several of which (sphingomyelin, PtdCho, PtdSer, phosphatidic

Part II

Table 3B: Published values of T2 relaxation times in different human tissues.

1150 Part II

Medical Uses

Part II

A

Fig. 4. Example spectra from cell extracts. All measurements were performed at 9.4 T. Spectra were plotted and analyzed using MestRe-C version 1.5.1 software. (A) 1 H decoupled 31 P MR spectrum of the water soluble fraction of Jurkat T-cell extracts (G3-P = glycerol-3-phosphate, PE = phosphoethanolamine, PC = phosphocholine, Pi = inorganic phosphate, GPE = glycerol-3phosphoethanolamine, GPC = glycerol-3-phosphocholine, NDPs = nucleotide diphosphates, NTPs = nucleotide triphosphates. NAD+ = nicotinamide adenine dinucleotide). Spectrum is the result of 1280 scans plotted with line broadening of 2 Hz. (B) 1 H decoupled 31P MR spectrum of the lipid fraction of Jurkat T-cell extracts. Spectrum is the result of 1280 scans plotted with line broadening of 0.1 Hz and zero filling to 64 K data points (Plasm.PtdEthn = plasmalogen phosphatidylethanolamine, PtdIns = phosphatidylinositol, Plasm.PtdCho = plasmalogen phosphatidylcholine, SM = sphingomyelin, PtdCho = phosphatidylcholine, PtdEthn = phosphatidylethanolamine).

31 P

MRS on Biopsy and In Vivo

31 P

MRS In Vivo 1151

Tissue biopsy required? Sample preparation required? Max measurement time Measurement co-registered with anatomical image? System disturbed? Capacity to follow kinetics Spectral resolution Sensitivity to motion Complexity of NMR sequence Field strength generally available Coil design / availability Safety issues

Tissue Extracts

In vivo

Yes Yes Unlimited Not usually

No No Approx. 1 h Yes

Yes No ∼0.01ppm N/A Decoupled pulse-acquire 7–14 T Good range of commercial coils Biopsy

No Yes ∼0.1 ppm Depends on tissue and method Requires localization 1.5–3 T Few commercial coils; most need to be custom-built MR Hazards (esp. RF heating from surface coils)

acid, phosphatidylglycerol and alkylacylphosphatidylcholine) predicted tumor characteristics such cellular infiltration, lymphatic invasion and necrosis [20]. In other studies biopsy specimens from patients with chronic ductopenic rejection [21] and liver cirrhosis [22] showed elevated PE and PC, and reduced GPE and GPC, reflecting altered phospholid metabolism [21]. In malignant hyperthermia-susceptible patients PCA extracts of vastus medialis muscle showed elevated GPC/(PCr+Pi) but no histological differences, suggesting GPC could be a marker of impairment [23]. PCA extracts of transurethral prostate samples show statistically different ratios of PE/totP, PC/totP and GPC/totP in prostate cancer samples compared with those from patients with benign prostatic hyperplasia [24]. In extracts of breast tumors no correlation was found between levels of GPC, GPE, PC, and PE and with indices of tumor proliferation, but higher PC levels were found in high grade tumors [25]. PCA and lipid extracts from kidney show that oncocytomas (benign tumors) have a biochemical composition between healthy and malignant tissue [26]. Blood samples are easier to obtain than tissue samples. One application is in thyroid cancer, where plasma levels of sphingomeylin and PtdCho have been found to be reduced in patients with remnants of thyroid tissue, and are even lower in patients with metastatic thyroid cancer, compared with patients in remission [27]. 31

spectroscopy measurements on the tissue in vivo is clearly very welcome. While the sensitivity and spectral resolution can never approach those obtained in vitro, the possibility of obtaining biochemical information from tissues non-invasively, repeatedly if necessary, provides the opportunity to monitor aspects of tissue biochemistry and response to treatment. Table 4 lists some of the main differences between measurements in vivo and ex vivo. A direct comparison of spectra of breast tumors from extracts and in vivo is given in [28].

Magnets Most 31 P MRS studies on human subjects are performed in clinical NMR systems using horizontal bore superconducting magnets at a field strength of between 1.5 and 3T. Since sensitivity increases with field strength even higher fields would be advantageous but are not yet widely available. Since clinical scanners are normally designed for operation at a single frequency (for 1 H MR imaging) it is necessary to add or modify the RF hardware (coils, frequency source, transmit amplifier, receiver preamplifier, filters etc.) to permit operation at the 31 P resonant frequency.

RF Coils for In Vivo Studies (see also Chapter “Radiofrequency Coils for Magnetic Resonance Spectroscopy”)

P MRS In Vivo

Introductory Comments Given the difficulty and limited patient acceptance of obtaining tissue biopsy samples, the option to perform the

Appropriate RF coils of high sensitivity are essential for acquiring good 31 P MRS data. Because in vivo 31 P MRS is largely a research activity, with users having different specifications and applications, at present the choice of commercial coils is very limited. Unlike 1 H coils, for

Part II

Table 4: Relative merits of tissue extracts and in vivo 31 P MRS measurements

1152 Part II

Medical Uses

Part II

which the vendors of the MR systems provide a good range of options, commercial 31 P coils are often supplied by specialist coil companies, while many researchers design and build their own. Some good reviews on coil design principles can be found in [29–32]. Again in contrast to 1 H coils, almost all 31 P coils are used in transmit/receive mode—principally because there is no 31 P body resonator to provide a uniform transmit field. Occasionally a small surface coil receiver (to optimize sensitivity) is used together with a larger transmit coil (to optimize transmit field uniformity). Surface Coils Apart from studies in the brain, 31 P MRS measurements are usually performed with surface coils because of their superior sensitivity in superficial tissues [33]. There is a huge variety of designs, with the main differences being shape (e.g. circular or rectangular; mounting on a flat or curved surface), number of turns, material for construction (e.g. wire or foil), type of tuning network (e.g. simple LC circuit or balanced match), and whether there is a complementary 1 H coil to provide imaging and dual resonance capability. Surface coils are characterized by a spatial variation in the RF field. During transmit this leads to non-uniform excitation within the field of view of the coil. For small regions of interest that are not too close to the coil conventional amplitude-modulated RF pulses are often used, calibrated to give the required flip angle at the centre of the volume. However larger volumes are often of interest. In this case it is necessary to use adiabatic RF pulses to achieve uniform excitation over the region of interest (see below). The non-uniform transmit field of surface coils also leads to non-uniform RF power deposition, which can be very high close to the coil itself. This is hard to measure experimentally, so sophisticated models (e.g. using finite element analysis) are required to calculate the expected power deposition [34]. International guidelines specify limits to RF power deposition for local and whole-body irradiation [35]. In practice the high RF power deposition near surface coils does limit the minimum repetition time or pulse amplitude in some studies, especially where double-resonant techniques are being used (see below). Volume Coils Volume coils are characterized by a more uniform RF field than surface coils, but usually at the expense of lower sensitivity. The most common designs for clinical studies are co-axial pairs of surface coils (often loosely called “Helmholtz pairs” but in fact having a separation larger than the radius), or birdcage coils [36]. Birdcage coils can be circularly polarized (which reduces RF power and improves receive sensitivity) and also double-resonant (1 H/31 P) [37].

Double-Resonance Coils When performing 31 P MRS studies a 1 H capability is also usually required to obtain localization MR images, to shim, and sometimes to acquire complementary 1 H MR spectra (sequential or interleaved) or for double resonance studies (NOE, decoupling, polarization transfer). The hardware required depends on whether both 1 H and 31 P need to be used simultaneously. Double Resonance Coil Options (a) Two separate coils—such as 1 H body coil and 31 P surface coil or co-axial surface coils [38,39]. It is necessary to ensure they do not interact, in particular that 1 H transmit fields do not induce currents in the 31 P coil (leading to signal drop-out and local heating). (b) Single coil, either with wide tuning range to permit sequential tuning to 1 H and then 31 P (e.g. 63 MHz and 25 MHz at 1.5 T), or that has two independent tuning networks may be switched to tune to each frequency. (c) Single dual-resonant coil, for which the tuning circuit has resonances at both 1 H and 31 P. Depending on the design, the 1 H and 31 P connections may be on the same input [40] or different ones [31]. System Options (a) Single channel with wide frequency range to permit transmit and receive operation at 1 H followed by 31 P in separate measurements. Some systems permit operation at specific frequencies only (e.g. at 1 H and 31 P). No double resonance experiments are possible. (b) As (a) but with rapid switching of the transmit frequency. This permits irradiation at the 1 H frequency during the recovery time of 31 P measurements, and hence 31 P signal enhancement by the nuclear overhauser effect (NOE). (c) Two-channel system that is able to transmit at one frequency (e.g. 1 H) while receiving on another (e.g. 31 P). Such systems can perform 1 H -decoupling (see below) as well as NOE studies. (d) Two-channel system with full transmit and receive capability on two channels simultaneously. This permits both sequential acquisition at different frequencies [41–43] and also more complex double-resonance measurements such as polarization transfer [44,45]. Other Hardware for Double Resonance Studies If there are no transmit pulses during signal acquisition then the only requirement is that the coils and frequencies do not interact. However for decoupling studies, where 1 H irradiation is applied during 31 P receive, it is necessary to apply a high level of isolation (∼80 dB) between

31 P

MRS on Biopsy and In Vivo

31 P

MRS In Vivo 1153

Adiabatic RF Pulses for In Vivo Studies with Surface Coils Adiabatic RF pulses have the useful property that once a threshold transmitter amplitude is exceeded they produce uniform excitation over the sensitive region of the RF coil. In general they are characterized by a modulation of the frequency (or phase) as well as of the amplitude. For example the sech/tanh pulse (“hyperbolic secant pulse”) widely used for slice-selective inversion [46] is described by Amp = 0 sech(βt) Freq = µβ tanh(βt) A variety of adiabatic pulses have been developed. Helpful reviews are available [47]. Other useful pulses include the tanh/tan pulse for 90◦ excitation [48] and the BIR4 or BIRP pulses for excitation by smaller flip angles to reduce partial saturation effects [48,49]. While there are adiabatic pulses suitable for many purposes some applications are particularly difficult, for example slice-selective excitation (for which the subtraction of two measurements is required [50]) and spin-refocusing [51]. The major disadvantages of adiabatic RF pulses compared with conventional amplitude-modulated pulses is that they are often (but not always) longer and/or requiring of a higher peak or total power.

Localization Methods for In Vivo Studies For most studies in vivo it is necessary to select signal from a specified volume, usually identified from MR images acquired on the same occasion. A schematic representation of the three classes of technique is shown in Figure 5. An ideal acquisition method would have a sharp edge profile matching the requested target volume on the MR images, acquire signal from the selected volume with 100% efficiency, and include no contamination at all from elsewhere. The relative merits of the different acquisition strategies available are discussed briefly below.

Part II

the channels. If this is not done the main problems are (1) saturating the receiver, leading to lower receiver gain, and (2) introducing extra noise into the receive channel, which again reduces the signal-to-noise ratio. Isolation is normally achieved in three steps, (i) use a coil that has a sharp resonance at the frequency of interest (e.g. 31 P) and negligible response at the other frequency (e.g. 1 H) (ii) a narrow band-width 31 P preamplifier, to minimize impact of 1 H irradiation, (iii) narrow bandwidth filters on both 1 H transmit and 31 P receive paths.

Fig. 5. Schematic illustrating the different methods of spectral localization.

Surface-Coil Localization The relative sensitivity of an RF coil in receive mode at a particular point in space is proportional to the relative RF field produced at the same position by the coil in transmit mode [52]. Thus the sensitivity falls away with distance from a surface coil, limiting the sensitive volume. An indication of this region is shown in Figure 5, although in practice the boundary is not well defined. However, the localization method is simple and has the maximum available sensitivity for a given coil. It is normally only used alone (i.e. without further localization) when the sensitive region is almost entirely occupied with the target tissue, and when the signal losses associated with additional forms of localization cannot be tolerated. Single Voxel Localization Methods Single voxel methods make use of a series of frequencyselective RF pulses in combination with slice-selective B0 field gradients. These define three intersecting slices that are usually but not necessarily orthogonal. Of the wide range of strategies that have been developed over the years, only three are in widespread use—PRESS [53], STEAM [54] and ISIS [55]. 31 P MRS has been demonstrated to work with STEAM [56] but ISIS has been the method of choice. While PRESS and STEAM have the great advantage of achieving localization in a single measurement, they both include echo times during which significant losses occur for nuclei which are characterized by short transverse time constants, such as are observed in most 31 P metabolites. As technology advances and shorter minimum echo times become possible this will become less of a problem. PRESS and STEAM. PRESS is conceptually the simplest single voxel method and uses a 90◦ − 180◦ − 180◦ double spin echo sequence (Figure 6A). The first 90◦ RF pulse excites a slice of spins (e.g. transverse). The first 180◦ refocus pulse affects an intersecting slice (e.g. sagittal), so that only spins with the column defined by the intersection

1154 Part II

Part II

A

Medical Uses

TE1

RF

TE2 Acquire

Gx

90°

180°

180°

Surface coils Volume coils Short T2 metabolites Single-shot Edge definition Sensitivity Use use for shimming

Gy

Gz

180° (OFF or ON)

B

Table 5: A summary of the relative merits of STEAM, PRESS, and ISIS for single voxel acquisition of 31 P MR spectroscopy data.

90°x

RF

STEAM

PRESS

ISIS

Poor Good Moderate

Poor Good Poor

Good Poor Good

Yes Good 1.2 has good sensitivity (92.8%) and specificity (100%) for predicting 3-year (5-year) survival, while a ratio >2.5 predicts for 5-year survival [91]. 31 P MRS is also used in experimental studies to improve the transplant technique, for example combined pretreatment with lipidsoluble and water-soluble antioxidants has been shown to lead to better restoration of energy phosphates in kidney subjected to ischaemia-reperfusion [92]. Cardiovascular disease is common in patients with chronic renal failure. Measurements have been performed of the phosphorylation potential of heart in renal dialysis patients and show that the PCr/ATP ratio is significantly reduced [93]. Exercise tolerance of CRF patients in muscle has also been performed [94]. 31 P MRS permits measurements of metabolite concentrations in normal and diseased kidneys in vivo [95]. In advanced renal cell carcinoma it has been found that plasma levels of lysophosphatidylcholine are significantly lower than in control subjects [96]. Renal stone patients have low potassium and magnesium status. 31 P MRS has been used to assess the effect on intracellular pH, cytosolic magnesium and phosphorylation potential (PCr/ATP) in muscle [97]. Renal stones have also been analyzed with ex vivo 31 P MAS [98]. Liver Liver is characterized by absence of PCr but the presence of a large number of sugar phosphates in the PME region [99].

31 P

MRS In Vivo 1157

In liver 31 P MRS has been used to assess severity of hepatitis C virus-related liver disease without biopsy [100]; the PME/PDE ratio can distinguish between mild, moderate and cirrhosis disease with specificity and sensitivity of 80% [100]. In addition cirrhosis is characterized by altered hepatic gluconeogenesis [101]. Energy metabolism and phospholipid biochemistry has been studied in the liver of patients with obstructive jaundice. Without treatment patients PME/total phosphate ratio is increased by 50% compared with controls [102]. Lung cancer patients have been found to have elevated PME in the liver, reported to reflect increased glucose flux and gluconeogenesis from alanine [103]. ATP infusion restores hepatic energy levels in patients with advanced lung cancer [104]. 31 P MRS has also been used to evaluate changes in liver metabolism after surgery [105,106]. Studies in experimental models include hypothermia to reduce hepatic failure under hypoxia, progression of chronic liver disease to cirrhosis, and models to assess damage in acute liver failure. Brain A wide range of brain disorders have been investigated using 31 P MRS. Elevated GPC, GPE and high energy phosphate (PCr) have been observed in the anterior cingulate of brains of patients with first-episode schizophrenia [107,108], while parts of the brain of patients with bipolar disorder are characterized by reduced pH [109]. Lithium, used to treat depression, changes levels of ATP and of PME [110,111]. The pH is also observed to be reduced, but this may be the effect of disease rather than lithium [112]. Lithium is presumed to reduce intracellular myo-inositol, and increase levels of inositol monophosphate precursors (included in PME but not resolved). In alcoholics lower concentrations and enhanced transverse relaxation of white matter phospholipids is observed, reflecting changes in membrane composition and rigidity [113]. While most studies look at resting levels of metabolites, some studies are now looking at the effects of visual activation on high-energy phosphates and pH in normal volunteers and in patients, for example those with mitochondrial disease [114]. A few studies have looked at spectral changes with age. In particular, it has been found that PE dominates in neonate, falling with age (as does PC). GPE, GPC, and PCr increase in concentration with postnatal age [115]. Heart For review see [116]. In patients with haemachromatosis, often associated with cardiomyopathy, the PCr/ATP ratio in the left ventricle is significantly reduced compared with healthy volunteers, possibly reflecting mitochondrial impairment due to iron overload [117].

Part II

35–40% below normal, suggesting defective oxidative phosphorylation in mitochondria of diseases muscles [83]; r Studies of muscle energy metabolism, including coupling of ADP levels to glycogenolysis to regenerate ATP [84], non-oxidative ATP metabolism and acidosis [85]; r McArdles syndrome, a lack of phosphorylase activity in skeletal muscle that prevents the production of lactic acid under ischaemic exercise, and is therefore elegantly demonstrated by an absence of acidification with 31P MRS [86]; r peripheral vascular disease—showing larger changes in PCr utilization owing to reduced oxygen supply, and correlation between PCr and NIRS recovery rate constants [87], but that there may be defective metabolism in mitochondria of claudicating calf muscle [88].

MRS on Biopsy and In Vivo

1158 Part II

Medical Uses

Part II

PCr/ATP is reported to be reduced with age [118], dilated cardiomypathy [119], multiple sclerosis [120] and aortic stenosis [119]. The PCr/ATP ratio recovers towards control values in the latter group following surgical valve replacement [121]. In hypertensive patients the unstressed PCr/ATP ratio is variously reported as being normal [119] or reduced [122]. No reduction in PCr/ATP is observed in the non-infarcted septal myocardium of patients compared with controls [123]. Heart transplant patients studied within 24 h of a biopsy showed reduced PCr/ATP compared with controls, but with no strong correlation either with biopsy scores or with subsequent biopsies [124]. Stress tests show that cardiac patients with some forms of ischaemia showed a decrease in PCr/ATP even under light exercise, while no decrease was observed in patients whose hearts were non-ischaemic [122]. (Note: 31 P MRS measurements in the heart are often contaminated by signal from blood. In particular signal from 2,3 diphosphoglycerate overlaps with Pi signal making it hard to quantify.)

31 P MRS has been used extensively in experimental studies on effects of drugs, optimization strategies (reversing pH gradient, carbogen breathing), radiation etc. on cell systems.

Drug Detection While 31 P MRS is normally used for study of endogenous compounds it also has the potential for monitoring uptake and metabolism in vivo of drugs that contain 31 P nuclei. The sensitivity is much lower than that of radionuclide labels, but the method has the advantages of being able to follow the metabolism of the drug, is not limited by a radioactive half-life, and makes use of the native compound without further chemistry required to perform the measurement. One example application has been the investigation of the alkylating agent ifosfamide. 31 P MRS studies have examined ifosfamide metabolites in body fluids [133], preclinical model systems [134,135], and in patients [136].

References Tumors A good review of the use of 31 P MRS in cancer is given by Negendank [125]. In particular 31 P MRS has been important in demonstrating that, contrary to expectation, the intracellular space of many cancers is alkaline [125]. In addition PE and PC have been shown to be elevated in many cancers. The ratio of PME/NTP is now being evaluated as an early indicator of response. There is also increasing evidence that both PME/NTP [126] and NTP/Pi [127] in the pretreatment spectrum may be able to predict response to treatment. Separate studies in patients with non-resectable soft tissue sarcoma have shown that changes in PME/ATP predict response to isolated limb perfusion to permit limb sparing [128]. Patients with non-small cell lung cancer have infusion of ATP to restrict weight loss; an increase in liver ATP (8.8 → 12.2%) is observed [104]. PME levels have also been observed to be higher in the weight-loss patients [129]. In breast cancer, there is an association between transformation and an increase in the phosphomonoesters (PMEs), while a decrease in PME content after treatment is associated with response to treatment as assessed by tumor volume [130]. In normal premenopausal breast a consistent pattern of changes is seen in the 31 P MR spectrum during the menstrual cycle, suggesting a potential application in assessing effects of exogenous hormones such as contraceptives or hormone replacement therapy which may increase breast cancer risk [131]. 31 P MRS can also be used as marker of drug action (based on results of model systems and leading into clinical trials, e.g. 17-AAG [132]).

1. Moon R, Richards J. J. Biol. Chem. 1973;248:7276. 2. Gadian DG. Nuclear Magnetic Resonance and Its Applications to Living Systems. Oxford University Press, 1995. 3. Cohn M, Hughes TR Jr. J. Biol. Chem. 1962;237:176. 4. Jung WI, Straubinger K, Bunse M, Schick F, Kuper K, Dietze G, Lutz O. Magn. Reson. Med. 1992;28:305. 5. Cox M, Nelson DL. Lehninger Principles of Biochemistry. Palgrave Macmillan, New York, 2004. 6. Stryer L, Berg JM, Tymoczko JL. Biochemistry W H Freeman, New York, 2002. 7. Podo F. NMR Biomed. 1999;12:413. 8. Cuadrado A, Carnero A, Dolfi F, Jimenez B, Lacal JC. Oncogene. 1993;8:2959. 9. Jimenez B, del Peso L. Montaner S. Esteve P. Lacal JC. J. Cell. Biochem. 1995;57:141. 10. Gonzelez-Mendez R, Litt L, Koretsky AP, Voncolditz J, Weiner MW, James TL. J. Magn. Reson. 1984;57:526. 11. Murphy EJ, Rajagopalan B, Brindle KM, Radda GK. Magn. Reson. Med. 1989;12:282. 12. Tyagi RK, Azrad A, Degani H, Salomon Y. Magn. Reson. Med. 1996;35:194. 13. Henke J, Willker W, Engelmann J, Leibfritz D. Anticancer Res. 1996;16:1417. 14. Meneses P, Glonek T. J. Lipid. Res. 1988;29:679. 15. Al-Saffar NM, Titley JC, Robertson D, Clarke PA, Jackson LE, Leach MO, Ronen SM. Br. J. Cancer. 2002;86:963. 16. Ronen SM, DiStefano F, McCoy CL, Robertson D, Smith TA, Al-Saffar NM, Titley J, Cunningham DC, Griffiths JR, Leach MO, Clarke PA. Br. J. Cancer 1999;80:1035. 17. Tate AR, Griffiths JR, Martinez-Perez I, Moreno A, Barba I, Cabanas ME, Watson D, Alonso J, Bartumeus F, Isamat F, Ferrer I, Vila F, Ferrer E, Capdevila A, Arus C. NMR Biomed. 1998;11:177. 18. Cheng LL, Lean CL, Bogdanova A, Wright SC, Ackerman JL, Brady TJ, Garrido L. Magn. Reson. Med. 1996;36:653.

31 P

References 1159

45. Payne GS, Leach MO, Magn. Reson. Med. 2000;43:510. 46. Silver MS, Joseph RI, Hoult DI. Physical Review A 1985; 31:2753. 47. Garwood M, Ugurbil K. In: P. Diehl, E. Fluck, H. Gunther, Kosfeld R, Seelig J (Eds). NMR: Basic Principles and Progress, Springer-Verlag, Berlin Heidelberg, 1992, p 109. 48. Garwood M, Ke Y. J. Magn. Reson. 1991;94:511. 49. Bottomley PA, Ouwerkerk R. J. Magn. Reson. Series A 1993;103:242. 50. Johnson AJ, Garwood M, Ugurbil K. J. Magn. Reson. 1989;81:653. 51. Ugurbil K, Garwood M, Rath AR, Bendall MR. J. Magn. Reson. 1988;78:472. 52. Hoult DI, Richards RE. J. Magn. Reson. 1976;24:71. 53. Bottomley P. Ann. NY Acad. Sci. 1987;508:333. 54. Frahm J, Merboldt K-D, Hanicke W. J. Magn. Reson. 1987;72:502. 55. Ordidge R, Connelly A, Lohman J. J. Magn. Reson. 1986; 66:283. 56. Merboldt KD, Chien D, Hanicke W, Gyngell ML, Bruhn H, Frahm J. J. Magn. Reson. 1990;89:343. 57. Brown T, Kincaid B, Ugurbil K. Proceedings of the National Academy of Sciences U.S.A. 1982;79:3523. 58. Koch T, Brix G, Lorenz WJ. J. Magn. Reson. Series B 1994;104:199. 59. Doyle VL, Howe FA, Griffiths JR. Phys. Med. Biol. 2000; 45:2093. 60. Schwarz AJ, Leach MO. Phys. Med. Biol. 2000;45:2105. 61. de Graaf RA, In vivo NMR spectroscopy, John Wiley & Sons, 1998. 62. Pohmann R, von Kienlin M. Magn. Reson. Med. 2001; 45:817. 63. Madden A, Leach MO, Collins DJ, Payne GS. Magn. Reson. Med. 1991;19:416. 64. Knijn A, de Beer R, van Ormondt D. J. Magn. Reson. 1992; 97:444. 65. ven den Boogaart A, Van Hecke P, Van Huffel S, GraveronDemilly D, de Beer R. In: European Society for Magnetic Resonance in Medicine and Biology: 13th Annual Meeting, Prague, p. 318, 1996. 66. Provencher SW. Magn. Reson. Med. 1993;30:672. 67. Buchli R, Meier D, Martin E, Boesiger P. Magn. Reson. Med. 1994;32:453. 68. Tofts PS, Wray S. NMR Biomed 1988;1:1. 69. Meyerhoff DJ, Karczmar GS, Matson GB, Boska MD, Weiner MW. NMR Biomed 1990;3:17. 70. Buchli R, Martin E, Boesiger P. NMR Biomed. 1994;7:225. 71. Buchli R, Duc CO, Martin E, Boesiger P. Magn. Reson. Med. 1994;32:447. 72. Murphy -Boesch J, Jiang H, Stoyanova R, Brown TR. Magn. Reson. Med. 1998;39:429. 73. Doyle VL, Payne GS, Collins DJ, Verrill MW, Leach MO. Phys. Med. Biol. 1997;42:691. 74. Argov Z, Lofberg M, Arnold DL. Muscle Nerve 2000;23: 1316. 75. Heerschap A, Houtman C, in ’t Zandt HJ, van den Bergh AJ, Wieringa B. Proc. Nutr. Soc. 1999;58:861. 76. Newcomer BR, Larson -Meyer DE, Hunter GR, Weinsier RL. Int. J. Obes. Relat. Metab. Disord. 2001;25:1309.

Part II

19. Payne GS, Chung Y-L, Vaidya SJ, Leach MO. In International Society for Magnetic Resonance in Medicine: Twelth Annual Meeting. Kyoto, Japan, 2004, p 68. 20. Merchant TE, Kasimos JN, Vroom T, de Bree E, Iwata JL, de Graaf PW, Glonek T. Cancer Lett. 2002;176:159. 21. Taylor-Robinson SD, Sargentoni J, Bell JD, Thomas EL, Marcus CD, Changani KK, Saeed N, Hodgson HJ, Davidson BR, Burroughs AK, Rolles K, Foster CS, Cox IJ. Gut. 1998;42:735. 22. Taylor-Robinson SD, Sargentoni J, Bell JD, Saeed N, Changani KK, Davidson BR, Rolles K, Burroughs AK, Hodgson HJ, Foster CS, Cox IJ. Liver. 1997;17:198. 23. Payen JF, Fouilhe N, Sam-Lai E, Remy C, Dupeyre R, Mezin P, Halsall J, Stieglitz P. Anesthesiology. 1996;84: 1077. 24. Cornel EB, Smits GA, Oosterhof GO, Karthaus HF, Deburyne FM, Schalken JA, Heerschap A. J. Urol. 1993;150:2019. 25. Smith TA, Bush C, Jameson C, Titley JC, Leach MO, Wilman DE, McCready VR. NMR Biomed. 1993;6:318. 26. Tosi MR, Tugnoli V, Bottura G, Lucchi P. Applied Spectroscopy 2001;55:908. 27. Moka D, Dietlein M, Raffelt K, Hahn J, Schicha H. Am. J. Med. 2002;112:634. 28. Smith TA, Glaholm J, Leach MO, Machin L, Collins DJ, Payne GS, McCready VR. Br. J. Cancer. 1991;63:514. 29. Schnall M. In: P Diehl, E Fluck, H Gunther, R Kosfeld, J Seelig (Eds). NMR Basic Principles and Progress 26. Springer-Verlag: Berlin Heidelberg, 1992, p 33. 30. Link J. In: P Deihl, E Fluck, H Gunther, R Kosfeld, J Seelig (Eds). NMR Basic Principles and Progress 26. SpringerVerlag: Berlin Heidelberg, 1992, p 1. 31. Hoult DI. Prog. NMR Spec. 1978;12:41. 32. Jin J. Electromagnetic Analysis and Design in Magnetic Resonance Imaging. CRC Press LLC, London, 1999. 33. Ackerman JJ, Grove TH, Wong GG, Gadian DG, Radda GK. Nature. 1980;283:167. 34. Prock T, Collins DJ, Leach MO. Phys. Med. Biol. 2002;47: 1805. 35. IEC. In: International Electrotechnical Commission (IEC), 2002. 36. Hayes CE, Edelstein WA, Schenck JF, Mueller OM, Eash M. J. Magn. Reson. 1985;63:622. 37. Murphy-Boesch J, Srinivasan R, Carvajal L, Brown TR. J. Magn. Reson. Series B 1994;103:103. 38. Klomp DW, Collins DJ, van den Boogert HJ, Schwarz A, Rijpkema M, Prock T, Payne GS, Leach MO, Heerschap A. Magn. Reson. Imaging. 2001;19:755. 39. Murphy-Boesch J, He L, Elsasser C, Brown TR. In: International Society for Magnetic Resonance in Medicine: Forth Annual Meeting. New York, USA 1996, p 1430. 40. Collins DJ, Prock T, Leach MO. In: International Society for Magnetic Resonance in Medicine: Eighth Annual Meeting, 2000, p 1417. 41. Yongbi NM, Payne GS, Leach MO. Magn. Reson. Med. 1994;32:768. 42. Arias-Mendoza F, Javaid T. Stoyanova R, Brown TR, Gonen O. NMR Biomed. 1996;9:105. 43. Gonen O, Hu J, Murphy-Boesch J, Stoyanova R, Brown TR. Magn. Reson. Med. 1994;32:104. 44. Gonen O, Mohebbi A, Stoyanova R, Brown TR. Magn. Reson. Med. 1997;37:301.

MRS on Biopsy and In Vivo

1160 Part II

Medical Uses

Part II

77. Tartaglia MC, Chen JT, Caramanos Z, Taivassalo T, Arnold DL, Argov Z. Muscle Nerve 2000;23:175. 78. Schunk K, Pitton M, Duber C, Kersjes W, SchadmandFischer S, Thelen M. Invest. Radiol. 1999;34:116. 79. Mattei JP, Bendahan D, Roussel M, Lefur Y, Cozzone PJ. FEBS Lett. 1999;450:173. 80. Smith SA, Montain SJ, Matott RP, Zientara GP, Jolesz FA, Fielding RA. J. Appl. Physiol. 1999;87:116. 81. Schunk K, Losch O, Kreitner KF, Kersjes W, SchadmandFischer S, Thelen M. Invest. Radiol. 1999;34:348. 82. Lodi R, Kemp GJ, Muntoni F, Thompson CH, Rae C, Taylor J, Styles P, Taylor DJ. Brain 122 (Pt 1) 121 (1999). 83. Park JH, Niermann KJ, Ryder NM, Nelson AE, Das A, Lawton AR, Hernanz-Schulman M. Olsen NJ. Arthritis Rheum 2000;43:2359. 84. Hsu AC, Dawson MJ. Magn. Reson. Med. 2003;49:626. 85. Roussel M, Mattei JP, Le Fur Y, Ghattas B, Cozzone PJ, Bendahan D. J. Appl. Physiol. 2003;94:1145. 86. Ross BD, Radda GK, Gadian DG, Rocker G, Esiri M, Falconer-Smith J, Engl N. J. Med. 1981;304:1338. 87. Kemp GJ, Roberts N, Bimson WE, Bakran A, Harris PL, Gilling-Smith GL, Brennan J, Rankin A, Frostick SP. J. Vasc. Surg. 2001;34:1103. 88. Pipinos, II, Shepard AD, Anagnostopoulos PV, Katsamouris A, Boska MD. J. Vasc. Surg. 2000;31:944. 89. Niekisch MB, Von Elverfeldt D, Saman AE, Hennig J, Kirste G. Transplantation 2004;77:1041. 90. Moller HE, Gaupp A, Dietl K, Buchholz B, Vestring T. Magn. Reson. Imaging. 2000;18:743. 91. Seto K, Ikehira H, Obata T, Sakamoto K, Yamada K, Kashiwabara H, Yokoyama T, Tanada S. Transplantation 2001;72:627. 92. Jonsson O, Lindgard A, Fae A, Gondalia J, Aneman A, Soussi B. Scand J. Urol. Nephrol. 2003;37:450. 93. Ogimoto G, Sakurada T, Imamura K, Kuboshima S, Maeba T, Kimura K, Owada S. Mol. Cell. Biochem. 2003;244: 135. 94. Sala E, Noyszewski EA, Campistol JM, Marrades RM, Dreha S, Torregrossa JV, Beers JS, Wagner PD, Roca J. Am. J. Physiol. Regul. Integr. Comp. Physiol. 280 R1240 (2001). 95. Kugel H, Wittsack HJ, Wenzel F, Stippel D, Heindel W, Lackner K. Acta Radiol. 2000;41:634. 96. Sullentrop F, Moka D, Neubauer S, Haupt G, Engelmann U, Hahn J, Schicha H. NMR Biomed. 2002;15:60. 97. Mairiang E, Hanpanich P, Sriboonlue P. Magn. Reson. Imaging. 2004;22:715. 98. Bak M, Thomsen JK, Jakobsen HJ, Petersen SE, Petersen TE, Nielsen NC. J. Urol. 2000;164:856. 99. Bell JD, Bhakoo KK. NMR Biomed. 1998;11:354. 100. Lim AK, Patel N, Hamilton G, Hajnal JV, Goldin RD, TaylorRobinson SD. Hepatology 2003;37:788. 101. Changani KK, Jalan R, Cox IJ, Ala-Korpela M, Bhakoo K, Taylor-Robinson SD, Bell JD. Gut 2001;49:557. 102. Mann DV, Lam WW, Magnus Hjelm N. So NM, Yeung DK, Metreweli C, Lau WY. Gut 2002;50:118. 103. Leij-Halfwerk S, Dagnelie PC, Van Den Berg JW, Wilson JH, Sijens PE. Clin. Sci. (London) 2000;98:167. 104. Leij-Halfwerk S, Agteresch HJ, Sijens PE, Dagnelie PC. Hepatology 2002;35:421. 105. Mann DV, Lam WW, Hjelm NM, So NM, Yeung DK, Metreweli C, Lau WY. Ann. Surg. 2002;235:408.

106. Mann DV, Lam WW, Hjelm NM, So NM, Yeung DK, Metreweli C, Lau WY. Hepatology 2001;34:557. 107. Jensen JE, Miller J, Williamson PC, Neufeld RW, Menon RS, Malla A, Manchanda R, Schaefer B, Densmore M, Drost DJ. Br. J. Psychiatry. 2004;184:409. 108. Bluml S, Tan J, Harris K, Adatia N, Karme A, Sproull T, Ross B. J. Comput. Assist. Tomogr. 1999;23:272. 109. Hamakawa H, Murashita J, Yamada N, Inubushi T, Kato N, Kato T. Psychiatry. Clin. Neurosci. 2004;58:82. 110. Yildiz A, Demopulos CM, Moore CM, Renshaw PF, Sachs GS. Biol. Psychiatry. 2001;50:3. 111. Silverstone PH, Rotzinger S, Pukhovsky A, Hanstock CC. Biol. Psychiatry. 1999;46:1634. 112. Kato T, Murashita J, Kamiya A, Shioiri T, Kato N, Inubushi T. Eur. Arch. Psychiatry Clin. Neurosci. 1998;248:301. 113. Estilaei MR, Matson GB, Payne GS, Leach MO, Fein G, Meyerhoff DJ. Alcohol Clin. Exp. Res. 2001;25: 1213. 114. Rango M, Bozzali M, Prelle A, Scarlato G, Bresolin N. J. Cereb. Blood. Flow. Metab. 2001;21:85. 115. Bluml S, Seymour KJ, Ross BD. Magn. Reson. Med. 1999;42:643. 116. von Kienlin M, Beer M, Greiser A, Hahn D, K Harre, H Kostler, W Landschutz, T Pabst, Sandstede J, Neubauer S. J. Magn. Reson. Imaging. 2001;13:521. 117. Schocke MF, Zoller H, Vogel W, Wolf C, Kremser C, Steinboeck P, Poelzl G, Pachinger O, Jaschke WR, Metzler B. Magn. Reson. Imaging. 2004;22:515. 118. Schocke MF, Metzler B, Wolf C, Steinboeck P, Kremser C, Pachinger O, Jaschke W, Lukas P. Magn. Reson. Imaging. 2003;21:553. 119. Beer M, Seyfarth T, Sandstede J, Landschutz W, Lipke C, Kostler H, von Kienlin M. Harre K, Hahn D, Neubauer S. J. Am. Coll. Cardiol. 2002;40:1267. 120. Beer M, Sandstede J, Weilbach F, Spindler M, Buchner S, Krug A, Kostler H, Pabst T, Kenn W, Landschutz W, von Kienlin M, Toyka KV, Ertl G, Gold R, Hahn D. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 2001;173: 399. 121. Beer M, Viehrig M, Seyfarth T, Sandstede J, Lipke C, Pabst T, Kenn W, Harre K, Horn M, Landschutz W, von Kienlin M, Neubauer S, Hahn D. Radiologe. 2000;40:162. 122. Butterworth EJ, Evanochko WT, Pohost GM. Ann. Biomed. Eng. 2000;28:930. 123. Beer M, Sandstede J, Landschutz W, Viehrig M, Harre K, Horn M, Meininger M, Pabst T, Kenn W, Haase A, von Kienlin M, Neubauer S, Hahn D. Eur. Radiol. 2000;10: 1323. 124. Buchthal SD, Noureuil TO, den Hollander JA, Bourge RC, Kirklin JK, Katholi CR, Caulfield JB, Pohost GM, Evanochko WT. J. Cardiovasc. Magn. Reson. 2000;2:51. 125. Negendank W. NMR Biomed. 1992;5:303. 126. Arias-Mendoza F, Smith MR, Brown TR. Acad. Radiol. 2004;11:368. 127. Zakian KL, Shukla-Dave A, Meyers P, Gorlick R, Healey J, Thaler HT, Huvos AG, Panicek DM, Koutcher JA. Cancer Res. 2003;63:9042. 128. Kettelhack C, Wickede M, Vogl T, Schneider U, Hohenberger P. Cancer 2002;94:1557. 129. Leij -Halfwerk S, van den Berg JW, Sijens PE, Wilson JH, Oudkerk M, Dagnelie PC. Cancer Res. 2000;60:618.

31 P

References 1161

154. Brown TR, Stoyanova R, Greenberg T, Srinivasan R. Murphy-Boesch J. Magn. Reson. Med. 1995;33:417. 155. Newcomer BR, Boska MD. Magn. Reson. Med. 1999;41: 486. 156. Bottomley PA, Ouwerkerk R. J. Magn. Reson. Series B 1994;104:159. 157. Yongbi NM, Leach MO, Payne GS, Collins DJ. NMR Biomed. 1992;5:95. 158. Buchthal SD, Thoma WJ, Taylor JS, Nelson SJ, Brown TR. NMR Biomed. 1989;2:298. 159. Luyten PR, Groen JP, Vermeulen JW, den Hollander JA. Magn. Reson. Med. 1989;11:1. 160. Thomsen C, Jensen KE, Henriksen O. Magn. Reson. Imaging. 1989;7:231. 161. Moller HE, Vermathen P, Rummeny E, Wortler K, Wuisman P, Rossner A, Wormann B, Ritter J, Peters PE. NMR Biomed. 1996;9:347. 162. Meyerspeer M, Krssak M, Moser E. Magn. Reson. Med. 2003;49:620. 163. Schindler R, Krahe T, Neubauer S, Hillenbrand H, Entzeroth C, Horn M, Lackner K, Ertl G. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1992;157: 452. 164. Neubauer S, Krahe T, Schindler R, Hillenbrand H, Entzeroth C, Horn M, Bauer WR, Stephan T, Lackner K, Haase A et al. Magn. Reson. Med. 1992;26:300. 165. Blackledge MJ, Oberhaensli RD, Styles P, Radda GK. J. Magn. Reson. 1987;71:331. 166. Straubinger K, Jung WI, Bunse M, Lutz O, Kuper K, Dietze G. Magn. Reson. Imaging. 1994;12:121. 167. Thomsen C, Jensen KE, Henriksen O. Magn. Reson. Imaging. 1989;7:557. 168. Jung WI, Widmaier S, Bunse M, Seeger U, Straubinger K, Schick F, Kuper K, Dietze G, Lutz O. Magn. Reson. Med. 1993;30:741. 169. Noggle JH, Schirmer RE. The Nuclear Overhauser Effect, Academic Press: New York and London, 1971. 170. Bachert P, Bellemann ME. J. Magn. Reson. 1992;100: 146. 171. Murphy-Boesch J, Stoyanova R, Srinivasan R, Willard T, Vigneron D, Nelson S, Taylor JS, Brown TR. NMR Biomed. 1993;6:173. 172. Luyten PR, Bruntink G, Sloff FM, Vermeulen JW, van der Heijden JI, den Hollander JA. Heerschap A. NMR Biomed. 1989;1:177. 173. Meninger M, Landschutz W, Beer M, Seyfarth T, Horn V, Pabst T, Haase A, Hahn D, Neubauer S, von Kienlin M. Magn. Reson. Med. 1999:41:657.

Part II

130. Ronen SM, Leach MO. Breast Cancer Res. 2001;3:36. 131. Payne GS, Dowsett M, Leach MO. The Breast 1994;3:20. 132. Chung YL, Troy H, Banerji U, Jackson LE, Walton MI, Stubbs M, Griffiths JR, Judson IR, Leach MO, Workman P, Ronen SM. J. Natl. Cancer Inst. 2003;95:1624. 133. Martino R, Crasnier F, Chouini-Lalanne N, Gilard V, Niemeyer U, De Forni M, Malet-Martino. J. Pharmacol. Exper. Therap. 1992;260:1133. 134. Rodrigues LM, Maxwell RJ, McSheehy PMJ, Pinkerton CR, Robinson SP, Stubbs M, Griffiths JR, Br. J. Cancer 1997;75:62. 135. Mancini L, Payne GS, Dzik-Jurasz AS, Leach MO. Magn. Reson. Med. 2003;50:249. 136. Payne GS, Pinkerton CR, Bouffet E, Leach MO. Magn. Reson. Med. 2000;44:180. 137. Robitaille P-M, Robitaille P, Brown GJ,. Brown G. J. Magn. Reson. 1991;92:73. 138. Govindaraju V, Young K, Maudsley AA. NMR Biomed. 2000;13:129. 139. Deicken RF, Calabrese G, Raz J, Sappey-Marinier D, Meyerhoff D, Dillon WP, Weiner MW, Fein G. Biol. Psychiatry 1992;32:628. 140. Hubesch B, Sappey-Marinier D, Roth K, Meyerhoff DJ, Matson GB, Weiner MW. Radiology 1990;174:401. 141. Jensen JE, Drost DJ, Menon RS, Williamson PC. NMR Biomed. 2002;15:338. 142. Harris RC, Hultman E, Nordesjo LO. Scand. J. Clin. Lab. Invest. 1974;33:109. 143. Bylund-Fellenius AC, Walker PM, Elander A, Holm S, Holm J, Schersten T. Biochem. J. 1981;200:247. 144. Dawson M, Allan RJ, Watson TR. Br. J. Clin. Pharmacol. 1982;14:453. 145. Taylor DJ, Styles P, Matthews PM, Arnold DA, Gadian DG, Bore P, Radda GK. Magn. Reson. Med. 1986;3:44. 146. Rees D, Smith MB, Harley J, Radda GK. Magn. Reson. Med. 1989;9:39. 147. Newham DJ, Cady EB. NMR Biomed. 1990;3:211. 148. Venkatasubramanian PN, Mafee MF, Barany M. Magn. Reson. Med. 1988;6:359. 149. Cady EB. J. Magn. Reson. 1990;87:433. 150. Roth K, Hubesch B, Meyerhoff DJ, Naruse S, Gober JR, Lawry TJ, Boska MD, Matson GB, Weiner MW, J. Magn. Reson. 1989;81:299. 151. Meyerhoff DJ, Boska MD, Thomas AM, Weiner MW. Radiology. 1989;173:393. 152. Blackledge MJ, Styles P. J. Magn. Reson. 1989;83:390. 153. Gruetter R, Fusch C, Martin E, Boesch C. Magn. Reson. Med. 1993;29:7.

MRS on Biopsy and In Vivo

1163

Boguslaw Tomanek Institute for Biodiagnostics, Winnipeg, Manitoba, Canada

The Requirements Localized magnetic resonance spectroscopy (MRS) as well as magnetic resonance imaging (MRI) require both maximal signal-to-noise ratio (SNR) and radio frequency (rf) field (B1 ) homogeneity. While MRI allows the application of post-processing methods, such as correction of image intensity to correct B1 inhomogeneities, MRS is more demanding from this point of view leaving very little opportunity for post-processing and requiring mostly hardware improvement. In addition, due to the smaller volume of interest (VOI) and particularly low metabolite concentration, there is less signal available in MRS than in MRI. The time required to obtain good spectra is long pushing the SNR requirements to the limits.

The Issues Unfortunately, when considering rf probe construction the requirements of high SNR and good B1 homogeneity usually conflict. The SNR depends on many factors such as the strength of the magnetic field, geometry, resistance and quality factor (Q) of the coil, sample size, filling factor of the coil, type of the reception (linear or circular), etc. The B1 field distribution depends on the geometry of the coil, its type (transmit and receive or transmit/receive), and the material used for the coil construction. When designing an rf coil all of these factors plus patient access to the coil and the direction of the magnetic field relative to B1 must be considered. Therefore whilst a coil may work very well, providing good SNR and B1 homogeneity, at a low frequency (say 1.5 T) it may not work so well (or even not at all) at a higher frequency (say 3 T). A trade-off amongst all these factors is inevitable and there is no simple, consistent recipe for making a perfect rf coil that fulfills all the requirements for MRS. With this knowledge of the problems we shall consider some of the requirements for the MRS coil: SNR, B1 distribution, coil geometry, method of reception as well as some other factors. For this discussion we will be ignoring other potential problems associated, for example, with relaxation times, pulse parameters or even eddy currents induced in the coil, gradients coil, or in the cryostat due to Graham A. Webb (ed.), Modern Magnetic Resonance, 1163–1170.
C 2008 Springer.

the short TE (fast gradient switching) required frequently in MRS. The first expression describing the NMR signal was derived by Abragam [1], who showed that the electromagnetic force (emf) induced by the nuclear magnetization in the receiving coil can be expressed as: rms ∝ K ηM0 (µ0 Qω0 Vc /4FkTc  f )1/2

(1)

where K is numerical factor depending on the coil geometry; η is the filling factor, i.e. a measure of the fraction of the rf field volume occupied by the sample; M0 is the magnetization (proportional to the magnetic field strength); µ0 is the permeability of free space; Q is the quality factor of the coil; ω0 is the Larmor frequency; Vc is volume of the coil; F is the noise figure of the preamplifier; k is Boltzmann constant; Tc is the probe temperature;  f is the bandwidth (in Hertz) of the receiver. As one can notice, the SNR is proportional to the 3/2 power of the magnetic field (M0 ∼ ω0 ). This is true assuming that other factors are independent of frequency. This is an important statement that will be analyzed in detail later. The above revelation explains the historical tendency to introduce stronger and stronger magnets in an effort to increase SNR. Therefore, magnets generating a magnetic field of 20 T were produced for ex vivo MRS, while whole body MRI systems have reached 7 T, and even higher field magnets are under construction. However, a closer analysis of the dependence of noise mechanism with frequency reveals a less favorable SNR relationship. The rule of thumb is a linear relationship between the SNR and the magnetic field strength. While the Equation (1) gives a reasonable estimate of the SNR, it does not consider other factors such as rf coil shape, preamplifier noise, and other electronic components whose performance depends on, and usually degrades with, the frequency. Thus, innocent engineers constructing rf coils for MRI systems are often blamed for insufficient increase in SNR as expected from MR theory. Probably the most comprehensive summary of the SNR and its relationship with MR hardware was given by Hoult et al. [2,3], who applied reciprocity theory and

Part II

Radio Frequency Coils for Magnetic Resonance Spectroscopy

1164 Part II

Medical Uses

Part II

considering the sources of noise they obtained: ψrms =

K (B1 )x y Vs N γλ¯2 I (I + 1) · 7.12kTs



p FkTc lζ  f

1/2

7/4

×

ω0 [µµ0 ρ(Tc )]1/4

(2)

where K (B1 )x y is the effective field over the sample volume produced by unit current flowing in the receiving coil, p is the perimeter of the conductor, l is the length of the conductor. The above factors are dependent only on coil geometry. Tc is the temperature of the coil, ρ(Tc ) is the resistivity of the material from which the coil is made, ζ is the proximity factor, F is the quality of the preamplifier. From the above equation one can see that SNR seems 7/4 to be proportional to ω0 . However, in the above equation Q and η have been replaced by a function ζ , that is also ω0 dependent. Hence, SNR is not always proportional to the 7/4 power of the frequency. A careful reader will also notice that SNR depends on the type and geometry of the coil (factors: K, l, ρ).

The Solenoid Coil and Saddle-Shape Coil As seen from Equation (1) SNR can be improved by an increase of the Q factor, which depends on the coil resistance and inductance. The Q of a solenoidal coil is propor1/2 tional to ω0 only well below self-resonance frequency of the coil, but when approaching the self-resonance frequency the Q drops away from the ω1/2 line and prevents further increase of the SNR as one could expect from Equation (2). The only solution is then to decrease the selfresonance frequency of the coil by the reduction of the coil inductance or/and resistance. This can be achieved by reducing the number of turns for a solenoidal coil. A detailed analysis of solenoidal and saddle-shape coils [2], the most common coils producing very homogenous rf field, yield to the conclusion that the solenoidal coil performance is about three times better that the saddle-shape coil with comparable dimensions. As shown above, a solenoidal coil would be a very good candidate for the rf probe providing good B1 homogeneity and good sensitivity within a reasonable range of frequencies. However, further consideration should be given when the direction of the main magnetic field is considered. To produce the NMR phenomenon (to flip the magnetization), the direction of the rf field must be perpendicular to the main magnetic field. This creates additional constraints when selecting an rf coil for MRS and MRI. Unfortunately, the direction of the rf field produced by the solenoidal coil (along its long axis) prevents almost all applications for horizontal bore magnets. Such a coil could not be made for a human head, for example, since it would require the head to bend by 90◦ .

Fig. 1. Surface coil for 3 T MR system (National Research Council, Institute for Biodiagnostics, Winnipeg, Canada).

However, a single loop solenoidal probe, namely a surface coil (Figures 1 and 2), overcomes this obstacle. In addition, in MRS we are interested in small voxels to detect small abnormalities. Thus a volume coil is usually not required or even not welcome. A surface coil (Figure 2A) provides a very good SNR produces but also a very inhomogenous B1 field (Figure 2B). Along the axis the B1 field can be calculated, neglecting interaction of the tissue with the rf field, based on Bio-Savart law: B1 (x) =

µ0 I a2  2 (a 2 + x 2 )3

(3)

where a is the radius of the coil, x is the axial distance from the center of the coil and I is the current flowing through the coil.

Surface Coil So far we have concluded that the best solution from the point of view of the SNR is the application of the √ surface coil. Further improvement in SNR by up to 2 can be achieved by the application of the quadrature surface coil [4,5]. However, the perfect quadrature rf coil requires the creation of equal and orthogonal B1 fields (Figure 3), which is difficult to achieve for a surface coil. The most common configuration for the quadrature surface coil [so-called circular polarized (CP) coil] is the use of a shaped butterfly coil and a standard surface coil to

Radio Frequency Coils for MRS

B0

B

Fig. 2. (A) Geometrical model of the surface coil: A current I flows in the ring of a radius. (B) The B1 field magnitude generated by the surface coil along the x-axis.

z

l

a

x

create two orthogonal or, more often—nearly orthogonal rf fields. While such a coil provides√larger field of view (FOV) it gives only modest (less than 2) increase in SNR due to the lack of perfect perpendicularity or/and unequal amplitude of B1 fields. The quadrature surface coil requires, however, less power when compared to the linear coil (LP coil). Let us now answer a question: what is the optimum diameter of the surface coil to maximize SNR within the specific VOI? In attempt to find the answer we will take a closer look at the sources of noise. Neglecting radiation losses, the noise picked up by the receiver (N )

x

is proportional to the sum of the effective resistance of the sample (Rs ) and the resistance of the coil itself (Rc ): N∝



Rs + Rc

(4)

The resistance of the sample comes from two sources: dielectric losses (that can be minimized using distributed capacitance and proper capacitor shielding) and inevitable inductive losses associated with rf induced eddy currents within the sample. Hoult and Lauterbur [3] showed that Rs ∝ ω02 B1 b5

(5)

where ω0 is the Larmor frequency, B1 is the rf field produce by unit current, b is the sample radius. One can see that the sample losses increase with frequency and become predominant at a higher frequency. It can be shown [5,6] that in such a case SNR per unit volume depends on the surface coil radius a as: SNR ∝ 1/a

Fig. 3. Two surface coils assembled into the quadrature probe. The two perpendicular B1 fields are marked with arrows (National Research Council, Institute for Biodiagnostics, Winnipeg).

(6)

This equation, as explained in detail by Hayes and Axel [6] is exact only at the surface of the coil, however, it gives a good estimate of the dependence of the SNR of the coil radius. Thus, the smaller the coil the higher SNR hence surface coils which are so popular in MRS. The diameter of a surface coil may be chosen to give maximum SNR at a particular point within the sample. If sample loses are dominant (above 1T), maximum SNR can be √ obtained for a = d/ 5 where a is the coil radius and d is the distance from the coil center [7,8]. When the coil losses are dominant [9]√then the optimum SNR occurs for the coil radius a = d 2. The exact optimum diameter depends on the frequency and the sample. For example, at 3T and 1 cm3 VOI placed at d = 2.5 cm from the coil the maximum SNR occurs, as found experimentally, when the coil diameter is 4.0 cm. This diameter lies between the coil loss dominant case (7.1 cm) and the sample loss dominant case (2.25 cm).

Part II

z

A

Surface Coil 1165

1166 Part II

Medical Uses

Part II

Superconducting rf Coils The noise in NMR comes also from the coil and depends on the temperature: Rc ∝

 µµ0 ω0 ρ(Tc )

(7)

This equation could be used to decrease noise thus increase SNR by decreasing the coil resistance. The recent introduction of high-temperature superconductive materials [10] allowed new possibilities for this increase of SNR. This new technology, however, brings new technical challenges, such as cryocooling devices, high-temperature superconducting material, patient safety, etc., and has not yet found clinical applications. It is mostly used in micro-MRI research[10,11] and in chemistry MR labs.

Phased Array The success of the surface coils in producing high SNR also, unfortunately, reduces FOV. To rectify this limitation a phased array technology was introduced to extend the FOV over the sample [12]. Each element of the array, coupled to an independent receiver channel, works as an independent surface coil maximizing SNR and increasing FOV (Figure 4). Recently, the so-called parallel imaging techniques using phased array have been introduced into MRI reducing the total acquisition time [13,14]. Despite the costs and complexity of the electronics 8, 16, or even 32-channel phased array coils are become a clinical standard. Because complexity of

Fig. 4. Phased array (Siemens, Germany).

the MRS experiment (e.g. requirements of the phase control of rf pulses) the phased array technology is only slowly being introduced into MRS.

B1 Homogeneity Vs. SNR In quantitative MRS, metabolite concentration cannot be properly evaluated when the rf coil produces an inhomogenous B1 field. Even excellent MRS techniques [15,16] require homogenous transmit and receive rf fields. The inhomogenous B1 field causes a variety of inconveniences, such as imperfect spatial localization [17,18], poor water suppression [19], and contamination from outside the voxel of interest [20]. B1 inhomogeneity also prevents the quantification of metabolites concentration as the spectra would then depend on the tissue–coil orientation. This would cause different spectra from the same tissue, which is of particular concern for heterogenous tissues such as tumors. Therefore, the application of rf coils producing spatially uniform B1 field is preferred. This conclusion contradicts the previous one we recorded, i.e. that the best SNR is produced by a surface coil, which unfortunately produces a very inhomogenous rf field. Volume coils, such as the saddle coil, the solenoidal coil, the Adelman-Grant, the Helmholtz pair, and birdcage (Figure 5), etc. produce very homogenous field. Unfortunately, they suffer from low sensitivity and

Fig. 5. Geometrical model of the 16-element birdcage rf coil.

Radio Frequency Coils for MRS

Diode switch

Tx Rx

Transmit-Only and Receive-Only Coils So far we have only considered coils that are transmit and receive (Tx /Rx ). In this case the excitation and receive rf profiles are identical. However, in most clinical systems the standard configuration is volume coil—transmit-only (Tx ), local coil—receive-only (Rx ). This arrangement comes mostly from manufacturer convenience, since in this case the sample is excited always with the same, usually body rf coil and SAR is calculated for this one coil only. This configuration allows also easy attaching other receive coils dedicated for any particular body part. The set-up also avoids problems with inhomogenous excitation, yet reception remains non-uniform. Other difficulties with separate Tx and Rx coils are associated with SAR limits: the large body coil requires a lot of rf power (typically for 3 T clinical systems up to 30 kW) due to the unnecessary excitation of organs outside the VOI. The coupling between the coils [24–26] is yet another problem. All these challenges increase rapidly with the field strength. Leaving aside SAR issues (which is particularly a pulse sequence issue) we will consider decoupling. The decoupling of the coils is needed to avoid any current being generated in the receive coil during the transmission. The problem here is that, during the transmission, mutual inductance between coils current is induced in the receiving coil, which generates an opposing B1 field. During the receiving period the noise currents produced by the larger Tx coil are also induced in the receive coil. Both effects cause the degradation of the SNR and artifacts in MRI. To separate transmit and receive coils, crossed diodes (Figure 6) are used in series with the Tx coil as well as in parallel with the tuning capacitor of the receive coil. The solution as proposed by Bendall and Edelstein et al. [24] is to detune the receive coil during the power transmission and detune the Tx coil during the receiving period. The cross diodes in the Tx coil conduct during the trans-

Fig. 6. Decoupling model of the transmit and receive coils. The cross diodes in the transmit coil conduct during the transmission while diodes in the receive coil are open during the receiving. The receive coil is actively decoupled using externally driven PIN diodes during the transmission.

mit (high voltage) thus the coil is resonating at the Larmor frequency; during the receive period, when the current is small, the open diodes prevent the current from flowing in the Tx coil. An alternative to using active switching may also be used: the receive coil is actively decoupled using externally driven PIN diodes. The diodes are placed at a certain distance from the capacitance so the cable creates an inductance between the diodes and the capacitance. During the transmission the diodes are switched on and create a short-circuit. This way the Rx coil is not active. The same assembly could also be used without an external drive (passive mode): during transmission the small current, induced in the Rx coil by Tx coil turns the diodes on and such a circuit has a high impedance and prevents current flow in the Rx coil. During the reception the current is low the diodes are by-passed by the signal. In addition the Tx and Rx coils can be design to be electrically orthogonal each to other if the space allows. At higher magnetic fields (>3T), when the rf wavelength within the sample becomes comparable with the sample size, and a body rf coil is used for the excitation, a new problem arises: the complex interaction between the body tissue and the rf field. This problem is discussed in detail by Hoult [27]. The other problem when using the separate transmit and receive coils is that the relative direction of B1 field may be different for transmit and receive coils. This may also degrade the signal [28–30] as well efficacy of the shimming procedures [31,32].

Part II

high power requirements compared to the surface coil. Because the surface coil, as stated before, generates a very inhomogenous rf field, it provides spatially dependent flip angles, which is particularly inconvenient when more than one pulse is used—as is common in MRS. There are few methods, none of these being perfect, that try to deal with this problem. One of them is the application of adiabatic rf pulses [21,22], that, however, are suitable only for a single-shot techniques. They also require high power [thus increased specific absorption ratio (SAR)], strong gradients, and they also limit echo time (TE). A very elegant solution was proposed by Bendall [23] who proposed the application of multiple inhomeogenous rf coils, that take advantage of the spatial dependence of flip angles to increase the localization and allow a larger FOV.

Transmit-Only and Receive-Only Coils 1167

1168 Part II

Medical Uses

Part II

Implanted rf Coils l1

l2

X

Fig. 7. Multi-ring surface coil assembly. Currents I1 , I2 flow in opposite directions to compensate B1 inhomogeneity of each separate surface coil.

Local rf Coils with Improved B1 Homogeneity and SNR An interesting solution [33,34] to generate identical fields for both excitation and receiving periods was proposed: the combination of surface coil sensitivity with volume coil homogeneity. This idea uses two or more inductively coupled rings of different diameter, carrying current in opposite directions (Figure 7). Such an assembly produces homogenous, identical excitation and receiving B1 in the desired region of interest while providing SNR comparable with the surface coil. While this coil seems to be perfect for MRS its major disadvantage is fixed VOI, what in practice means that for each VOI another rf coil should be constructed. The possible applications of this coil include skin or thyroid MRS, where the VOI varies little between subjects. Another solution to increase SNR is the application of rf shields [35]. This concept works particularly well for breast MR imaging and spectroscopy. If the B1 field is perpendicular to the chest wall, a Helmholtz pair probe gives good field homogeneity over the breast and good chest penetration, but significant chest power dissipation with concomitant loss of SNR. Weak B1 fields covering the large volume could generate substantially larger losses than bigger fields confined to a smaller volume. Thus, the combination of the Helmholtz pair with the rf shielding of the chest wall away from the breast brought a substantial improvement of up to a factor of two when compared with the standard breast rf coil.

For imaging of some small, deeper organs, for example, the spinal cord, the rf surface coils are used but due to their field drop with the distance from the coil, they bring even more challenges in the coil design. Because of that the SNR within the VOI is compromised due to the low and inhomogenous B1 field. The noise source from the “disturbing” tissue (e.g. fat) between the coil and the VOI, where the B1 is strong, is yet another source of noise. One of the possible ways of dealing with this problem is the application of implanted rf coils [36,37]. Briefly, the implanted coil consists of a loop of wire (the “actual” coil generating and receiving rf signal) tuned to the Larmor frequency, and a matching ring inductively coupled to the loop. Thus, no electrical connections between the implanted coil and the external matching network are required. Such an implanted rf coil provides better SNR than the surface coils, thanks to their close proximity to the VOI. It also brings some disadvantages: the procedure can no longer be considered non-invasive, and as such its application to human MRS becomes questionable. In addition, implanted coils are vulnerable to various failure modes including mechanical and electrical. The blood and body fluids can change the coil loading or even its position, thus SNR during the MRS experiment. In some cases matching also becomes difficult due to the restricted access of the matching coil and the loop.

Microcoils When using ultra-high fields and ultra-low samples (∼1 µl) new challenges arise such as susceptibility, low filling factors, increased coil noise, skin depth, etc. As shown [38,39] for the coils of diameter of order of millimeters or less the major source of the noise comes from the coil. Therefore, special attention must be placed to the resistance of the coil. A so-called single and multilayer rf microcoil, using metallic foil rather than wires was used to enable very high field ex vivo MRS [40]. The performance of the standard rf surface coil can also be surpassed by a microstrip rf surface coil for extremely high-field MRI as reported by Zhang et al. [41]. This design is based on the microstrip transmission line concept that does not require the use of lumped elements. The microstrip surface coil can be used up to 500 MHz. Such a coil gives about 20% better SNR when compared to the standard surface coil [41]. This concept can also be used with any other configurations such as quadrature volume or phase array rf coils.√Such an assembly resonates at the frequency f = c/2L εeff where c is the speed of light, L length of the coil, εeff dielectric constant of the material. This concept has not yet found a clinical use but may be promising.

Radio Frequency Coils for MRS

Quantification of the tissue metabolites by heteronuclear MRS can be achieved using dual frequency rf coils. The dual frequency coils allow simultaneous measurement of two nuclei frequencies, without disruptions associated with exchanging coils for imaging and spectroscopy examinations. One is usually 1 H while the other is phosphorus (31 P), sodium (23 Na), or other nuclei (23 Na). The proton frequency is often used for high-resolution imaging that allows lesion localization while the other frequency is used for spectroscopic lesion analysis. The proton frequency also provides the sensitivity needed for shimming. Dual frequency coils have to deliver SNR of the single frequency rf coil as well as produce homogenous rf field over the sample volume. To achieve these needs double-tuned quadrature birdcage coils were developed. One example is a design based on two coaxial birdcage coils resonating at 31 P and 1 H frequencies each. Such a coil provides excellent 31 P performance although lower SNR for 1 H when compared to a single frequency 1 H coil. This design allows altering the 31 P coil length thus can provide better SNR sacrificing however B1 [42]. Another solution is to use the same sized coils and tune every other leg of the coil to the desired frequency, using trap circuits and inductive matching to optimize the coil [43,44]. An elegant design based on transmission line resonators for a volume quadrature, double-tuned coil was also proposed by Vaughan [45]. In our work to even further increase SNR for two or more frequencies, a dual frequency surface coil was designed using frequency splitting circuitry [46]. This construction allows concurrent 1 H and 31 P MRS and provides the same B1 homogeneity for both frequencies almost without reducing SNR when compared to the single frequency rf coils. Such a design was further improved by combining a multi-ring coil concept [47] with a splitting circuit. This coil provides better SNR than a volume coil within the specified VOI, but about 70% of the SNR of the single frequency surface coil, however, it creates homogenous and identical B1 fields for both frequencies over the desired VOI. The dual frequency coils require dual frequency receive and transmit channels, which are still not standard in clinical MRI systems, hence their limited application.

Summary This chapter describes the most common rf coils for MRS as well as possible methods of improving SNR and B1 homogeneity. As concluded, there is no perfect rf coil for MRS. Compromises must be made among SNR, B1 homogeneity, voxel size, total experiment time, size, and position of VOI. The choice of the coil depends on the application, field strength, available MRI system, and the

patient position. The volume coil—transmit and surface or local coil receive-only seem to be the most effective, yet not perfect rf coil configuration.

Glossary of Terms SNR B1 M0 µ0 Q K η ω0 Vc F k Tc f

signal-to-noise ratio rf field magnetization permeability of free space quality factor of the coil numerical factor depending on the coil geometry filling factor Larmor frequency volume of the coil noise figure of the preamplifier Boltzmann constant probe temperature bandwidth (in Hertz) of the receiver

References 1. Abragam A. The Principles of Nuclear Magnetism. Clarendon Press: Oxford, 1961, p 82. 2. Hoult DI, Richards RE. J. Magn. Reson. 1976;24:71. 3. Hoult DI, Lauterbur, PC. J. Magn. Reson.1979;34:425. 4. Chen CH, Hoult DI, Sank VJ. J. Magn. Reson. 1983;54: 324. 5. Hyde JS, Jesmianowicz A, Grists TM, Froncisz W, Kneeland JB. Magn. Reson. Med. 1987;4:179. 6. Hayes CE, Axel L. Med. Phys. 1985;12(5):604. 7. Edelstien WA, Foster TH, Schenk JF. The relative sensitivity of surface coils to deep lying tissue. Proceedings of the SMRM 4th Annual Meeting, London, 1985, p 964. 8. Wang J, Reykowski A, Dickas J. IEEE Trans. Biomed. Eng. 1995;42:908. 9. Chen CN, Hoult DI. Biomedical Magnetic Resonance Technology. Adam Hilger: Bristol, 1989, p 161. 10. Hill HDW. Solid State Commun. 1997;102:169. 11. Darrasse L, Ginefri JC. Biochemie. 2003;85:915. 12. Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. Magn. Reson. Med. 1990;16:192. 13. Sodickson DK, Manning WJ. Magn. Reson. Med. 1997;38: 591. 14. Pruessmann KP, Wieger M, Scheidegger MB, Boesiger P. Magn. Reson. Med. 1999;42:952. 15. Bottomley PA. Selective volume method for performing localized NMR spectroscopy. US Patent 4,480,228, 1984. 16. Frahm J, Merboldt KD, Hanicke W. J. Magn. Reson. 1987;72:502. 17. van Zijl PCM, Moonen CTW, Alger JR, Cohen JS, Chesnick SA. Magn. Reson. Med. 1989;10:256. 18. Lawry TJ, Kaczmar GS, Weiner MW, Matson GB. Magn. Reson. Med. 1989;9:299. 19. Moonen CTW, van Zijl PCM. J. Magn. Reson. 1990;88:28. 20. Lawry TJ, Kaczmar GS, Weiner MW, Matson GB. Magn. Reson. Med. 1989;9:299.

Part II

Dual Frequency rf Coils

References 1169

1170 Part II

Medical Uses

Part II

21. de Graaf RA, Nicolay K, Garwood M. Magn. Reson. Med. 1996;35:652. 22. de Graaf RA, Nicolay K, Garwood M. J. Magn. Reson. B. 1996;113:35. 23. Bendall MR. Chem. Phys. Lett. 1983;99(4):310. 24. Edelstein WA, Hardy CJ, Mueller OM. J. Magn. Reson. 1986;67:156. 25. Haase A. J. Magn. Reson. 1985;61:130. 26. Barberi EA, Gati JS, Rutt BK, Meonon RS. Magn. Reson. Med. 2000;43:284. 27. Hoult DI. J. Magn. Reson. Imaging. 2000;12:46. 28. Crowley MG, Evelhoch JL, Ackerman JJH. J. Magn. Reson. 1985;64:30. 29. Styles P, Smith MB, Briggs RW, Radda GK. J. Magn. Reson. 1985;62:397. 30. Chen CH, Hoult DI. Biomedical Magnetic Resonance Technology. Adam Hilger: Bristol, 1989, p 157. 31. van Zijl PCM, Moonen CTW, Alger JR, Cohen JS, Chesnick SA. Magn. Reson. Med. 1989;10:256. 32. Crozier S, Field J, Brereton IM, Moxon LN, Shannon GF, Doddrell DM. J. Magn. Reson. 1991;94:123. 33. Tomanek B, Ryner L, Hoult DI, Kozlowski P, Saunders JK. Magn. Reson. Imaging. 1997;15: 1199. 34. King S, Ryner L, Tomanek B, Sharp J, Smith I. Magn. Reson. Med. 1999;42:655.

35. Tomanek B, Hoult DI, Chen X, Gordon R. Magn. Reson. Med. 2000;43:917. 36. Silver X, Xu W, Mercer EV, Beck BL, Bossart EL, Inglis B, Mareci TH. Magn. Reson. Med. 2001;46:1216. 37. Arnder LL, Shattuck MD, Black RD. Magn. Reson. Med. 1996;35:727. 38. Cho Z, et al. Med. Phys. 1998;15:815. 39. Peck T, Magin R, Lautenbur P. J. Magn. Reson. B. 1995;108: 114. 40. Grant SC, Murhpy LA, Magin RL, Friedman G. IEEE Trans. Magn. 2001;37(4). 41. Zhang X, Ugurbil K, Chen W. Magn. Reson. Med. 2001;46: 443. 42. Fitzsimmons JR, Beck BL, Brooker HR. Magn. Reson. Med. 1993;30:107. 43. Matson GB, Vermathen P, Hill TC. Magn. Reson. Med. 1999;42:173. 44. Shen GX, Boada FE, Thulborn KR. Magn. Reson. Med. 1997;38:717. 45. Vaughan JT, Hetherington HP, Out JO, Pan JW, Pohost GM. Magn. Reson. Med. 1994;32:206. 46. Schnall MD, Subramanian H, Leigh JS, Chance B. J. Magn. Reson. 1985;65:122: 47. Volotovskyy V, Tomanek B, Corbin I, Buist R, Tuor UI, Peeling J. Magn. Reson. Eng. Concepts in Magnetic Resonance, Part B 2003;17B:11.

1171

M. Albert Thomas1 , Amir Huda1,2 , Hyun-Kyung Chung1 , Nader Binesh1 , Talaignair Venkatraman1 , Art Ambrosio1 , and Shida Banakar1 1 Department

of Radiological Sciences, University of California, Los Angeles, CA 90095, USA; and 2 Department of Physics, California State University, Fresno, CA 93740, USA

Introduction Acquisition of three-dimensional (3D) spatially localized, water-suppressed one-dimensional (1D) 1 H MR spectra has been optimized in human tissues over the last two decades [1–6]. Introduction of automated MRS a decade ago has enabled a 1–2 min pre-scan to calibrate the flipangle using the transmitter power, to adjust the static field homogeneity (B0 ) using the linear gradient shim coils, and to optimize water suppression, which used to be an at least 15 min operation, facilitating the integration of clinical MRS protocols into routine MRI examinations [7]. In 1 H MR spectroscopy of human brain, spectral resonances due to methyl, methylene, and methine protons of N-acetylaspartate (NAA), glutamate/glutamine (Glx), creatine (Cr), choline (Ch), myo-inositol (mI), GABA, aspartate (Asp), N-acetylaspartyl-glutamate (NAAG), and other metabolites have been identified in the region of 0–4.5 ppm [2–12]. One-dimensional MRS in vivo is hampered by the fact that it is difficult to resolve a multitude of peaks existing over a small spectral range of approximately 300 Hz at 1.5 T. Spectral overlap of macromolecules, GABA, and glutathione with the methyl resonances of NAA, Cr, NAAG, and methylene resonances of Glx has also been reported [8,14–15]. Attempts have been made to quantify Glx [16], glucose [17], and other metabolites with only minimal success due to the difficulty in extracting this information from a region with many overlapping resonances. Using spectral-editing techniques, it is possible to select a particular J -coupled metabolite [2,14,18– 20]. However, one major drawback of editing techniques is that only one metabolite is optimized at a time assuming that the multiplets of the J -coupled metabolites are well separated. Using the LC-Model post-processing algorithm, quantification of several metabolites in human brain in vivo has been reported recently using a basis-set of 10–20 different metabolite spectra in vitro [21–23]. Unlike the spectral editing techniques, which target one metabolite at a time, two-dimensional (2D) MRS Graham A. Webb (ed.), Modern Magnetic Resonance, 1171–1183.
C 2008 Springer.

can unambiguously resolve many overlapping peaks nonselectively in vitro as shown by Ernst and co-workers two decades ago using the vertical bore NMR spectrometers [24–25]. 2D-MR spectroscopy enables converting a crowded overlapping 1D MR spectrum to a betterresolved 2D spectrum through the addition of a spectral dimension. Instead of a standard 1D spectrum plotting intensity vs. a single-axis (i.e. chemical shift), 2D MR spectroscopy techniques produce a 2D spectrum plotting intensity vs. two axes, the dimensions of which depend on the specific 2D MR technique [25]. Better dispersion of several metabolite peaks and improved spectral assignment make the proposed technique more attractive. It seems natural to explore the clinical potentials of 2D MRS techniques of human tissues in vivo. There have been several attempts during the last 15 years in the implementation and evaluation of 2D NMR spectroscopy on in vivo MRI scanners and MR spectrometers [26–30]. The aim of this chapter is to discuss the recent progress on the implementation of selected localized 2D MRS sequences on the whole body 1.5 T and higher field MRI scanners. This includes a summary of 1D spectroscopy with and without chemical shift imaging (CSI), followed by single- and multi-voxel based 2D spectroscopy dealing with coherent interactions such as chemical shift and J -coupling. A brief overview of potential artefacts in 2D spectroscopy and simulation is also included.

Single- and Multi-voxel Based 1D 1 H MR Spectroscopy Clinical 1D MR spectroscopy has some important differences and challenges compared to MR imaging. Often how the challenges are resolved define the particular technique of spectroscopy with its limitations and advantages. Some of these basic challenges are pertinent to water and lipid suppression, spatial localization of the volume of interest (VOI) including single vs. multiple volume elements (voxel) as shown in Figure 1.

Part II

Spatially Resolved Two-Dimensional MR Spectroscopy in vivo

1172 Part II

Medical Uses

Part II

WATER SUPPRESSION (CHESS or STIR)

OUTER VOLUME SUPPRESION (OVS)

Volume localization STEAMSV or PRESSSV or STEAMCSI or PRESSCSI

Data Acquisition

Spoiler gradients

Signal Averaging

TR

Fig. 1. Different modules of 1D MRS protocol.

Water and Lipid Suppression What makes MR imaging quite sensitive is the large concentration of water in biological tissue. However, the dominance of water limits spectroscopy since the metabolites are tens of thousands of times lower in concentration than water. Thus, in order to obtain a visible signal from the metabolites, MRS techniques must incorporate ways of minimizing or eliminating the water signals. This water suppression can be extended from changes in hardware for data acquisition (digitizers), pulse sequences, to postprocessing software. The primary contribution stems from an addition of three frequency selective (water as the target frequency) radio frequency (rf) pulses with arbitrary flip-angles followed by a dephasing gradient to the pulse sequence. These pulses are commonly called the chemical shift selective saturation (CHESS) pulses [5,31]. Instead of water saturation using CHESS, an inversion recovery scheme can also be employed with a 180◦ frequency selective pulse on water such that excitation of the metabolites can be prolonged until the crossing of the water signal at the null point with known T1 values [32]. Quite often the lipid signals from either within the VOI or outside of the VOI hold a similar challenge like water. In order to suppress the lipids, the same two approaches utilized in water suppression in terms of pulse sequence modification can also be applied here [33]. The most common approach is the use of frequency selective (targeted at lipids) pulses in conjunction with spatial selection around the edges of the brain called the outer volume saturation or suppression (OVS) slabs. Of course, post-processing software can also contribute to elimination of lipid signals.

are attributed to the spatial shifts introduced via the application of the frequency encoding gradients during signal sampling to reproduce the selected slice within a plane or a data point in volumetric MRI. A frequency encoding gradient cannot be used in spectroscopy during sampling because the differences in frequencies of metabolites sought after are due to the small chemical shifts and a frequency encoding gradient would corrupt this basic information. However, the demand for volume definition remains, whether single voxel or multiple voxel spectroscopy is done since the metabolite concentration information is sought from a particular voxel and not the whole brain. The most common localization procedure is the use of slice selection and phase encoding gradients. These serve to demarcate the dimensions of the voxel and are operated for a very short duration. The spins then return to the influence of the external magnetic field.

Single Voxel Spectroscopy The two most popular pulse sequences for defining a small (ranging from 1 cm3 to 8 cm3 ) volume within the anatomy are termed as STEAM and PRESS [3,9,34]. Both PRESS (90◦ ss −180◦ ss −180◦ ss ) and STEAM (90◦ ss −90◦ ss −90◦ ss ) use frequency selective RF pulses along with magnetic field gradients to isolate a single volume, called STEAMSV and PRESSSV; however, the timing and the sequence of the pulses has consequences on the sensitivity and the strength of the signal [35].

Multivoxel Spectroscopy or Chemical Shift Imaging

Spatial Localization of the Volume There is a fundamental difference between the localization done in imaging and spectroscopy. Imaging relies on acquisition of the entire proton signal (primarily water and fat) and the differences in frequencies of the signal

This technique known as spectroscopic imaging (SI) or CSI is challenged with distilling both chemical shift as well as spatial information from the acquired region of interest [36–37]. The data processing in this case requires both spatial and chemical shift dimensions. For CSI, the

3D Localized 2D MR Spectroscopy

Single Volume Localized 2D 1 H MR Spectroscopy For MR spectroscopy of a particular nucleus, namely hydrogen (1 H), one counts on the differences in internal spin interactions which are due to different interand intra-molecular environments. These differences in the molecular environment are of several types and are, hence, responsible for different features of a spectrum or an image. For example, the translational molecular motion is exploited in a diffusion weighted image. While the chemical shift expresses the local electronic environment and is predominantly an intra-molecular property, molecular motion also influences this shift significantly. A primary manifestation and measure of the chemical bonds between the nuclear spins, i.e. the indirect spinspin coupling communicated through the covalent bonds, is the exclusively intra-molecular J -coupling, which is independent of the applied magnetic field. The motive for introducing the 2D MRS sequences to whole body MR

spectroscopy is simple—the extraction of more information out of the MR spectrum to aid in the biochemical analysis of human tissue in vivo. 2D spectroscopy can do this by converting a crowded, overlapping 1D MR spectrum to a better-resolved 2D spectrum through the addition of a spectral dimension [24,25]. The problems due to the overlap in the 1D spectrum introduced by the multiplicity of various peaks due to J -coupling could be removed by separating the interactions due to chemical shift and indirect spin-spin coupling (J ) along the two dimensions in 2D MR spectroscopy. Instead of a standard 1D spectrum plotting intensity vs. a single-axis (i.e. chemical shift), 2D NMR spectroscopy techniques produce a 2D spectrum plotting intensity vs. two axes, the dimensions of which depend on the specific 2D NMR technique. The Nobel prizes were awarded to Prof. Richard Ernst in 1991 and Prof. Kurt Wuthrich in 2002 for their innovative contributions focused on coherent and incoherent intra- and inter-molecular interactions using multidimensional MR spectroscopy in vitro [38]. Exploitation of this interaction is the crucial link to the molecular chemistry of the metabolites in human tissues. Before applying these multidimensional MR techniques in human tissues, each sequence needs to be converted first into localizing a VOI using three orthogonal slice-localizing rf pulses. A second task is to minimize the number of rf pulses essential for creation of Hahn’s spin-echo or coherence transfer echoes. Similar to the 1D MRS protocol, shown in Figure 2 are different compartments of 2D MRS protocol. In addition to repeated spectral acquisition for signal averaging, the sequence has to be repeated multiple times for encoding the second spectral dimension.

Second spectral sampling

CHESS

OVS

JPRESS COSY SECSY CT-COSY CT-PRESS

Signal Averaging

TR Fig. 2. Different modules of 2D MRS protocol.

Data Acquisition

Spoiler gradients

Part II

PRESS and STEAM sequences described above can be used including the field gradients modified for spatial encoding (STEAMCSI and PRESSCSI). The phase encoding gradients are repeated several times between the remainder of the rf pulses each time repeating the entire sequence but with the phase encoding gradient amplitude changing slightly. This successive change is done in a stepwise fashion with the gradient amplitude changing; thus, several samples of a frequency are produced. The data can then be Fourier transformed to allocate signal intensities of different frequencies to appropriate voxels.

Single Volume Localized 2D 1 H MR Spectroscopy 1173

1174 Part II

Medical Uses

Part II

2D J-Resolved MR Spectroscopy (JPRESS) As discussed earlier, the conventional PRESS sequence uses three slice-selective rf pulses (90◦ , 180◦ , and 180◦ ) along three orthogonal directions with the second spinecho originating from a voxel at the intersection of three orthogonal slices. The original 2D J -resolved homonuclear MR spectroscopic sequence is a modified Hahn spinecho sequence [90◦ -t1 /2-180◦ -t1 /2-acquire(t2 )], where signal is acquired during t2 and the incremental period t1 is the additional time domain to sample the second spectral dimension of a 2D MR spectrum [25]. In 1995, Ryner et al. coded the first 3D-localized 2D J -resolved sequence (JPRESS) by modifying the PRESS sequence to allow simultaneous incrementation of the two periods before and after the last 180◦ rf pulse [39,40] as shown in Figure 3. The localized 2D JPRESS sequence has been implemented on both GE and Siemens 1.5 and 3 T MRI scanners [39–46]. Shown in Figure 4 are 2D JPRESS spectra of a 100 mM Lac phantom using a Siemens 3T Trio MRI scanner before and after rotating the entire 2D dataset by 45◦ . A voxel size of 27 ml and a head transmit/receive coil were used. The time period  was kept to the minimum necessary to play out rf and gradient pulses. The evolution time (t1 ) was incremented in 10-ms steps in order to achieve a spectral window of ±50 Hz that is well within the range of homonuclear (1 H–1 H) J -coupling. A digital resolution of 1024 complex points along t2 over a sweep width of 2000 Hz and TR of 2000 ms was used. The total duration was 4.3 min using two averages per t1 and 64 increments. However, a longer acquisition time of 17 min will be necessary for signal averaging the metabolite signals at the physiological concentrations. Water suppression was achieved using CHESS. The refocusing of the chemical

90°

∆

180°

∆

..t1/2..

∆'

180°

∆'

..t1/2..

t2...

RF

X

Y

Z Fig. 3. Localized 2D J-resolved MRS sequence (JPRESS) consisting of three slice-selective rf (90◦ , 180◦ , 180◦ ) pulses.

shift by the last 180◦ RF pulse resulted in only J -coupling information along the F1 dimension. A 45◦ rotation of the 2D data matrix results in J -couplings and chemical shifts along orthogonal axes F1 and F2 as shown in Figure 4B. As reported earlier [40,41], additional cross peaks due to strong coupling have been observed even at 3 T. The peaks due to the non-coupled spins are also retained in the 2D J -resolved spectrum unlike the 1D edited spectra. Another advantage of this sequence is that 100% of the available magnetization is retained, neglecting losses caused by T2 relaxation and spin diffusion. Hence, it is more advantageous than some of the filtered spectral editing techniques, in which 50% or more of the available magnetization is lost. The 2DJ -resolved pulse sequence is versatile such that it can be used for the acquisition of a standard 1D PRESS spectrum by simply removing the incrementation of the t1 -period. A homonuclear broadband decoupled 1 H spectrum can also be retrieved after projecting the 2D JPRESS cross peaks onto the F2 dimension as demonstrated in Figure 4B. This technique has been further extended to human brain and prostate tumours recently to select certain metabolites [42–43]. An over-sampled 2D J -resolved sequence without any water suppression has also been demonstrated [44–46].

Chemical Shift Correlated MR Spectroscopy (L-COSY) Figure 5 shows a localized 2D correlated spectroscopy (L-COSY) sequence [47]. Similar to JPRESS, the VOI was localized in ‘one-shot’ by a combination of three slice-selective rf pulses (90◦ -180◦ -90◦ ), a new MRS volume localization sequence called CABINET (coherence transfer based spin-echo spectroscopy). The last sliceselective 90◦ RF pulse acts also as a coherence transfer pulse for the 2D spectrum In addition to the slice-selective 180◦ rf pulse, there are B0 gradient crusher pulses before and after the last 90◦ rf pulses. The incremental period for the second dimension can be inserted at two different locations: first, immediately after the formation of the Hahn spin-echo (t1 ) and second, after the last gradient crusher column (k ∗ t1 ), where k ∗ t1 can be 4 µs or minimum allowed by the scanner hardware. The L-COSY sequence can be considered a single-shot technique in terms of simultaneous volume localization and coherence transfer. In contrast to PRESS, the CABINET sequence retains only 50% net signal from the localized volume due to a selection of only N-type echo, enabled by the B0 gradient crusher pulses [47]. If a surface coil is used for both transmission/reception, one has to consider the influence of flip-angle errors due to the delivery of the inhomogeneous rf pulses delivered by a surface coil. In the CABINET sequence with the

3D Localized 2D MR Spectroscopy

Single Volume Localized 2D 1 H MR Spectroscopy 1175

flip-angle distribution of (ϕ ◦ , 2ϕ ◦ , ϕ ◦ ), the total signal amplitude from the VOI will be scaled by a factor of sin4ϕ. The maximum attenuation coefficient for the PRESS sequence (ϕ ◦ , 2ϕ ◦ , 2ϕ ◦ ) will be sin5ϕ. The diagonal and cross peaks of an L-COSY spectrum have mixed phases

along the F1 axis. In contrast to the amplitude modulation in conventional COSY, the phase modulation in L-COSY is caused by the evolution during the B0 gradient pulse before the last 90◦ rf pulse. Pure phase L-COSY spectra can be recorded using a quadrature detection method along the

Part II

Fig. 4. Localized 2D JPRESS spectra of 100 mM Lac phantom. (A) before and (B) after 45◦ rotation of the 2D data matrix.

1176 Part II

Medical Uses

Part II

90°

∆

180°

∆

...t1...

∆'

90°

∆'

...k ×t1..t2...

RF

X

Y

Z Fig. 5. The 2D L-COSY sequence consisting of three sliceselective rf (90◦ , 180◦ , 90◦ ) pulses.

F1 axis described by Doddrell and co-workers [48,49] that will require two separate P- and N- type spectral acquisition and recombination of the two datasets. Relaxation during the gradient pulses before the last 90◦ rf pulse will cause further losses in signal intensities. Two-dimensional L-COSY has been implemented recently on GE and Siemens 1.5 and 3 T MRI scanners [41,47,50] and the test-retest reliability has been demonstrated in vitro as well as in vivo [41,51]. Shown in Figure 6 are the L-COSY spectra recorded in a) a grey matter brain phantom using a Siemens 3 T MRI scanner and b) the occipito-parietal white matter region of a 38-year-old healthy human subject using a GE 3 T MRI scanner. The brain phantom had the following metabolites at physiological concentrations: 8.9mM NAA, 0.7 mM GABA, 2.1 mM Asp, 0.9 mM Ch, 7 mM Cr, 1 mM Glc, 12.5 mM Glu, 2.5 mM Gln, 2 mM GSH, 4.4 mM mI, 0.4 mM Lac, 3.6 mM PCr, 0.6 mM PCh, 1.8 mM Tau, 0.3 mM Thr, 1 mM PE, and 1 mM DSS in a phosphate buffer solution to maintain pH of 7.2. 2D L-COSY spectra were recorded using the following parameters: 27 ml voxel, TE = 20–30 ms, TR = 2 s, total number of scans = 768 (96 t1 increments and 8NEX/t1 ). The total duration for each 2D scan with water suppression was approximately 24 min. Typical experimental parameters used for 2D L-COSY are listed in Table 1. Applications of this technique in brain and breast tumours, and brain disorders have also been reported recently [52–54]. Other researchers have also reported the implementation of different versions of 2D COSY [30,49,55–57]. A gradient enhanced COSY in combination with volume localized spectroscopy (VOSY) used the steam sequence for volume localization [49]. Due to additional signal losses, a human brain 2D COSY spectrum recorded in a gross occipital volume of 5 × 6 × 8 cm3 using a 2 T MRI scanner took a total sampling duration of 1 h and 42 min.

Simultaneous acquisition of COSY from multiple volumes of interest was proposed by Delmas and co-workers (55) and two-voxel localized COSY spectra of rat brain were recorded in a total duration of 42 min using a Bruker 4.7 T MRI scanner. A different pulse-sequence (ISISCOSY) was also proposed by Welch et al. (57), where the volume was localized by outer-volume suppressed ISI (OSIRIS). Two major drawbacks of ISIS-COSY compared to L-COSY were: (1) eight shots are necessary to achieve the VOI and (2) five rf pulses are necessary to record the single voxel localized COSY leading to more dependence on the B1 -field inhomogeneity.

Spin-Echo Correlated MR Spectroscopy (SECSY) A second variation of COSY, namely spin echo correlation spectroscopy (SECSY) was originally reported by Nagayama et al. [58]. Localized SECSY works the same way as L-COSY as shown in Figure 3 except the second incremental period (k∗ t1 ) should be kept the same as the first period with k = 1. Compared to L-COSY, the diagonal peaks of SECSY lie on (F2 = F1 = 0) and the J -cross peaks are symmetrically disposed above and below the diagonal. A localized SECSY spectrum recorded in the anterior cingulate of a 24-year-old healthy subject is shown in Figure 7. 1024 complex points along the t2 dimension, 64 points along the t1 dimensions, a 27 ml voxel and 16 averages per t1 were acquired. The resultant 2D spectrum was displayed in the magnitude mode. Typical experimental parameters for 2D SECSY are included in Table 1. An advantage of SECSY over COSY is that a smaller sweep width is needed along F1 , however there is additional T∗2 weighting during the second incremental period.

Constant-Time Based Point Resolved Spectroscopic Sequence (CT-PRESS) Constant time (CT)-PRESS based on constant time (Tc ) chemical shift encoding was recently implemented by Dreheret al. [59]. In addition to three rf pulses required for PRESS, an additional 180◦ rf pulse was used. 2D CTPRESS spectra were recorded in rat brain with improved spectral resolution using a Bruker 4.7 T MRI scanner. Two different versions of modified CT-PRESS were recently implemented on a whole body 1.5 T MRI scanner including only three-slice selective (90◦ − 180◦ − 180◦ ) rf pulses for volume localization and the CT chemical shift encoding as an integral part of volume localization [60]. In the first version, CT was inserted prior to the first spin-echo and in the second version, prior to the second spin-echo. A 2D CT-PRESS spectrum recorded in a phantom containing six metabolites, namely 12.5 mM NAA, 10 mM Cr, 3 mM Ch, 7.5 mM mI, 12.5 mM Glu and 5 mM Lac at pH = 7.3 using a GE 1.5 T MRI scanner is shown

3D Localized 2D MR Spectroscopy

Residual Water

GSH

Fig. 6. 2D L-COSY spectra of (A) a grey-matter brain phantom (Siemens) and (B) occipito-parietal white matter region of a 38-year-old healthy human brain (GE) using 3 T MRI scanners.

NAA

Cr Cr Ch mI

Asp NAA

GIx

PCh mI-Ch PE mI Asp

Tau

NAA GABA

GIx

Thr/Lac

NAA

B Residual Water

PCh GSH

Ch

Cr mI

Cr Asp

mICh PE mI

Tau

Asp NAA

GIx GABA/MM

Thr/Lac

NAA GIx

lipids

Part II

A

Single Volume Localized 2D 1 H MR Spectroscopy 1177

1178 Part II

Medical Uses

Constant-time Based Correlated MR Spectroscopy (CT-COSY)

TR /TE

2000/20–30 ms

Number of averages/t1 Number of increments (t1 ) RF pulse widths (90◦ /180◦ ) Digital resolution (t2 ) Voxel size Duration of crusher gradients Spectral width (F2 /F1 )

8–16 64–128 3.2/5.2 ms Complex 1024–2048 18–27 ml 4–6 ms (2000–5000)/(625–1250) Hz

Constant time (CT)-COSY can be easily derived from CTPRESS by converting the last 180◦ into 90◦ rf pulse [61]. Four versions of CT-COSY sequence with the following slice-selective (ss) rf pulses can be easily conceived as described below: (A) 90◦ss − (Tc − t1 ) − 180◦ss − t1 − 90◦ss −Acquire (t2 ), (B) 90◦ss − (Tc + t1 )/2 − 180◦ss − (Tc − t1 )/2 − 90◦ss − Acquire(t2 ), (C) 90◦ss − t1 − 180◦ss − (Tc − t1 ) − 90◦ss − Acquire (t2 ), and (D) 90◦ss − (Tc − t1 )/2 − 180◦ss − (Tc + t1 )/2 − 90◦ss − Acquire (t2 ), where Tc is defined as the constant time between the 90◦ss pulses. Shown in Figure 9 is a localized 2D CT-COSY spectrum recorded in a phantom containing six metabolites, namely 12.5 mM NAA, 10 mM Cr, 3 mM Ch, 7.5 mM mI, 12.5 mM Glu, and 5 mM Lac at pH = 7.3. The following parameters were used: TR = 2 s, TE = 132 ms, Tc = 125 ms, t1 = 0.8 ms, t2 = 0.8 ms, spectral widths of 2500 Hz and 625 Hz along the two spectral axes (F2 and F1 ); 1024 complex points along t2 and 128 points along t1 dimensions; the number of excitations (NEX) per t1 between 8. Compared to L-COSY, a fixed interval of Tc (> t1max ) separating the preparation and mixing periods is chosen to be of the order of (1/2J ). The precession under J -coupling is not affected by the 180◦ rf pulse during Tc . Hence, a

in Figure 8. Global water suppression was achieved using CHESS. One major advantage of 2D CT-PRESS is that a homonuclear broadband-decoupled COSY-type spectrum along the F1 -dimension.can be recorded without a second rf-channel for decoupling. Also, compared to the basic localized COSY spectra, a further increase in SNR was obtained by CT-PRESS for coupled resonances since there was no coherence transfer of magnetization between the J -coupled protons leading to disappearance of crosspeaks. Two major drawbacks of the optimized 2D CTPRESS are: 1) The signal amplitude in CT-PRESS depends on Tc . 2) The spectrum has to be acquired in the 2D mode, which requires long acquisition time.

Residual Water NAA Ch mI

PDP NAA Glx

20 0 −40

mI Asp

mICh

GABA PE GIx

NAA

Fig. 7. 2D localized SECSY spectrum of recorded in the occipito-parietal grey matter region of a 24-year-old healthy volunteer using a 1.5 T GE MRI scanner.

Thr/Lac

4.8

4.4

4.0

3.6

3.2

2.8

F2 (ppm)

2.4

2.0

1.6

1.2

F1 (Hz)

40

60

80

Cr

Cr

−80

Part II

Table 1: Experimental parameters for 2D MRS

3D Localized 2D MR Spectroscopy

NAA Glu NAA

Ch Cr

mI

Cr

mI

Cr NAA

Ch

NAA GIu Lac

Residual Water

mI

Cr

Ch

Cr

NAA

Glu

NAA

mICh

mI

NAA GIu

Lac

Fig. 9. A CT-COSY spectrum of recorded in a phantom containing six metabolites using a 1.5 T GE MRI scanner.

Part II

Fig. 8. A CT-PRESS spectrum of recorded in a phantom containing six metabolites using a 1.5 T GE MRI scanner.

Residual Water Cr

Single Volume Localized 2D 1 H MR Spectroscopy 1179

1180 Part II

Medical Uses

Part II

further increase in SNR was obtained by CT-PRESS for coupled resonances since there was no coherence transfer of magnetization between the J -coupled protons. Secondly, collapse of multiplets along the second dimension was evident without using a second rf channel to achieve broadband decoupling.

Artefacts in Localized 2D MRS and Simulation In this section, a brief discussion of the commonly observed artefacts in localized 2D MR spectra in presented. First, the flip-angle errors such as an imperfect 180◦ rf pulse, can lead to ghost artefacts in 2D JPRESS. A fourstep phase-cycling scheme, namely EXORCYCLE can eliminate such artefactual peaks [25]. Second, strong coupling effect also complicates 2D JPRESS as discussed by Ryner et al. [40]. These artefacts can also impact CTPRESS and CT-COSY. Third, the asymmetric cross peaks of L-COSY/SECSY are disposed on both sides of the diagonal peaks. The asymmetry has been reported to be more severe with the upper cross peaks of NAA at (F2 = 2.5 ppm, F1 = 4.35 ppm) above the diagonal [47]. A GAMMA (general approach to magnetic resonance mathematical analysis) software package can be used to simulate different metabolite spectra [62]. The GAMMA library [62] uses a density matrix description of the spin system and provides the simulation of MRS experiments. Density matrices can be calculated for each metabolite using the reported chemical shifts and J -coupling [10]. The GAMMA codes have been extended to simulate the JPRESS, L-COSY/SECSY, CT-COSY, and CT-PRESS spectra including water suppression. As shown in Figure 10, The first 90◦ rf pulse of the L-COSY pulse se¯ to mimic quence is replaced by a binomial rf pulse (1 3¯ 3 1) the water suppression pulse sequence [63]; where the numerals give the relative pulse lengths, the bars indicate a 180◦ phase shifted pulse and equal delays τ between

90°

Multi-Voxel Based 2D 1 H MR Spectroscopy Long data acquisition has been one of the major drawbacks of single-voxel localized 2D MRS. There will be a further increase of time when conventional SI is added to 2D MRS [36–37]. In contrast to reconstructing metabolite images from the 2D spatial and 1D spectral data, the metabolite images can be reconstructed from the 2D spatial and 2D spectral data projecting the cross peak volumes into the corresponding spatial images. This methodology was termed as “cross peak imaging” (CPI) by Metzler and co-workers [65–66]. An obvious drawback of CPI is extended total acquisition time. For example, a combination of 16 × 16 spatial, 64 spectroscopic encoding steps and TR of 1 s will result in a total duration of 4.6 h [66]. Recently, implementation of circular sampling combined

180° ∆

90°

pulses. Thus 13¯ 31¯ corresponds to 11.25◦ (x)-τ -33.75◦ (x)-τ -33.75◦ (x)-τ -11.25◦ (−x), τ = 1/2f, and f is the offset frequency [63]. The phase shift is optimized based on the target a null at the water frequency. Shown in Figure 11 are (A) experimental 2D L-COSY spectrum of a phantom containing NAA, Cr, Ch, mI, Lac, and Glu and (B) simulated composite 2D COSY spectra of the same metabolites using GAMMA [64]. The parameters used for simulation were identical to that of experiments (TE of 30 ms for L-COSY, 1024 complex points along F2 and 128 points along F1 ). Another type of water-suppression artefact is due to the t1 -ridge running vertically at F2 = 4.8 ppm as evident in the experimental COSY spectrum shown in Figure 11A. This is mainly due to the scanner hardware fluctuations during the acquisition of the 2D raw matrix [25]. Also, more artefacts can arise due to wrong choice of post-processing parameters and one needs to use caution in selecting optimal apodization filters along both F2 and F1 dimensions during post-processing the 2D data.

∆

90° ...t1...

∆'

∆'

...k×t1... t2 ...

90°

90°

Acquire °

θ°

3θ°

3θ°

θ° Acquire

Fig. 10. A 2D COSY sequence used for simulation.

3D Localized 2D MR Spectroscopy

Multi-Voxel Based 2D 1 H MR Spectroscopy 1181

Part II

4.0

A

3.0

mI-Ch

NAA

F1 (ppm)

2.0

mI

1.0

GIu

Lac

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

F2 (ppm)

4.0

B

3.0

mI-Ch

2.0

F1 (ppm)

mI

NAA

1.0

GIu

Lac

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

F2 (ppm)

Fig. 11. 2D L-COSY spectra of (A) a brain phantom containing six metabolites and (B) simulated spectra for the same metabolites (Reproduced with permission from Banakar et al. [64]). (See also Plate 91 on page XVII in the Color Plate Section.)

1182 Part II

Medical Uses

Part II

with 2D J -resolved spectroscopy was demonstrated by Renshaw and co-workers using a 4 T whole body MRI scanner with a total duration of 2 h [67]. A drastic reduction of total duration has been proposed by two different multidimensional MRS sequences. First, Adalsteinsson and Spielman introduced the first application of time-varying gradients during the read-out time to acquire multi-voxel based 2D J -resolved spectral data in phantom solutions containing lactate and ethanol using a 1.5 T MRI scanner [68]. Spiral-based k-space trajectories were used in conjunction with a 2D J -resolved sequence. By preceding the spiral read-out with a 2D preparation scheme, 128 spectral encoding with TR of 2 s resulted in a total duration of 17 min. An asymmetric variant of echo-planar spectroscopic imaging (EPSI) consisting of a train of trapezoidal readout gradients, each followed by short refocusing pulse was implemented by Meyer et al. [69]. A constant-time based COSY was used and the total acquisition was 34 min. Application of this technique was demonstrated in rat brain using a 4.7 T MRI scanner and the cross peak images of mI and Tau were recorded only. In an alternative approach, COSY was combined with spectroscopic ultra-fast low angle rapid acquisition with relaxation enhancement (U-FLARE) [70]. CT along the two spectral dimensions (evolution and detection) facilitated complete effective homonuclear decoupling. A spiral imaging combined CT-PRESS was also recently proposed [71].

Summary Localized 2D MRS is still at its infancy and different techniques discussed earlier need further improvements in terms of voxel sizes, coil design, etc. Non-selective excitation of the J -coupled metabolites is facilitated by different 2D MRS techniques. 2D cross peaks of more than ten J -coupled metabolites have been observed using L-COSY recorded in the frontal and occipital regions of human brain [41,47,54]. In contrast to ISIS-COSY [57], the 2D techniques discussed in this chapter are one-shot based. Even though it has been shown recently that 1D MR spectra processed using the LC-Model algorithm [21], 2D L-COSY shows the unambiguous presence of 2D cross peaks originating from several J -coupled metabolites. However, further work is necessary in order to ascertain the superiority of 2D L-COSY over the other methodology. Due to the differential T2∗ -weighting during the incremental duration of 2D L-COSY, absolute quantitation of several metabolites is further complicated by lack of T2 and T1 values in the literature for different proton groups (methyl, methylene, methine, etc.) in any given metabolite. Regarding the application of these multidimensional MRS techniques in different pathologies, targeting a specific metabolite or selected metabolites

in different pathologies may be more advantageous than non-selective excitation/detection of several metabolites. Another major concern is that the 2D spectra are complicated by artefacts contributed by strong coupling of various metabolite protons unavoidable at both 1.5 and 3 T MRI scanners. As discussed earlier, the artefacts in these multi-voxel based 2D MRS sequences need to be further evaluated. Further work is necessary to demonstrate the clinical potential of both single and multi-voxel based multidimensional MR SI methods.

Acknowledgment This work was supported by the grants from the National Institute of Health (NIH) and the US Army Breast and Prostate Cancer Research Programs. Authors acknowledge the scientific support of Drs. Kenneth Yue and Lawrence Ryner in the earlier implementation of 2D MR sequences, and Ashwin Thomas and Tai Dou during the preparation of the chapter.

References 1. Gadian DG. Nuclear Magnetic Resonance and Its Applications to Living Systems. Oxford University Press: New York, 1982. 2. Rothman DL, Behar KL, Hetherington HP, Shulman RG. Proc. Natl. Acad. Sci. U.S.A. 1984;81:6430. 3. Moonen CTW, Kienlin MV, van Zijl PCM et al. NMR Biomed. 1989;2:201. 4. Kreis R, Ross BD, Farrow N, Ackerman Z. Radiology 1982;182:9. 5. Michelis T, Merboldt KD, Bruhn H et al. Radiology 1983;187:219. 6. Gruetter R, Weisdorf SA, Rajanayagan V et al. J. Magn. Reson., 1998;135:260. 7. Webb PG, Sailasuta N, Kohler SJ et al. Magn. Reson. Med. 1994;31:365. 8. Rothman DL, Petroff OAC, Behar KL, Mattson RH. Proc. Natl. Acad. Sci. U.S.A. 1993;90:5662. 9. Bruhn H, Frahm J, Gyngell ML et al. Magn. Reson. Med. 1989;9:126. 10. Govindaraju V, Young K, Maudsley AA. NMR Biomed. 2000;13:129. 11. Renshaw PF, Lafer B, Babb SM et al. Biol. Psych. 1997;41:837. 12. Pouwels PJW, Frahm J. Magn. Reson. Med. 1998;39:53. 13. Barker PB, Hearshen DO, Boska MD. Magn. Reson. Med. 2001;45:765. 14. Trabesinger AH, Weber OM, Duc CO, Boesiger P. Magn. Reson. Med. 1999;42:283. 15. Prichard JW. Curr. Opin. Neur. 1997;10:98. 16. Pan JW, Stein DT, Telang F et al. Magn. Reson. Med. 2000;44:673. 17. Choi IY, Lei H, Gruetter R. J. Cereb. Blood Flow Metab. 2002;22:1343. 18. Hardy DL, Norwood TJ. J. Magn. Reson. 1998;133:70.

3D Localized 2D MR Spectroscopy

46. Adalsteinsson E, Hurd RE, Mayer D et al. Neuroimage 2004;22:381. 47. Thomas MA, Yue K, Binesh N et al. Magn. Reson. Med. 2001;46:58. 48. Brereton IM, Crozier S, Field J, Doddrell DM. J. Magn. Reson. 1991;93:54. 49. Brereton IM, Galloway GJ, Rose SE, Doddrell DM. Magn. Reson. Med. 1994;32:251. 50. Binesh N, Thomas MA. Proc. Intl. Soc. Magn. Reson. Med. Kyoto, Japan, 2004, p 2303. 51. Binesh N, Yue K, Fairbanks L, Thomas MA. Magn. Reson. Med. 2002;48:942. 52. Prescot AP, Leach MO, Saran F, Collins DJ. Proc. Intl. Soc. Magn. Reson Med. 2004, p 2493. 53. Thomas MA, Yue K, Binesh N, Debruhl N. J. Magn. Reson. Imag. 2001;14:181. 54. Binesh N, Bugbee M, Fairbanks L et al. Proc. Intl. Soc. Magn. Reson. Med. Hawaii, 2002, p 699. 55. Delmas F, Beloeil JC, van der Sanden BPJ et al. J. Magn. Reson. 2001;149:119. 56. Ziegler A, Gillet B, Beloeil JC et al. Magn. Reson. Mat. Phys. Biol. Med. 2002;14:45. 57. Welch JWR, Bhakoo K, Dixon RM et al. NMR Biomed. 2003;16:47. 58. Nagayama K, Wuthrich K, Ernst RR. Biochem. Biophys. Res. Commun. 1979;90:305. 59. Dreher W, Leifbritz D. Magn. Reson. Imag. 1999;17:141. 60. Chung HK, Banakar S, Thomas MA. Proc. Intl. Soc. Magn. Reson. Med., Kyoto, Japan, 2004, p 687. 61. Chung HK, Banakar S, Thomas MA. Proc. Intl. Soc. Magn. Reson. Med. Toronto, 2003, p 1143. 62. Smith SA, Levante TO, Meier BH, Ernst RR. J. Magn. Reson. A 1994;106:75. 63. Hore PJ. J. Magn. Reson. 1983;55:283. 64. Banakar S, Venkatraman TN, Yue K et al. Proc. Int. Conf. Math. Engg. Tech. Med. and Biol. Sci., Las Vegas, 2002, Vol II, p 500. 65. Metzler A, Izquierdo M, Ziegler A et al. Proc. Natl. Acad. Sci. U.S.A. 1995;92:11912. 66. Ziegler A, Metzler A, Kockenberger W et al. J. Magn. Reson. 1996;B112:141. 67. Jensen JE, Frederic BD, Wang L et al. Proc. Intl. Soc. Magn. Reson. Med. 2004, p 2301. 68. Adalsteinsson E, Spielman DM. Magn. Reson. Med. 1999;41:8. 69. Mayer D, Dreher W, Leibfritz D. Magn. Reson. Med. 2000;44:23. 70. D. Mayer, Dreher W, Leibfritz D. Magn. Reson. Med. 2003;49:810. 71. Mayer D, Kim DH, Adalsteinsson E, Spielman DM. Proc. Intl. Soc. Magn. Reson. Med., Kyoto, Japan, 2004, p 678.

Part II

19. de Graaf AA, Luyten PR et al. Single and Double Quantum Lactate Imaging of the Human Brain at 1.5 T. 9th SMRM, New York 1990, p 138. 20. Crozier S, Brereton IM et al. Magn. Reson. Med. 1990;16:492. 21. Ith M, Bigler P, Scheurer E, et al. Magn. Reson. Med. 2002;48:915. 22. Kumar A, Thomas MA, Lavretsky H et al. Am. J. Psych. 2002;159:630. 23. Pfeuffer J, Tkac I, Provencher SW, Gruetter R. J. Magn. Reson. 1999;141:104. 24. Aue WP, Bartholdi E, Ernst RR. J. Chem. Phys. 1976;64:2229. 25. Ernst RR, Bodenhausen G, Wokaun A. Principles of NMR Spectroscopy in One and Two Dimensions. Oxford Publications: Oxford, 1987, pp 283–489. 26. Crozier S, Brereton IM et al. Magn. Reson. Med. 1990;16:492. 27. Sotak CH, Freeman DM, Hurd RE. J. Magn. Reson., 1988;78:355. 28. Desmoulin F, Seelig J. Magn. Reson. Med. 1990;14:160. 29. van Zijl PCM, Chesnick AS, Despres D. et al. Magn. Reson. Med. 1993;30:544. 30. Kreis R, Boesch C. J. Magn. Reson. 1996;B113:103. 31. Ogg RJ, Kingsley PB, Taylor JS. J. Magn. Reson. 1994;B103:1. 32. Patt SL, Sykes BD. J. Chem. Phys. 1972;56:3182. 33. Tran TK, Vigneron DM, Sailasuta N et al. Magn. Reson. Med. 2000;43:23. 34. Bottomley PA. Ann. NY. Acad. Sci. 1987;508:333. 35. Salibi N, Brown MA. Clinical MR Spectroscopy: First Principles. Wiley-Liss: New York, 1998, p 75. 36. Brown TR, Kincaid BM, Ugurbil K. Proc. Natl. Acad. Sci. U.S.A. 1982;79:3523. 37. Maudsley AA, Hilal SK, Perman WH, Simon HE. J. Magn. Reson. 1983;51:147. 38. Boesch C. J. Magn. Reson. Imag. 2004;16:177. 39. Ryner LN, Sorenson JA, Thomas MA. J. Magn. Reson. 1995;B107:126. 40. Ryner LN, Sorenson JA, Thomas MA. Magn. Reson. Ima. 1995;13:853. 41. Thomas MA, Hattori N, Umeda M, et al. NMR Biomed. 2003;16:245. 42. Thomas MA, Ryner LN, Mehta MP et al. J. Magn. Reson. Imag. 1996;6:453. 43. Yue K, Binesh N, Marumoto N et al. Magn. Reson. Med. 2002;47:1059. 44. Hurd RE, Gurr D, Sailasuta N. Magn. Reson. Med. 1998;40:343. 45. Swanson MG, Vigneron DB, Tran TK et al. Magn. Reson. Med. 2001;45:973.

References 1183

Part II

Part II, Section 2: Applications in Pharmaceutical Science

1187

Foreword to Applications in Pharmaceutical Science

NMR has a long history of applications in the pharmaceutical sciences. Initially it was almost exclusively used for the study of small molecules, i.e., drug leads or drugs themselves. As NMR technology developed and became amenable for the study of macromolecules and their complexes NMR has become an increasingly valuable as a tool in the drug discovery process. The 30 articles comprising this section of the Handbook of Modern Magnetic Resonance reflect the major contributions that NMR is making in this field. The articles cover the broad spectrum of activities that NMR encompasses, ranging from the early phases of drug discovery to lead optimization and even to the monitoring of drug effects within organisms in the newly emerging field of metabonomics. The chapters are written by leading experts in their respective fields, who have highlighted specific applications with examples from their own laboratories. In keeping with the concept of a Handbook, most of the articles focus on a particular NMR technique. However several

of the articles cover broad fields of application, as exemplified by the chapters on antimicrobial peptides and toxin structures. Both classes of molecules have obvious applications in the pharmaceutical sciences. In the short space available in this introduction it is not possible to give a detailed overview of all of the articles. I would just like to highlight one area where I think NMR has shown particularly exciting potential and has had a major impact in the pharmaceutical sciences over the last few years. That area is NMR-based screening. There are now a number of complementary NMR approaches that can be used to discover new ligands by using NMR as the detector in screening assays. Reflecting the importance of this field four chapters that address various aspects of NMR screening are included in this Handbook and many of the other chapters also address the area. David Craik October 2004

1189

Key Terms for Applications in Pharmaceutical Science

Overview of NMR in the Pharmaceutical Sciences

Flow injection analysis NMR

Drug design

Flow probe

LC NMR

LC-NMR

Ligand based screening

LC-NMR-MS

Metabonomics

Microcoil NMR

Flow NMR

Receptor based screening SAR by NMR Screening

Developments in NMR Hyphenation for Pharmaceutical Industry

Structure-based design

Cryogenic probes

Structural genomics

Drug impurity profiles LC-NMR

Applications of Cryogenic NMR Probe Technology for the Identification of Low-Level Impurities in Pharmaceuticals

LC-NMR/MS

Correlation experiments

Stop flow LC-NMR

Natural product screening Solid phase extraction

Cryogenic NMR probes Degradation products Metabolite identification

LC-NMR in Dereplication and Structure Elucidation of Herbal Drugs

Metabonomic studies

Crude plant extracts

Natural products

Dereplicaion

Pharmaceutical impurities

HMBC

Protein characterization

HSQC

Small volume NMR probes

LC-NMR

Structure characterization

Structure elucidation of natural products WET pulse sequence

Flow NMR Techniques in the Pharmaceutical Sciences

DI-NMR

New Approaches to NMR Data Acquisition, Assignment and Protein Structure Determination: Potential Impact in Drug Discovery

Direct injection NMR

Automated structure determination

FIA-NMR

Assignment

Cold probes

1190 Key Terms Pharmaceutical Sciences

Computer-assisted sequential assignment

Drug membrane-interactions

Fast multidimensional NMR

Model membranes Orientation

Transferred Cross-Correlated Relaxation: Application to Drug/Target Complexes Bound conformations

Partial Alignment for Structure Determination of Organic Molecules

Cross-correlated relaxation

Liquid crystal phase

Drug receptor complexes

Partial alignment

Epothilone/tubulin complex

Organic molecules

Transferred cross-correlated relaxation

Residual dipolar couplings

Transferred NOE

Stretched polymer gels

Novel Uses of Paramagnets to Solve Complex Protein Structures

Measurement of Residual Dipolar Couplings and Applications in Protein NMR

Domain orientation

Aligned proteins

J-modulation

Conformational substrates

Long-range restraints

Echo-anti-echo manipulation

Paramagnetic labelling

Protein folding

Protein structure

Protein NMR

Pseudocontact shift

Residual dipolar coupling

Residual dipolar coupling

Fast Assignments of 15 N-HSQC Spectra of Proteins by Paramagnetic Labeling

Using Chemical Shift Perturbations to Validate and Refine the Docking of Novel IgE Antagonists to the High-Affinity IgE Receptor Chemical shift

HSQC spectrum

Docking

Ligand binding

High-affinity binding

Paramagnetic labelling

IgE receptor

PLATYPUS

Peptides

Proteins

Simulated annealing

Pseudocontact shift Resonance assignment

Dual-Region Hadamard-Encoding to Improve Resolution and Save Time

Phospholipid Bicelle Membrane Systems for Studying Drug Molecules

Adiabatic gHMBC

Bicelles

HMBC

Conformation

Hadamard spectroscopy

Cyclosporine-A

Key Terms 1191

Region-selection

Mono-cadmium

Resolution enhancement

Tautomer

Sensitivity enhancement

Thiomandelate

Nonuniform Sampling in Biomolecular NMR

NMR Spectroscopy in the Analysis of Protein–Protein Interactions

Maximum entropy reconstruction Nonuniform data acquisition Sensitivity Spectrum analysis Resolution

Chemical exchange Cross saturation Deuterium labelling of proteins Mapping protein interfaces Large proteins

Structural Characterization of Antimicrobial Peptides by NMR Spectroscopy

Protein complexes

Antimicrobial peptides

Protein-protein interactions

Diffusion NMR

Protein structure

Disulfide cross-linking

Residual dipolar coupling

Membrane-peptide interactions

Titrations

Membrane insertion of peptides

TROSY NMR

Paramagnetic nuclei

Peptide solution structure Solid-state NMR

Identification and Characterization of Ternary Complexes Using NMR Spectroscopy

Pharmaceutical Applications of Ion Channel Blockers: Use of NMR to Determine the Structure of Scorpion Toxins

Borate

Low abundance samples Nano probes Scorpion toxins Small sample probes Toxin structure

Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β-Lactamases

Dihydrofolate reductase Inorganic ions Interligand NOE Nuclear Overhauser Effect Transferred NOE Trypsin

The Transferred NOE Fast exchange

Cadmium NMR

Non-specific binding

Cooperativity

Off rate

Imidazole

On rate

Metallo-β-lactamases

Saturation transfer difference (STD)

Minimum chemical shift approach

Slow exchange

1192 Key Terms

Pharmaceutical Sciences

Spin diffusion

Physiological variation

Transferred NOE

Proteomics Transcriptomics

NMR Kinetic Measurements in DNA Folding and Drug Binding DNA hairpin

Protein Misfolding Disease: Overview of Liquid and Solid-State High Resolution NMR Studies

DNA quadruplex

Amyloid

Drug binding kinetics

Alzheimers disease

Fast chemical exchange

Intrinsically unstructured proteins

Line shape analysis

Non-native states of proteins

Magnetisation transfer NMR

Natively unfolded proteins

Quadruplex-ligand interactions

Prion proteins

Slow conformational equilibria

Protein misfolding disease Structural biology

The Use of NMR in the Studies of Highly Flexible States of Proteins: Relation to Protein Function and Stability

F NMR Spectroscopy for Functional and Binding High-Throughput Screening

Amide exchange rates

Competition binding experiments

Flexibility

Drug discovery

Growth hormone

19

Insulin

3-FABS

Order parameters

FAXS

Protein dynamics

High-throughput screening

Trifluoroethanol

IC50 measurement

19

F NMR spectroscopy

KD measurement

NMR-based Metabonomics Techniques and Applications

Ligand libraries Protein binding

Atherosclerosis

Screening

Biofluids Clinical trials Diagnosis

Applications of Receptor-Based NMR Screening in Drug Discovery

Drug safety

Core replacement

Efficacy

Drug discovery

Genomics

Fragment-based screening

Metabonomics

Lead validation

Pharmaceutical attrition

Receptor-based screening

Key Terms 1193

NMR SHAPES Screening Compound libraries Drug scaffolds High-throughput screening Lead optimisation

HSQC Lead discovery NMR-based screening SAR-by-NMR STD NMR

SHAPES screening

NMR-Based Screening Applied to Drug Discovery Targets

NMR and Structural Genomics in the Pharmaceutical Sciences High throughput structure determination

Chemical shift perturbations

HSQC

Competition STD NMR

Structural genomics

Fragment-based screening

X-ray crystallography

1195

Horst Joachim Schirra and David J. Craik∗ Institute for Molecular Bioscience, University of Queensland, Brisbane 4072, Australia

Abbreviations: 3-FABS, Fluorine atoms for biochemical screening; CRINEPT, cross relaxation-enhanced polarization transfer; D, deuterium; DFT, discrete Fourier transformation; DI-NMR, direct injection NMR; FIANMR, flow injection analysis NMR; FAXS, Fluorine chemical shift anisotropy and exchange for screening; FT, Fourier transformation; GFT, G-matrix Fourier transformation; HSQC, heteronuclear single quantum coherence; ILOE, inter-ligand NOE; INEPT, insensitive nuclei enhanced by polarization transfer; LC, liquid chromatography; ME, maximum entropy reconstruction; MS, mass spectrometry; NMR-SOLVE, structurallyoriented library valency engineering; NOE, nuclear Overhauser effect; PR-NMR, projection-reconstruction NMR; RDC, residual dipolar coupling; SAR, structureactivity relationship; SEA-TROSY, solvent-exposed amides with TROSY; STD, saturation transfer difference spectroscopy; TROSY, transverse-relaxation optimised spectroscopy; Water-LOGSY, Water-ligand observed via gradient spectroscopy Abstract NMR spectroscopy has a wide range of applications in the pharmaceutical sciences. This chapter gives an overview of these applications and attempts to draw together information from the specialist articles on individual topics covered in this handbook of modern magnetic resonance.

Introduction Since its inception NMR has been an important tool for the characterization of molecules that have relevance in the pharmaceutical sciences. Initial interest focused on small organic molecules that were either drugs themselves or were leads in the drug discovery process. As the technology developed over the last few decades molecules of increasing complexity have been examined and it is now possible to study not only small-molecule drugs, but also their macromolecular targets, including proteins and nucleic acids. NMR spectroscopy is now a powerful tool for the study of structure, dynamics, and interactions of biomolecules. In particular, the study of protein– ligand interactions forms the basis for the use of NMR Graham A. Webb (ed.), Modern Magnetic Resonance, 1195–1202.
C 2008 Springer.

spectroscopy in the process of drug design and discovery, ranging from the simple identification of whether a particular compound binds to a target protein on the one hand to the full determination of a protein–ligand complex’s three-dimensional structure on the other. An outline of the general process of structure-based drug design is given in Figure 1. The iterative cycle is generally started by determining the three-dimensional structure of a target protein or the structure of a complex between the target protein and a ligand, identified through screening a library of compounds. This structural information is then used to design new ligands, which are subsequently synthesized and screened for binding to the target protein and/or for biochemical activity. The structures of promising ligands are then determined in complex with the target protein and used to suggest improvements during further iterations of the design cycle. The cycle is repeated until suitable lead molecules with the desired affinity, specificity, activity, and pharmacokinetic properties have been identified. NMR spectroscopy is ideally suited to provide critical information at several points of this iterative cycle. Together with X-ray crystallography it is the second major experimental technique used for determining the threedimensional structures of proteins and protein–ligand complexes in atomic detail, thus providing accurate and detailed descriptions of the regions of a protein involved in ligand binding, the conformation of the bound ligand and the key interactions between the two molecules that determine binding affinity and specificity. By contrast with X-ray crystallography NMR is typically carried out in solution (for an overview on solid-state NMR spectroscopy applications see the related section of Handbook of Modern Magnetic Resonance), thus removing the need for suitable crystals. In addition, NMR spectroscopy is unique in providing crucial information about dynamics of the target protein as well as the bound ligand. Molecular motion is important in determining the kinetics of protein–ligand complex formation and dissociation. Upon binding of a ligand, the dynamics of the reactive site of the target protein is usually perturbed. In addition, the mobility of the active site also determines the range of ligands that can be bound and whether multiple conformations of the ligand can exist. Furthermore, NMR spectroscopy can be used to study solvent accessibility and study and identify

Part II

Overview of NMR in the Pharmaceutical Sciences

1196 Part II

Part II

1

Pharmaceutical Sciences

Structure of the Target Protein

6

Clinical Trials, Product Launch

Structure of the Target-Ligand Complex 3

Ligand Design

4

2

Screening/Assay

Lead Molecule

5 Ligand Synthesis

Fig. 1. Schematic outline of the process of structure-based drug design and discovery. In an iterative cycle the structures of protein targets and/or protein-ligand complexes are determined, ligands are designed using the structural information, screened for binding to the target and assayed for activity. The interactions of suitable ligands with the target proteins are structurally investigated and the information is used for improving the ligands in further iterations of this cycle until a suitable lead compound has been generated. NMR spectroscopy is involved in all aspects of this cycle: (1) Protein structure determination; (2) ligand screening; (3) structural analysis of protein ligand-complexes; (4) ligand design (as a conformational analysis tool); (5) ligand synthesis (as a structure verification tool); and (6) quality control of the pharmaceutic production process and metabonomic studies during clinical trials and after release of the drug.

bound water molecules. Finally, NMR spectroscopy can be used to determine binding constants and other thermodynamic data as well as ionization states of the protein and/or ligand. In this overview article we will briefly discuss the applications of NMR spectroscopy techniques in drug design and discovery. More detailed discussions are presented in the individual chapters of this section of the Handbook of Modern Magnetic Resonance. A number of other review articles covering various aspects of the application of NMR spectroscopy in drug design and discovery have been published in recent years [1–8] and in addition to the specialist articles contained in this edition of Handbook of Modern Magnetic Resonance the reader is referred to these publications. The review is divided into four sections. In the first section, we review the latest technical developments in NMR spectroscopy, ranging from hardware developments such as the use of cryoprobes and coupled techniques like LC-NMR to improvements in the acquisition and measurement of NMR spectra. The second section deals with the structural characterization of target molecules and ligands and discusses labeling strategies, NMR experiments suited for large molecules (TROSY) and the use of non-NOE-based NMR data such as residual dipolar couplings (RDC) and paramagnetic labels. The third section is dedicated to applications of NMR spectroscopy in the

screening of ligands, and reviews target-based methods as well as ligand-based methods. While the first three sections focus on applications of NMR to obtain structural information, the fourth section focuses on NMR applications unique to pharmacy and pharmacology, from drug monitoring, quality control, and pharmacokinetics to the new field of metabonomics. The chapter finishes with a short section on future developments and directions.

Technical Developments Several significant technical developments have been made in NMR in hardware, software, and experimental techniques in recent years. The most significant development in terms of NMR hardware is the introduction of cryoprobes. Cryoprobes reduce the thermal noise in the detection circuit by cooling the receiver coil and electronic parts of the probe (but not the sample) with liquid helium. Consequently, the sensitivity of a cryoprobe is two to four times higher than conventional roomtemperature probes [9,10]. This sensitivity improvement benefits pharmaceutical applications in two ways: on one hand, the improved sensitivity can be utilized to investigate samples of much lower concentration than before, which is especially useful in natural product discovery or screening for low-concentration impurities in pharmaceutical samples during quality control, as

NMR in the Pharmaceutical Sciences

relaxation-induced polarization transfer [28], a highly effective transfer mechanism for molecules with a molecular weight above 200 kDa. Thus, TROSY and CRINEPT represent complementary techniques that have revolutionized NMR spectroscopy within a short period of time by making it applicable to proteins larger than the “traditional” 30 kDa limit [29]—reports of NMR data on a 900 kDa complex have been widely cited [30]. In addition, modifications of the TROSY experiment, like solvent-exposed amides with TROSY (SEA-TROSY), have been developed [31] that are specifically geared towards screening studies as part of structure-based drug design approaches. Techniques for improving the study of large proteins and protein complexes with NMR spectroscopy are reviewed in more detail in Chapter 19 [32]. Other developments with respect to experimental NMR techniques focus on modifying and speeding up data acquisition. In general, they involve replacing the uniform sampling of time-space with other sampling schedules and consequently replacement of the discrete Fourier-transformation (DFT) with other transformations. One approach is multiplexing, by encoding more than one resonance frequency in one dimension, as realized in G-matrix FT (GFT) [33,34] or Hadamard encoding [35,36]. Other approaches are projection-reconstruction NMR (PR-NMR) [37] and maximum entropy reconstruction (ME) [38]. These techniques are discussed in more detail in Chapters 6, 14, 15 [39–41].

Structure-based Design As already outlined, the first step in the cycle of structurebased drug design is the structural characterization of the target receptor molecule. Table 1 shows a collection of

Table 1: NMR technologies used for structural characterization of receptors, ligands, and ligand–receptor complexes Target

Technique

Information

Receptor

2-4D NMR RDC Relaxation

3D structure 3D structure Macromolecular dynamics

Ligand

1D/2D NMR RDC Chemical shift Lineshape/relaxation analysis Transfer NOE

Solution conformation Solution conformation Charge/tautomeric state Solution dynamics Bound conformation

Receptor–ligand complex

3D/4D NMR RDC Shift perturbation (HSQC) Line width Relaxation

3D structure of complex Ligand orientation in complex Mapping of binding site, Stoichiometry of complex, K D Binding kinetics Dynamics of complex

Part II

reviewed in more detail in Chapter 2 by Martin [11]. On the other hand, the improved sensitivity can be utilized to acquire data of the same quality in a much shorter time—up to 16 times faster—which is especially useful in high-throughput screening applications. Other developments on the hardware side include the integration of NMR spectroscopy with other analytical techniques such as liquid chromatography (LC) or mass spectrometry (MS) to yield techniques such as LC-NMR [12] or LC-NMR-MS [13,14]. In addition, developments such as direct-injection NMR (DI-NMR) [15] and flowinjection analysis NMR (FIA-NMR) [16] aim at utilizing NMR spectroscopy in support of pharmaceutical applications. DI-NMR has already become a routine technology in pharmaceutical laboratories and in other industrial applications [17–21]. These techniques are reviewed in Chapters 3–5 [22–24]. The most significant advance with respect to new experimental techniques is the development of transverserelaxation optimized spectroscopy (TROSY) by Pervushin et al. [25], in which the line-broadening effects of transverse relaxation are eliminated by selecting specifically the line of a heteronuclear multiplet for which the effects of dipole–dipole relaxation and chemical shift anisotropy compensate each other, yielding improved spectral resolution and sensitivity. However, TROSY is applied only during evolution periods in NMR experiments, and transverse relaxation, which is still active during periods of magnetization transfer, becomes a limiting factor for molecular weights beyond 100 kDa when magnetization is transferred via spin-spin couplings with the insensitive nuclei enhanced by polarization transfer (INEPT) technique [26]. This limitation can be overcome by replacing INEPT steps with cross relaxation-enhanced polarization transfer (CRINEPT) [27], which combines INEPT with cross-correlated

Structure-based Design 1197

1198 Part II

Pharmaceutical Sciences

Part II

NMR techniques that can be employed for this purpose as well as for the structure determination of protein–ligand complexes. Isotopic labeling is essential for the structure determination of target proteins that are not small in NMR terms (i.e. larger than 10 kDa). Labeling is also a prerequisite to harvest the benefits of the TROSY effect. The stable isotopes most commonly used for labeling are 15 N, 13 C, and 2 H (deuterium), and they can be employed in different labeling schemes [42,43]. Uniform labeling with 15 N and 13 C allows for sequential assignment and structure determination of the target protein with multidimensional triple resonance experiments [44]. For larger proteins, additional substitution of non-exchangeable protons with deuterium ameliorates relaxation effects. In this case, TROSY experiments are usually employed to obtain spectral information. In large multidomain proteins individual domains can be labeled by segmental labeling thereby simplifying the information in the NMR spectrum [45]. Another method of simplifying spectral information, which is especially useful for screening studies with large proteins, is the selective labeling of individual amino acids or of amino acid types [46,47]. The former approach can also be achieved with amino-acid specific pulse sequences [48,49]. After NMR spectra have been recorded, and all resonances have been assigned, the three-dimensional structure of the target protein is determined mostly from information based on NOE connectivities and dihedral restraints [50]. Recently, additional sources of structural information have been made accessible, most important among them RDCs [51] and shift and relaxation changes associated with paramagnetic nuclei [52,53]. Both methods have the distinct advantage of providing long-range constraints, which is a useful complementation for the short-range information derived from NOEs and dihedral restraints. Furthermore, they are able to provide structural information in the case of large proteins that have been extensively labeled with deuterium and thus have a reduced number of observable proton-proton NOEs. Labeling, RDC and paramagnetic methods are discussed in more detail in Chapters 8, 9, 11, 12, and 19 [32,54–57]. Similar methods can be employed to determine the three-dimensional structures of ligand molecules. Although ligand molecules are easier to examine because of their small size and can generally be studied by standard homonuclear NMR methods, the use of more advanced techniques such as RDCs has been demonstrated to be beneficial e.g. for obtaining structural information on sugars and small organic molecules as reviewed in Chapter 11 [56]. Other examples of the application of NMR spectroscopy for structure determination of small organic molecules are antimicrobial peptides and scorpion toxins as discussed in Chapters 16 and 17 [58,59].

A major milestone in structure-based drug design is a detailed determination of the three-dimensional structure of a ligand-target protein complex. If the structure of the target protein has been solved, but the ligand binding site is unknown then a quick determination of the binding site can be obtained by chemical shift mapping: a labeled sample of the target protein is titrated with the ligand, and due to the change in chemical environment associated with ligand binding amide proton resonances at, or close to, the binding site change their position in a HSQC spectrum. Mapping of the shift perturbations on the structure of the target protein reveals the ligand binding site. Similarly, in the case of a small ligand protein binding to a large receptor the binding face of the ligand can be determined by titrating labeled ligand with unlabeled receptor and chemical shift mapping. To determine the detailed nature of protein–ligand interactions the observation of intermolecular NOEs between target protein and ligand is crucial. The aforementioned combination of labeled protein and unlabeled ligand (or vice versa) is helpful here, and for example, “filtered” experiments can be used to selectively observe NOE signals either within the protein, within the ligand, or most importantly the desired intramolecular signals between protein and ligand [60]. RDCs offer an alternative approach for determining the orientation of ligands within the target protein’s binding site. Upon binding, the ligand adopts the alignment tensor of the target protein, and thus a comparison of ligand RDCs in the free and bound state makes it possible to determine the conformation of the ligand in the bound state in the complete absence of ligand-protein or ligand– ligand NOEs [61–63]. A further discussion of techniques for determining the structures of protein–ligand complexes, including some case examples, is given in Chapters 13, 18, 19, and 20 [32,64–66]. Transferred NOE and cross-correlated relaxation methods are also particularly important for determining the conformation of bound ligands, provided certain kinetic conditions are met. These approaches are described in Chapters 7 and 21 [67,68]. One of the distinct advantages of NMR spectroscopy is the possibility of obtaining information about the dynamics of a protein or a protein–ligand complex. The dynamics of the binding site is of fundamental importance in determining affinities and binding kinetics. In addition, it is a not an uncommon observation that the binding site of a protein exhibits considerable flexibility, but becomes more rigid upon binding of the ligand [69,70]. Phenomena like this can be readily investigated with NMR relaxation experiments, which can determine molecular motions on a nanosecond up to millisecond timescale [71]. In addition, lineshape analysis offers an avenue into further determining events on the millisecond timescale like chemical exchange, including the measurement of rate constants of the exchange process [72]. Finally, NMR

NMR in the Pharmaceutical Sciences

NMR Screening The screening of libraries of compounds is becoming an increasingly useful adjunct to structure-based drug design. Table 2 gives an overview of the different screening strategies that can be employed. In general, screening techniques can either be target-based—i.e. changes in the macromolecular target protein upon ligand binding are observed, or ligand-based—i.e. changes in NMR properties of the ligand upon binding are observed. In general target-based screening techniques are based on monitoring chemical shifts with a HSQC spectrum. For large target proteins specific labeling strategies and TROSY experiments are employed, but the principle stays the same. A recent modification of the TROSY experiment specifically geared for application in screening programs with large target proteins is the SEA-TROSY experiment [31]. The experiment is based on the concept that only groups on the solvent-exposed surface of the protein are likely to be involved in ligand binding, while residues in the protein interior are unlikely to be in contact with the ligand. SEA-TROSY observes by magnetization transfer from the bulk water only solvent-exposed amide protons in fully deuterated and 15 N-labeled proteins that are dissolved in water. Thus the SEA-TROSY spectrum contains significantly fewer signals than conventional spectra, alleviating the problem of spectral overlap for large target proteins. Table 2: NMR technologies used for high-throughput screening of ligand–receptor complexes Target

Technique

Reference

Receptor-based

SAR by NMR SEA-TROSY NMR-SOLVE

[75] [31] [47]

Ligand-based, Direct

Relaxation Diffusion coefficients STD ILOE Transfer NOE NOE pumping Water-LOGSY SHAPES

[78,79] [78] [83] [65,85] [68,80] [81,82] [84] [87]

Ligand-based, Indirect

FAXS 3-FABS STD with spy-molecule

[86] [96] [86]

The by now “classic” strategy employed in NMR target-based screening is structure–activity relationships (SAR) by NMR [75]. It aims at the identification of two ligands with moderate binding affinities that bind to two different but adjacent binding sites on the target protein. Crosslinking of the two ligands via chemical synthesis then produces a first-generation lead molecule with substantially improved binding affinity. It has been responsible for the development of numerous novel drug leads, as outlined in the article by Hajduk in this Handbook [76]. NMR-Structurally orientated library valency engineering (NMR-SOLVE) is a new target-based screening approach that has been developed specifically for the postgenomic era to provide guidance for the construction of libraries against multiple related target proteins [47]. One key element of NMR-SOLVE is to use highly selective isotopic labeling (e.g. 13 C-1 H labeled CH3 groups in otherwise uniformly deuterated proteins) to observe only a small number of key resonances in a binding site, thus yielding information in the absence of complete assignments. The NMR-SOLVE strategy is based on the fact that several enzyme families have multiple binding sites—one that is common for all family members, and one that is specific for each protein. In a first-step NMRSOLVE uses a known reference ligand that binds to the common binding site to map and identify the specifically labeled groups that are around this binding site. In the next steps mimics of the reference ligand that could bind to the common binding site of most family members of the protein family are investigated. Finally, in an act of core-expansion structural information is used to design a linker that could connect the core of ligand-mimetic fragments binding in the common binding site with additional ligand fragments that would bind to the secondary binding sites, unique to each protein. Thus, a library of bivalent ligands is obtained that is specifically geared towards a given protein family. The NMR-SOLVE strategy has been demonstrated with target proteins as large as 170 kDa [47]. Further discussion on target-based screening approaches can be found in Chapters 27 and 29 [76,77]. A plethora of ligand-based screening methods has been developed targeting a wide variety of observable NMR parameters, as indicated in Table 2. Parameters that have been utilized include longitudinal, transverse, and double-quantum relaxation [78,79], diffusion coefficients [78], and intramolecular and intermolecular transfer of magnetization, such as transfer NOE [68,80], NOE pumping, and reverse NOE pumping [81,82], saturation transfer experiments [83], Water-ligand observed via gradient spectroscopy (water-LOGSY) [84] and inter-ligand NOEs (ILOE) [65,85]. Individual techniques are discussed in detail in Chapters 21, 26, and 29 [68,77,86] and are reviewed here only briefly. The main advantage of monitoring the small ligands is that there is no need for isotopic labeling of the target

Part II

titrations can be used to determine dissociation constants of a protein–ligand complex. These topics are further reviewed in Chapters 18, 21, 22, and 23 [66,68,73,74].

NMR Screening 1199

1200 Part II

Pharmaceutical Sciences

Part II

protein, and thus limitations of protein yields in minimal media or molecular size of the target protein do not apply. Furthermore, the identity of a binding ligand in a mixture can be readily established, and screening methods based on 1D-NMR experiments are generally faster than a 2D-HSQC spectrum. However, ligand-based methods provide no direct information about ligand binding sites, and generally require the ligand to be in fast or medium exchange with the target protein. Ligand-based screening methods can be divided in two classes: direct methods, in which the NMR signals of the ligand are observed directly, and indirect methods, where signals of a “reporter” or “spy” molecule that binds weakly to the target protein are observed. The indirect methods are especially suitable for the investigation of strongly binding ligands that are not in fast exchange with the target protein. As the ligand binds competitively to the protein, it replaces the reporter molecule, and the changes in the NMR signal of the reporter molecule are observed. A broadly applicable approach that utilizes ligandbased screening is SHAPES [87]. This drug design strategy is based on a diverse library of small organic fragment molecules, whose molecular shapes represent those most commonly found in the structures of established drugs. Unlabeled target proteins are screened against this library of weakly binding water-soluble library, and hits are used to construct more complex, high-binding lead compounds. The strategy is discussed in more detail in Chapter 28 [88].

Studies of Drug Effects NMR spectroscopy has not only applications within the iterative cycle of structure-based drug design, as outlined in the previous section, but can also be applied in later stages once a lead molecule has been identified. One obvious example is the use of NMR spectroscopy as a routine analytic technique for the quality control during synthesis and production. This includes the profiling of drug impurities or the analysis of drug stability. Another application is pharmacokinetics and the screening for drug metabolites in biological fluids or extracts. The use of coupled analytical techniques such as LC-NMR and new technical developments such as cryoprobes have already been mentioned and are reviewed in more detail in Chapters 2, 3, and 4 [11,22,23]. A most exciting development in this respect is the application of NMR spectroscopy in metabonomics [89]. Metabonomics is formally defined as the quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification [90]. In other words, in metabonomics biofluids and tissues are monitored with various analytical techniques, chiefly among them 1D-NMR

spectroscopy, for changes in the levels of metabolites. These changes can occur as a consequence of disease (vs. healthy state) or after administering of a drug. The immense value of metabonomics in all stages of drug development from first assessments of pharmacology, through clinical trials and later on for monitoring of patients and epidemiological studies is becoming increasingly apparent. NMR-based metabonomic techniques and applications are discussed in detail in Chapter 24 [90].

Future Directions A remaining frontier for both NMR spectroscopy and X-ray cystallography is the structure elucidation of membrane proteins. Membrane proteins are often involved in signal transduction and consequently many targets of pharmaceutical interest involve membrane-bound or membrane-associated proteins. Recent developments in the field of solid-state NMR have made it possible to study the structure of membrane proteins [91,92]. Similar progress has been made in solution NMR spectroscopy with the development of phospholipid bicelle systems that mimic the environment of the cell membrane and are mobile enough to allow the study of drug-membrane or protein-membrane interactions with solution NMR. These developments are further detailed in Chapter 10 [93]. Another challenge/opportunity for NMR spectroscopy in the pharmaceutical sciences comes from structural genomics programs. In the postgenomic era it is becoming increasingly important to obtain information about ligand-target protein interactions not just for one target protein, but for whole classes of related proteins. More often than not, full assignments of the target protein(s) are not available or cannot be obtained within the constraints of time. Thus the development of techniques, such as NMR-SOLVE [47], that can provide information about ligand–protein interactions without full NMR characterization of the target protein is very important. A discussion of NMR applications in structural genomics is provided in Chapter 30 [94]. One of the interesting features that is emerging from such programs is that the average size and amino acid composition of genomically encoded proteins is substantially different from folded proteins in the PDB. There are many proteins now being discovered that are intrinsically unfolded and hence unsuitable for structure determination. Chapter 25 [95] describes NMR approaches to the study of such proteins and to the related issue of protein misfolding in disease.

Acknowledgments DJC is an Australian Research Council Professorial Fellow. HJS is an Australian Research Council Postdoctoral Fellow.

NMR in the Pharmaceutical Sciences

1. Pellecchia M, Sem DS, W¨uthrich K. Nat. Rev. Drug Discov. 2002;1:211. 2. Homans SW. Angew. Chem. Intl. Ed. 2004;43:290. 3. Heller M, Kessler H. Pure Appl. Chem. 2001;73:1429. 4. Meyer B, Peters T. Angew. Chem. Intl. Ed. 2003;42:864. 5. Stockman BJ. Prog. Nuc. Mag. Reson. Spectrosc. 1998;33:109. 6. Roberts GCK. Drug Discov. Today. 2000;5:230. 7. Craik DJ, Scanlon MJ. Annu. Rep. NMR Spectrosc. 2000;42:115. 8. Stockman BJ, Dalvit C. Prog. Nuc. Mag. Reson. Spectrosc. 2002;41:187. 9. Styles P, Soffe NF, Scott CA, Cragg DA, Row F, White DJ, White PCJ. J. Magn. Reson. 1984;60:397. 10. Styles P, Soffe NF, Scott CA. J. Magn. Reson. 1989;84:376. 11. Martin GE. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1179–1192. 12. Dorn HC. In: DM Grant, RK Harris (Ed). Encyclopedia of Nuclear Magnetic Resonance, John Wiley & Sons, Ltd: West Sussex, England, 1996, p 2026. 13. Burton KI, Everett JR, Newman MJ, Pullen FS, Richards DS, Swanson AG. J. Pharm. Biomed. Anal. 1997;15:1903. 14. Corcoran O, Spraul M. Drug Discov. Today 2003;8:624. 15. Keifer PA, Smallcombe SH, Williams EH, Salomon KE, Mendez G, Belletire JL, Moore CD. J. Comb. Chem. 2000;2:151. 16. Keifer PA. Magn. Reson. Chem. 2003;41:509. 17. Keifer PA. Curr. Opin. Chem. Biol. 2003;7:388. 18. Lewis K, Phelps D, Sefler A. Am. Pharm. Rev. 2000;3:63. 19. Kalelkar S, Dow ER, Grimes J, Clapham M, Hu H. J. Comb. Chem. 2002;4:622. 20. Stockman BJ, Farley KA, Angwin DT. Methods Enzymol. 2001;338:230. 21. Kraus WE, Houmard JA, Duscha BD, Knetzger KJ, Wharton MB, McCartney JS, Bales CW, Henes S, Samsa GP, Otvos JD, Kulkarni KR, Slentz CA. N. Engl. J. Med. 2002;347:1483. 22. Keifer PA. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1187–1193. 23. Spraul M. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1195–1202. 24. Karagianis G, Waterman PG. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1203–1209. 25. Pervushin K, Riek R, Wider G, W¨uthrich K. Proc. Natl. Acad. Sci. U.S.A 1997;94:12366. 26. Morris GA, Freeman R. J. Am. Chem. Soc. 1979;101: 760. 27. Riek R, Wider G, Pervushin K, W¨uthrich K. Proc. Natl. Acad. Sci. U.S.A 1999;96:4918. 28. Br¨uschweiler R, Ernst RR. J. Chem. Phys. 1992;96:1758. 29. Riek R, Pervushin K, W¨uthrich K. Trends Biochem. Sci. 2000;25:462. 30. Fiaux J, Bertelsen EB, Horwich AL, W¨uthrich K. Nature. 2002;418:207. 31. Pellecchia M, Meininger D, Shen AL, Jack R, Kasper CB, Sem DS. J. Am. Chem. Soc. 2001;123:4633. 32. Gell DA, Mackay JP. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1313– 1320.

33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63.

Kim S, Szyperski T. J. Am. Chem. Soc., 2003;125:1385. Ding K, Gronenborn AM, J. Magn. Reson. 2002;156:262. Kupce E, Freeman R. J. Magn. Reson. 2003;163:56. Kupce E, Freeman R. J. Magn. Reson. 2003;162:300. Kupce E, Freeman R. J. Am. Chem. Soc. 2003;125:13958. Wernecke SJ, Daddario LR. IEEE Transact. Comput. 1977;26:351. Crouch R. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1273–1278. Maciejewski MW, Stern AS, King GF, Hoch JC. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1279–1285. Pineda-Lucena A. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1213–1220. Kay LE, Gardner KH, Curr. Opin. Struct. Biol. 1997;7:722. Goto NK, Kay LE. Curr. Opin. Struct. Biol. 2000;10:585. Kay LE, Ikura M, Tschudin R, Bax A. J. Magn. Reson. 1990;89:496. Xu R, Ayers B, Cowburn D, Muir TW. Proc. Natl. Acad. Sci. U.S.A 1999;96:388. Griffey RH, Redfield AG, Loomis RE, Dahlquist FW. Biochemistry. 1985;24:817. Pellecchia M, Meininger D, Dong Q, Chang E, Jack R, Sem DS. J. Biomol. NMR 2002;22:165. Schubert M, Smalla M, Schmieder P, Oschkinat H. J. Magn. Reson. 1999;141:34. Schubert M, Oschkinat H, Schmieder P. J. Magn. Reson., 2001;148:61. W¨uthrich K. NMR of Proteins and Nucleic Acids. John Wiley & Sons: New York, 1986. Bax A, Kontaxis G, Tjandra N. Methods Enzymol. 2001; 339:127. Ubbink M, Worrall JA, Canters GW, Groenen EJ, Huber M. Annu. Rev. Biophys. Biomol. Struct. 2002;31:393. Pintacuda G, Keniry MA, Huber T, Park AY, Dixon NE, Otting G. J Am. Chem. Soc. 2004;126:2963. Byrd RA, Fowler CA, McFeeters RL, Gaponenko V. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1229–1235. Ding K, Gronenborn AM. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1261–1265. Luy B, Kessler H. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1253– 1259. Pintacuda G, Huber T, Keniry MA, Park AY, Dixon NE, Otting G. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1237–1243. Delepierre M, Possani LD. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1299–1304. Nguyen LT, Prenner EJ, Vogel HJ. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1289–1297. Otting G, W¨uthrich K. J. Magn. Reson. 1989;85:586. Shimizu H, Donohue-Rolfe A, Homans SW. J. Am. Chem. Soc. 1999;121:5815. Bolon PJ, Al-Hashimi HM, Prestegard JH. J. Mol. Biol. 1999;293:107. Al-Hashimi HM, Bolon PJ, Prestegard JH. J. Magn. Reson. 2000;142:153.

Part II

References

References 1201

1202 Part II

Pharmaceutical Sciences

Part II

64. Starovasnik MA, Fairbrother WJ. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1267–1272. 65. London RE. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1321–1330. 66. Damblon C, Roberts GCK. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1305–1311. 67. Carlomagno TMR, Griesinger C. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1221–1228. 68. Williamson MP. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1331–1336. 69. Hodsdon ME, Cistola DP. Biochemistry 1997;36:1450. 70. Hodsdon ME, Cistola DP. Biochemistry 1997;36:2278. 71. Palmer 3rd AG. Curr. Opin. Struct. Biol. 1997;7:732. 72. Nieto PM, Birdsall B, Morgan WD, Frenkiel TA, Gargaro AR. Feeney J. FEBS Lett. 1997;405:16. 73. Kristensen SM, Kasimova MR, Led JJ. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1343–1350. 74. Searle MS, Balkwill G. Williams HEL, Gavathiotis E. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1337–1341. 75. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Science 1996;274:1531. 76. Hajduk PJ. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1375–1381. 77. Gesell JJ, McCoy MA, Senior MM, Wang Y-S, Wyss DF. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1393–1402. 78. Hajduk PJ, Olejniczak ET, Fesik SW. J. Am. Chem. Soc. 1997;119:12257. 79. Dalvit C, Bohlen JM. Annu. Rep. NMR Spectrosc. 1999; 37:203.

80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96.

Ni F. Prog. Nuc. Mag. Reson. Spectrosc. 1994;26:517. Chen A, Shapiro MJ. J. Am. Chem. Soc. 1998;120: 10258. Chen AD, Shapiro MJ. J. Am. Chem. Soc. 2000;122:414. Mayer M, Meyer B. Angew. Chem. Intl. Ed. 1999;38: 1784. Dalvit C, Pevarello P, Tato M, Veronesi M, Vulpetti A, Sundstrom M. J. Biomol. NMR 2000;18:65. Li D, DeRose EF, London RE. J. Biomol. NMR 1999;15: 71. Dalvit C, Veronesi M. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1367– 1373. Fejzo J, Lepre CA, Peng JW, Bemis Ajay GW, Murcko MA, Moore JM. Chem. Biol. 1999;6:755. Lepre CA, Moore JM. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1383– 1391. Nicholson JK, Lindon JC, Holmes E. Xenobiotica. 1999;29:1181. Lindon JC, Holmes E, Nicholson JK. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1351–1359. Watts A. Curr. Opin. Biotech. 1999;10:48. Opella SJ. Nat. Struct. Biol. 1997;4:845. Guo J, Tian X, Pavlopoulos S, Makriyannis A. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1245–1251. Cemazar M, Craik DJ. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1403– 1408. Schwalbe H, Wirmer J. In: Webb GA (Ed). Modern Magnetic Resonance, Springer, The Netherlands, 2006, pp 1361– 1365. Dalvit C, Ardini E, Fogliatto GP, Mongelli N, Veronesi M. Drug Discov. Today. 2004;9:595.

Part II

Instrumentation

1205

Gary E. Martin Pfizer Global Research & Development, Michigan Structure Elucidation Group, Kalamazoo, MI 49001-0199

Abstract Following the detection of low-level impurities in pharmaceuticals, structural characterization of these components represents a continuing challenge. Irrespective of whether the impurity in question arises from side reactions of the chemical processes being employed for the synthetic elaboration of the active pharmaceutical, through degradation chemistry processes, or through accidental contamination, the identification of low-level impurities is a mandatory component of the preparation and marketing of pharmaceuticals. The development of cryogenic NMR probes and the corresponding increases in sensitivity that they provide has significantly impacted this phase of pharmaceutical development. The role of cryogenic NMR probe technology in the characterization of lowlevel impurities from pharmaceuticals is discussed.

improving sensitivity, is also a well-established approach to enhancing NMR performance [4–7]. Unlike conventional NMR probe developments, which were commercially realized after prototype design and testing, the commercial manufacture of cryogenic NMR probes proved to be a more daunting task requiring a number of years for even the first prototypes to be delivered to customer laboratories. In part this was due to design considerations necessary to cool the internal components to temperatures for the rf coil in the range of 10–25 K and preamps housed in the probe body to temperatures near 50 K without destroying those components during thermal cycling that the probe must occasionally be subjected to. Nevertheless, numerous commercial cryogenic NMR probes have been delivered and have begun to have a significant impact on the way in which various disciplines perform NMR-based chemical structure investigations.

Introduction

Cryogenic NMR Probes

Multiple creative approaches have been utilized in the quest for improved NMR sensitivity. NMR spectroscopy, despite the wealth of chemical structure information that it can provide, has always had an Achilles’ heel in the form of the low inherent sensitivity of the technique. Approaches that have been used to improve sensitivity have included the development of new pulse sequences, such as the proton- or “inverse”-detected heteronuclear 1D-, 2D-, and 3D-NMR experiments that are now in wide use. Another approach that has been heavily exploited has been that of “shrinking” the size of NMR probes and hence the required sample [1,2]. Although small volume probes have been known for a considerable period of time, contemporary small volume probe development effectively began in 1991 with the first reports of 3 mm NMR probe applications, followed shrinking of probe diameters to 1.7 mm, then 1 mm, and finally µ-coil probes with sample volumes as small as a few µl. This area was the subject of a recent review and the interested reader is referred there for more details [3]. Cooling the rf coils of an NMR probe to cryogenic temperatures to reduce noise, thereby

Commercial examples of cryogenic NMR probes have been developed in a limited number of configurations. Typically, they have been 3 or 5 mm diameter probes and most commonly, inverse-detection triple resonance gradient designs, although a cryogenic probe for direct carbon detection has also been developed. Cryogenic NMR probes have also been developed for flow applications. Perhaps one of the most interesting of these is the interchangeable flow cell (IFC) design by Varian, which allows the user to change between NMR tubes and a flow cell without having to remove the cryoprobe from the magnet. Although some early cryogenic probes were developed and tested using an open configuration of a liquid helium dewar, to the best of the author’s knowledge, all commercial examples use a closed loop system to deliver the helium refrigerant to the probe at temperatures in the vicinity of 8–10 K. The helium refrigerant is used to cool the rf coils to their operating temperature as well as the associated electronic circuitry housed in the probe body such as preamps, after which the helium refrigerant is cycled back and re-chilled. Typically,

Graham A. Webb (ed.), Modern Magnetic Resonance, 1205–1212.
C 2008 Springer.

Part II

Applications of Cryogenic NMR Probe Technology for the Identification of Low-Level Impurities in Pharmaceuticals

1206 Part II

Pharmaceutical Sciences

Part II

the rf coil temperature for cryogenic probe installations is 25 K although in the author’s laboratory 500 and 600 MHz cryogenic gradient triple resonance inverse probes are both operated at 20 K to provide improved sensitivity. From a serviceability standpoint, the closed loop helium chillers perform reliably for intervals ranging from 10 to 14 months in the author’s experience before requiring a rebuild. This type of required “annual” maintenance, and the associated expense, should be taken into consideration when any NMR laboratory weighs the sensitivity advantages of a cryogenic NMR probe installation vs. a high sensitivity small volume conventional NMR probe.

Sample Preparation In the author’s experience, one factor that has a considerable impact on cryogenic NMR probe performance is the uniformity of sample preparation. While conventional NMR probes are very forgiving in terms of sample-tosample variation in column height, etc. the same cannot be said for cryogenic NMR probes. Figure 1 shows an example of the variability of performance that can be seen simply by varying column height in a conventional NMR tube over the range from 35 to 20 mm. Given that the utilization of cryogenic NMR probes will generally be in those cases where time or sample are scarce, the impact of sample preparation on the quality of the data generated should receive adequate attention. Another less obvious factor associated with sample preparation when using a cryogenic NMR probe is the selection of tube size. While it would seem to make sense to use a 5 mm NMR tube for samples to be run

Fig. 1. The effect of solvent column height on lineshape and s/n characteristics for a 4 mm NMR tube with a sample of d4 -methanol. The spectra of the solvent multiplet were acquired in a Varian 5 mm 500 MHz triple resonance gradient-inverse detection Cold ProbeTM . Each column height was first gradient and then manually shimmed to afford the best possible lineshape and sensitivity.

in a 5 mm cryogenic NMR probe, thereby avoiding filling factor losses, reality is quite the opposite. Almost counterintuitively, substantially better performance can be obtained with extremely scarce samples by preparing them carefully in a 3 mm NMR tube. An illustration of attainable s/n ratios for a limited quantity of the simple antibiotic clindamycin in 3, 4, and 5 mm NMR tubes is presented in Figure 2. These results can be explained when the noise sources in a cryogenic NMR probe are considered. In reality, the sample is the single largest source of noise in a cryoprobe NMR experiment. Hence, as the diameter of the NMR sample itself is decreased in going from 5 to 4 and then 3 mm sample tubes, the noise contributed to the experiment correspondingly diminishes as the isolation of the rf coil from the noise source improves with smaller diameter NMR tubes [8]. Thus, despite potential filling factor losses associated with smaller diameter NMR tubes run coaxially in a 5 mm probe, performance still improves since the impact of the sample noise is larger than filling factor losses.

Identification of Degradants To illustrate what would be a typical pharmaceutical industry application of cryogenic NMR probe sensitivity, a sample of a complex alkaloid studied more than a decade ago that was kept in a sealed 5 mm NMR tube was used as a model system. The sample consisted of 2.5 mg of the complex spiro nonacyclic indoloquindoline alkaloid cryptospirolepine (1) that was stored in DMSO-d6 (Scheme 1). During the life of the sample, the color changed from an initial red-orange to a dark brown. When the sample tube was cut open and the sample subjected to HPLC analysis, the chromatogram shown in Figure 3 was obtained. By LC/MS methods, there was no trace of the starting alkaloid left in the sample among the 26 degradant peaks observed in the chromatogram. An investigator dealing with a typical pharmaceutical would not be faced with total degradation. Instead, most of the drug would be intact and impurity peaks ranging from a few tenths of a percent to a few percent would be observed in the chromatogram. Nevertheless, this still affords a useful example of the capabilities engendered in cryogenic NMR probe technology. The two largest chromatographic peaks in the HPLC of the degraded sample of 1 were identified as cryptolepinone (2) and cryptoquindolinone (3). Cryptolepinone (2) is a simple molecule and was readily identified from just a COSY and 1 H–13 C GHSQC spectra. Cryptoquindolinone (3) is a more complex molecule and required considerably more effort to characterize. Approximately 100 µg of 3 was isolated chromatographically and, using a 600 MHz spectrometer with a state-of-the-art 3 mm triple resonance gradient inverse-detection NMR probe afforded a good 1 H–13 C HSQC spectrum overnight (18 h). Given the relative differences in sensitivity of the HSQC

Applications of Cryogenic NMR Probe Technology 1207

Applications of Cryogenic NMR Probe Technology

Part II

A 5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0 ppm

B 5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0 ppm

C 5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0 ppm

Fig. 2. Comparison spectra of 11.6 µg samples (0.0027 µmoles) of the simple antibiotic clindamycin prepared: (A) 500 µl solve in a 5 mm NMR tube; (B) 292 µl solvent in a 4 mm NMR tube; or (C) 163 µl of solvent in a 3 mm NMR tube (identical column heights of 30 mm). Each sample was gradient and then hand shimmed to afford optimum lineshape and performance, after which a spectra was recorded using 8 transients. The s/n ratio was then measured using the anomeric proton of the sugar with a 200 Hz standard “noise” window downfield of the lowfield anomeric proton doublet. The measured s/n for three samples in this limited sample scenario was: (A) 14.4:1 s/n; (B) 20.8:1; and (C) 21.5:1.

and HMBC experiments, a long-range 1 H–13 C spectrum was anticipated to require from 3 to 4 days. In contrast, using the available sample, the 6 Hz optimized HMBC [9] shown in Figure 4 was acquired overnight at 500 MHz using a 5 mm triple resonance gradient inverse-detection cryogenic NMR probe. These data were necessary to assemble the structure of the degraded molecule. Using the same sample, it was also possible over a weekend to acquire a 3–6 Hz optimized long-range 1 H–15 N CIGARHMBC spectrum of 3 in which correlations to three of the four nitrogens in the structure were observed, including a correlation to the N10 resonance at 230 ppm, which confirmed the presence of the C=N double bond as shown in Scheme 1. Without where indicated access to a cryogenic NMR probe, long-range 1 H–15 N data for a sample of this size would consume a week or more of spectrometer time, precluding the acquisition of these data in most laboratories. The example just described is typical of many encountered within the pharmaceutical industry. While it

is certainly possible to isolate larger samples and solve impurity and degradant structures using conventional NMR probe technology there are several disadvantages to doing so. First the chromatography required to isolate larger samples generates large volumes of chromatographic mobile phase that must be disposed of or recycled. Second, the time required to isolate larger samples for conventional spectroscopy is proportionally longer and undesirable whenever tight timelines are imposed on a pharmaceutical development effort for whatever reason.

Applications of Cryogenic NMR Probe Technology There are several possible ways to break down applications of cryogenic NMR probe technology that have appeared in the literature thus far. One logical subdivision would be to separate applications on the basis of the type

34.789

25.103

26.083

22.936 23.510

21.396

18.588 19.064

16.887 17.238

14.16813.969 14.685 15.549 15.701

12.741

11.557

10.087

7.693

18.123

16.591

< 4.000

Scheme 1.

Fig. 3. HPLC chromatograph showing the result obtained after chromatographing a 2.5 mg sample of cryptospirolepine (2) stored in DMSO-d6 for more than 10 years. Based on LC/MS measurements, none of the starting alkaloid remained in the sample.

Applications of Cryogenic NMR Probe Technology

Applications of Cryogenic NMR Probe Technology 1209

Part II

Fig. 4. Phase-sensitive 6 Hz optimized non-gradient HMBC spectrum of a sample of ∼100 µg of cryptoquindolinone (4) dissolved in 150 µl DMSO-d6 in a 3 mm NMR acquired overnight using a Varian 500 MHz 5 mm triple resonance gradient inverse-detection cryogenic NMR probe.

of cryoprobe employed. However, the vast majority of applications utilize inverse-detection cryogenic probes so subdivision on the basis of probe type is not particularly informative. A more logical way of sorting applications is on the basis of the type of study reported. It is this basis that will be used for this final segment of this contribution. Reports that have appeared in the literature will be subgrouped into small molecule applications, which include both natural products and pharmaceuticals; metabolic and metabonomic studies; and finally protein and biochemical applications, in which NMR screening is included.

Small Molecule Applications Among the earliest of the small molecule applications reported in the literature was a report by Logan and

co-workers in 1999 that demonstrated the results that could be obtained with a 5 mm 1 H cryogenic probe in the acquisition of NOE difference spectra on small samples of the natural product taxol [10]. In 2000, the author and co-workers reported a comparative study using conventional and cryogenic 3 mm gradient inverse NMR probes using a 40 µg sample of the indole alkaloid strychnine [11]. This comparison demonstrated an improvement in performance for the cryogenic probe relative to the conventional 3 mm probe by approximately a factor of 3.5X. Liu and co-workers presented the results of their application of cryogenic NMR probe capabilities in the study of several mass-limited pharmaceutical samples [12]. With samples ranging from 10 to 80 µg this group of authors was able to record data that allowed the characterization of the samples. In an example that utilized a 13 C detection cryogenic NMR probe, Bringmann and co-workers [13] reported the results of a study of the biosynthetic

1210 Part II

Pharmaceutical Sciences

Part II

pathway leading to acetogenic isoquinoline alkaloids. The 13 C-13 C INADEQUATE experiments employed to study the incorporation of [13 C2 ] acetate were greatly facilitated by the enhanced sensitivity provided by the cryoprobe. The significant performance advantage of cryogenic NMR probes for the acquisition of long-range 1 H–15 N correlation data was demonstrated using the small oxazolidinone antibiotic eperezolid by the author and coworkers in 2001 [14]. Using a 2 mg sealed sample, data were acquired in both a conventional 3 mm gradient inverse triple-resonance probe and a 5 mm cryogenic probe. Data with a s/n ratio of ∼100:1 were acquired in less than 30 min using the cryogenic NMR probe while it took ∼4 h to achieve only half that s/n ratio in the conventional 3 mm probe. In 2002, Gustafson and co-workers reported the elucidation of the complex polyketide-derived macrolactam pecillastrin A from a deep water Caribbean sponge [15]. It was necessary for these authors to resort to a 500 MHz gradient inverse triple resonance cryogenic probe to acquire the necessary long-range correlation data on the 800 µg sample of the C82 natural product to complete the structure elucidation. Later in 2002, the author and co-workers [16] reported the utilization of a 500 MHz gradient triple resonance cryoprobe in the acquisition of long-range 1 H– 13 C and 1 H–15 N correlation data in the characterization of one of the isolated degradants, 3, shown in Scheme 1. Cook and co-workers put the sensitivity enhancement of cryogenic NMR probe technology to excellent use in their study of the Boreal forest fulvinic acid [17]. This group of authors acquired both 1 H–13 C HSQC and HMBC on this complex mixture, which gave >300 direct correlation responses in the HSQC spectrum. In a study that represents an interesting potential technique for natural products characterization, Griesinger and co-workers reported the measurement of long-range 1 H–13 C coupling constants of natural products at natural abundance using orienting media [18]. Sandvoss and co-workers, exploited the sensitivity of a cryoprobe to facilitate the acquisition of HSQCTOCSY data for limited samples of two complex new asterosaponins from the starfish Asterias rubens [19]. Cryogenic NMR probe capabilities were also of importance in the acquisition of HSQC-TOCSY and 1 H–15 N long-range data acquired during the characterization of the complex indoloquinoline dimeric alkaloid quindolinocryptotackieine, which relied on Computer-Assisted Structure Elucidation (CASE) methods because of the extensive overlap in both the proton and carbon frequency domains [20]. Another example of the use of cryogenic NMR probe capabilities for small molecules is found in the work of Exarchou and co-workers [21] who used a combination of solid phase extraction (SPE) and cryoprobe technology for the identification of constituents present in Greek oregano.

Further applications are surveyed in a recently published review [8].

Metabolic and Metabonomic Applications In the first application of cryogenic NMR probe technology of which the author is aware, Pease and co-workers reported the results obtained with a 2.5 mm cryoprobe in the characterization of several metabolites [22]. Using a spectrometer equipped with a 13 C detection cryogenic NMR probe, Keun and co-workers investigated the applicability of 13 C spectral data as a compliment to routinely utilized 1 H data for metabonomic studies [23]. Griffin [24] reported an overview of ongoing metabonomic studies of xenobiotic toxicity and disease states that relied on cryoprobe data. Corcoran and co-workers [25] characterized the sensitivity advantages afforded by cryogenic NMR flow probes in LC-NMR-MS applications in a study of acetaominophen metabolites in urine. St¨ockigt and coworkers [26] described the results of an interesting study in which alkaloid metabolism in hybrid Rauwolfia plant cell cultures was followed using a 500 MHz cryogenic NMR probe. The authors contrast the cryoprobe results with conventional 800 MHz NMR data on the same samples. Cryogenic NMR probe capabilities were also utilized by Bertini and co-workers [27] to study the degradation of aromatic compounds by Pseudomonas putida. More recent applications are surveyed in a recent review [8].

Protein and Biochemical Applications In an early application in the field of protein NMR, D¨otsch and co-workers [28] reported the development of several new 13 C detected protein NMR experiments designed to exploit the sensitivity of a 13 C detection cryoprobe. Wand and co-workers [29] described their efforts in the optimization of the application of cryogenic NMR probe technology in the study of proteins. In one of the early applications of cryoprobe technology involving proteins used in NMR-based screening of pharmaceuticals, Fesik and co-workers [30] compared the results obtained in NMR screening studies using specifically 13 C labeled amino acids in proteins with the more traditionally employed 1 H–15 N HSQC NMR screening methods. The authors demonstrated that with cryogenic NMR probes, this approach gave ∼3X the sensitivity of the more traditional proton-nitrogen based screening approach. After the initial flurry of cryoprobe applications in the field of protein NMR, there was somewhat of a gap before papers again began to appear. This interval corresponds to initial vendor-investigator collaborations during the early development of cryogenic NMR probes and

Applications of Cryogenic NMR Probe Technology

Conclusions Cryogenic NMR probes are continuing to go through what will probably be a long series of iterative design improvements that should move sensitivity incrementally higher. At present, cryogenic probes offer a very significant gain in sensitivity over conventional NMR probes, typically threefold or higher at observation frequencies ranging from 500 to 900 MHz. The pharmaceutical industry is already heavily invested in cryogenic NMR probes, which have significant reduced sample isolate requirements for the characterization of impurities and degradants of pharmaceuticals in the possessive author’s—laboratories. Similar advantage has been gained in laboratories devoted to the characterization of metabolites of pharmaceuticals. As investigator access to cryogenic NMR probe-equipped spectrometers continues to improve, there will undoubtedly be many new applications reported in the literature in the fields of natural

product characterization, biosynthesis, forensic sample identification, and many other disciplines.

References 1. Claridge TDW. High Resolution NMR Techniques in Organic Chemistry. Pergammon Press: Amsterdam, 1999, pp 221– 58. 2. Martin GE, Hadden CE, Russell DJ. In: G Gauglitz, T Vo-Dinh (Eds). Handbook of Spectroscopy, Vol. 1. Wiley-VCH: New York, 2003, pp 234–54. 3. Martin GE. In: DM Grant, RK Harris (Eds). Encyclopedia of Nuclear Magnetic Resonance, Vol. 9. Wiley and Sons: Chichester, 2002, pp 98–112. 4. Styles P, Soffe NF, Scott CA, Cragg DA, Row F, White D, White PCJ. J. Magn. Reson. 1983;60:397. 5. Styles P, Soffe NF, Scott CA. J. Magn. Reson. 1989;84: 376. 6. Anderson WA, Brey WW, Brooke AL, Cole B, Delin KA, Fuks LH, Hill HDW, Johanson ME, Kotsubo VY, Nast R, Withers RS, Wong WH. Bull. Magn. Reson. 1995;17:98. 7. Hill HDW. IEEE Trans. Appl. Superconduct. 1997;7: 3750. 8. Martin GE. In: GA Webb (Ed). Annual Reports on NMR Spectroscopy, Vol. 56. Elsevier: Amsterdam, 2005, pp 1–99. 9. These data were acquired without gradients. In the author’s experience, for 1 H–13 C long-range correlation experiments with very small samples, i.e. 0) and thus the angle θ is between 0 and the magic angle θ magic or between 180◦ and 180◦ − θ magic . In case that θ lies in the remaining region (3 cos2 θkl,mn − 1 < 0) the outer lines become narrower and the inner lines become broader. The c case for positive cross-correlated relaxation rate kl,mn is represented in Figure 1. Often, line width measurements suffer from inaccuracies, especially since there are many contributions to the line width. Therefore, it is desirable to perform product operator transformations that create operators exclusively based on cross-correlated relaxation that otherwise could not be created. For that, we write down the product operator transformations that are effected by dipole/dipole cross-correlated relaxation. As pictorially demonstrated in Figure 1, we then find [2]: c kl,mn t

c Ik,x Im,y —−→ Ik,x Im,y cosh kl,mn t − 4Ik,x Im,y Il,z In,z c sinh kl,mn t

(2)

Similarly we obtain for the evolution of an antiphase operator: c t kl,mn

2Ik,x Im,y Il,z —−→ 2Ik,x Il,z Im,y c × cosh kl,mn t − 2Ik,y Im,x In,z c × sinh kl,mn t

(3)

Obviously, operators are formed due to the evolution of cross-correlated relaxation that would not be formed without. The trick to perform now quantitative  experiments lies in the observation that the simultaneous evolution of the Jkl and Jmn couplings will lead to the same transformation as in Equation (3) but with known transfer amplitude provided the size of the coupling constants is known: Jkl ,Jmn

2Ik,x Im,y In,z —−→ 2Ik,x Im,y Il,z sin π Jkl t sin π Jmn t − 2Ik,y Im,x In,z cos π Jkl t cos π Jmn t + additional terms

(4)

Thus, two experiments can be performed, one that transforms the operator 2Ik,x Im,y Il,z into 2Ik,y Im,x In,z with c sinh kl,mn t in a so-called quantitative  cross experiment and a second experiment, the reference experiment, in which this transformation is effected by evolution of J

Transferred Cross-Correlated Relaxation

Cross-Correlated Relaxation for the Measurement 1249

Part II

c

b

70.9

C3

H2’

H2” 13C

2.60

2.72

2.48

2.36

2.72

2.60

2.48

2.36

1

H

a) y

y 1

H

∆ ∆ 2 2

∆ ∆ 2 2 ϕ1

13

C

t1 2

τ´ 2

τ´-t1 2

ϕ3 ϕ2 τM ∆’ τM ∆’ τM ∆’ τM ∆’ τ˝ + − − + 4 4 4 b 2 4a 4 4 4 4

τ˝ 2

t2, ϕrec

ϕ4 Decoupling

Gz

c Fig. 2. Quantitative CH,CH HCCH experiment (a) and spectra from reference (b) and cross experiments (c) applied to the H2 (H2 )C2C3H3 moiety of epothilone when bound to tubulin. From the comparison of the peak intensities of the respective DD,DD D D,D D peaks the cross-correlated relaxation rates (C2H2  ,C3H3 ) = 2.1. + / − 0.3 and C2H 2”,C3H 3 = 3.8 + / − 0.4) can be derived and related to the torsional angle C1–C2–C3–C4 of −146◦ .

couplings with the efficiency: sin π Jkl t sin π Jmn t. Provided the coupling constants are known or can be measured by well-known methodology, it is possible by comparing the intensities of a cross peak in the cross to the reference experiment to determine the cross-correlated relaxation rate.

Measurement and Application of the Quantitative Γ Method for Cross-Correlated Relaxation Rates Quantitative  methods have been introduced originally by Felli et al. [3] and Pelupessy et al. [4] for dipole– dipole cross-correlated relaxation by transferring an initial

operator to another operator by the evolution of the desired cross-correlated relaxation rate. Figure 2 shows the c quantitative  sequence that measures the CH,CH crosscorrelated relaxation rate. In this  HCCH experiment, which has been applied to aliphatic side chains or to sugars to measure the pseudorotation phase cross-correlated relaxation Cc i Hi ,C j H j of the double- and zero-quantum coherence (DQ/ZQ) 4Hi z Ci x C j y generated at time point a creates the DQ/ZQ operator 4H j z C j x Ci y . In the second part of the experiment, the operator 4H j z C j x Ci y is transferred via a 90◦ y-pulse applied to 13 C nuclei further to the proton H j to give rise to a cross peak at (ωH j , ωCi ). CCSA,DD i ,C j H j and HCSA,DD cross-correlated relaxation are refocused by i ,C j H j application of the 180◦ carbon and proton pulses during

1250 Part II

Pharmaceutical Sciences

Part II

the mixing time τ M . The following transfers are achieved in the sequence of Figure 2:   4Hi z Ci x C j y → 4Hi z Ci x C j y cosh Cc i Hi ,C j H j τM  
 × cos2 π JCH  −sinh Cc i Hi ,C j H j τM


× sin2 π JCH  −4H j z C j x Ci y  
 × sinh Cc i Hi ,C j H j τM cos2 π JCH   


+ cosh Cc i Hi ,C j H j τM sin2 π JCH 

stems from the fact that the cross-correlated relaxation is active also in the reference experiment. It is obvious that a comparison of the relative intensities based on the knowledge of the respective coupling constants yields the cross-correlated relaxation rate. The 8C1x C2z C3x H3z term is then transformed into H3x for detection. The delays are optimized for optimal refocusing and defocusing of the couplings. Selective π pulses on the carbons prevent loss of magnetization to other passively coupled carbons. The complete coherence transfers are given in the following: 1J C1 C2

π/2(C)

(5) The last term produces a cross peak at (ωCi , ωH j ) due to coherence transfer between 4Hi z Ci x C j y and 4H j z C j x Ci y , while the first term produces a cross peak at (ωCi , ωH j ) and can therefore be distinguished. In the experiment with  = 0, the intensity of the cross peak (I cross ) is proportional to sinh(Cc i Hi ,C j H j τ M ) whereas for  = 1/2JCH , the intensity of the cross peak (I ref ) is proportional to cosh(Cc i Hi ,C j H j τ M ). By comparing the intensity of the cross peak measured in the two experiments one can determine:   I cross c = tanh  τ M C H ,C H i i j j I ref

(6)

This is applied in Figure 2B and C to the torsion angle defined by C1–C2–C3–C4 of epothilone and an angle of −143◦ is obtained. An extension of this method is the measurement of cross-correlated relaxation between a methyl group and a remote HC moiety in a C1 H3 C2 C3 H fragment [5] which yields the torsion angle about the C2 C3 bond. The pulse sequence used for the measurement of the CH,CH rate between the C3 H vector and the C1 H3 vectors is shown in Figure 3A. Two experiments are recorded, yielding a cross and a reference spectrum similarly as for the HCCH case except for the fact that the transfer requires one further transfer step and the double-quantum and zero-quantum coherence is excited between geminal instead of directly bound carbons. The CCR rate is extracted from the ratio of peak intensities in the cross and reference experiment. The pulse sequence is optimized for maximum sensitivity of the cross experiment: the 8H1z C1y C2z C3y coherence present at point c is transformed 1 2 3 3 by CDD 1 H,C3 H into 8C x Cz C x Hz with efficiency equal to DD 3 sinh(C1 H,C3 H Trel ) cosh2 (CDD 1 H,C3 H Trel ). For the reference experiment, the same operator transformation is achieved with the efficiency: 3 sin(π JC1 H /2) sin(π JC3 H /2) cosh2 (π JC1 H /2) cosh3 (CDD The last term 1 H,C3 H Trel ).

1J C2 C3

a 2H1z C1y —−→ b 4H1z C1y C2x —−→ c 8H1z C1y C2z C3y π/2(C)

π/2(C)

CCR(cross)

———−→ d 8C1x C2z C3x H3z —−→ 8C1z C2x C3z H3z 1 J (reference) CH 1J C1 C2

1J C2 C23

INEPT

—−→ e 4C2z C3y H3z —−→ f 2C3x H3z —−→ g H3x π/2(C)

(7)

Transferred Cross-Correlated Relaxation Due to the fact that cross-correlated relaxation depends linearly on the correlation time it can be used to determine the conformation of ligands when bound to target molecules, provided the off rate is fast enough to enable detection of the cross-correlated relaxation rate via the free ligand [6, 7]. The conditions, under which such an experiment can be performed, are similar to those found for transferred NOEs [8] and for K d s normally exceeding 10−6 M, ligands are in an equilibrium between a protein-bound and a free form. The period during which the molecule is in the bound conformation will contribute to relaxation to the largest extent and hence will heavily weight experiments that are based on relaxation. The population-averaged cross-correlated relaxation rate for dipole, dipole cross-correlated relaxation between two sites with fractions pL of the free ligand and pML = (1 − pML ) of the bound ligand is given in Equation (8): 

2 γC2 γH2 µ02 h¯ 2 CDD,DD = m Hm ,Cn Hn 5 (4π)2 rC3 m Hm rC3 n Hn

3 cos2 θmn,L − 1 2 × Smn,L pL τc,L 2 2 + Smn,M L pM L

3 cos2 θmn,M L − 1 τc,M L 2



(8)

Transferred Cross-Correlated Relaxation

Cross-Correlated Relaxation for the Measurement 1251

Part II

a)

φ3

b)

22/23

Fig. 3. (A) Pulse sequence for the measurement of dipolar–dipolar cross-correlated relaxation between the (CH)ave. vector and the (CH)3 vector in epothilone.  = 1/(21 JCH ); T = 1/(41 JCC ) = 7 ms; Tc = 1/ (41 JCC ) = 7 ms; T1 = 1/(81 JCC ) = 3.5 ms; T1c = Tc . Delays 1–4 are equal to Trel /4 (Trel = 1/1 JCC = 28 ms) in the cross experiment; in the reference experiment delay 1 = Trel /4 + /8 = delay 3 and delay 2 = Trel /4 − /8 = delay 4. φ 1 = y; φ 2 = x, −x; φ 3 = 2(y), 2(−y); φ 5 = 4(x), 4(−x); φ 6 = 8(y), 8( −y); φ 8 = x; φ 10 = y.; φ 11 = 2(x,−x), 4(−x,x), 2(x, −x); φ 4 = φ 7 = φ 9 = x. The pulses with phase φ 4 and φ 9 are Q3 pulses of 768 ms duration, centered at 63 ppm, to selectively invert the C3 and C4 (C3 resonates at 70.9 ppm and C4 at 53.0 ppm); the pulse with phase φ 7 is a Q3 pulse of 512 ms duration centered at 37 ppm, to selectively invert the C22/23 and C4 (C22 resonates at 19.1 ppm and C23 at 21.7 ppm). The proton carrier was at 4.78 ppm and the carbon carrier at 37 ppm; spectral widths were 4807.69 Hz for proton and 8333.33 Hz for carbon. Carbon decoupling in acquisition consisted of a GARP modulated pulse train at 2.38 kHz field strength. Reference (b) and cross (c) spectra acquired with the sequence in (a). The peaks visible in the reference spectrum are at the chemical shift of C22 (19.1 ppm) and C23 (21.7 ppm) in the carbon dimension and at the chemical shift of H3 (4.27 ppm) in the proton dimension. The cross experiment was four times longer than the reference experiment. 1024 and 64 complex points were acquired in t2 and t1 , respectively. The final matrices were 2048 × 128 points. The gray lines represent negative contours. The sample contained 0.5 mM epothilone A and 5 µM tubulin dissolved in D2 O. (D) Dependence of cross-correlated relaxation rates on the torsion angle about the C3–C4 bond. The two cross-correlated relaxation rates determine the angle to be −45◦ . (E) Numbering scheme of epothilone A and location of the torsion angles studied (indicated by bars drawn along the relevant bonds. (See also Plate 94 on page XVIII in the Color Plate Section.)

1252 Part II

Pharmaceutical Sciences

Part II

If the correlation time of the free ligand τc,L is much smaller than the correlation time of the bound ligand τc,M L , the observed averaged cross-correlated relaxation is dominated by the bound conformation rate CDD,DD m Hm ,Cn Hn and therefore, precise information on the conformation of a bound ligand can be obtained. This concept has been introduced on the determination of the conformation of a phosphotyrosine peptide weakly bound to STAT-6 [7] []and of a tRNA analog bound to EF-Tu· GDP complex [6]. Here we will show the power of the method on the determination of the epothilone conformation in the tubulinbound state.

Application to the Epothilone/Tubulin Complex In this review, we describe the application of trCCR measurements for the determination of the conformation of epothilone when bound to tubulin. Epothilones are microtubule stabilizers and inhibit cell proliferation through a mechanism of action which is analagous to r that of the renowned clinical anti-cancer drug Taxol
(paclitaxel).[9] Epothilones exhibit extraordinary antiproliferative activity in vitro and they efficiently induce cell death in paclitaxel-resistant tumor cell lines at up to 5000-fold lower concentrations than paclitaxel [9–11]. In addition, they possess higher water solubility than paclitaxel [10,12] which allows delivery in vivo with non cremaphor-containing formulation vehicles, thus eliminating formulation-based side effects [major hypersensitivity reactions (HSR)] [13] The potential clinical utility of epothilones is supported by in vivo experiments with epothilone B in a variety of nude mouse human tumor models [10,14]. Fortunately, the kinetic and thermodynamic properties of the complex between epothilone A and tubulin are in the desirable range for trNOE and trCCR experiments. Indeed, the trNOE and trCCR data are found in agreement with a dissociation constant K d in the range of 10– 100 µM. Evidence of specific binding of epothilone A to tubulin is provided by the restoration of the NOE spectrum of free epothilone A upon addition of epothilone B to the mixture, which proves the quantitative displacement of epothilone A from tubulin by the tighter binder epothilone B (not shown). The existence of specific and transient binding of epothilone A to tubulin enables the structural analysis of the active conformation of the epothilones by NMR. The tubulin-bound conformation of epothilone A was calculated from 46 inter-proton distance restraints and seven torsion angle restraints measured for a 0.5 mM solution of epothilone A in water in the presence of 5 µM tubulin. The distance restraints were derived from transferred NOE experiments. To filter out spin-diffusion mediated

peaks, only those signals were taken into account which were present with opposite sign to the diagonal peaks in a transferred ROESY experiment. The dihedral angle restraints were obtained by measuring CH–CH dipolar– dipolar and CH–CO dipolar-CSA transferred CCR rates for 60–70% 13 C labeled epothilone A. The trCCR experiments [6,7] were indispensable to obtain a unique description of the bound conformation, as more than one structure of the macrolide ring is compatible with the same H–H distance set (NOE intensities). Seven trCCR rates defined the torsion angles C1–C2–C3–C4, C5–C6–C7– C8, C12–C13–C14–C15, C13–C14–C15–O11, and C14– C15–C16–C17 were all determined using the quantitative  HCCH. The C2–C3–C4–C5 was determined from the ratio of the cross-correlated relaxation rates between the C22 H3 and C23 H3 , respectively, with the C3 H dipole. The O11–C1–C2–C3 was determined from CSA/DD trCCR (not described here) [15]. Figure 3B and C shows a representation of the methyl group experiment with the cross and reference experiments, respectively. In the cross spectrum the C23 (ω1)–H3 (ω2) peak is missing, indicating that the dipolar–dipolar CCR between the (CH)23 ave. vector and the (CH)3 vector is close to zero. The quantitative value for the two CCR rates can be extracted according to the equation: Icross /Iref = tanh(Trel )/sin2 (πJ CH /2) cos2 × (π JCH /2)

(9)

where Icross and Iref are the intensities of the peaks in the cross and reference experiment, respectively and  indicates the cross-correlated relaxation rate. The measured CCR rates between the (CH)3 vector and the (CH)23 ave. and (CH)22 ave. vectors are 0.1 ± 0.1 and −1.4 ± 0.4, respectively. The dihedral angle θ that complies with the ratio between the two CCR rates is −45◦ ± 5◦ . We used the ratio of the two CCR rates, instead of the rates themselves, to determine the dihedral angle because this approach is independent from both the τ c and the internal motion order parameter S 2 , if equal internal reorientation of all the CH vectors is assumed. The change in the θ dihedral angle from the gauche + conformation in the free form to the gauche—conformation in the bound form is corroborated by transfer-NOE data. The tubulin-bound conformation of epothilone A [15] is shown in green in Figure 4 and is compared with the free (unbound) conformation of epothilone A determined by X-ray crystallography, [15,16] which is shown in gray. We choose to compare the tubulin-bound conformation of epothilone with the X-ray structure and not with the solution structure available in CD2 Cl2 , because of the extensive flexibility of epothilone in solution in absence of tubulin [17]. However, the most populated conformer in

Transferred Cross-Correlated Relaxation

References 1253

Part II

b f

f

Fig. 4. Comparison of the conformations of epothilone as a stereoview. Arrows indicate free epothilone A (f: X-ray) and epothilone A bound to tubulin (B). Note the inversion of the orientation of the thiazole ring as well as the change in position of the 3-OH group. (See also Plate 95 on page XVIII in the Color Plate Section.)

solution is indeed very similar to the X-ray conformer. The position of the thiazole nitrogen (blue in Figure 4) and 3-OH (red in Figure 4), which are important for the delineation of a pharmocophore model, change significantly upon binding. The structure of epothilone as derived from transferred NOE and cross-correlated relaxation is not identical with the structure derived from an electron crystallography study on Zn induced tubulin sheets [18]. The difference may be due to differences between the conformation of tubulin when assembled in Zn induced flat sheets as opposed to soluble tubulin or the low resolution of the electron crystallography that required extensive modeling of the epothilone conformation.

Conclusion In conclusion, we have shown that transfer NOE and trCCR rates can be used for the determination of the conformation of epothilone when bound to tubulin in physiological conditions. So far, it has not been possible to obtain crystal structures of tubulin or microtubules. The structure differs from the one proposed from electron crystallography on Zn induced tubulin sheets which could be

due to differences in the preparation of the samples under investigation or due to the low resolution of the electron crystallography structure.

References 1. Reif B, Hennig M, Griesinger C. Science. 1997;276:1230– 3; Reif B, Diener A, Hennig M, Maurer M, Griesinger C. J. Magn. Reson. 2000;143:45–68. 2. Schwalbe H, Carlomagno T, Junker J, Hennig M, Reif B, Richter C, Griesinger C. Methods Enzymol. 2001;338:35– 81. 3. Felli IC, Richter C, Griesinger C, Schwalbe H J. Am. Chem. Soc. 1999;121:56–7. 4. Pelupessy P, Chiarparin E, Ghose R, Bodenhausen G. J. Biomol. NMR. 1999;13:375–80. 5. Carlomagno T, S´anchez V, Blommers MJJ, Griesinger C. Angew. Chem. 2003;115:2619–21; Angew. Chem. Int. Ed. 2003;42:2515–7 6. Carlomagno T, Felli IC, Czech M, Fischer R, Sprinzl M, Griesinger C. J. Am. Chem. Soc. 1999;121:1945–8. 7. Blommers MJJ, Stark W, Jones CE, Head D, Owen CE, Jahnke W. J. Am. Chem. Soc. 1999;121:1949–53. 8. Clore GM, Gronenborn AM. J. Magn. Reson. 1982;48:402– 17; Clore GM, Gronenborn AM. J. Magn. Reson. 1983;53: 423–42; Ni F. Prog. NMR Spectrosc. 1994;26:517; Lian LY,

1254 Part II

Part II

9. 10. 11. 12.

Pharmaceutical Sciences

Barsukov IL, Sutcliffe MJ, Sze KH , Roberts GCK. Methods Enzymol. 1994;239:657–707 Bollag DM, McQueney PA, Zhu J, Hensens O, Koupal L, Liesch J, Goetz M, Lazarides E, Woods CM. Cancer Res. 1995, 55, 2325–33. Altmann K-H, Wartmann M, O’Reilly T. BBA-Rev. Cancer 2000;1470:M79–91. Kowalski J, Giannakakou P, Hamel E. J. Biol. Chem. 1997;272:2534–41. H¨ofle G, Bedorf N, Steinmetz H, Schomburg D, Gerth K, Reichenbach H. Angew. Chem. 1996;108:1671–2; Angew. Chem. Int. Ed. 1996;35:1567–9.

13. Rowinsky EK. Ann. Rev. Med. 1997;48:353–74. 14. Altmann K-H, Bold G, Caravatti G, End N, Florsheimer A, Guagnano V, O’Reilly T, Wartmann M, Chimia 2000;54:612– 21. 15. Carlomagno T, Blommers MJJ, Meiler J, Jahnke W, Schupp T, Petersen F, Schinzer D, Altmann K-H, Griesinger C. Angew. Chem. 2003;115:2615–9, Angew. Chem. Int. Ed. 2003;42:2511–5. 16. Rihs G, Walter HR. unpublished results. 17. Taylor RE, Zajicek J. J. Org. Chem. 1999;64:7224–8. 18. Nettles JH, Li H, Cornett B, Krahn JM, Snyder JP, Downing KH. Science 2004;305:866–69

1255

R. Andrew Byrd, C. Andrew Fowler, Robert L. McFeeters, and Vadim Gaponenko Structural Biophysics Laboratory; Center for Cancer Research; National Cancer Institute; Frederick, MD 21702-1201, USA

Abbreviations: PCS, pseudocontact shift; RDC, residual dipolar coupling; pmiRDC, paramagnetic induced residual dipolar coupling; PRE, paramagnetic relaxation enhancement; PAS, paramagnetic anisotropic susceptibility.

Introduction NMR is a critical tool in current structural biology, and plays a significant role in the understanding of pharmacologically significant proteins, as well as the development of modulators for these proteins. The general approach for structure determination has been well established; however, as the systems of interest become larger (e.g. >25 kDa) and less well behaved in solution (e.g. solubility limits of 50–200 µM), new approaches are required to enable NMR to remain a key player. The introduction of cryogenic probes has dramatically impacted our ability to study lower concentration and/or lower stability proteins. Nevertheless, there has been a need to develop new approaches to provide long-range structural restraints to solve structures of highly deuterated proteins and to determine the orientation of domains in multidomain proteins, or in multimeric complexes. This article will focus on our recent efforts in this area of research, specifically the introduction of paramagnetic metal ions into non-metalloproteins and the unique and powerful structural restraints that can be obtained from this approach. The ability of paramagnetic centers to provide valuable structural information for proteins has been recognized for a long time [1,2]. The principal significance of ˚ or this information is the long-range distance (10–40 A) global (distance and orientation) nature of the restraints. By augmenting traditional short-range, scalar restraints (such as NOEs) with long-range restraints, more accurate protein structures can be obtained. The relative proximity and/or orientation of protein domains or proteins within a complex can also be investigated. There are three main types of data that are obtainable: paramagnetic relaxation enhancement (PRE), pseudocontact shifts (PCSs), and residual dipolar couplings (RDCs). Paramagnetic relaxation enhancements provide long-range distance information [3–6]; however, while Graham A. Webb (ed.), Modern Magnetic Resonance, 1255–1261.
C 2008 Springer.

relaxation enhancement is a powerful tool, we will focus our discussion here on the use of PCSs and RDCs. Paramagnetic induced residual dipolar couplings (pmiRDCs) and cross-correlation rates between dipolar and Curie relaxation pathways offer unique ways for obtaining orientational information on bond vectors in protein molecules [7–10]. PCSs provide both distance and orientational information that is very important for protein structure determination [11–14]. However, until recently, observation of these valuable structural parameters was limited to naturally metal binding proteins. Investigators have now begun to examine ways to introduce paramagnetic centers into non-metal binding proteins. We will illustrate two different approaches to achieve paramagnetic tagging and discuss new methods that enhance this general approach by providing more precise measurements of RDCs. Finally, we will demonstrate the impact of PCSs and pmiRDCs on solving difficult protein structures (>25 kDa) and on defining domain orientation in two pharmacologically significant proteins, STAT4 and hepatocyte growth factor (HGF).

Methods to Bind Paramagnets to Non-Metalloproteins One way to introduce paramagnetic probes into nonmetalloproteins is through chemical modification that introduces a metal-chelating site. A facile modification site is either a unique, natural cysteine residue or an engineered cysteine residue within the protein of interest. Two early examples of this type of protein modification with paramagnetic agents involved the attachment of nitroxide spin labels to lysozyme and bovine pancreatic trypsin inhibitor [15,16], allowing the measurement of relaxation enhancements. One limitation of this approach is evident for proteins with multiple, structurally important cysteine residues, wherein complications due to cross-linking, aggregation or refolding of the unmodified protein may occur. In this case, alternative approaches to attachment of the metal binding ligand can be utilized (vide infra). More recently, EDTA based ligands were successfully employed for paramagnetic labeling of macromolecules. Proteins modified with

Part II

Novel Uses of Paramagnets to Solve Complex Protein Structures

1256 Part II

Pharmaceutical Sciences

Part II

S-(2-pyridylthio)-cysteaminyl-EDTA have been shown to bind Co2+ and allow measurement of PCSs and pmiRDCs. This approach has been demonstrated on barnase and the N-terminal domain of STAT4 [17,18]. Other paramagnetic ions, e.g. lanthanides, can be bound to the EDTA ligands and yield a wealth of PCS and pmiRDC data. It has been shown that under some conditions the EDTA-metal complex can exist in two forms, believed to be stereoisomers. The two forms result in two different paramagnetic ansisotropic susceptibility (PAS) tensors and two sets of shifted resonances, as observed for two proteins, trigger factor and ArgN [19,20]. Further inspection indicates that the multiple forms may be due to the protonation state of the EDTA moiety at the time of complexation (Gaponenko et al., to be published). Additionally, a different cysteine-ligation paramagnetic tagging reagent has been proposed [21]; however, multiple isomers were observed and the reagent may prove problematic. To address the problem of multiple forms, a slightly modified and enantiometrically pure EDTA-derived tag has been proposed [19]. This new ligand is anticipated to facilitate broader use of the cysteine-ligation approach, and it is clear that further advances in ligand design will be forthcoming soon. A related approach employed cysteaminyl-EDTA-derivatized 4-thio-deoxythymidine incorporated into DNA that was subsequently used to form a protein-DNA complex. By binding Mn2+ to the EDTA moiety, long-range intermolecular distances were derived from PREs of protein resonances [22,23]. Since PCSs are dependent on the distance and orientation of nuclei from the paramagnet, it is likely that not all segments of a molecule will reveal PCS data using a single ligation site. However, it is straightforward to serially engineer multiple ligand attachment sites, thereby obtaining PCS data for all segments of secondary structure (Figure 1) [17]. The multiple ligand attachment sites also yield separate, distinct alignment tensors, relative to the molecular coordinate system. This is particularly attractive, since it is often difficult to reduce the degeneracy for RDC solutions with one alignment [24,25], while multiple alignments can be difficult to achieve due to protein interactions with the alignment media or

Fig. 1. Co2+ ligation at three distinct sites on STAT4NT . (A) wt-C107, (B) T50C, and (C) K92C. The vectors illustrate the PCSs observed for HN protons.

correlation of the tensor orientations (C.A. Fowler et al., in preparation). A second method of achieving paramagnetic binding to proteins is through the use of chelating peptide tags. Several peptides that can be cloned at either terminus of an expressed protein have been reported to bind paramagnetic metals. ATCUN (amino terminal Cu2+ (Ni2+ ) binding) motifs [26] with bound Cu2+ have been used to provide distance restraints by PRE in both ubiquitin [27] and several calmodulin-peptide complexes [28]. Another short peptide tag, HHP (His-His-Pro), has been cloned at the N-terminus of thioredoxin [29], allowing the measurement of PRE in the presence of Ni2+ . Longer peptides can also chelate metals. A 17 residue lanthanide binding peptide tag has been cloned at the N-terminus of ubiquitin and used to obtain partial orientation and observe pmiRDCs [30]. Finally, chelated metals can be bound using a combination approach. A 26 residue, calmodulin binding peptide was attached via a short linker to the Cterminus of dihydrofolate reductase (DHFR) and Tb3+ saturated calmodulin was added and observed to bind to DHFR, transferring a degree of paramagnetic alignment and allowing the measurement of pmiRDCs [31]. We have proposed a simpler, common peptide tag that is used in immobilized metal affinity chromatography (IMAC) [32–35]. IMAC is a technique generally used to purify recombinant proteins expressed with a chelating peptide tag. The tag binds to partially chelated metal ions in the column resin, allowing contaminants to be washed away followed by elution of the purified protein. One of the most common tags used for IMAC is hexahistidine (the so called his-tag) [36,37], which can be added at either the N- or C-terminus of a protein. Since his-tags can bind resin-associated metal ions, it follows that they can also bind free metal ions in solution. It has recently been shown that several his-tagged proteins are capable of binding free cobalt ions in solution. The paramagnetic anisotropy of the cobalt-histag complex allows measurement of PCSs and also leads to partial alignment of the sample in a magnetic field, allowing the observation of pmiRDCs. The PCSs and pmiRDCs measured from the protein-his6 -Co2+ species

Novel Uses of Paramagnets

PCS Assignment and Use of PCSs and pmiRDCs as Structural Restraints 1257

Part II

Table 1: Magnitudes of pmiRDCs observed for various protein-metal complexes Protein and chelating tag NK1-his6 Ubiquitin-his6 (N-terminus) Ubiquitin-his6 (C-terminus) Ubiquitin-LBP Ubiquitin-LBP Ubiquitin-LBP DHFR-CBP YmoA-his6 NusB-his6 STAT4-EDTA Trigger factor-EDTA-carbonic acid

Metal or metal chelate

Largest magnitude 1 H-15 N RDC (Hz)

Co2+ Co2+ Co2+ Tb3+ Dy3+ Tm3+ calmodulin(Tb3+ )4 Co2+ Co2+ -NTA Co2+ Dy3+

−3.0 −2.5 −0.7 −7.6 −6.6 +4.5 −7.4 +2.0 +2.0 −4.0 +8.0

LBP, lanthanide binding peptide; DHFR, dihydrofolate reductase; CBP, calmodulin binding peptide; NTA, nitrilotriacetic acid

have been shown to be relatively small; however, they are structurally significant and readily obtained (vide infra). Table 1 lists the magnitudes of observable pmiRDCs for several different chelating tags and conditions.

of the paramagnetic center with the protein atoms [2,12]. The pseudocontact shift is described by the following equation: δpc =

PCS Assignment and Use of PCSs and pmiRDCs as Structural Restraints First, we will consider the PCSs. The presence of a metal ion introduces a PAS that results in the observation of PCSs for the NMR resonances. PCSs are measured as differences (δ pc ) in chemical shift between the paramagnetic and diamagnetic species. Such differences are readily measured in the metal-free and metal-bound states. PREs can degrade the sensitivity of NMR experiments for sites close to the metal. For this reason, we have chosen to use Co2+ , which provides less relaxation enhancement than lanthanides or Mn2+ . Assignment of PCSs can be achieved using traditional NMR assignment strategies [38]. An alternative assignment strategy utilizes a preliminary structure of the macromolecule to estimate the PAS tensor orientation from several unambiguous PCSs [11], which have been assigned by inspection of HSQC spectra. The subsequent assignment of the remaining PCSs is accomplished by an iterative procedure of predicting the PCS values, using the PAS tensor orientation and the preliminary structure coordinates, and then re-optimizing the PAS based on new assignments. We have found that this procedure can assign all observable PCSs, including shifts up to 650 ppb, which were not previously assignable by inspection or other methods (Gaponenko et al. in preparation, 2005). If the metal is chelated only by the EDTA or His6 ligand, then there is no Fermi-contact contribution to δpc , since there is no direct overlap of the molecular orbitals

 Prh Pax
3 cos2 θ − 1 + 3 sin2 θ cos(2ϕ). r3 r

(1)

where Pax and Prh are the axial and rhombic components of the PAS, r is the distance between the unpaired electron and the observed nucleus, and θ and φ are the polar angles describing the orientation of the electron-nucleus vector in the PAS tensor frame. PCSs follow the same functional form as RDCs and the structural restraints are very similar to homonuclear dipolar couplings [39]. However, the PCS restraints are more useful than homonuclear dipolar couplings because the sign information is readily available. It is a straightforward procedure to incorporate PCS restraints into structure calculation programs, e.g. XplorNIH, that already have utilities for RDC refinement. The restraint is expressed in terms of the actual PCS and does not have an explicit input distance or orientation. Since many PCS restraints are included (for each tagged site) and combined with other NOE and torsion-angle restraints, a unique solution can be obtained. The distance of the atom, corresponding to the PCS, from the paramagnetic center is determined from the final structures. The ˚ for the corresponding distances range from 10 to 40 A two cases described below. The use of RDCs in protein structure determinations has become very significant and has been recently reviewed [40]. In the present application, partial alignment of the protein arises due to the paramagnetic susceptibility of the tagged protein. The alignment is known to be small [9]; nevertheless, the pmiRDCs are readily obtained without the need for steric alignment media,

1258 Part II

Pharmaceutical Sciences

Part II

and have shown to be extremely valuable, both in terms of structure determination and domain orientation (vide infra). The pmiRDCs obtained from paramagnetic alignment are used in structure calculations exactly as has been described for steric alignment induced RDCs. They have also been useful for cross-validation.

New Approaches to Measurement of Small, Paramagnetically Induced RDCs Paramagnetic alignment can induce relatively small RDCs (±2 to ±8 Hz for HN -N), and it is important to have accurate methods to measure these couplings. Typical procedures for measuring RDCs involve the use of direct frequency measurements of resolved multiplet components, often augmented with the IPAP procedure [41]. These methods can suffer from measurement precision for small couplings. We have recently introduced a suite of experiments, based on the principal of J-modulation, that facilitate the measurement of HN -N, Hα-Cα, Cα-C’, and Cα-Cβ RDCs [58]. The experiments encode the coupling of interest into peak modulation within familiar two-dimensional HN -N (equivalent to HSQC spectra) or HN-C’ correlation spectra. Briefly, J-modulation can be incorporated into a pulse sequence in several different forms (see Figure 2), and the underlying principal remains that the coupling of interest is evolved for an amount of time, , which can be varied. The observed resonance intensity is modulated by the coupling of interest in a sinusoidal manner dependent on . Different types of J-modulation can be introduced into pulse sequences and result in different modulation functions of the coupling of interest. Simple, non-constant time J-modulation (Figure 2A) measures coupling for a total time 2, but the intensity is a composite of J-modulation and I-spin relaxation, which complicates data analysis. Constant-time J-modulation incorporates the J-modulation delay into a constant time period, keeping relaxation effects constant and providing a means to decouple a third spin (Figure 2B). Chemical shift evolution can be reintroduced to constant time Jmodulation rather easily (Figure 2C), thus expanding the repertoire of experiments that can be designed. Acquiring a series of spectra in which  is incremented allows the coupling of interest to be extracted from a plot of resonance intensity vs. total evolution time (Figure 2D). Examples of simple J-modulation [42–45], and constanttime J-modulation [46] have appeared, and we have extended these concepts and developed a suite of experiments to facilitate the rapid and precise measurement of a complete set of RDCs. With adequate signal-to-noise and a sufficient number of data, the precision of J-modulated couplings can exceed those of direct and E.COSY measurements. For a detailed discussion of these methods

and the overall suite of experiments see McFeeters et al. [58]. In general, the precision of RDC measurement using the J-modulation experiments can be of the order of ±0.1 Hz and is significantly improved compared to IPAP methods or three-dimensional methods used to decrease spectral overlap. The convenience of spectral analysis for the familiar HSQC-like HN -N 2D spectra facilitates a rapid, accurate measurement of multiple RDCs for samples with even weak alignment, such as found for paramagnetic induced alignment. These advances ensure that the maximum amount of structural data can be obtained from the paramagnet-tagged protein system.

Structural Applications of PCSs and pmiRDCs We have addressed two difficult problems in protein structure determination: (1) improving structural accuracy for large, perdeuterated proteins, and (2) orientation of protein domains (or separate molecules) without the use of interdomain/intermolecular NOEs. Each of these problems reflects the use of NMR methods in larger, more biologically and pharmacologically significant proteins. The use of deuteration and selective methyl-protonation to enable structural studies of proteins ≥25 kDa is becoming common and has been reviewed [47]. Structural accuracy can be significantly improved by combining observed HN (Figure 1) and methyl-1 H PCSs with the limited NOE-restraints available from an ILVlabeled protein to calculate the three-dimensional structure, as demonstrated for the N-terminal domain of STAT4 [17]. When PCSs were included in structure calculations the RMSD for backbone atoms in secondary structure elements between the solution structure and the crystal ˚ while the qualstructure decreased from 2.9 to 2.0 A, ity factor, Q [48], calculated from pmiRDCs not used in structure refinement decreased from 0.6 to 0.2. Furthermore, PCSs restraints for methyl groups of Ile, Leu, and Val afforded improvement in accuracy of the sidechain positions for these residues. The RMSD values for all heavy atoms in Ile, Leu, and Val sidechains relative to ˚ while the crystal structure decreased from 2.97 to 2.43 A, the precision within the ensemble improved from 2.37 ˚ Using the N-terminal domain of STAT4, we to 1.66 A. have also demonstrated that it is possible to break magnetic symmetry in homodimers, including the difficult case of head-to-head dimers, thus yielding unique structural information for each monomeric domain within the dimer [38] By binding subequimolar amounts of the paramagnet, only one monomer was tagged with paramagnet, and the PCSs to otherwise symmetric residues rendered the two sites magnetically inequivalent. The asymmetric system provided structural restraints, in the form of PCSs and pmiRDCs for each monomer, which enabled the

Structural Applications of PCSs and pmiRDCs 1259

Novel Uses of Paramagnets

B

C

D

Fig. 2. Improved measurement of RDCs via J-modulation. (A) Basic pulse sequence element for non-constant time J-modulation. Intensity as a function of , is a composite of both relaxation and modulation according to sin(2π JIS ) rexp(-R2). (B) Constanttime J-modulation removes relaxation damping resulting in modulation by sin(2π JIS ). (C) Frequency labeling can be reintroduced into constant-time J-modulation with no alteration in the form of modulation. (D). Typical modulation data, Hα-Cα couplings in an isotropic (solid line with squares) and aligned (dashed with circles) sample.

positioning of the two monomers with respect to one another (Gaponenko, unpublished). The improvements in structure calculations derive directly from the fact that PCSs, unlike one-bond RDCs, contain translational (distance) information in addition to orientational information. This fact heightens the value and future impact of using PCS restraints in general for protein structure determinations. The second problem that we address is the ability to orient domains relative to one another within a multidomain protein or complex in the absence of traditional, NOE-based distance restraints, which is especially pertinent where deuteration is employed. This is a general form of the problem relating to the positioning of monomers within a homodimer structure (see above). The general approach assumes that any relatively rigid molecular fragments can be oriented relative to one another, provided enough data can be measured to determine the alignment tensor for each fragment. If the overall molecule is determined to behave as a single rigid unit, then each fragment must experience the same degree of alignment. Superimposing the alignment tensors allows the fragments to be correctly oriented with respect to one another. This technique has been successfully applied to fragments as small as a few residues [49,50], as well as to the problem of orienting entire protein domains. Recent examples of

domain orientation using RDCs include the B and C domains of barley lectin [51], the domains of maltodextrin binding protein [52], and sequential domains in polyubiquitin chains [53–55]. The tensorial information may also be obtained from relaxation analyses [56,57]; however, we will focus on the use of RDCs and PCSs in this discussion. Although useful, domain orientation using RDCs poses two significant challenges. The first of these is overcoming the well known “inversion problem” [25]. Although an alignment tensor can be determined using as few as five RDCs, it is impossible to determine which direction along each axis is positive and which is negative. This leads to four degenerate orientations, and additional independent data is required to overcome this inversion problem. A second problem is that while domains can be properly oriented, RDCs provide no translational (or distance) information and additional data must be used to position the two domains relative to one another. We have recently combined PCSs with RDCs to orient the two domains in the NK1 fragment of HGF (C.A. Folwer et al., in preparation, 2005). RDCs for the two domains were measured in phospholipid bicelles, yielding four possible domain orientations. Several additional alignment media were screened to find a second, independent alignment tensor; however, the protein either interacted with the media or the alignment tensor was

Part II

A

1260 Part II

Pharmaceutical Sciences

Part II

Fig. 3. Solution structure of NK1 refined against bicelle RDCs and cobalt PCSs (PCSs illustrated as orange and purple vectors to the N and K1 domains, respectively). (See also Plate 96 on page XIX in the Color Plate Section.)

not sufficiently different to solve the inversion problem. The protein was then re-cloned with a his-tag at the Cterminus, and shown to bind cobalt specifically to the his-tag. A saturating amount of cobalt was added to the sample, resulting in the observation of PCSs (±80 ppb) for amide protons in both domains. Furthermore, partial alignment of NK1 in the magnetic field, due to paramagnetic anisotropy, allowed the observation of HN -N pmiRDCs on the order of ±3 Hz. The alignment tensor determined from the pmiRDCs was unique relative to the steric alignment tensors and was used to resolve the inversion problem, leading to a unique orientation of the two domains. The observed HN PCSs for both domains were subsequently used as restraints and combined with RDCs measured from bicelle alignment, in a structure calculation to position the two domains of NK1 with respect to each other. The PCSs provided the requisite translational (distance) restraints to achieve a unique structural solution, and the significance is clearly seen in the long-range ˚ that correspond to the PCS restraints distances (10–40 A) (Figure 3). The cobalt pmiRDCs were withheld from the structure calculations and subsequently used for crossvalidation, which confirmed the structure. The final structure was obtained without the use of any intermolecular NOEs, and it is currently being analyzed to improve our understanding of the interaction of HGF with the cMet receptor.

Conclusion We have summarized some of the exciting new applications of paramagnets in non-metalloprotein systems. The structures generated using the newly accessible restraints

are of high quality and hold the potential for more accurate structures, access to larger structures, and more rapid structure determinations. While, to date, a limited number of applications of paramagnetic labeling have been described in the literature, it is already clear that this approach offers a wealth of valuable structural information. Improved tagging techniques and better software for data analysis should further enhance the advantages offered by this methodology for NMR structure determination and studies of molecular dynamics.

References 1. La Mar GN, Overkamp M, Sick H, Gersonde K. Biochem. 1978;17:325. 2. Bertini I, Luchinat C, Piccioli M. Meth. Enzymology. 2001;339:314. 3. Gillespie JR, Shortle D. J. Mol. Biol. 1997;268:170. 4. Gillespie JR, Shortle D. J. Mol. Biol. 1997;268:158. 5. Battiste JL, Wagner, G. Biochem. 2000;39:5355. 6. Gaponenko V, Howarth JW, Columbus L, Gasmi-Seabrook G, Yuan J, Hubbell WL, Rosevear PR. Prot. Sci. 2000;9:302. 7. Hus JC, Marion D, Blackledge M. J. Mol. Biol. 2000;298:927. 8. Ghose R, Prestegard JH. J. Magn. Reson. 1997;128:138. 9. Tolman JR, Flanagan JM, Kennedy MA, Prestegard JH. Proc. Natl. Acad. Sci. USA. 1995;92:9279. 10. Banci L, Bertini I, Huber JG, Luchinat C, Rosato A. J. Am. Chem. Soc. 1998;104:12903. 11. Allegrozzi M, Bertini I, Janik MBL, Lee YM, Liu G, Luchinat C. J. Am. Chem. Soc. 2000;122:4154. 12. Gochin M. J. Biomol. NMR. 1998;12:243. 13. La Mar GN, Del Gaudio J, Frye JS. Biochim. Biophys. Acta. 1977;498:422.

Novel Uses of Paramagnets

35. K˚agedal L. In: J-C Janson and L Ryd´en (Eds). Protein Purification: Principles, High-Resolution Methods, and Applications, 2nd ed. Wiley-VCH, Inc.: New York, 1998, p. 311. 36. Hochuli E. In: JK Setlow (Ed.). Genetic Engineering; Vol. 12, Plenum Press: New York, 1990, p. 87. 37. Hochuli E, Bannwarth W, D¨obeli H, Gentz R, St¨uber D. Biotechnology. 1998;6:1321. 38. Gaponenko V, Altieri AS, Li J, Byrd RA. J. Biomol. NMR. 2002;24:14S3. 39. Tjandra N, Marquardt J, Clore GM. J. Magn. Reson. 2000;142:393. 40. Bax A. Protein Sci. 2003;12:1. 41. Ottiger M, Delaglio F, Bax A. J. Magn. Reson. 1998;131:373. 42. Tjandra N, Grzesiek S, Bax A. J. Am. Chem. Soc. 1996;118:6264. 43. Wirmer J, Schwalbe H. J. Biomol. NMR. 2002;23:47. 44. Evenas J, Mittermaier A, Yang DW, Kay LE. J. Am. Chem. Soc. 2001;123:2858. 45. Mittermaier A, Kay LE. J. Am. Chem. Soc. 2001;123:6892. 46. Hitchens TK, McCallum SA, Rule GS. J. Magn. Reson. 1999;140:281. 47. Kay LE, Gardner KH. Curr. Opin. Struct. Biol. 1997;7:722. 48. Ottiger M, Bax A. J. Biomol. NMR. 1999;13:187. 49. Fowler CA, Tian F, Al-Hashimi HM, Prestegard JH. J. Mol. Biol. 2000;304:447. 50. Delaglio F, Kontaxis G, Bax A. J. Am. Chem. Soc. 2000;122:2142. 51. Fischer MWF, Losonczi JA, Weaver JL, Prestegard JH. Biochem. 1999;38:9013. 52. Skrynnikov NR, Goto NK, Yang D, Choy WY, Tolman JR, Mueller GA, Kay LE. J. Mol. Biol. 2000;295:1265. 53. Fushman D, Varadan R, Assfalg M, Walker O. Prog. Nucl. Magn. Reson. Spectrosc. 2004;44:189. 54. Varadan R, Walker O, Pickart C, Fushman D. J. Mol. Biol. 2002;324:637. 55. Varadan R, Assfalg M, Haririnia A, Raasi S, Pickart C, Fushman D. J. Biol. Chem. 2004;279:7055. 56. Bruschweiler R, Liao X, Wright PE. Science. 1995;268: 886. 57. Fushman D, Xu R, Cowburn D. Biochem. 1999;38:10225. 58. McFeeters RL, Fowler CA, Gaponenko VV, Byrd RA. J. Biomol. NMR. 2005;31:35.

Part II

14. Ubbink M, Ejdeback M, Karlsson BG, Bendall DS. Structure. 1998;6:323. 15. Schmidt PG, Kuntz ID. Biochem. 1984;23:4261. 16. Kosen PA, Scheek RM, Naderi H, Basus VJ, Manogaran S, Schmidt PG, Oppenheimer NJ, Kuntz ID. Biochem. 1986;25:2356. 17. Gaponenko V, Sarma SP, Altieri AS, Horita DA, Li J, Byrd RA. J. Biomol. NMR. 2004;28:205. 18. Dvoretsky A, Gaponenko V, Rosevear PR. FEBS Lett. 2002;528:189. 19. Ikegami T, Verdier L, Sakhaii P, Grimme S, Pescatore B, Saxena K, Fiebig KM, Griesinger C. J. Biomol. NMR. 2004;29:339. 20. Pintacuda G, Moshref A, Leonchiks A, Sharipo A, Otting G. J. Biomol. NMR. 2004;29:351. 21. M. Prudencio, Rohovec J, Peters JA, Tocheva E, Boulanger MJ, Murphy ME, Hupkes HJ, Kosters W, Impagliazzo A, Ubbink M. Chemistry. 2004;10:3252. 22. Iwahara J, Schwieters CD, Clore GM. J. Am. Chem. Soc. 2004;126:5879. 23. Iwahara J, Anderson DE, Murphy EC, Clore GM. J. Am. Chem. Soc. 2003;125:6634. 24. Ramirez BE, Bax A. J. Am. Chem. Soc. 1998;120:9106. 25. Al-Hashimi HM, Valafar H, Terrell M, Zartler ER, Eidsness MK, Prestegard JH. J. Magn. Reson. 2000;143:402. 26. Harford C, Sarkar B. Acc. Chem. Res. 1997;30:123. 27. Donaldson LW, Skrynnikov NR, Choy WY, Muhandiram DR, Sarkar B, Forman-Kay JD, Kay LE. J. Am. Chem. Soc. 2001;123:9843. 28. Mal TK, Ikura M, Kay LE. J. Am. Chem. Soc. 2002;124:14002. 29. Jensen MR, Lauritzen C, Dahl SW, J. Pedersen, Led JJ. J. Biomol. NMR. 2004;29:175. 30. W¨ohnert J, Franz KJ, Nitz M, Imperiali B, Schwalbe H. J. Am. Chem. Soc. 2003;125:13338. 31. Feeney J, Birdsall B, Bradbury AF, Biekofsky RR, Bayley PM. J. Biomol. NMR. 2001;21:41. 32. Ueda EKM, Gout PW, Morganti L. J. Chromatogr. A. 2003;988:1. 33. Lindner P, Guth B, W¨ulfing C, Krebber C, Steipe B, M¨uller F, Pl¨uckthun A. Methods. 1992;4:41. 34. Porath J, Carlsson J, Olsson I, Belfrage G. Nature. 1975;258:598.

References 1261

1263

Guido Pintacuda1,2 ,Thomas Huber3 , Max A. Keniry1 , Ah Young Park1 , Nicholas E. Dixon1 , and Gottfried Otting1 1 Research

School of Chemistry, Australian National University, Canberra ACT 0200, Australia 2 Karolinska Institute, Tomtebodav¨ agen 6, S-17177 Stockholm, Sweden 3 Departments of Mathematics and Biochemistry, The University of Queensland, Brisbane QLD 4072, Australia

Abstract 15

15

N-HSQC spectra of N-labeled proteins are widely used to identify ligand-binding sites on protein surfaces, providing a way to assess protein–protein interactions as well as to screen for the binding of small chemical compounds. The method is particularly powerful if the cross peaks in the 15 N-HSQC spectrum have been sequencespecifically assigned. The present article reviews the use of paramagnetic labeling to obtain these resonance assignments quickly with the use of 15 N-labeled protein, provided the crystal structure of the protein is known and the protein can be labeled with a paramagnetic ion. The method also yields the anisotropy tensor of the magnetic susceptibility, which can be used to model the orientation of the ligand with respect to the protein.

Introduction 15 N-HSQC spectra of uniformly 15 N-labeled proteins play a unique role in pharmaceutical research by providing a two-dimensional fingerprint of the protein, where each amino acid residue, except proline, is represented by a peak from its backbone amide proton [1,2]. The cross peaks serve as sensitive reporters of ligand binding, as changes in the local chemical environment affect their chemical shifts and line widths, providing positive evidence of binding and information about the location of interacting surfaces. The binding of small molecules and interactions with macromolecular partners can both be assessed in this way. The method presents a tool for target validation (e.g. where do natural interaction partners bind?) but is even more attractive for the screening of chemical compound libraries (which compounds bind?) [3,4], since weak and strong binding can be detected equally well. Assignment of the spectral changes observed in 15 NHSQC spectra to specific binding sites (where does the compound bind?) requires the assignment of the cross

Graham A. Webb (ed.), Modern Magnetic Resonance, 1263–1269.
C 2008 Springer.

peaks to the individual amino-acid residues in the protein. Traditional ways of assigning the 15 N-HSQC spectrum of a protein of molecular weight above 15 kDa require the preparation of doubly- (15 N and 13 C) or triply labeled (15 N, 13 C, and 2 H) samples and the recording of three-dimensional NMR spectra. These experiments are much less sensitive and more time consuming than recording of a 15 N-HSQC spectrum. Although uniformly labeled protein samples can, in principle, readily be prepared by expressing the protein in cell cultures grown in isotope-labeled media, labeling with isotopes other than 15 N proves prohibitively expensive in many cases. Therefore, resonance assignments are restricted to proteins which can be expressed in high yields and which are sufficiently soluble to record the NMR spectra necessary for resonance assignment. As a result, NMR resonance assignments seem to be out of reach for many target proteins of pharmaceutical interest, even when crystal structures and 15 N-HSQC spectra may be available. The present article describes a different strategy for assigning 15 N-HSQC spectra of 15 N-labeled proteins. It requires prior knowledge of the three-dimensional protein structure and a lanthanide-binding site in the protein. The strategy has been demonstrated for the 30 kDa complex of the subunits ε (N-terminal segment with residues 2–186) and θ of Escherichia coli DNA polymerase III, where ε had been selectively labeled in three separate samples with 15 N-labeled Leu, Val, and Phe [5]. Resonance assignments are obtained from the comparison of two highly sensitive 15 N-HSQC spectra, recorded with and without a paramagnetic lanthanide ion bound to the complex.

ε186–θ Complex The crystal structure of the N-terminal domain of ε (ε186) has been solved by X-ray crystallography [6]. The same fragment has been studied by NMR spectroscopy, both free and in complex with θ, using doubly and triply

Part II

Fast Assignments of 15N-HSQC Spectra of Proteins by Paramagnetic Labeling

1264 Part II

Pharmaceutical Sciences

Part II

isotope-labeled samples [7,8]. Two Mn2+ ions are bound to the active site of the enzyme in the crystal structure. In analogy to the related domain of DNA polymerase I, these two ions can be replaced by a single lanthanide [9].

Paramagnetic Restraints Derived from 15 N-HSQC Spectra of Paramagnetic and Diamagnetic ε186–θ Complexes N-HSQC spectra of the complex between ε186 and θ with a Dy3+ ion bound (the paramagnetic complex) and of the complex without an ion bound or with La3+ bound (the diamagnetic complex) were recorded at a 1 H NMR frequency of 600 MHz. Spectra recorded in the presence or absence of La3+ were very similar; either spectrum provided an adequate diamagnetic reference for the measurement of paramagnetic restraints. The cross peaks recorded for the paramagnetic complex are strongly shifted with respect to those of the diamagnetic spectrum (Figure 1). The chemical shift difference between the paramagnetic and diamagnetic spectrum presents the pseudocontact shift (PCS). The PCS values, measured in ppm, are almost the same in the 15 N and 1 H dimension of the spectrum. Consequently, the paramagnetic signals are shifted along approximately parallel lines, enabling the pairwise association of diamagnetic and paramagnetic cross peaks. Ambiguities resulting from situations 15

Fig. 1. 15 N-HSQC spectra of 15 N-Leu labeled ε186 complexed to unlabeled θ. Two spectra are superimposed, one recorded with paramagnetic Dy3+ and the other with diamagnetic La3+ complexed to ε186. No decoupling was applied during the acquisition time. The cross peaks appear as antiphase doublets split by the one-bond 1 H–15 N coupling constant. Positive and negative contours are plotted, respectively, with blue and yellow lines for the spectrum of the paramagnetic complex and black and red lines for the spectrum of the diamagnetic complex. Straight lines connect pairs of diamagnetic and paramagnetic cross peaks. Dashed lines identify the two most strongly shifted paramagnetic cross peaks for which the diamagnetic partners could not be identified by simple visual inspection. The spectra were recorded with 0.5 mM solutions of the complex at 25◦ C and pH 7.0, using total recording times of 5 and 16 h for the diamagnetic and paramagnetic samples, respectively, on a Varian Inova 600 MHz NMR spectrometer. Reproduced with permission from c 2004 Am. Chem. Soc. Ref. [5].


of cross-peak overlap (Figure 1) can often be resolved by the assignment algorithm discussed below. The 15 N-HSQC spectra are best recorded without decoupling during the acquisition time [5], resulting in a shorter pulse sequence and hence, reduced loss of magnetization by relaxation. More importantly, analysis of the doublet fine structure provides access to paramagnetic parameters, which would not be observable in a decoupled spectrum (Figure 2). For example, differential line broadening of the doublet components reflects the cross-correlated relaxation (CCR) between dipole–dipole (with the amide proton and amide nitrogen as the dipolarcoupled spin system) and Curie relaxation of the paramagnetic center [10–12]. In addition, the coupling constant displayed by the doublet contains the residual dipolar coupling (RDC) which arises from the partial alignment of the paramagnetic molecule with the magnetic field [13,14]. Finally, the overall broadening of the doublet components reflects the paramagnetic relaxation enhancement (PRE) by the paramagnetic center. The PRE effect is a function of the distance from the Dy3+ ion [15,16]. In practice, ˚ from the cross peaks from amide protons closer than 14 A metal ion were broadened beyond detection. PCS, RDC, CCR, and PRE effects [17] present four paramagnetic parameters arising from different structural information (Figure 3). Of these four effects, PCS data of protons are most easily measured with high accuracy. Their dependence on the anisotropy of the magnetic

115 δ1(15N) ppm 120

125

130 13

11

9 δ2(1H)/ppm

7

15 N-HSQC Spectra of Proteins by Paramagnetic Labeling

PLATYPUS Algorithm for Resonance Assignments from Paramagnetic Restraints

∆v'

∆v''

δ2(1H)/ppm 7.5

7.4

7.3

7.2

Fig. 2. Cross section through a cross peak from the spectrum of Figure 1 with two Lorentzian lineshapes fitted to the doublet components (dashed lines). The fit provides the accurate chemical shift of the doublet, the line widths of the individual doublet components, and the coupling constant of the doublet. Reproc 2004 Am. Chem. Soc. duced with permission from Ref [5].


susceptibility tensor also provides stringent geometric restraints that are relatively insensitive with respect to local structural variations. In contrast, in our experience PRE effects yield the least accurate data, partly because line widths can vary between samples due to aggregation effects and chemical exchange phenomena [5]. The values of all four paramagnetic effects can be predicted provided

In our assignment strategy, the resonance assignment is obtained from a comparison of the experimentally measured paramagnetic data with the data predicted from the three-dimensional structure of the protein. An exhaustive evaluation of all n! possible assignments, however, is not viable when the number of cross peaks, n, is large. Consequently, a novel algorithm named PLATYPUS (paramagnetic labeling for assignment with terrific yields of proteins using their structures) was developed to derive the resonance assignments from the paramagnetic data automatically (Figure 5). Since the parameters defining the susceptibility anisotropy tensor are not known a priori, the algorithm employs a grid search with a large number of different tensors. A target function, defined as the sum of the squared deviations between experimentally measured and back-calculated paramagnetic data, is minimized to determine the best-fitting resonance assignment. Starting from each grid point, PLATYPUS uses the “Hungarian method for minimal cost assignment” [18] to arrive at resonance assignments that are continuously updated, while the tensor parameters and rotational correlation time are optimized by gradient minimization of the target function (Figure 5). In principle, the grid search over different possible susceptibility anisotropy tensors is five-dimensional, since five parameters are required to define the shape and orientation of the tensor. In practice, the algorithm

Fig. 3. Dependence of paramagnetic effects on local geometry. (a) Pseudocontact shifts (PCS) depend on the location of the nuclear spin with respect to the susceptibility anisotropy tensor. They result in chemical shift changes of the NMR signal. (b) Residual dipolar couplings (RDC) depend on the angle formed between the vector connecting the coupled spins and the axes of the alignment tensor resulting from alignment of the paramagnetic protein in the magnetic field. They are manifested in altered multiplet splittings. (c) Cross-correlated relaxation (CCR) between the Curie spin of the paramagnetic protein and the dipole–dipole interaction between two nuclear spins depends on the angle between the internuclear vector and the vector between the nuclear and electronic spins. This relaxation mechanism results in different line widths of the two doublet components. (d) Paramagnetic relaxation enhancements (PRE) depend on the distance from the paramagnetic center. The relaxation enhancement results in equal line broadening of both doublet components.

Part II

that the three-dimensional structure of the protein, the position of the metal ion, and the susceptibility anisotropy tensor are known (Figure 4) [17].

J+D

7.6

PLATYPUS Algorithm for Resonance Assignments 1265

1266 Part II

Pharmaceutical Sciences

Part II

Fig. 4. Iso-surfaces indicating pseudocontact shifts of ±1, ±0.5, and ±0.2 ppm, respectively, superimposed onto the crystal structure of ε186 (drawn in a ribbon representation). Blue and red colors identify positive and negative pseudocontact shifts, respectively. The N–H groups of Leu (blue), Phe (pink), and Val (orange) residues are displayed in a balland-stick representation. The figure was plotted with Molmol [39]. Reproduced c 2004 with permission from Ref. [5].
Am. Chem. Soc. (See also Plate 97 on page XIX in the Color Plate Section.)

converges even when starting from any plausible combination of axial and rhombic components of the tensor and using increments as large as 20◦ between different tensor orientations. PLATYPUS presents a very fast algorithm because the search space is limited to the optimization of the assignments initially suggested for each of the grid points. Its efficiency allows the extension of the search to test in a systematic way all conceivable pairings of diamagnetic and paramagnetic cross peaks, even in cases where the pairings are ambiguous due to cross-peak overlap (Figure 1). For those cross peaks, minimization of the target function yields not only the resonance assignment of the paramagnetic cross peaks but also the identification of their diamagnetic partners.

Results Obtained with Selectively Labeled ε186–θ Complexes ε186 contains 17 leucine residues. All their 15 N-HSQC cross peaks were observable in the diamagnetic ε186–θ complex containing selectively 15 N-Leu labeled ε186. In the presence of Dy3+ , however, the 15 N-HSQC cross peaks of only eight of them could be observed; the other leucine

residues were too close to the paramagnetic center and their resonances were broadened beyond detection. Despite the limited number of cross peaks which could be evaluated and ambiguities in pairing some of the paramagnetic cross peaks with diamagnetic partners (Figure 1), PLATYPUS yielded assignments in complete agreement with those determined previously in the conventional manner with triply isotope-labeled protein preparations [8]. This result is particularly remarkable, since the description of the shape and orientation of the susceptibility anisotropy tensor alone requires five independent variables, i.e. information from at least five different 15 N–1 H cross peaks. The assignment of 15 N-HSQC cross peaks broadened beyond detection in the presence of Dy3+ can be achieved by the use of a different, less strongly paramagnetic lanthanide ion. In the case of the ε186–θ complex, all 17 leucine residues yielded observable cross peaks in the presence of Ce3+ . Paramagnetic effects from Ce3+ are limited to the closer vicinity of the metal ion, providing complementary data to those obtained with Dy3+ [5]. Besides resonance assignments, PLATYPUS also provides a measure of the reliability of the individual assignments by checking for their resilience against variations of the three-dimensional structure. This is achieved by

15 N-HSQC Spectra of Proteins by Paramagnetic Labeling

Outlook 1267

(a)

(b)

(c)

(d)

Fig. 5. Pictorial representation of progressing assignments in PLATYPUS. Different orientations and shapes of the susceptibility anisotropy tensor are systematically varied in a grid search. The figure illustrates the procedure for the case of pseudocontact shifts only. For each grid point the paramagnetic effects are calculated and the back-calculated paramagnetic data sorted by magnitude (lower lines). Comparison with experimental values, also sorted by magnitude (top lines), generates an assignment (a). The tensor parameters are subsequently optimized by gradient minimization. Recalculation of the predicted paramagnetic effects with the new tensor parameters may require some reassignments to maintain the match of predicted versus experimental data when sorted by magnitude (b and c). The assignment with the lowest residual target function (the sum of the squared differences between experimental and back-calculated data) obtained from any of the starting grid points is presented as the correct assignment (d). The Hungarian method for minimal cost assignment [18] is used to perform the sorting and minimization in the m-dimensional space spanned by m different paramagnetic parameters.

the deliberate addition of structural noise to the coordinates of the protein structure, including variations in the orientations of chemical bonds and changes in the locations of protein atoms and of the paramagnetic ion, followed by repetition of the assignment procedure. As expected, availability of more than a single selectively 15 Nlabeled sample improves the reliability of the resonance assignment.

Two other protocols have been reported [19,20] which use a similar approach to determine NMR resonance assignments with the help of an available three-dimensional protein structure. These protocols back-calculate residual dipolar couplings and diamagnetic chemical shifts from the structure and use combinatorial optimization algorithms for the assignment. However, they require doubly 15 N/13 C-labeled samples and the time-consuming measurement of two or three residual dipolar couplings per amino acid residue. Paramagnetic data are much easier to evaluate from straightforward 15 N-HSQC spectra. Furthermore, PCSs are less sensitive than residual dipolar couplings to small differences between protein structures in solution and in single crystals. Finally, PLATYPUS yields the magnetic susceptibility anisotropy tensor simultaneously with the resonance assignment (Figure 4). With Dy3+ , sizable ˚ from the PCSs can be observed for distances up to 40 A metal ion [21]. This opens unique opportunities for the study of intermolecular complexes with ligands.

Outlook The assignment strategy outlined here provides access to resonance assignments with unprecedented speed. For comparison, 56 days of NMR spectrometer time were used to assign the resonances of ε186 in complex with θ[7,8]. Although our strategy provides resonance assignments of only a few 15 N-HSQC cross peaks for each selectively labeled sample, complete resonance assignments of all cross peaks observable for a uniformly 15 N-labeled protein are often not necessary in pharmaceutical applications. The main drawback of the method is the requirement of a lanthanide-binding site. Most proteins are diamagnetic and do not possess a specific lanthanide-binding site. In principle, lanthanide-complexing reagents have been reported which can be covalently attached to side-chain amine, carboxylate, or sulfhydryl groups on the protein surface [22,23]. The most generally applicable technique for site-specific attachment of a lanthanide-complexing reagent may be via formation of a disulfide bond with a single Cys residue on the protein surface. A suitable reagent needs to have a short linker [24] to avoid poorly defined positioning of the lanthanide ion with respect to the protein surface [25,26]. Furthermore, formation of different diastereomeric complexes must be suppressed which would otherwise increase the number of NMR signals [27–29]. Corresponding reagents are not yet commercially available. It has been demonstrated that fusion proteins can be designed with lanthanide-binding peptide motifs

Part II

Alternative Methods

1268 Part II

Pharmaceutical Sciences

Part II

[30–34]. Lanthanide-binding peptide motifs avoid the generation of different diastereomers, but restrict the choice of attachment sites to the terminal ends of the protein. Furthermore, many of the paramagnetic effects are reduced when the linker to the lanthanide-binding peptide is flexible. A paramagnetic tag attached to a selected site on the protein surface offers a powerful tool for the detection and structural characterization of ligand binding. The attachment of TEMPO spin labels has been shown to enhance the sensitivity of NMR methods for the screening of compound libraries by several orders of magnitude [35]. TEMPO spin labels have also been shown to provide longrange structural information in large biomolecules as well as access to the study of transient interactions in partially or completely unfolded proteins. The relaxation enhancement effect of TEMPO spin labels is, however, governed by a 1/r 6 distance dependence, whereas PCSs induced by lanthanide ions follow a 1/r 3 distance law, resulting in effects observable over longer distances. Furthermore, the isotropic nature of the magnetic susceptibility tensor in TEMPO prohibits the observation of the PCS, CCR, and RDC effects that contain the most valuable structural information (Figure 3). Therefore, more information can be obtained from site-specifically attached lanthanides. In addition, lanthanide ions could not only detect the presence of ligand binding, but also provide a detailed picture of the relative orientation of the ligand with respect to the magnetic susceptibility anisotropy tensor and hence, with respect to the protein. This information could be obtained with a minimal number of resonance assignments of the protein. The present version of PLATYPUS was designed for the assignment of selectively 15 N-labeled proteins. For some proteins, cell-free protein expression systems offer a cost-effecitve and fast route to protein samples with selectively labeled amino acids [36–38]. Selective labeling of proteins becomes expensive when the proteins cannot be expressed in high yields. For those situations, a modified strategy would be desirable, which provides reliable cross-peak assignments for uniformly 15 N-labeled proteins. Work towards this goal is in progress.

Acknowledgments Complete resonance assignments of ε186 or ε186 in complex with θ were not available at the time we published the 15 N-HSQC assignments of Leu, Val, and Phe labeled ε186 in complex with θ. We thank Drs. Eugene F. DeRose and Robert E. London for making their resonance assignments available to us afterwards to confirm the correctness of the assignments. G.P. thanks the EU for a postdoctoral fellowship within the Research Training Network on Cross Correlation (HPRN-CT-2000-00092). G.O. thanks

the Australian Research Council for a Federation Fellowship. Financial support by the Australian Research Council is gratefully acknowledged.

References 1. Pellecchia M, Sem DS, W¨uthrich K. Nat. Rev. Drug Discov. 2002;1:211. 2. van Dongen M, Weigelt J, Uppenberg J, Schultz J, Wikstr¨om M. Drug Discov. Today 2002;7:471. 3. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Science 1996;274:1531. 4. Hajduk PJ, Meadows RP, Fesik SW. Q. Rev. Biophys. 1999;32:211. 5. Pintacuda G, Keniry MA, Huber T, Park AY, Dixon NE, Otting G. J. Am. Chem. Soc. 2004;126:2963. 6. Hamdan S, Carr PD, Brown SE, Ollis DL, Dixon NE. Structure 2002;10:535. 7. DeRose EF, Li D, Darden T, Harvey S, Perrino FW, Schaaper RM, London RE. Biochemistry 2002;41:94. 8. DeRose EF, Darden T, Harvey S, Gabel S, Perrino FW, Schaaper RM, London RE. Biochemistry 2003;42:3635. 9. Brautigam CA, Aschheim K, Steitz TA. Chem. Biol. 1999; 6:901. 10. Boisbouvier J, Gans P, Blackledge M, Brutscher B, Marion D. J. Am. Chem. Soc. 1999;121:7700. 11. Pintacuda G, Hohenthanner K, Otting G, M¨uller N. J. Biomol. NMR 2003;27:115. 12. Ghose R, Prestegard JH. J. Magn. Reson. 1997;128:138. 13. Tolman JR, Flanagan JM, Kennedy MA, Prestegard JH. Proc. Natl. Acad. Sci. USA 1995;92:9279. 14. Barbieri R, Bertini I, Cavallaro G, Lee YM, Luchinat C, Rosato A. J. Am. Chem. Soc. 2002;124:5581. 15. Gu´eron M. J. Magn. Reson. 1975;19:58. 16. Vega AJ, Fiat D. Mol. Phys. 1976;31:347. 17. Bertini I, Luchinat C, Parigi G. Solution NMR of Paramagnetic Molecules: Applications to Metallobiomolecules and Models. Elsevier: Amsterdam, 2001. 18. Kuhn HW. Nav. Res. Logist. Q. 1955;2:83. 19. Hus JC, Prompers JJ, Br¨uschweiler R. J. Magn. Reson. 2002;157:119. 20. Zweckstetter M. J. Biomol. NMR 2003;27:41. 21. Allegrozzi M, Bertini I, Janik MBL, Lee YM, Lin GH, Luchinat C. J. Am. Chem. Soc. 2000;122:4154. 22. Liu S, Edwards DS. Bioconjugate Chem. 2001;12:7–34. 23. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB. Chem. Rev. 1999;99:2293. 24. Selvin PR. Annu. Rev. Biophys. Biomol. Struct. 2002;31: 275. 25. Voss J, Wu J, Hubbell WL, Jacques V, Meares CF, Kaback HR. Biochemistry. 2001;40:3184. 26. Purdy MD, Ge P, Chen J, Selvin PR, Wiener MC. Acta Crystallogr. D. 2002;58:1111. 27. Dvoretsky A, Gaponenko V, Rosevear PR. FEBS Lett. 2002;528:189. 28. Ikegami T, Verdier L, Sakhaii P, Grimme S, Pescatore B, Saxena K, Fiebig KM, Griesinger C. J. Biomol. NMR, 2004;29:339.

15 N-HSQC Spectra of Proteins by Paramagnetic Labeling

35. Jahnke W, Rudisser S, Zurini M. J. Am. Chem. Soc. 2001; 123:3149. 36. Kigawa T, Muto Y, Yokoyama S. J. Biomol. NMR 1995;6: 129. 37. Guignard L, Ozawa K, Pursglove SE, Otting G, Dixon NE. FEBS Lett. 2002;524:159. 38. Yabuki T, Kigawa T, Dohmae N, Takio K, Terada T, Ito Y, Laue ED, Cooper JA, Kainosho M, Yokoyama S. J. Biomol. NMR 1998;11:295. 39. Koradi R, Billeter M, W¨uthrich K. J. Mol. Graph. 1996;14: 51.

Part II

29. Pintacuda G, Moshref A, Leonchiks A, Sharipo A, Otting G. J. Biomol. NMR, 2004;29:351. 30. Ma C, Opella SJ. J. Magn. Reson. 2000;146:381. 31. Biekofsky RR, Muskett FW, Schmidt JM, Martin SR, Browne JP, Bayley PM, Feeney J. FEBS Lett. 1999;460:519. 32. Feeny J, Birdsall B, Bradbury AF, Biekofsky RR, Bayley PM. J. Biomol. NMR 2001;21:41. 33. Wohnert J, Franz KJ, Nitz M, Imperiali B, Schwalbe H. J. Am. Chem. Soc. 2003;125:13338. 34. Caravan P, Greenwood JM, Welch JT, Franklin SJ., Chem. Commun. 2003;20:2574.

References 1269

1271

Jianxin Guo1 , Xiaoyu Tian1 , Spiro Pavlopoulos2 , and Alexandros Makriyannis1 1 Center

for Drug Discovery, Bouve College of Health Sciences, Northeastern University, 360 Huntington Avenue, Boston, MA 02115, USA; and 2 Department of Pharmaceutical Sciences, University of Connecticut, U-3092, 69 North Eagleville Road, CT 06269, USA

Abbreviations: NMR, nuclear magnetic resonance; DSC, differential scanning calorimetry; DPPC, 1,2dipalmitoyl-sn-glycero-3-phosphocholine; DMPC, 1,2dimyristoyl-sn-glycero-3-phosphocholine; DHPC, 1,2dihexanoic-sn-glycero-3-phosphocholine; CHAPSO, 3(cholamidopropyl) dimethylammonio-2-hydroxy-1propanesulfonate; SDS, sodium dodecyl sulfate; DPC, dodecylphosphocholine; 8 -THC, (−)-8 -tetrahydrocannabinol; Me-8 -THC, (−)-O-methyl-8 -tetrahydrocannabinol

Introduction The design of drugs that modulate membrane bound receptors has always been of prime importance to the pharmaceutical industry. It is increasingly being recognized that the cellular membrane plays an important role in drug action and that understanding the manner in which drug molecules interact with lipid bilayers can enhance our abilities to design and develop improved medications. It has been proposed that lipophilic drugs reach their sites of action through cell membrane penetration and lateral diffusion within the membrane leaflet [1–4]. The protein supporting lipid membranes may be capable of a high degree of structural discrimination by positioning the ligand in a proper location, orientation, and active conformation for a productive interaction with the receptor. Therefore, acquiring detailed knowledge of drug and ligand physical properties that underlie their interaction profiles with the cellular membrane is of great interest in the discovery, design, and delivery of novel therapeutic agents. Information related to drug–membrane interactions, however, can rarely be derived directly from experiments with complex and labile natural membranes. Thus, various membrane models with different levels of complexity have been developed in order to exploit all the biophysical techniques available [5–7]. Whether it is in the form of liposomes, micelles, or other membrane mimetics, the interplay between the ligand and membranes is central to many aspects in pharmaceutical research. Graham A. Webb (ed.), Modern Magnetic Resonance, 1271–1277.
C 2008 Springer.

Liposomes (in most cases, multilamellar vesicles) have been extensively utilized to study the manner in which drug molecules perturb cellular membranes as well as the location, conformation, and preferred orientation of a drug molecule within the membrane bilayer using differential scanning calorimetry (DSC), nuclear magnetic resonance (NMR) spectroscopy, and small angle Xray diffraction [3,8–15]. Among these techniques, NMR spectroscopy holds a prominent place by being particularly sensitive to the fine structural and dynamic details of drug molecules within the bilayer system. However, liposomes are not suitable model membrane systems for exploring the conformational properties of embedded ligands since even the smallest lipid vesicles are too large to yield a quality high-resolution NMR spectrum [16]. Alternatively, fast tumbling micelles consisting of various detergents have been widely used to mimic the membrane environment in high-resolution NMR conformational analysis [5,17–19], even though it has long been a concern that micelles may not accurately represent cellular membrane bilayers [20,21]. Recently, phospholipid bilayered-micelles (often referred to as bicelles) have received much attention as promising membrane-mimetic systems [22–27]. Work in our laboratory has demonstrated that bicelle systems can serve as excellent bilayer membrane models for studying drug–membrane interactions using high-resolution NMR. We have developed an isotropic bicelle system for studying the conformational properties of small- and medium-sized ligands without the need of isotopic labeling [21]. We also used magnetically oriented bicelles to determine the preferred orientations of several lipophilic drugs from deuterium NMR experiments [28]. In this chapter, we briefly review the commonly used model membrane systems for studying drug–membrane interactions using NMR with an emphasis on phospholipid bicelles. To highlight the advantages of using bicelle systems for such work, we have sought to determine the conformational properties as well as the preferred orientation of two lipophilic ligands in different membranemimetic systems. A comparison of the ligand spectra in

Part II

Phospholipid Bicelle Membrane Systems for Studying Drug Molecules

1272 Part II

Pharmaceutical Science

Part II

the different media clearly demonstrates the value of phospholipid bicelles in such studies.

Membrane Systems for NMR Studies Cell membranes consist of a variety of glycerolphospholipids, sphingolipids, and cholesterol with local heterogeneities and segregated microdomains. Although much work has been dedicated to understand their structural and dynamic properties, many questions still remain unanswered [29]. Opportunities for studying the drug–membrane interactions with native membranes are presently confined to the use of solid-state NMR methods. Only few published examples can be found of the interactions between small- or medium-sized molecules with native membranes where the ligand is isotopically labeled in specific positions for effective NMR observation [30,31]. Simplified model membranes composed of certain synthetic or native lipids have become the most commonly used systems for such work. It can be argued that notwithstanding their great functional diversity, membranes share the common features of an anisotropic bilayer system consisting of a hydrophobic interior and a hydrophilic headgroup region. This fundamental structure defines many of the key structural and functional properties of biological membranes. For this reason, it is possible to use simplified lipid systems as legitimate mimics of the cellular membrane.

Liposomes Liposomes are the most used and best-characterized model membrane system. They can be produced in a variety of sizes and lipid compositions including multilamellar vesicles and small or large unilamellar vesicles [32,33]. Liposomes are generally prepared from glycerolphospholipids, mainly phosphatidycholines such as DPPC, DMPC, and POPC, while cholesterol and charged lipids can also be added. Because multilamellar vesicles are relatively easy to prepare and multilayering makes it possible to incorporate larger amounts of drug, they have been extensively employed for studying drug–membrane interactions using various solid-state NMR techniques, such as stationary 2 H, 31 P, and 13 C NMR as well as the cross polarization magic angle spinning (CP/MAS) experiments [11,34,35]. Liposomes in the L α liquid crystalline phase have been used to determine the preferred orientations of native and xenobiotic molecules and the effects of these ligands on the membrane system [8,9,11,12,14,36– 39]. Liposomes are usually not suitable for studying the conformation of a ligand using solution NMR because of the relatively slow tumbling rates resulting from their large size. For such applications, other model membrane systems consisting of much smaller aggregates must be used.

Organic Solvent Mixtures The simplest membrane-mimetic environment resembling the membrane interior and interfacial regions can be obtained by carefully mixing two organic solvents of low and intermediate polarity [40]. However, such systems are overly simplified membrane models and are of limited value.

Micelles The most commonly used spherical micelle preparations can be obtained from SDS, DPC, or DHPC. Because of effective motional averaging, the small and uniformly sized micelles can yield satisfactory high-resolution NMR spectra of an enclosed ligand. Micelles have been extensively utilized as membrane-mimetic media for studying the conformational properties of small molecules and peptides [18,19]. However, as micelles do not consist of natural cell membrane lipids and have a highly curved surface, their value as a reliable model membrane system is somewhat reduced [21,26].

Phospholipid Bicelles An optimal membrane-mimetic system for solution NMR experiments should consist of a relatively smaller lipid aggregate with a bilayer structure. With a short correlation time, such a membrane preparation enables more detailed structural studies of drugs in a bilayer environment. In order to form such a lipid aggregate, the mean chemical potential µ0N must initially decrease and reach its minimum at a certain aggregation number N . If there were no free energy contributions, such small lipid aggregates would be entropically favored. However, because of the large cohesive binding energies of lipid molecules, discrete planar bilayer systems have never been observed when only one type of lipids is present. For a hypothetical planar bilayer disc, the free energy for molecules sequestered along the rim is expected to be much higher than that of molecules in the planar center. The mean free energy per molecule in such a planar bilayer disc is [41]: µ0N = µ0∞ + αkT /N 1/2 where µ0∞ is the mean free energy in an aggregate of infinite size, k is the Boltzmann constant and T is the temperature. As µ0N decreases monotonically with N , the expectation is that an infinite number of lipid molecules will aggregate spontaneously at critical micelle concentrations while single component lipid bilayers with exposed edges are not favored. However, discrete planar bilayer disc assemblies can be obtained if detergents are added to stabilize the energetically unfavorable edges [23,26]. Such systems,

Phospholipid Bicelle Membrane Systems

Isotropic Bicelles

Fig. 1. Isotropic (left) and magnetically oriented bicelles (right).

Magnetically Orientated Bicelles

commonly referred to as phospholipid bicelles, are typically composed of DMPC with either DHPC or a bile-salt derivative CHAPSO in 0.1 M KCl solution [26]. Above the phase transition temperature (Tm ), the DMPC-rich domain resembles the liquid crystalline L α phase multilamellar membrane bilayers [22,42]. Phospholipid bicelles combine the attractive structural and dynamic properties of both micelles and lipid vesicles and maintain most of the key bilayer properties that are absent in the micellar systems. Additionally, compared to the liposome preparations, bicelles are relatively small, noncompartmentalized, effectively monodisperse and ideally suited for highresolution NMR studies. Phospholipid bicelles (Figure 1) are a versatile model membrane system whose morphology and dynamic properties can be adjusted over wide ranges of lipid to detergent ratio (q), total lipid concentration, buffer pH, and temperature [26]. With appropriate manipulation, bicelles can provide isotropic membrane systems suitable for studying the conformational properties of embedded ligands using high-resolution NMR [21,43]. Alternatively, bicelles can serve as anisotropic membrane preparations capable of orienting in an applied magnetic field [23,25].

Optimizing Isotropic Bicelles for Drug Conformational Studies At a DMPC to DHPC molar ratio (q) of 0.5 and an overall lipid concentration of 15% (w/v), the bicelle system has been shown to be at the isotropic state and the bicelles are small enough for high-resolution experiments NMR [43]. This is promising in that without isotopic labeling,

such a preparation can be used to explore the conformational properties of incorporated drugs. However, such bicelles may not serve as an ideal membrane environment for studying our lipophilic drugs. As the diame˚ ter of the bicelle is estimated to be approximately 80 A [44], the embedded drugs may have much higher chances of interacting with the stabilizing detergents rather than with the lipids. In addition, small bicelles are incapable of incorporating adequate quantities of lipophilic drugs and precipitation has been frequently observed in our research. For these reasons, we have sought to develop larger isotropic bicelles that would better represent cellular membrane bilayer systems. Our approach to maximizing bicelle size while maintaining isotropic conditions was to systematically adjust the lipid to detergent ratio (q) as well as the overall lipid concentration. Additionally, preparations free of KCl allow us to use higher values of q and the stability of bicelles can be significantly improved by adding trace amount of amphiphiles. We found that a preparation of a much higher DMPC/DHPC ratio (q = 2.0) and an overall lipid concentration of 8% (w/v) is optimal for not only allowing effective drug incorporation, but also providing an isotropic system suitable for high-resolution NMR studies. In such a preparation, no precipitation of either the drug or lipid was observed over the course of several days. Well-resolved proton resonances reflect the isotropic nature of the optimized bicelle membrane media within the NMR time scale. This can be attributed to the ample spacing between the individual bicelles, which allow for faster motional averaging [25]. Although a certain degree of broadening in the 1 H resonances is still observed, the spectral resolution is adequate for standard 2D-NMR experiments such as DQF-COSY

Part II

B0

Phospholipid Bicelle Membrane Systems 1273

1274 Part II

Pharmaceutical Science

Part II

9

9

8

10

7

10a

OH 2

6a

4

10

7

10a

O 2

D

6a 2'

O 4a

8

3'

1'

2'

4' O 4a

5'

4

1'

4' 3'

5'

D Fig. 2. The structure of 8 -THC (left) and Me-8 -THC (right).

and NOESY from which ligand conformational properties can be obtained. To evaluate the experimental benefits of our isotropic bicelle preparation, we compared the 1 H spectra of 8 THC (Figure 2) in chloroform, SDS micelles, and two different isotropic bicelle preparations with q = 0.5 and q = 2.0, respectively. Figure 3 shows the NOE patterns

A

observed between the aromatic ring protons and protons of the pentyl tail. In chloroform, only two NOEs were observed between the H2, H4 aromatic protons, and the pentyl tail H1 protons. Additional NOEs were observed between the aromatic H4 resonance and resonances due to protons beyond H1 of the pentyl sidechain with their relative intensities varying depending on the membrane

ppm

B

ppm 6.0

6.0

6.5

6.5 2.5

2.0

1.5

1.0

ppm

C

2.5

2.0

1.5

1.0

ppm

2.5

2.0

1.5

1.0

ppm

D ppm

ppm

6.0

6.0

6.5

6.5 2.5

2.0

1.5

1.0

ppm

Fig. 3. Expansions of NOESY spectra showing NOEs between the aromatic ring and the pentyl-tail of 8 -THC in (A) CDCl3 ; (B) SDS micelles; (C) q = 0.5 bicelle, and (D) q = 2.0 bicelle.

Phospholipid Bicelle Membrane Systems

by: U = −N χ B02 where N is the number of long acyl chain DMPC lipids within one bicelle disc, and B0 is the external magnetic field, when χ is less than zero, the energically favorable orientation of a bicelle aggregate will be with its principal axis perpendicular to the applied magnetic field. In earlier work, we have described how the preferred orientation of a drug molecule within the membrane bilayer system can be obtained from the quadrupolar splittings (ν Q ) of 2 H-labels introduced in specific sites of the molecule [9,11]. We have, thus, sought to explore whether such data could be obtained in a magnetically oriented bicellar system. Figure 4 compares three deuterium spectra of 2,4-D2 -Me-8 -THC (Figure 2) obtained in DHPC micelles, DPPC multilamellar vesicles, and magnetically oriented DMPC/DHPC bicelle preparations, respectively. The spectrum acquired in a DHPC micelle solution displays a single isotropic deuterium peak, while that from the DPPC multilamellar vesicles corresponds to two pairs of overlapping Pake patterns from which the quadrupolar splittings can be extracted. Conversely, in magnetically oriented bicelles, the spectrum is reduced to two pairs of doublets representing the two and four deuterium atoms in Me-8 -THC. The sharp deuterium resonances indicate that the bicelles are completely aligned in the magnetic field. Due to the fact that the bicelles with the incorporated drug molecules are magnetically oriented with their

Magnetically Aligned Bicelles for Studying Drug Orientation The formation of magnetically oriented phospholipid bicelles under various conditions and the corresponding properties of the bicelles have been reviewed [22,45]. By using a series of titration studies, it has been found that the liquid crystalline L α DMPC/DHPC bicelles with q ranging from 2.0 to 3.5 are well aligned in the applied magnetic field as indicated by their 31 P and 2 H NMR spectra, while addition of KCl can drastically improve the degree of orientation. The orientation may be explained by the anisotropy of the diamagnetic susceptibility tensor χ of the lipid molecule. If we assume an axis of rotational symmetry, the diamagnetic anisotropy χ will be the difference between the susceptibilities parallel and perpendicular to this axis. For the bicelle forming long acyl chain lipid DMPC in the L α phase, χ was found to be (−1.0 ± 0.1) × 108 erg cm−3 G−2 [46]. Since the magnetic potential energy U can be expressed

10

0

-10

kHz 

Fig. 4. A comparison of deuterium spectra of 10% 2,4-D2 -Me8 -THC in DHPC micelles (top), DPPC multilamellar bilayers (middle), and magnetically oriented DMPC/DHPC bicelles (bottom).

Part II

preparations used. In SDS micelles, the NOESY spectrum from 8 -THC is comparable to that observed in chloroform where strong NOEs were observed between the H2, H4 aromatic protons, and the H1 protons in the pentyl group. Two weak NOEs were also observed between H4 and H2 , and H3 , H4 overlapping protons, suggesting a conformational preference for orientation of the pentyl chain towards the H4 side of the tricyclic ring system. In the q = 0.5 bicelle solution, the H4– H2 and H4–H3 /H4 NOEs are similar in intensity to the H4–H1 , H2–H1 NOEs; while in the q = 2.0 solutions, the former set of NOEs are significantly stronger than the latter. This observation shows that higher proportions of DMPC within the bicelle lead to an increased preference for an “amphipathic” conformation of the pentyl sidechain towards the H4 side of the tricyclic ring system. The above results indicate that the choice of model membrane systems can greatly influence the structural and dynamic properties of the incorporated drugs. It is also worth noting that in both the SDS micelle and the q = 0.5 bicelle preparations, 8 -THC was observed to precipitate over 3–6 h whereas it was well solubilized in the q = 2.0 bicelle solution. Our results obtained from the q = 2.0 bicelle preparation are congruent with previous studies using DPPC multilamellar membrane bilayers [11,14,15], which we attribute to the more bilayer-like morphology of the q = 2.0 lipid bicelles compared to the SDS micelle and q = 0.5 bicelle solutions. Therefore, the q = 2.0 bicelles represent a satisfactory membrane bilayer system suitable for high-resolution NMR conformational studies.

Phospholipid Bicelle Membrane Systems 1275

1276 Part II

Pharmaceutical Science

Part II

bilayer normal perpendicular to the external magnetic field, the observed splittings in the 2 H spectrum correspond to the 90◦ edges of the respective Pake patterns from the semi-solid hydrated DPPC sample. However, the ν Q values are much reduced in the bicelle system, reflecting a lower order parameter (Smol ) when compared to that from the multilamellar preparations. The calculated orientation of Me-8 -THC (to be reported elsewhere) closely matches our result from the multilamellar vesicles, thus attesting to the reliability of the bicelle system as a model membrane. In contrast to micelle preparations where quadrupolar couplings are completely averaged out due to fast isotropic tumbling, magnetically oriented bicelles can serve as an alternative to multilamellar membrane preparations for studying the orientation of a ligand in an anisotropic membrane system. In addition, the experiment described above was carried out under solution NMR conditions and offers the advantages of higher sensitivity and more accurate direct measurement of the ν Q values from the sharp 2 H doublets.

Conclusion Bicelles can serve as satisfactory membrane systems for isotropic and anisotropic NMR experiments aimed at studying the conformational properties and preferred orientation of lipophilic drug molecules in the cell membrane. Using the lipophilic 8 -THC as an example, we have demonstrated that high-resolution NOESY spectra of this ligand in optimized isotropic bicelles can be obtained without the need for isotopic labeling. Our results clearly showed that isotropic bicelle systems are better membrane bilayer models than the commonly used micelles. Arguably, the specific “amphipathic” conformation of 8 -THC obtained in the optimized bicelles may be necessary for a productive interaction with its receptor. We also demonstrated that the preferred orientation of the embedded cannabinoids can be determined by adjusting bicelle conditions to produce larger bicelle aggregates capable of aligning in the magnetic field. Phospholipid bicelles offer many opportunities for use as successful membrane mimetics. The preparation may be further improved by using lipids with certain negative charges and some degree of unsaturation, and by incorporating appropriate amounts of cholesterol to more closely resemble native cell membranes. A most recent application of bicelles is for determining the relative stereochemistry of multiple stereocenters [47]. Also, crystallization of membrane proteins has been achieved from bicelleforming lipid/detergent mixtures [48], an experiment which may be extended to study the drug/membrane– protein complex.

References 1. Herbette LG, Chester DW, Rhodes DG. Biophys. J. 1986;49: 91–3. 2. Mason RP, Rhodes DG, Herbette LG. J. Med. Chem. 1991;34: 869–77. 3. Makriyannis A. In: RG Pertwee (Ed). Cannabinoid Receptors. Academic Press: London, UK, 1995, pp 87– 115. 4. Xie XQ, Melvin LS, Makriyannis A. J. Biol. Chem. 1996;271: 10640–7. 5. Henry GD, Sykes BD. Bull. Can. Biochem. Soc. 1987;24:21– 6. 6. Scrimin P, Tecilla P. Curr. Opin. Chem. Biol. 1999;3:730–5. 7. Milhaud J. Biochem. Biophys. Acta 2004;1663:19–51. 8. Makriyannis A, Banijamali A, Van Der Schyf C, Jarrell H. NIDA Res. Monogr. 1987;79:123–133. 9. Makriyannis A, Banijamali A, Jarrell HC, Yang DP. Biochem. Biophys. Acta 1989;986:141–5. 10. Mavromoustakos T, Yang DP, Charalambous A, Herbette LG, Makriyannis A. Biochim. Biophys. Acta 1990;1024:336– 44. 11. Yang DP, Banijamali A, Charalambous A, Marciniak G, Makriyannis A. Pharmacol. Biochem. Behav. 1991;40:553– 7. 12. Makriyannis A, Yang DP, Mavromoustakos T. NIDA Res. Monogr. 1991;112:106–28. 13. Yang DP, Mavromoustakos T, Makriyannis A. Life Sci. 1993;53:117–22. 14. Makriyannis A, Yang DP. In: BN Dhawan, RC Sirmal, R Raghubir, RS Rapaka (Ed). Recent Advances in the Study of Neurotransmitter Receptors. Central Drug Research Institute: Lucknow, India, 1994, pp 329–48. 15. Mavromoustakos T, Yang DP, Makriyannis A. Biochim. Biophys. Acta 1995;1237:183–8. 16. Henry GD, Sykes BD. Methods Enzymol 1994;239:515–35. 17. Opella SJ, Kim Y, McDonnell P. Meth Enzymol 1994;239:536–60. 18. Mavromoustakos T, Theodoropoulou E, Yang DP, Lin SY, Koufaki M, Makriyannis A. Chem. Phys. Lipids 1996;84:21– 34. 19. Naito A, Nishimura K. Curr. Top. Med. Chem. 2004;4:135– 45. 20. Sanders CR, Oxenoid K. Biochem. Biophys. Acta 2000;1508: 129–45. 21. Guo J, Pavlopoulos S, Tian X, Lu D, Nikas SP, Yang DP, Makriyannis A. J. Med. Chem. 2003;46:4838–46. 22. Sanders CR, Schwonek JP. Biochemistry 1992;31:8898– 905. 23. Sanders CR, Hare BJ, Howard KP, Prestegard JH. Prog. Nucl. Magn. Reson. Spectrosc. 1994;26:421–44. 24. Howard KP, Prestegard JH. J. Am. Chem. Soc. 1996;118: 3345–53. 25. Tjandra N, Bax A. Science 1997;278:1111–4. 26. Sanders CR, Prosser RS. Structure 1998;6:1227–34. 27. Glover KJ, Whiles JA, Wu G, Yu N-J, Deems R, Struppe JO, Stark RE, Komives EA, Vold RR. Biophys. J. 2001;81:2163– 71. 28. Guo J. Ph.D. Thesis. University of Connecticut, Storrs, 2004.

Phospholipid Bicelle Membrane Systems

39. Yamaguchi S, Huster D, Waring A, Lehrer RI, Kearney W, Tack BF, Hong M. Biophys. J. 2001;81:2203–14. 40. Girvin ME, Rastogi VK, Abildgaard F, Markley JL, Fillingame RH. Biochemistry 1998;37:8817–24. 41. Israelachvilli JN, Mitchell DJ, Ninham BW. J. Chem. Soc. Faraday Trans. II 1976;72:1525–68. 42. Sanders C. Biophys. J. 1993;64:171–81. 43. Vold RR, Prosser RS, Deese AJ. J. Biomol. NMR 1997;9:329– 35. 44. Vold RR, Prosser RS. J. Magn. Reson. B 1996;113:267–71. 45. Sternin E, Nizza D, Gawrisch K. Langmuir 2001;17:2610–6. 46. Scholz F, Boroske E, Helfrich W. Biophys. J. 1984;45:589– 92. 47. Yan J, Delaglio F, Kaerner A, Kline AD, Mo H, Shapiro MJ, Smitka TA, Stephenson GA, Zartler ER. J. Am. Chem. Soc. 2004;126:5008–17. 48. Faham S, Bowie JU. J. Mol. Biol. 2002;316:1–6.

Part II

29. Dowhan W, Bogdanov M. In: DE Vance, JE Vance (Ed). Biochemistry of Lipids, Lipoproteins and Membranes. Elsevier Press: Amsterdam, 2002, pp 1–35. 30. Watts A, Ulrich AS, Middleton DA. Mol. Membr. Biol. 1995;12:233–46. 31. Kisselev OG, Kao J, Ponder JW, Fann YC, Gautam N, Marshall GR. Proc. Natl. Acad. Sci. USA 1998;95:4270–5. 32. Bloom M, Evans E, Mouritsen OG. Q. Rev. Biophys. 1991;24:293–397. 33. Bangham AD. Chem. Phys. Lipids 1993;64:275–85. 34. Tang P, Yan B, Xu Y. Biophys. J 1997;72:1676–82. 35. Gutierrez ME, Garcia AF, Africa de Madariaga M, Sagrista ML, Casado FJ, Mora M. Life Sci. 2003;72:2337– 60. 36. Davis J. Biochim. Biophys. Acta 1983;737:117–71. 37. Makriyannis A, Rapaka RS. Life Sci. 1990;47:2173–84. 38. Bechinger B. Biochem. Biophys. Acta 1999;1462:157–83.

References 1277

1279

Burkhard Luy and Horst Kessler TU M¨unchen, Department Chemie, Lehrstuhl f¨ur Organische Chemie II, Lichtenbergstrasse 4, D-85747 Garching, Germany

Introduction NMR spectroscopy is the most important method for structure elucidation in solution. Once the atomic connectivity is established (constitution of the molecule) the proof of stereochemistry is the next step. NMR as an achiral method provides only relative stereochemical information unless chiral environments (solvents, interaction with shift reagents, chemical modification via chiral auxiliaries) are used. For determining the relative stereochemistry, the classical parameters are NOE effects and J coupling constants. Only recently, two different new sources of stereochemical information have been introduced for biomolecules: cross-correlated relaxation [1] and residual dipolar couplings (RDCs) [2,3]. We will here consider the application of RDCs for small molecules such as drugs or drug-like molecules.

Residual Dipolar Couplings Magnetically active nuclei in molecules interact through space, as they are magnetic dipoles. These couplings, which depend on the relative orientation of the dipoles and their distance, are on the order of several kHz and cause immense numbers of splittings in solid-state NMR. In solution, the rapid isotropic molecular tumbling averages out all dipolar couplings, resulting in the narrow line width of high-resolution NMR spectroscopy. With the removal of all anisotropic interactions, however, there is also a significant loss in potential structural information. Therefore, an intermediate state between solid and liquid is desirable in which the narrow line width is maintained, but anisotropic interactions are re-established and measurable. In general, there are two ways to reach this intermediate state: if the sample of interest is available in a microcrystalline phase, high-speed magic angle spinning can effectively reduce the dipolar interactions to give resolved resonances similar to the liquid state; in contrast, a sample in solution can be transferred into a directional molecular lattice, which then partially aligns the molecule to have measurable anisotropic interactions. For example, if the molecules are oriented to a low percentage, let us say on the order of 0.1%, the dipolar coupling is also reduced Graham A. Webb (ed.), Modern Magnetic Resonance, 1279–1285.
C 2008 Springer.

to ∼0.1% of its original size. A molecule containing a CH fragment, which would normally exhibit a 1 JCH coupling of about 130 Hz in isotropic solution will give rise to an additional RDC that adds or subtracts to this coupling depending on the tiny preferred orientation. If one measures the 1 JCH coupling constant with and without partial orientation, the difference directly yields the RDC.

The Alignment Tensor In the following section, a very brief and handwavy introduction to the alignment tensors is given and includes only basic equations. For a more detailed and exact derivation of key equations, we refer to [4]. In the static case, the dipolar coupling constant is given by [5]   κ 1 D = 3 cos2 θ − , (1) R 3 where θ is the angle between an internuclear vector and the magnetic field B0 . The term κ=−

3 γ I γ S µ0h¯ 8π 2

(2)

only depends on physical constants [5]: the gyromagnetic ratios γI and γS of the two vector-spanning spins I and S, respectively, the Planck constant h¯ = h/2π and the permeability of vacuum µ0 . For example, for 1 H–1 H and 13 ˚ 3 , reC–1 H spin pairs, κ = −360.3 and −90.6 kHz A spectively. However, in the case of a tumbling molecule, the dipolar coupling is further averaged in the way that   κ 1 D¯ = 3 cos2 θ − (3) R 3 represents the so-called RDC constant between the two spins, which depends on the average alignment of the molecule. In the case of isotropic tumbling as in conventional liquid-state NMR, cos2 θ = 1/3 in all three dimensions

Part II

Partial Alignment for Structure Determination of Organic Molecules

1280 Part II

Pharmaceutical Sciences

Part II

and the dipolar interaction is quenched. In a partially oriented molecule instead, the term cos2 θ generally is a three-dimensional tensor. This tensor is equivalent to the probability tensor P to find the axis of the magnetic field B0 (t) along the internuclear vector r expressed in a reference frame, which is fixed to the molecule. This (3 × 3) probability tensor is symmetric with respect to the diagonal and the diagonal elements must fulfill the condition Px x + Pyy + Pzz = 1, which implicitly leads to five independent tensor components. In reverse, this implies that in an arbitrary case a minimum of five non-identical RDCs have to be measured to unambiguously identify a probability tensor of a given sample. Although the concept of the probability tensor P is very intuitive and sufficient to completely describe RDCs in rigid molecules, it is more common in the NMR literature to express the partial orientation of a molecule by the so-called alignment tensor A. This tensor is related to the probability tensor P via 1 A=P− 1 3

(4)

and is therefore proportional to the anisotropy of the magnetic field distribution in the molecular frame. The eigenvalues of A are usually denoted as A x x , A yy , and A zz and fulfill the condition A x x + A yy + A zz = 0. For further definitions like the axial and rhombic components of an alignment tensor of the Saupe matrix, we refer to the appendix of [4].

Alignment Media So far, we have not discussed how to partially orient the molecules. When a molecule has considerable magnetic susceptibility anisotropy, the molecule orients partially under the strong magnetic field. But this effect—although already observed in the eighties [6]—is normally too small to be observed. On the other hand, orientation can also be induced by an oriented molecular lattice, which then partially aligns the molecule of interest via steric or electrostatic interactions. All known alignment media can basically be assigned to one of two classes: liquid crystalline phases and stretched polymer gels.

Liquid Crystalline Phases In 1963, the first spectrum of benzene partially oriented in the nematic mesophase of 4,4 -di-nhexyloxyazoxybenzene was reported, a spectrum with at least 30 resonances and a width of the multiplet pattern of approximately 2500 Hz [7]. Following this result, a large number of liquid crystals for partially aligning small molecules was found in the sixties and seventies (see,

for example, reviews [8–10]), but it turned out that the orientation introduced by liquid crystalline phases generally is very strong, yielding numerous large splittings by RDCs that can hardly be interpreted for more complex organic molecules. Poly-γ-benzyl-l-glutamate, dissolved in solvents like dichloromethane, chloroform, or dimethylformamide, was introduced by Panar and Phillips [11]. It was one of the least orienting liquid crystals known at that time and was used for the first successful measurements of RDCs in an organic solvent to obtain structural information of a small molecule [12–14]. Currently, new liquid crystalline phases for organic solvents have been developed that can achieve lower degrees of anisotropies. 4-n-Pentyl-4 -cyanobiphenyl [15] and polyγ-ethyl-l-glutamate [16] seem to be two promising candidates for the measurement of RDCs in small molecules. The existence of liquid crystalline phases with very low induced anisotropies was proven in the last 7 years in the field of biomolecular NMR. Several lipid/detergent mixtures [2,3], filamentous phage [17], and other liquid crystalline phases [18,19] were used to successfully measure RDCs for structure refinement of proteins and nucleic acids. These alignment media are, of course, also applicable to small molecules in aqueous solutions [20–26].

Stretched Polymer Gels Deloche and Samulski showed in their pioneering work in 1981 that partial alignment can also be achieved by stretching polymer gels [27], usually mechanically. This technique now is a standard approach in polymer NMR to obtain information about polymer properties. Not until 2000 was the use of these stretched polymer gels for aligning molecules dissolved inside the gel to obtain RDCs for structural investigations realized. The new approach is called “strain-induced alignment in a gel” and is demonstrated on stretched polyacrylamide [28–31] and polyacrylamide/acrylate [32] copolymers. Several ways of aligning by using a shigemi plunger [28], teflon funnels [33], or glass capillaries [32] have been developed and successfully applied to proteins as well as small molecules. Recently, we were able to show that stretched polymers can also be used to partially align small molecules in organic solvents [34]. Cross-linked polystyrene (PS) sticks were simply swollen in NMR tubes, where they are automatically stretched by the boundaries of the glass walls (cf. Figure 1). In the meantime, the approach was successfully applied to a number of other cross-linked polymers like polymethylmethacrylate (with no polymer signals in the aromatic region, unpublished results) and polydimethylsiloxane (with strongly improved spectral quality [35]). Very recently, the gap of alignment media for polar

Structure Determination of Organic Molecules

RDC Measurement 1281

Relative Size of RDCs

Fig. 1. Photographs of a cross-linked PS stick designed for partial alignment in different states of swelling. From left to right: unswollen polymer stick in standard 5-mm NMR tube, polymer stick directly after polymerization, free polymer stick completely swollen, and polymer stick swollen in the NMR tube.

solvents was closed by stretched gels that swell in DMSO [36,37] and other solvents like methanol, acetonitrile, or acetone [36]. The main advantage of stretched polymer gels compared to liquid crystalline phases is that no lower limit of alignment is imposed because an unstretched gel principally does not show any alignment. The induced anisotropy is generally field independent, but as in liquid crystals, a significant temperature dependence is observed (35,72). Despite the microheterogeneity of the gels, line widths below 1 Hz can be achieved. Very importantly, the strength of alignment can be adjusted in many ways, like changing the diameter of the polymer sticks, the amount of cross-linking agent, or the polymerization conditions [34– 36]. Of course, swelling properties and thus the strength of alignment are different in different solvents. It is also interesting to note that the orientation and overall form of the alignment tensor not only depends on the polymer chosen but also on the solvent used [36, 72].

RDC Measurement NMR experiments on partially aligned samples are special with respect to some details in the setup. The lock, for example, is usually split by the quadrupolar coupling of the deuterium nuclei. As a consequence, the spectrometer might lock on one or the other half of the doublet and built-in spectrometer frequency references that are based on the defined lock signal typically lead to wrongly referenced chemical shifts. Also, shimming is affected: since the lock signal is split, it cannot be used for shimming purposes. Instead, it is necessary to shim on the integral of the FID. Having learned how to deal with these specialties, the acquisition of NMR experiments is straightforward. Some general considerations concerning the practicability of measuring certain RDCs should, however, be taken

Equations (1) and (2) contain all the information necessary to estimate the relative size of different types of RDCs. The dipolar coupling is proportional to the gyromagnetic ratios γI and γS and to the inverse cube of the internuclear distance r of an arbitrary pair of spins. Since the RDCs also depend on the angle θ, their size generally cannot be predicted without knowing the alignment tensor. Nevertheless, the maximum possible RDC values can be estimated and we will concentrate on these maximum values in the following text. In Figure 2, the relative size of maximum 1 H-X-RDCs is shown for different nuclei with respect to the internuclear distance r . It turns out that typical DHH RDCs over two or three bonds and one bond DCH RDCs show the strongest couplings, while long-range DCH or direct DCC RDCs (one-fourth of the DCH couplings for a given distance r ) are one order of magnitude smaller in size. Because of potential line broadening and to avoid strong coupling or second order effects in NMR spectra, the largest occurring RDCs should be ideally limited to be ≤∼30 Hz by adjusting the alignment medium. In this range, one bond DCH and DHH couplings can be accurately measured and provide valuable structural information. All other RDCs to carbon nuclei, however, are very small in size and require highly accurate techniques to obtain structural information. Therefore, the goal must be to measure as many strong RDCs as possible and only in cases where the easily obtained information is not sufficient, more advanced techniques should be used to measure the weaker RDCs, which generally have a larger relative error. In addition to proton and carbon bound RDCs, one bond DNH couplings should provide useful structural information with RDCs of intermediate size. Whenever present, DHF and DFF RDCs behave very similar to DHH RDCs. Although having a high gyromagnetic ratio and 100% natural abundance, RDCs involving 31 P nuclei are usually small because of the large internuclear distance to the next NMR-active nucleus.

Experimental Methods After having the assignment, RDCs are in principle measurable in 1D experiments. Proton-coupled 13 C spectra provide useful information in this regard. In fact, if the alignment is so strong that RDCs are on the order of 1 JCH coupling constants, one-dimensional experiments provide the only viable way to measure RDCs [12]. However, if the largest RDCs are in the desired range of ±30 Hz

Part II

into account. We therefore have a quick look at the relative size of different types of RDCs before we give an overview of a number of available experiments.

1282 Part II

Pharmaceutical Sciences

Part II

Fig. 2. Relative size of maximum dipolar couplings for protons for various NMRactive nuclei with respect to the internuclear distance r [see Equations (1) and (2)]. 13 C–13 C dipolar couplings are onefourth of the 1 H–13 C dipolar couplings.

1H-1H 1H-19F

106

1H-31P

Max. dip. Coupling [Hz]

1H-13C 1H-15N 13C-13C

105

104

103 1

2

3

4

5

6

7

Distance [A]

more advanced NMR experiments with high sensitivity and better resolution can be applied. The most easily measurable RDCs are one bond DCH couplings, which can be measured using standard coupled HSQC or HMQC experiments [12–14,25,26,31,34–36] (cf. Figure 3, for example, spectra of strychnine in PS/CDCl3 gel). Because of the variation in (1 JCH + DCH )-effective couplings, a procedure in which every multiplet component is phased

separately should be applied to get reliable coupling constants [25]. If the resolution of standard HSQC/HMQC experiments is not sufficient, a large number of specialized techniques developed in the field of biomolecular NMR spectroscopy are available to accurately measure one bond DCH and DNH couplings. Spin-state selective excitation [38,39] and coherence transfer [40], IPAP [41],

Fig. 3. Region of the 1 H,13 C-correlation spectra (HSQC) acquired on ∼50 mg strychnine in CDCl3 (left) and in a PS gel swollen in CDCl3 (right) at 300 K. 1 JHC and 1 JHC + DHC couplings, respectively, are given next to the corresponding cross peaks. As can be seen from the 1D slice at 37.8 ppm DCH RDCs are easily measured. The broad signals of the PS polymer are indicated with an asterisk (∗ ).

Structure Determination of Organic Molecules

gyromagnetic ratio and 100% natural abundance is perfectly suited for RDC measurements. DCF couplings should be measurable the same way as DCH RDCs. In the case when a rather strong JHF coupling is present, specially designed E-COSY and S3 E-E-COSY type experiments allow the very sensitive measurement of DHF , DHH , and DFF couplings [69,70].

Applications The first applications to obtain structural information from RDCs for sugars were reported in 2000 and for mediumsized molecules in organic solvents in 2003. Considering this short time-span, the demonstrated uses of RDCs are quite remarkable. Maybe the most elegant application of RDCs is found in six-membered chair-like rings, where RDCs can be used directly to distinguish axial and equatorial protons without having the need to derive an alignment tensor [25]. The method uses the fact that all axial C–H vectors are oriented in the same direction and therefore must have virtually identical DCH couplings. The assignment then simply can be achieved by looking at the occurrence of identical RDCs: all protons with very similar DCH RDCs are axial while all others are equatorial (cf. Figure 4). But even for arbitrary molecules the prochiral assignment of methylene groups can be achieved in most cases without having to rely on NOE connectivities as long as the alignment tensor is known [12,13]. This is especially important for isolated CH2 groups, which have few neighbors for NOE identification. A major application can also be found in the determination of Z /E configurations [14] and the determination of the relative chirality of stereochemical centers [26,35] even in different parts of a molecule that might be spatially distant. For this technique, structural models of all possible configurations have to be built first. The measured RDCs are then compared with RDCs backcalculated from the different structures. The quality of the fits expressed in

Fig. 4. Measured DCH RDCs for menthol in a PS/CDCl3 gel. Axial and equatiorial protons can easily be distinguished without knowing the alignment tensor.

Part II

J modulation [42], JE-TROSY [43], and quantitative J [44] experiments are some useful examples. For methylene groups, the so-called SPITZE-HSQC provides a separation of multiplet components in four spectra that allow RDC determination in more crowded regions [45]. It should also be mentioned that in almost all experiments, the use of BIRD filtering techniques can further improve spectra by selectively decoupling long-range C–H couplings [46–48]. The experiments designed for measuring DCH RDCs should also be easily adaptable to the DNH case. Due to the fast averaging DCH RDCs of methyl groups cannot be used directly as structural information. However, they can be converted to DCC RDCs to the attached carbon via the formula DCC = 3 3 /rCC ) [13]. DCH3 (−3γC /γH )(rCH The measurement of DHH couplings is more difficult: in simple cases (J + D) splittings can be measured directly in a one-dimensional experiment. However, since the sign of the splitting is not known and J and D are of comparable size, two possible RDC values must be considered (e.g. J = 5 Hz, |J + D| = 7 Hz leads to either D = 2 or −12 Hz). A sign-sensitive measurement of coupling constants is therefore highly desirable, which can be achieved in principle by E-COSY [49–51] and S3 EE-COSY [39,52] methods. Unfortunately, the applicability of at least the homonuclear E-COSY methods might be strongly reduced because of the broadened lines of partially aligned samples. A more sophisticated homonuclear approach to accurately measure RDCs is the SIAMTOCSY [53,54], which uses in-phase and anti-phase components in TOCSY-type experiments in combination with the Keeler–Titman fitting procedure [55]. Integrationbased techniques like the CT-COSY [56,57] or even the integration of conventional COSY cross peaks [58] offer an alternative for measuring the size of RDCs. The sign of couplings might then be determined with additional pulse sequences that use recently developed Hartmann–Hahn like multiple pulse elements [59]. The sign of the dipolar coupling might also be measurable by slightly changing the strength of alignment in the same alignment medium. The difference in measured RDCs should then lead to the sign of the dipolar coupling relative to the scalar coupling. Long range DCH couplings can basically obtained in two ways: if the carbon atom has a proton attached, S3 E-E-COSY type methods [52,60] provide an efficient, sign-sensitive procedure; in the case that the carbon of interest does not have a proton attached, recently developed HMBC-based methods can be applied that use a sophisticated fitting procedure to obtain the size of the coupling [13,61,62]. Methods for measuring DCC couplings at natural abundance are the well-known INADEQUATE [63,64] and ADEQUATE [65,66] experiments, which inherently have very low sensitivity. RDCs to 31 P nuclei have been measured in nucleic acids [67,68]. Fluorine with its high

Applications 1283

1284 Part II

Pharmaceutical Sciences

Part II

χ 2 or the correlation factor R 2 leads to the best structural model. A challenging application is the determination of the sugar pucker with RDCs. Freedberg [24] has shown that the RDCs measured for sucrose cannot be explained by a single conformation. On the other hand, a large number of potential sugar puckers can be ruled out, giving more insight into the population of structures and the dynamic behavior of sugars. Dynamic regions in molecules are generally difficult to handle and there is no exception in the case of RDCs. RDCs in dynamic regions are averaged over all populated states and might lead to misinterpretations if the populated structures differ significantly. However, the use of several different alignment media should make it possible to unambiguously identify flexible regions in a molecule and together with the improved prediction of alignment tensors [71], the determined RDCs might be fitted to a small ensemble of structures in the future. A potential application of RDCs as upper distance restraints surprisingly has not yet been reported. RDCs over ˚ have been measured. For a given distances larger than 7A measured RDC, an upper distance restraint can be immediately specified, as long as the axial component of the alignment tensor is known.

Conclusion The development of partial alignment media over the last years allows for the first time the broad use of anisotropic parameters like RDCs or residual chemical shift anisotropy in high-resolution NMR spectroscopy. Liquid crystalline phases and stretched polymer gels with scalable alignment properties for practically all useful NMR solvents are available. The RDCs of these samples can be measured by simply calculating the difference of couplings measured in isotropic solution and couplings measured in aligned samples. So far mostly solutes in aqueous solutions have been studied. Recently developed alignment techniques in stretched polymer gels will open the measurement of RDCs to structural studies in all NMR relevant solutions. Quite a number of applications with small organic molecules have been demonstrated, including the assignment of axial/equatorial or prochiral methylene protons, and the determination of relative stereochemistry and configuration. The method significantly widens the potential of modern liquid-state NMR spectroscopy and it will be interesting to see new developments in this promising young field.

References 1. Reif B, Hennig M, Griesinger C. Science. 1997;276:1230. 2. Tjandra N, Bax A. Science. 1997;278:1111.

3. Prestegard JH. Nat. Struct. Biol. 1998;5:517. 4. Kramer F, Deshmukh MV, Kessler H, Glaser SJ. Concepts Magn. Reson. Part A 2004;21A:10. 5. Ernst RR, Bodenhausen G, Wokaun A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions. Oxford University Press: New York, 1987. 6. Bothnerby AA, Domaille PJ, Gayathri C. J. Am. Chem. Soc. 1981;103:5602. 7. Saupe A, Englert G. Phys. Rev. Lett. 1963;11:462. 8. Emsley JW, Lindon JC. NMR Spectroscopy Using Liquid Crystal Solvents. Pergamon Press: Oxford, 1975. 9. Diehl P, Henrichs PM. Specialist Report on NMR Spectroscopy. Chemical Society, Vol. 1, 1972. 10. Diehl P, Niederberger W. Specialist Report on NMR Spectroscopy. Chemical Society, Vol. 3, 1974. 11. Panar M, Phillips WD. J. Am. Chem. Soc. 1968;90:3880. 12. Thiele CM, Berger S. Org. Lett. 2003;5:705. 13. Verdier L, Sakhaii P, Zweckstetter M, Griesinger C. J. Magn. Reson. 2003;163:353. 14. Aroulanda C, Boucard V, Guibe F, Courtieu J, Merlet D. Chem. Eur. J. 2003;9:4536. 15. Bendiak B. J. Am. Chem. Soc. 2002;124:14862. 16. Thiele CM. J. Org. Chem. 2004;69:7403. 17. Hansen MR, Mueller L, Pardi A. Nat. Struct. Biol. 1998; 5:1065. 18. R¨uckert M, Otting G. J. Am. Chem. Soc. 2000;122:7793. 19. Fleming K, Gray D, Prasannan S, Matthews S. J. Am. Chem. Soc. 2000;122:5224. 20. Martin-Pastor M, Bush CA. Abstracts of Papers of the Am. Chem. Soc. 2000;220:U117. 21. Martin-Pastor M, Bush CA. J. Biomol. NMR 2001;19: 125. 22. Neubauer H, Meiler J, Peti W, Griesinger C. Helvetica Chim. Acta. 2001;84:243. 23. Azurmendi HF, Bush CA. Carbohydr. Res. 2002;337:905. 24. Freedberg DI. J. Am. Chem. Soc. 2002;124:2358. 25. Yan JL, Kline AD, Mo HP, Shapiro MJ, Zartler ER. J. Org. Chem. 2003;68:1786. 26. Yan JL, Delaglio F, Kaerner A, Kline AD, Mo HP, Shapiro MJ, Smitka TA, Stephenson GA, Zartler ER. J. Am. Chem. Soc. 2004;126:5008. 27. Deloche B, Samulski ET. Macromolecules. 1981;14:575. 28. Tycko R, Blanco FJ, Ishii Y. J. Am. Chem. Soc. 2000; 122:9340. 29. Ishii Y, Markus MA, Tycko R. J. Biomol. NMR 2001;21:141. 30. Sass HJ, Musco G, Stahl SJ, Wingfield PT, Grzesiek S. J. Biomol. NMR 2000;18:303. 31. Mangoni A, Esposito V, Randazzo A. Chem. Commun. 2003; 154. 32. Meier S, Haussinger D, Grzesiek S. J. Biomol. NMR 2002;24: 351. 33. Chou JJ, Gaemers S, Howder B, Louis JM, Bax A. J. Biomol. NMR 2001;21:377. 34. Luy B, Kobzar K, Kessler H. Angew. Chem. Int. Ed. 2004; 43:1092. 35. Freudenberger C, Spitteler P, Bauer R, Kessler H, Luy B. J. Amer. Chem. Soc. 2004;126:14690. 36. Freudenberger C, Kobzar K, S. Kn¨or, Heckmann D, Paululat T, Kessler H, Luy B. Angew. Chem. Int. Ed. 2005;44:423. 37. Haberz P, Farjon J, Griesinger C. Angew. Chem. Int. Ed. 2005;44:427.

Structure Determination of Organic Molecules

54. M¨oglich A, Wenzler M, Kramer F, Glaser SJ, Brunner E. J. Biomol. NMR 2002;23:211. 55. Titman JJ, Keeler J. J. Magn. Reson. 1990;89:640. 56. Tian F, Bolon PJ, Prestegard JH. J. Am. Chem. Soc. 1999;121:7712. 57. Wu ZR, Bax A. J. Magn. Reson. 2001;151:242. 58. Delaglio F, Wu ZR, Bax A. J. Magn. Reson. 2001;149:276. 59. Kramer F, Jung A, Brunner E, Glaser SJ. J. Magn. Reson. 2004;169:49. 60. Meissner A, Sørensen OW. Chem. Phys. Lett. 1997;276:97. 61. Edden RAE, Keeler J. J. Magn. Reson. 2004;166:53. 62. Schulte-Herbr¨uggen T, Meissner A, Papanikos A, Meldal M, Sørensen OW. J. Magn. Reson. 2002;156:282. 63. Bax A, Freeman R, Kempsell SP. J. Am. Chem. Soc. 1980;102:4849. 64. Meissner A, Sørensen OW. Concepts Magn. Reson. 2002;14: 141. 65. Reif B, K¨ock M, Kerssebaum R, Kang H, Fenical W, Griesinger C. J. Magn. Reson. Ser. A 1996;118:282. 66. Reif B, Køck M, Kerssebaum R, Schleucher J, Griesinger C. J. Magn. Reson. Ser. B 1996;112:295. 67. Wu ZR, Tjandra N, Bax A. J. Biomol. NMR 2001;19:367. 68. Carlomagno T, Hennig M, Williamson JR. J. Biomol. NMR 2002;22:65. 69. Luy B, Marino JP. J. Biomol. NMR 2001;20:39. 70. Luy B, Barchi JJ, Marino JP. J. Magn. Reson. 2001;152:179. 71. Zweckstetter M, Bax A. J. Am. Chem. Soc. 2000;122:3791.

Part II

38. Meissner A, Duus JO, Sørensen OW. J. Biomol. NMR 1997; 10:89. 39. Meissner A, Duus JO, Sørensen OW. J. Magn. Reson. 1997; 128:92. 40. Sørensen MD, Meissner A, Sørensen OW. J. Biomol. NMR 1997;10:181. 41. Ottiger M, Delaglio F, Bax A. J. Magn. Reson. 1998;131: 373. 42. Tjandra N, Bax A. J. Magn. Reson. 1997;124:512. 43. Luy B, Marino JP. J. Magn. Reson. 2003;163:92. 44. Vuister GW, Bax A. J. Am. Chem. Soc. 1993;115:7772. 45. Carlomagno T, Peti W, Griesinger C. J. Biomol. NMR 2000;17:99. 46. Garbow JR, Weitekamp DP, Pines A. Chem. Phys. Lett. 1982; 93:504. 47. Uhrin D, Liptaj T, Kover KE. J. Magn. Reson. Ser. A 1993; 101:41. 48. Feher K, Berger S, Kover KE. J. Magn. Reson. 2003;163:340. 49. Griesinger C, Sørensen OW, Ernst RR. J. Am. Chem. Soc. 1985;107:6394. 50. Griesinger C, Sørensen OW, Ernst RR. J. Chem. Phys. 1986; 85:6837. 51. Griesinger C, Sørensen OW, Ernst RR. J. Magn. Reson. 1987; 75:474. 52. Meissner A, Sørensen OW. Magn. Reson. Chem. 2001;39:49. 53. Prasch T, Gr¨oschke P, Glaser SJ. Angew. Chem. Int. Ed. 1998;37:802.

References 1285

1287

Keyang Ding and Angela M. Gronenborn Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892

Abstract Residual dipolar couplings (RDCs) provide important constraints for the determination and refinement of protein NMR structures. Based on echo–anti-echo manipulation or IPAP principle, a suite of sensitivity enhanced experiments are described for measuring backbone 1 HN –15 N, 15 N–13 C , 1 HN –13 C , 13 C –13 Cα, 13 Cα– 1 α 15 H , N(i)–13 Cα(i), 1 HN (i)–13 Cα(i), 15 N(i)–13 Cα(i − 1), and 1 HN (i)–13 Cα(i − 1) dipolar couplings in proteins. The accuracy of the measured couplings can be assessed by comparing the experimentally obtained values with those predicted based on high resolution structures. Even for very small RDCs, such as the 15 N(i)–13 Cα(i − 1) couplings that are smaller than 0.3 Hz, a correlation coefficient of 0.83 is obtained, attesting to the accuracy of couplings obtained with these sensitivity-enhanced IPAP experiments. We also present a novel application for the use of RDCs. Under certain conditions, the folded state of a protein comprises detectable, conformational sub-states. Such sub-states at local sites, so-called melting hot spots, are characterized by re-orienting bond vectors. Determination of RDCs allows for efficient and easy detection of such hot spots.

Introduction The importance of being ordered has been highly recognized in the field of protein NMR spectroscopy [1]. When a protein is placed into a magnetic field, it will experience a small degree of alignment. This causes any pair of spins within the protein to exhibit a dipolar interaction in addition to the scalar interaction that is manifested in the J coupling. It was noted earlier [2,3] that these small dipolar couplings are a direct reflection of the protein structure, since they provide a measure of the distance and the orientation of the vector connecting the two nuclei. Most proteins lack sufficient intrinsic magnetic susceptibility anisotropies for practical alignment purposes, however, alignment of any biological macromolecule can be achieved using liquid crystalline media. Generally, there Graham A. Webb (ed.), Modern Magnetic Resonance, 1287–1291.
C 2008 Springer.

are two major aspects that need to be addressed in the residual dipolar coupling (RDC)-related techniques: selection and optimization of the alignment media to yield measurable dipolar couplings and the development of accurate methods for measuring these small couplings. Any alignment medium should be inert to the solutes (proteins) under investigation and needs to be stable over reasonably wide pH and temperature ranges. In addition, it should not cause any adverse effects on the NMR line widths, so that the NMR spectra in the alignment media remain at comparable resolution to those in the isotropic aqueous solution. The first alignment medium ever used was a lyotropic liquid crystal consisting of binary mixtures of dihexanoyl phosphatidylcholine and dimyristoyl phosphatidylcholine [3]. At present, a large variety of different alignment media are in use, including suspensions of charged, rod-shaped viruses, such as tobacco mosaic virus and filamentous bacteriophages fd/M13 and Pf1, quasi-ternary systems of surfactant/salt/alcohol or aqueous alkyl-poly(ethylene glycol)/alcohol mixtures, known to form stacked lamellar phases, and vertical or radial strained polyacrylamide gels [4–10]. Such gels are generally believed to be the most inert media for proteins. The degree of alignment can be tuned by adjusting the concentration of the liquid crystalline materials or the degree of strain/compression in case of the polyacrylamide gels. A suitable degree of alignment results in RDC values for 1 DNH between 15 and 25 Hz. The most commonly investigated dipolar couplings for proteins are the five backbone one-bond 15 N–1 HN , 15 N– 13  13  13 α C , C – C , and 1 Hα–13 Cα and two-bond 1 HN –13 C RDCs [11]. In addition, one-bond 13 Cα–13 Cβ and 1 Hβ– 13 β C , and one-bond 13 C–13 C and 1 H–13 C couplings in methyl groups are also of interest [12,13]. Long-range 1 H–1 H RDCs of protons close in space or connected via multiple bonds may yield important information with respect to quantitative NOE connectivity or local conformation [14]. Accurate measurement of these small dipolar couplings is still a challenge in protein NMR spectroscopy. Generally, all presently available methods for RDC determination can be classified into two major categories.

Part II

Measurement of Residual Dipolar Couplings and Applications in Protein NMR

1288 Part II

Pharmaceutical Sciences

Part II

Fig. 1. Schematic representation of the echo–anti-echo manipulation. The upper left box displays the normal area of a 2D 15 N–1 HN HSQC spectrum. Only one 15 N–1 HN cross peak is shown, split by the coupling JNX and JHX along the 15 N and 1 H dimensions, respectively. Using echo–anti-echo manipulation, a spectrum such as shown in the right box is obtained. One peak of the pair remains at its original position, while the other one is flipped around the zero 15 N frequency line. In this manner, coupled 15 N–1 HN cross peaks are separated. The splitting JNX and JHX can be the active coupling JNH , or the passive coupling from a third spin X.

One methodology is called the quantitative J correlation and is based on intensity analysis [15]. Here, the coupling constant is often estimated from two points in the period during which this coupling evolves. If the measurement is based on a series of points, the resulting experiment is equivalent to a three-dimensional (3D) experiment. The accuracy of quantitative J correlation experiments relies critically on the uniqueness of the selected magnetization transfer pathways and the model employed to de-convolute the coupling from all other complex relaxation processes. The second methodology for the determination of dipolar couplings involves measuring the corresponding splitting directly from coupled multidimensional spectra. In these spectra the number of peaks is doubled due to coupling splitting, resulting in a reduction of resolution in these coupled spectra compared to their decoupled counterparts. In some cases, the splitting may actually be too small to be resolved with respect to the digital resolution of the spectra. It therefore is highly desirable to separate the two peaks within a split doublet. In the following section, we describe techniques for separating the two peaks within a coupled doublet. They result in a suite of sensitivity-enhanced experiments for accurately and efficiently measuring backbone RDCs in proteins [16–20]. In addition, we describe some of the applications with respect to using RDCs to study protein folding and unfolding [21].

Measurement of Backbone Residual Dipolar Couplings in Proteins Generally, there are two methods to separate the couplings within a split pair of resonances in a coupled

spectrum: echo–anti-echo manipulation [16–18] and spinstate-selective technique [22–24]. The echo–anti-echo manipulation is schematically represented in Figure 1. The two peaks within a doublet are manipulated into P-type and N-type peaks [18], respectively. In a series of FIDs, the t1 -modulation alternates between echo and anti-echo for one peak and between anti-echo and echo for the other peak within the doublet. In this way, sensitivity enhancement is achieved naturally and directly. The loss in resolution of such coupled spectra, when compared to the decoupled spectra, can be easily compensated for by setting the spectral width in the 15 N dimension to twice the chemical shift range and by implementing TPPI [25,26] to shift the carrier frequency to the edge of the 15 N chemical shift range. In special cases, either JNX or JHX may be zero. Based on the echo–anti-echo manipulation, four sensitivity-enhanced experiments [16,17] were devised for measuring one-bond 1 HN –15 N dipolar couplings and one sensitivity-enhanced experiment [18] for simultaneously measuring one-bond 15 N–13 C and two-bond 1 HN – 13  C dipolar couplings. We demonstrated the accuracy of the developed experiments using the protein GB1 aligned in liquid crystalline Pf1 (15 mg/ml) in 95% H2 O/5% D2 O at pH 7 and 25 ◦ C and measured a large set of RDCs. Comparing the experimentally obtained RDCs with those calculated from a previously obtained high resolution model structure [27] (PDB code: 3GB1), excellent agreement between these values can be noted [16–18]. For onebond 15 N–1 HN , 15 N–13 C and two-bond 1 HN –13 C RDCs the correlation coefficients are 0.998, 0.98, and 0.96, respectively. Interestingly, the presence of cross-correlated relaxation between 15 N–1 HN and 1 HN –1 Hα dipolar interactions causes one-bond 15 N–1 HN couplings measured in the 15 N dimension to be different from those measured in

Measurement of Residual Dipolar Couplings

Applications of Residual Dipolar Couplings in Proteins 1289

2D series number Raw dataset 4n − 3 4n − 2 4n − 1 4n New dataset 1 (2n − 1) = (4n − 3) + (4n − 2) 2n = (4n − 1) + (4n) New dataset 2 (2n − 1) = (4n − 3) − (4n − 2) 2n = (4n − 1) − (4n)

Real part

Imaginary part

cos(πkJ MX t1 )cos(ωH t2 − ωN t1 ) sin(π kJ MX t1 )sin(ωH t2 − ωN t1 ) cos(π kJ MX t1 )cos(ωH t2 + ωN t1 ) −sin(π kJ MX t1 )sin(ωH t2 + ωN t1 )

−cos(π kJ MX t1 )sin(ωH t2 − ωN t1 ) sin(π kJ MX t1 )cos(ωH t2 − ωN t1 ) −cos(π kJ MX t1 )sin(ωH t2 + ωN t1 ) sin(π kJ MX t1 )cos(ωH t2 + ωN t1 )

cos[ωH t2 − (ωN + π kJ MX )t1 ] cos[ωH t2 + (ωN + π kJ MX )t1 ]

−sin[ωH t2 − (ωN + π kJ MX )t1 ] −sin[ωH t2 + (ωN + π kJ MX )t1 ]

cos[ωH t2 − (ωN − π kJ MX )t1 ] cos[ωH t2 + (ωN − π kJ MX )t1 ]

−sin[ωH t2 − (ωN − π kJ MX )t1 ] −sin[ωH t2 + (ωN − π kJ MX )t1 ]

the 1 H dimension [17]. Therefore, it is in general preferable to measure the amide RDCs in the 15 N dimension in these spectra, although the orientational information is identical in both dimensions. The spin-state-selective technique combined with sensitivity enhancement and interleaved data acquisition is summarized in Table 1 [19]. This approach is especially useful for the cases where the coupling is to be measured and the 15 N chemical shift evolve during different time periods. In these cases, it is often necessary to record a 3D experiment, which in turn is reduced in dimensionality from 3 to 2 [28–30] using the accordion principle [31,32]. Generally, a scaling factor of k can be introduced to scale the JMX coupling evolution. By carefully setting the t1 modulation according to Table 1, sensitivity-enhanced experiments were implemented for measuring one-bond 13  13 α C – C and 1 Hα–13 Cα dipolar couplings [19]. In this way, two sub-spectra with resolution comparable to the decoupled spectra are obtained and the corresponding peak displacement along the 15 N dimension provides a direct measure of the one-bond 13 C –13 Cα and 1 Hα−13 Cα dipolar couplings. Again, the applicability of these experiments was demonstrated for protein GB1 aligned in liquid crystalline Pf1 (15 mg/ml) in 95% H2 O/5% D2 O at pH 7 and 25 ◦ C. Excellent correspondence between measured and calculated one-bond 13 C –13 Cα and 1 Hα– 13 α C dipolar couplings was found [19], with correlation coefficients of 0.96 and 0.96, respectively. In a further application of the spin-state-selective techniques, sensitivity-enhanced experiments for measuring 15 N(i)–13 Cα(i), 1 HN (i)–13 Cα(i), 15 N(i)–13 Cα(i − 1), and 1 HN (i)–13 Cα(i − 1) dipolar couplings were also developed [20]. Since the 15 N(i)–13 Cα(i) and 15 N(i)– 13 α C (i − 1) coupling evolution and the 15 N chemical shift evolution are in a common time period, there is no need for the accordion principle. Every cross peak in the two sub-

spectra exhibits E.COSY splittings from the intra-residue Cα spin. Because one-bond 15 N–13 Cα J couplings are small, ranging from 8.5 to 13.5 Hz, very high digital resolution in 15 N dimension is necessary for resolving the E.COSY doublet. Consequently, the application of this type of experiment is limited to small proteins with an apparent 15 N line width ≤8.5 Hz. The comparison between measured and calculated one-bond and two-bond 15 N–13 Cα, as well as two-bond and three-bond 1 HN –13 Cα RDCs for the protein GB1 aligned in liquid crystalline Pf1 (15 mg/ml) in 95% H2 O/5% D2 O at pH 7 and 25 ◦ C, yields correlation coefficients of 0.983, 0.833, 0.976, and 0.944, respectively. We wish to point out that the twobond 15 N–13 Cα RDCs are extremely small, namely ≤ 0.3 Hz, still yielding a remarkable correlation coefficient of 0.833 [20]. This implies that the experimentally measured values have to be highly accurate and that the quality of the model structure is also very good. 13

Applications of Residual Dipolar Couplings in Proteins The most direct application of RDCs in protein NMR is their use as orientational constraints for 3D structure determination. For a rigid molecule, the orientations of all the two-spin vectors are fixed with respect to the molecular frame. Therefore, the corresponding RDC is proportional to the scalar product between a 3 × 3 traceless order tensor, global to any pair of spins, and the secondorder tensor of the two-spin vector [33]. It is this relationship between the order tensors that is the basis for using RDCs in NMR structure determination. If, however, the directions of some two-spin vectors reorient with respect to the molecular frame, the proportionality to the scalar product is broken and the corresponding RDCs cannot

Part II

Table 1: Modulation of the coupling JMX evolution and 15 N chemical shift frequency ωN to the raw and manipulated FIDs in the 2D series for values of n from 1 to TD1 /2

Pharmaceutical Sciences

15

A

10

K13

5

L12

T11 0 −5

T25 A24

−10 −15

gel/pH5.5/25°C −15 −10 −5

0

5

Measured DNH (Hz)

15

5

0 −5 −10 −15

gel/pH3.0/37°C −15 −10 −5

0

5

10 15

Calculated DNH (Hz)

provide meaningful structural constraints. They can, on the other hand, reveal important information about motions, either relating to inter-domain motions [34] or local melting events [21]. Local melting hot spots are easily identified since local order tensor parameters will be different from the global tensor, caused by the local reorientation motion at this site. For example, we examined the structure and dynamics of a series of core mutants of the protein GB1. For one of these, S14, local melting was discovered under destabilizing conditions based on RDC measurements. The identification of the melting hot spots is illustrated in Figure 2 [21]. RDCs were measured in a stretched polyacrylamide gel at temperatures between 25 and 37 ◦ C and at pH 5.5 they correlate well (Pearson correlation coefficient = 0.99) with calculated values based on the refined NMR structure of wild-type GB1 (PDB code: 3GB1). However, at pH 3.0 and 37 ◦ C, the data points for residues T11, L12, K13, A24, and T25 deviate from the diagonal correlation line and are scattering along a horizontal line (Figure 2B). Under those experimental conditions S14 exists in a folded and unfolded state, as evidenced by a second set of resonances, not belonging to native state. For the unfolded protein, no correlation with the native, folded structure is preserved

K13 T25

−5 −10 −15

gel/pH3.0/37°C 0

5

10 15

Calculated DNH (Hz) 15

5

L12

T11

−15 −10 −5

C

10

A24

0

10 15

DNH (Hz) at pH5.5

15

B

10

Calculated DNH (Hz)

Measured DNH (Hz)

Part II

Fig. 2. Characterization of melting hot spots in a destabilized mutant of GB1 (S14) by RDC measurement. The measured DNH for S14 against the predicted values are shown in (A) and (B), and data points associated with local melting hot spots are displayed as filled triangles (T11, L12, K13) and filled circles (A24, T25). (C) Measured DNH of S14 in the molten state vs. those predicted from the folded structure. (D) Correlation between measured DNH of folded S14 at 25 ◦ C in a stretched gel at pH 5.5 and at pH 3.0. (Data points for the residues comprising the melting hot spots are excluded.)

Measured DNH (Hz)

1290 Part II

D

10 5 0 −5 −10

gel/25°C

−15 −15 −10 −5

0

5

10 15

DNH (Hz) at pH3.0

and all experimental RDCs scatter along the same horizontal line (Figure 2C) as noted before for the outliers of the folded state. We also established that the changes in RDCs are not caused by a pH change on the gel, since the couplings of the folded GB1 at pH 3.0 and 5.5 are in excellent agreement (Figure 2D). In conclusion, our application of RDC measurements to the GB1 folding landscape clearly demonstrates their power for identifying melting hot spots in proteins.

Discussion There is now ample data available (see above) that indicate the importance of RDC derived constraints for NMR structure determination of proteins as well as for identifying local areas of mobility. Proteins are also rich in 1 H–13 C spin pairs in the side chains. Therefore long-range 1 H– 1 H RDCs will probably attract more and more attention since they will be able to provide quantitative distance constraints. Once accurate and efficient methods are developed for measuring these side-chain 1 H–13 C and the long-range 1 H–1 H dipolar couplings, NMR structural biology will move into a new area again.

Measurement of Residual Dipolar Couplings

This work was supported in part by the Intramural AIDS Targeted Antiviral Program of the Office of the Director of the National Institutes of Health to AMG.

References 1. Gronenborn AM. C. R. Biol. 2002;325:957–66. 2. Tolman JR, Flanagan JM, Kennedy MA, Prestegard JH. Proc. Natl. Acad. Sci. U.S.A. 1995;92:9279–83. 3. Tjandra N, Bax A. Science. 1997;278:1111–4. 4. Clore GM, Starich MR, Gronenborn AM. J. Am. Chem. Soc. 1998;120:10571–2. 5. Hansen MR, Mueller L, Pardi A. Nat. Struct. Biol. 1998;5: 1065–74. 6. Prosser RS, Losonczi JA, Shiyanovskaya IV. J. Am. Chem. Soc. 1998;120:11010–1. 7. Barrientos LG, Dolan C, Gronenborn AM. J. Biomol. NMR. 2000;16:329–37. 8. Ruckert M, Otting GJ. Am. Chem. Soc. 2000;122:7793–7. 9. Tycko R, Blanco FJ, Ishii Y. J. Am. Chem. Soc. 2000;122: 9340–1. 10. Sass HJ, Musco G, Stahl SJ, Wingfield PT, Grzesiek SJ. Biomol. NMR. 1998;18:303–9. 11. deAlba E, Tjandra N. Prog. Nucl. Magn. Reson. Spectrosc. 2002;40:175–97. 12. Ottiger M, Bax A. J. Am. Chem. Soc. 1999;121:4690–5. 13. Chou JJ, Bax A. J. Am. Chem. Soc. 2001;123;3844–5. 14. Wu Z, Bax A. J. Am. Chem. Soc. 2002;124:9672–3. 15. Bax A, Vuister GW, Grzesiek S, Delaglio F, Wang AC, Tschudin R, Zhu G. Methods Enzymol. 1994;239:79–105. 16. Ding K, Gronenborn AM. J. Magn. Reson. 2002;158:173–7.

17. Ding K, Gronenborn AM. J. Magn. Reson. 2003;163:208– 14. 18. Ding K, Gronenborn AM. J. Am. Chem. Soc. 2003;125: 11504–5. 19. Ding K, Gronenborn AM. J. Magn. Reson. 2004;167:253–8. 20. Ding K, Gronenborn AM. J. Am. Chem. Soc. 2004;126: 6232–3. 21. Ding K, Louis JM, Gronenborn AM. J. Mol. Biol. 2004;335: 1299–307. 22. Ottiger M, Delaglio F, Bax A. J. Magn. Reson. 1998;131: 373–8. 23. Cordier F, Dingley AJ, Grzesiek S. J. Biomol. NMR. 1999;13: 175–80. 24. Lerche MH, Meissner A, Poulsen FM, Sorensen OW. J. Magn. Reson. 1999;140:259–63. 25. Drobny G, Pines A, Sinton S, Weitekamp D, Wemmer D. Faraday Div. Chem. Soc. Symp. 1979;13:49. 26. Bodenhausen G, Vold RL, Vold RR. J. Magn. Reson. 1980; 37:93–106. 27. Kuszewski J, Gronenborn AM, Clore GM. J. Am. Chem. Soc. 1999;121:2337–8. 28. Szyperski T, Yeh DC, Sukumaran DK, Moseley HNB, Montelione GT. Proc. Natl. Acad. Sci. U.S.A. 2002;99:8009–14. 29. Ding K, Gronenborn AM. J. Magn. Reson. 2002;156:262–8. 30. Kozminski W, Zhukov I. J. Biomol. NMR. 2003;26:157– 66. 31. Bodenhausen G, Ernst RR. J. Magn. Reson. 1981;45:367– 73. 32. Bodenhausen G, Ernst RR. J. Am. Chem. Soc. 1982;104: 1304–9. 33. Tolman JR, Al-Hashimi HM, Kay LE, Prestegard JH. J. Am. Chem. Soc. 2001;123;1416–24. 34. Braddock DT, Cai ML, Baber JL, Huang Y, Clore GM. J. Am. Chem. Soc. 2001;123:8634–5.

Part II

Acknowledgments

References 1291

1293

Melissa A. Starovasnik and Wayne J. Fairbrother Department of Protein Engineering, Genentech Inc., South San Francisco, CA 94080, USA

The binding of IgE to its high-affinity receptor, FcεRI, is fundamental to allergic disease. Molecules that block this interaction could therefore act as useful therapeutics for the treatment of asthma, allergic rhinitis, and other forms of atopy. To this end, binding selections using the extracellular portion of the α-chain of FcεRI (FcεRIα) and polyvalent peptide-phage libraries have yielded two distinct classes of peptide ligands that antagonize IgE binding to its receptor and prevent downstream IgE-mediated signaling events in basophils [1,2]. NMR spectroscopy has been used to characterize the structures of these peptide antagonists and their modes of binding to FcεRIα. Structure determination of the ∼32-kDa peptide/ receptor complexes by NMR methods was not feasible for these systems because the heavily glycosylated receptor protein could not be bacterially expressed and thus could not be isotopically labeled (at least not readily). The high carbohydrate content of the receptor also makes these systems challenging for structure determination using X-ray crystallography. As described below, however, large 1 H chemical shift perturbations observed for the isotopically labeled peptide ligands upon binding to the receptor protein in solution could be used qualitatively to validate hypothesis-driven models of ligand binding, and quantitatively to refine these models. The resulting structural models will be useful in the development of novel IgE antagonists.

Hairpin Peptide Structure The first family of peptides identified by phage display have the general form X4 CX2 GPX4 CX4 and bind FcεRIα with affinities in the 1–10 µM range [1]. Analysis of one of these peptides, IGE06, by NMR spectroscopy revealed that the peptide has a well-defined three-dimensional (3D) structure in solution comprising two β-strands (residues 5–8 and 11–14) connected by a type I β-turn centered at residues Pro9 and Trp10 (Figure 1A). The three Nterminal residues are not as well defined by the NMR data and appear to be more flexible in solution than residues Cys5-Cys14. Residues required for structure and/or funcGraham A. Webb (ed.), Modern Magnetic Resonance, 1293–1298.
C 2008 Springer.

tion were identified by analysis of peptide variants using both NMR and activity measurements. Importantly, NMR analysis of peptides having alanine substitutions for Gly8, Pro9, and Trp10 indicate that these substitutions do not disrupt the hairpin structure even though they have a significant impact on FcεRIα-binding affinity (Figure 1A).

Zeta Peptide Structure Using an expanded set of peptide-phage libraries, a second family of peptides with a simpler motif, X2 CPX2 CYX, was identified [2]. When synthetic peptides were assayed for inhibition of IgE binding to cell-surface FcεRI, however, IC50 values of 250 µM or greater were obtained. Surprisingly, this activity improved over time and stabilized at 64 µM when assayed similarly 10 days after solubilization. One-dimensional (1D) and two-dimensional (2D) NMR analysis of a freshly prepared nine-residue monomeric disulfide-bonded peptide demonstrated at least four conformationally distinct forms, with the major form showing no evidence of stable structure (Figure 2). Analysis of 1D NMR spectra over time showed changes in the relative populations of the different forms (Figure 2), and after 7 days at room temperature ∼90% of the peptide had converted to a single new form that also showed significantly longer retention time than the original sample in an analytical HPLC run. This form had a stable structure in solution as evidenced by significant chemical shift dispersion relative to “random coil,” and extreme values for backbone and side chain 3 J coupling constants. Mass spectra of the new peptide form showed it to be a covalent disulfide-linked dimer. Comparison of analytical HPLC chromatograms and NMR spectra of synthetic parallel or antiparallel dimeric peptides confirmed that the active state of the phage-derived peptide was an antiparallel dimer with two intermolecular disulfide bonds. Subsequent linking of the monomeric peptides with a three-residue linker sequence to make a 21-residue “single-chain dimer” species, and further optimization on phage, resulted in a 32 nM-affinity peptide antagonist (e131).

Part II

Using Chemical Shift Perturbations to Validate and Refine the Docking of Novel IgE Antagonists to the High-Affinity IgE Receptor

1294 Part II

Pharmaceutical Sciences

Part II

to contact IgE. This aromatic-rich IgE interaction site is centered around two tryptophan residues on the receptor (Trp87 and Trp110), between which Pro426 of bound IgE is buried [3]. In order to characterize further the receptorbinding modes of the different peptide classes, 13 C/15 Nlabeled peptides were obtained by expression of protein A Z-domain fusion proteins [4] in Escherichia coli using a previously described alkaline phosphatase induction system [5], followed by cleavage (using trypsin or CNBr for the zeta or hairpin peptide, respectively) and purification of the peptides [6]. Isotopic labeling of FcεRIα was not practical because the heavily glycosylated extracellular domain (average molecular mass ∼30 kDa) could be produced only via baculoviral infection of insect cells [6]. Fig. 1. Solution structures of representative (A) hairpin and (B) zeta peptides. Shown are ribbon diagrams of the first structure from the ensembles in PDB files 1JBF (hairpin) [1] and 1KCO (zeta) [2]. Side chains are colored to indicate those residues that are important for maintaining the structure of the peptide (green; disulfide bonds in yellow), and those that are important for receptor binding, but are not important for maintaining a stable structure of the peptide (from most important to least, red > orange > blue). The backbone of residues 2–4 of the hairpin peptide is colored to indicate that truncation of this region results in significant loss in receptor-binding affinity. (See also Plate 98 on page XIX in the Color Plate Section.)

NMR structural analysis of e131 showed that the peptide adopts a stable structure comprised of two small antiparallel 310 helices connected by two disulfide bonds and a flexible linker region (Figure 1B). The helical backbone conformation is stabilized by packing of two Tyr side chains (Tyr8 and Tyr20) “below” the disulfide bonds, with each Tyr hydroxyl donating a hydrogen bond across the “dimer interface” to the carbonyl oxygen of a Cys residue. The backbone structure is pseudo-symmetric with residues 1–9 adopting essentially the same conformation as residues 13–21. Peptides of this class were designated “zeta” peptides due to a resemblance between the structurally conserved regions and the Greek letter zeta (ζ). As with the hairpin class of peptides, residues in the zeta peptide required for structure and/or function were identified from analysis of peptide analogs both by NMR and activity measurements. Again, the most important residue for high-affinity receptor binding was found to be a proline (Pro16) (Figure 1B).

Receptor Binding Despite their significantly different structures, the twopeptide classes were shown to compete for binding at the same site on the receptor. Alanine-scanning mutagenesis of FcεRIα demonstrated further that the peptides bind to a region of the receptor that had been shown previously

NMR Analysis of Zeta Peptide/Receptor Complex The 23-residue zeta peptide selected for NMR structural studies (e117) was shown previously to bind FcεRI and block IgE binding with an IC50 of 80 nM [2]. The resonances of the free zeta peptide were assigned readily by standard analysis of 2D 1 H/15 N-HSQC, 3D 1 H/15 NTOCSY-HSQC, 2D 1 H/13 C-HSQC, 2D CBCA(CO)NH, and 2D 1 H/13 C-HCCH-TOCSY spectra [7]. Resonance assignments of the receptor-bound zeta peptide were complicated, however, by the overall reduced sensitivity of J -correlated spectra for the ∼32 kDa complex, and by large variations in line widths, with those residues at the peptide termini and in the GGH “linker” being sharper and more intense than those from the rest of the peptide. Nevertheless, unambiguous assignments were obtained for a majority of the zeta peptide resonances based on analysis of 3D 15 N-edited and 13 C-edited NOESY spectra. Comparison of the resonance assignments of the free and receptor-bound peptide showed that many of the peaks from residues within the “disulfide-bonded core” of the peptide experienced significant chemical shift perturbations; the most dramatic chemical shift changes upon receptor binding were seen for the 1 H resonances of Pro16 (Hα, Hβ, Hγ, and Hδ all shifted upfield by 1.7–3.1 ppm; Figure 3A). Shifts of this magnitude observed for all proton resonances within the proline side chain can be explained if the proline is packed between aromatic rings in the protein/peptide complex, in much the same way as Pro426 of IgE is found “sandwiched” between two tryptophan residues in the crystal structure of the Fc fragment of IgE in complex with FcεRIα [3]. By contrast, the sharper peaks from the bound state of the zeta peptide underwent only minimal changes in chemical shift relative to the free state (Figure 3A), consistent with this group of residues being of minimal importance for receptor binding [2]. In addition to the chemical shift perturbations indicating that the zeta peptide binds proximal to aromatic residues on the receptor, 13 C-edited/12 C-filtered NOESY

Chemical Shift Refinement of IgE Receptor/Peptide Complexes

Receptor Binding 1295

Part II

Fig. 2. Spontaneous rearrangement of a monomeric 9-mer peptide into a covalent disulfide-linked dimer monitored by 1D NMR spectroscopy. Amide and aromatic region of a series of 1D 1 H NMR spectra acquired at 500 MHz, 25 ◦ C, on a sample of peptide e101 (ALCPAVCYV-NH2 ), originally synthesized and purified as a monomeric peptide with an intramolecular disulfide bond. Lyophilized peptide was resuspended in water and spectra were acquired intermittently over the course of a week, with the time after resuspension indicated. The population of well-folded dimeric peptide present over time is estimated based on the relative intensities of the two most upfield peaks shown, corresponding to the Tyr8 1 Hε resonance in the ordered dimeric state and in the monomeric or intermediate states. Note that if the peptide existed in a single conformation in solution, that no more than 11 peaks would be expected in this region of the spectrum.

spectra [8] showed NOE correlations between at least two distinct aromatic groups on the receptor and peptide residues Val1, Leu14, and Val18. Peptide residues Val1, Gln2, and Pro4 also had NOE correlations with a pair of peaks resonating at 0.75 and 0.73 ppm that were assigned tentatively to a pair of methyl groups on the receptor based on resonance intensity and frequency. No NOE correlations were observed from Pro16 of the peptide, despite the large chemical shift changes observed for this residue, due to the increased line widths in the receptor-bound state. The resonance broadening is most likely due to the large chemical shift gradient in which the proline is located; even small changes in the orientation of Pro16 with respect to the aromatic residues on the surface of the receptor would cause chemical shift changes resulting in the observed resonance broadening. Receptor contacts involving aromatic and methylcontaining side chains are consistent with the zeta peptide binding to the IgE-binding site as identified by receptor mutagenesis. The peptide was therefore docked manually

into the IgE-binding site of FcεRIα using coordinates of the free peptide and the previously published crystal structure of the IgE/FcεRIα complex [3]. The initial peptide placement was based on the assumption that the functionally important proline residue, Pro16, which experienced the most significant 1 H chemical shift perturbations upon receptor binding (Figure 3A), would bind between Trp87 and Trp110 of the receptor, in approximately the same way as observed for Pro426 of IgE. The resulting peptide orientation appeared consistent with the observed intermolecular NOE correlations; in particular, intermolecular NOE correlations between zeta peptide residues Val1, Gln2, and Pro4 and the unidentified pair of receptor methyl groups could be assigned tentatively to Leu158 of the receptor because these are the only methyl groups in this region of the receptor. The aliphatic 1 H chemical shifts of the manually docked peptide were calculated using the program SHIFTS 4.1 (X. Xu and D.A. Case, The Scripps Research Institute, La Jolla, CA; http://www.scripps.edu/case/

Fig. 3. (A) Plot of the net change in zeta peptide 1 Hα chemical shifts upon binding FcεRIα (δ = [δ12Hα ]1/2 ). (B) Correlation between observed and calculated chemical shifts (δ = δtotal − δrandom coil ) following refinement of the zeta peptide/FcεRIα complex structure using AMBER 6.0 (rms error = 0.038 ppm, R 2 = 0.9978). (C) Superposition of 20 models of the zeta peptide/FcεRIα complex structure obtained by docking and refinement against the 1 H chemical shifts of the peptide (average rms error = 0.041 ± 0.005 ppm). (D) Plot of the net change in hairpin peptide 1 Hα chemical shifts upon binding FcεRIα. (E) Correlation between observed and calculated chemical shifts following refinement of the hairpin peptide/FcεRIα complex structure using AMBER 6.0 (rms error = 0.090 ppm, R 2 = 0.9954). (F) Superposition of 20 models of the hairpin peptide/FcεRIα complex structure obtained by docking and refinement against the 1 H chemical shifts of the peptide (average rms error = 0.12 ± 0.02 ppm). (See also Plate 99 on page XX in the Color Plate Section.)

Chemical Shift Refinement of IgE Receptor/Peptide Complexes

where rc, m, and el, refer to ring current, magnetic anisotropy, and electrostatic contributions, respectively. The δ const term is an empirical “fitting” parameter that includes contributions from δ m and δ el that are present in the random coil state [14]. Good general agreement was found between the calculated and experimentally observed chemical shifts, although differences suggested that the manually docked model could be refined further. The docked model was therefore refined by optimizing the agreement between the calculated and observed 1 H chemical shifts of the receptor-bound peptide using functionally equivalent code within the molecular modeling package AMBER 6.0 (http://amber.scripps.edu/doc6/). All the unambiguously assigned aliphatic 1 H chemical shifts were incorporated as restraints (force constant = 25 kcal/·mol/·ppm) in a simple simulated annealing protocol. For initial rounds of refinement the chemical shifts of prochiral methylene and methyl protons were averaged. The chemical shifts of all peptide protons were subsequently back-calculated using SHIFTS 4.1 and compared to the experimentally determined chemical shifts. In some cases the chemical shifts of prochiral proton pairs were sufficiently different, and the agreement between the calculated and experimentally observed values was sufficiently close, to allow for stereospecific assignments that could be incorporated into subsequent rounds of refinement. In this way stereospecific assignments were possible for Pro4 1 Hβ, Leu14 1 Hβ, and Pro16 1 Hβ and 1 Hγ. In addition to the chemical shift restraints, distance and dihedral angle restraints derived from those used to calculate the solution structure of the free peptide were incorporated into the refinement to ensure that the peptide structure was maintained. The coordinates of the receptor were restrained to be nearly fixed during the simulated annealing and subsequent restrained minimization calculations using a harmonic potential with a force constant of 5 kcal/·mol. A weak harmonic potential (force constant = 0.5 kcal/·mol) was also used to restrain the disulfidebonded core (residues Cys3-Tyr8 and Ala13-Tyr20) of

the zeta peptide. Finally, tentatively assigned intermolecular NOE distance restraints were incorporated iteratively into the refinement calculations as upper-bound distance ˚ only if they were clearly satisfied in restraints of 5.8 A, the current models; in total, 17 intermolecular distance restraints were included in the final simulated annealing calculation. Importantly, the intermolecular distance restraints were not required for convergence of the refined models. Indeed, trial calculations in which the starting zeta ˚ peptide and receptor molecules were separated by ∼20 A still converged to essentially the same peptide/receptor orientation as those starting from manually docked models, even in the absence of intermolecular distance restraints (W.J. Fairbrother, unpublished results). The resultant agreement between the calculated and experimentally observed 1 H chemical shifts was excellent (Figure 3B), suggesting that the refined model of the zeta peptide/receptor complex (Figure 3C) represents a good approximation of the structure. In this model Pro16 packs between Trp87 and Trp110 on the receptor, and the N-terminal region of the peptide makes additional ˚ hydrophobic contacts near Leu158. A subsequent 3 Aresolution X-ray crystal structure of the complex between a closely related zeta peptide (e131) and FcεRIα was in very close agreement with the NMR-based model [6], confirming the validity of the approach.

NMR Analysis of Hairpin Peptide/Receptor Complex Resonances of the free 15-residue hairpin peptide, IGE32, were assigned by analysis of 2D 1 H/13 C-HSQC, 2D CBCA(CO)NH, and 2D 1 H/13 C-HCCH-TOCSY spectra [7]. Assignments of the receptor-bound state were obtained by analysis of 2D 1 H/13 C-HSQC and 3D 13 C-edited NOESY spectra. However, as observed for Pro16 of the zeta peptide, the resonances of the functionally important Pro9 were broad and did not show observable cross peaks in the 13 C-edited NOESY spectrum; assignment of these resonances was thus confirmed by analysis of a 2D 1 H/13 C-HSQC spectrum of a sample prepared with 13 C-Pro9 specifically labeled hairpin peptide generated by chemical synthesis. Similar dramatic chemical shift changes to those found for the zeta peptide were observed when the hairpin peptide was added to FcεRIα (Figure 3D). In particular the 1 H resonances of residues Thr6 and Pro9 were shifted upfield by 1.7–3.2 ppm upon binding the receptor. As for the zeta peptide, these large chemical shift changes imply that the hairpin peptide packs closely with aromatic rings in the complex. In this case, a total of 40 intermolecular NOE correlations involving peptide residues Asn1, Leu2, Pro3, Thr6, Val13, and Met15, were identified. These included NOE correlations between peptide

Part II

qshifts/qshifts.htm), in which the “structural” effects on chemical shifts due to contributions of ring currents of aromatic groups [9,10], magnetic anisotropy from peptide groups [11], and electrostatic effects arising from charges and dipoles [12], are calculated using empirical functions [13–15]. The total chemical shift of a given resonance is defined as the sum of the “random coil” value, as observed in short disordered linear peptides [16,17], and the conformation-dependent contributions (δ). The conformation-dependent contribution to the total chemical shift, as calculated by SHIFTS 4.1, is thus defined as:   δ = δtotal − δrandom coil ≈ δrc + δm  + δel + δconst (1)

Receptor Binding 1297

1298 Part II

Pharmaceutical Sciences

Part II

residues Val13 and Met15 and a pair of methyl groups on the receptor; as with the zeta peptide complex, the receptor methyl resonances could be assigned tentatively to Leu158. This assignment allowed for manual docking of the hairpin peptide so that the C-terminal residues were proximal to Leu158 on the receptor while, at the same time, Pro9 packs between Trp87 and Trp10 in the IgEbinding site. The resulting docked model was refined in a similar fashion to the refinement of the zeta peptide complex, as outlined above. In the case of the hairpin peptide calculations, however, the weak harmonic potential (force constant = 0.25 kcal/·mol) used to restrain the peptide structure was applied only during the initial iterations, and was never applied to peptide residues Asn1, Leu2, Thr6, Trp10, or Trp12. Conformational freedom of these residues was found to be necessary in order to satisfy the 1 H chemical shift restraints; in particular the side chain conformation of Trp10 changed significantly during the refinement. Stereospecific 1 H resonance assignments were possible for Pro9 1 Hβ and 1 Hδ, and Val13 1 Hγ, and using the same criteria as for the zeta peptide complex a total of 22 intermolecular distance restraints could be included in the final simulated annealing calculations. The final correspondence between the calculated and observed 1 H chemical shifts was again excellent (Figure 3E) indicating that the resulting model (Figure 3F) is a good approximation of the structure of the hairpin peptide/receptor complex. As was found for the zeta peptide complex, and consistent with original assumptions, the functionally important Pro9 is sandwiched between Trp87 and Trp110 on the receptor. Thr6 is proximal to Trp156 on the receptor, which accounts predominantly for the upfield shifts observed for resonances of this residue. Comparison of the zeta and hairpin peptide/FcεRIα complex models reveals that even though the phagederived peptides are structurally distinct they make remarkably similar contacts with the receptor. For instance, in addition to the zeta peptide Pro16 and the hairpin peptide Pro9 being “sandwiched” between Trp87 and Trp110 of the receptor, Pro4 and Phe6 of the zeta peptide occupy essentially the same space on the receptor as Val13 and Thr6, respectively, of the hairpin peptide (Figure 3C and F). Interestingly, the latter interactions are not utilized by the natural ligand, IgE, and thus, together with the proline pocket defined by Trp87 and Trp110, represent a novel binding epitope for IgE antagonists.

Conclusion In the examples presented above the readily obtained 1 H chemical shifts of peptide ligands were used to validate docked models of the peptide/receptor complexes that were based originally on assumption and limited experimental data, such as mutational analysis. More extensive

data, such as receptor chemical shift assignments, and consequently assigned intermolecular NOE correlations, were not available. The 1 H chemical shifts of the bound peptides, and in particular the large upfield shifts of 1 H resonances observed for some of the peptide residues, could be accounted for quantitatively by directly refining the models against the observed chemical shifts. Some caveats of the approach are worth mentioning. First, this is not a de novo docking method, but rather a docking validation method, and thus requires some degree of independent data in order to localize the binding site. Starting models could come from de novo docking algorithms, or could be based on experimental data such as obtained from mutational analysis, as used here. Secondly, reasonably accurate structural models of the ligand and receptor are necessary and conformational changes upon binding should be minimal (although as noted in the hairpin peptide example some degree of conformational flexibility can be accommodated and might be necessary in order to obtain a good fit). Finally, the most sensitive restraints of ligand orientation are the ring current contributions of aromatic groups. Thus, the presence of aromatic groups that cause significant perturbations in the ligand chemical shifts is likely a prerequisite for successful application of this approach.

References 1. Nakamura GR, Starovasnik MA, Reynolds ME, Lowman HB. Biochemistry. 2001;40:9828. 2. Nakamura GR, Reynolds ME, Chen YM, Starovasnik MA, Lowman HB. Proc. Natl. Acad. Sci. U.S.A. 2002;99:1303. 3. Garman SC, Wurzburg BA, Tarchevskaya SS, Kinet JP, Jardetzky TS. Nature. 2000;406:259. 4. Dennis MS, Roberge M, Quan C, Lazarus RA. Biochemistry. 2001;40:9513. 5. Reilly D, Fairbrother WJ. J. Biomol. NMR. 1994;4:459. 6. Stamos J, Eigenbrot C, Nakamura GR, Reynolds ME, Yin J, Lowman HB, Fairbrother WJ, Starovasnik MA. Structure. 2004;12:1289. 7. Cavanagh J, Fairbrother WJ, Palmer AG III, Skelton NJ. Protein NMR Spectroscopy: Principles and Practice. Academic Press: San Diego, 1995. 8. Zwahlen C, S´ebastien PL, Vincent JF, Greenblatt J, Konrat R, Kay LE. J. Am. Chem. Soc. 1997;119:6711. 9. Johnson CE, Bovey FA. J. Chem. Phys. 1958;29:1012. 10. Haigh CW, Mallion RB. Prog. NMR Spectrosc. 1980;13:303. 11. McConnell HM. J. Chem. Phys. 1957;27:226. 12. Buckingham AD. Can. J. Chem. 1960;38:300. ¨ 13. Osapay K, Case DA. J. Am. Chem. Soc. 1991;113:9436. ¨ 14. Osapay K, Case DA. J. Biomol. NMR. 1994;4:215. 15. Dejaegere AP, Bryce RA, Case DA. In: JC Facelli, AC de Dios (Eds). Modeling NMR Chemical Shifts. American Chemical Society: Washington, 1999, p 194. 16. Bundi A, W¨uthrich K. Biopolymers. 1979;18:285. 17. Merutka G, Dyson HJ, Wright PE. J. Biomol. NMR. 1995;5:14.

1299

Ronald Crouch Varian NMR Systems, NC 27513, USA

The advent of indirect detection has revolutionized the world of NMR spectroscopy in all areas of chemistry. For the purpose of assembling fragments of moieties as defined by NMR resonances into meaningful structural fragments or even complete structures, HMBC [1] is one of the most powerful tools available to the modern NMR spectroscopist. In situations of low sensitivity brought about most commonly by the availability of vanishingly small amounts of sample to analyze, direct observation of 13 C resonances can be difficult or impossible. Accordingly, the sensitivity afforded by indirect detection with HMBC and variants provides a tool that has the potential to access the key and difficult quaternary 13 C atoms while at the same time assembling the puzzle pieces into a coherent structure. The strength of 1 H detection, thereby casting the 13 C chemical shift information into the indirect dimension, is also the weakness of HMBC. This is especially true in the more difficult situations where there are a large number of 13 C resonances spread out over a large chemical shift range. At a 1 H observe frequency of 500 MHz, 225 ppm of 13 C chemical shift represents nearly 28,000 Hz of frequency space. Resolving two 13 C resonances separated by10 Hz would require ∼2800 t1 increments (or ∼1000 t1 increments with high enough signal to noise to allow reliable linear prediction) in F1! Region selection in the 13 C dimension with HMBC is a very old idea [2–3]. Utilization of excitation sculpting [4–5] added considerable robustness to the method. Combination of excitation sculpting with a simple Hadamard-encoding step [6] provides a simple and reliable methodology to greatly reduce the number of t1 increments required for a needed resolution. This encoding methodology affords a mean to either record lower resolution HMBC information in less time or much higher resolved spectra in equivalent time relative to traditional sampling methods. Consider the simplest two-element Hadamard matrix +, + and +, −. Addition of the two elements leads to one enhanced signal to noise in one region, whereas subtraction results in cancelation of the first region and enhanced signal to noise in the second region. Bear in mind

Graham A. Webb (ed.), Modern Magnetic Resonance, 1299–1304.
C 2008 Springer.

that with modern NMR spectrometers, given an actual C spectrum from which to build the needed shaped 13 C pulses, a Hadamard matrix can automatically be created to record and construct a complete 2D HMBC spectrum in n increments, where n represents the number of 13 C resonances [7–10]. A major advantage of using a simple two-element matrix is that the regions can be quite wide and general and not require a precise understanding of carbon chemical shifts. It is quite reasonable to simply divide the carbon regions by half, and immediately obtain a time saving of a factor of two, because by employing Hadamard-encoding, the number of t1 increments is cut in half to maintain exactly the same digital resolution as in a full single-region HMBC. The magic is in the encoding pulses and it is quite practical to create whatever is needed on the fly to obtain the two required pulses for a simple two-element matrix. Alternatively, a pair of pulses appropriate for a routine experiment with any desired frequency bounds can be created, and the pulse width and power parameters remembered for future use. A pulse sequence optimized for long-range 1 H–X correlation spectroscopy with an option for single or multiple X-band selection is presented in Figure 1 [11]. As an introduction to the illustration of the method, consider the gHMBC spectrum acquired with a sample of cyclosporine-A shown in Figure 2. The F1 window required to observe all 13 C resonances is not particularly large. A 13 C chemical shift range from 0 to 190 ppm is more than enough. Inspection of the data in Figure 2 reveals that is quite practical to consider splitting the F1 window into two ∼80 ppm regions; 0–80 and 105– 185 ppm. Simply acquiring a simultaneous dual F1 region-selected long-range 1 H/13 C chemical shift correlation experiment with the pulse sequence shown in Figure 1, immediately affords a time saving of a factor of two over the full single-region data set, simply because the number of t1 increments for equivalent digital resolution in the 13 C dimension can be cut in half. A direct comparison of a single-region gHMBC and the simultaneous dual F1 region-selected gHMBC variant is presented in Figure 3, panels A–D. The two data sets 13

Part II

Dual-Region Hadamard-Encoding to Improve Resolution and Save Time

1300 Part II

Part II

1

Pharmaceutical Sciences

φ4 H δ1

X

δ2

∆

1/nJch

φ5

φ3

T2

| T1 | S1

S2

gradient G4

G5

Fig. 1. Gradient coherence-selected pulse sequence for HMBC with adiabatic swept refocusing. Narrow bars denote 90◦ pulses. The wide bar and two selective pulses (S1 and S2) are 180◦ pulses with the t1 evolution time split by the 1 H 180. All of the data presented in this note were acquired with a constant adiabaticity cosine2 (WURST2i) 180◦ pulse [12–13]. Gradients G4 and G5 are for coherence selection and can be in ratios of −3/5 or 5/−3 for N or P type selection of 13 CHn resonances. The δ1 and δ2 delays in the low pass filter are optimized for 1Jmax of 165 Hz and 1 Jmin of 130 Hz, as described by Sorensen and co-workers [14]. The delay for long-range XH coherence selection () can be set as desired. The gradients in the low pass filter are in the ratios of 9, −6, and −3. The basic phase cycling is as follows: φ3 = x, x, −x, −x; φ5 = x, −x; φ4 = x, −x, −x, x. Pulses S1 and S1 are adiabatic shaped pulses with downfield to upfield inversion sweep.

were obtained in precisely the same time with 128 t1 increments in each. The dual-region data (panels B and D) were acquired with two scans per t1 increment (with the two required band-selected shapes) and the single-region data (panels A and C) used four scans. Clearly the digital resolution and overall quality of data are improved for the dual-region Hadamard-encoded data presented in Figure 3, panels B and D. A simple single-region band-selected experiment can be performed with the pulse sequence shown in Figure 1 to obtain extremely high digital resolution with efficient selection of the desired F1 region and rejection of other responses. Figure 4 shows a very high 13 C resolution long-range 1 H–13 C correlation experiment wherein the correlations between the protons and only the carbonyl 13 C resonances have been selected. The time to acquire the data was once again ∼20 min. Comparison with the data in Figure 2 (or Figure 3, panel C) reveals the much higher digital resolution afforded by confining the F1 sampling solely to the carbonyl region of the 13 C spectrum. At this point, it is useful to address the details of the experiment in terms of the requirements imposed by the band-selected pulses. A good means for illustration would be to graphically compare the pulses required for the acquisition of the single-region data presented in Figure 4, with those required to acquire the simultaneous dual-region data presented in the Figure 3 pan-

els. This comparison is shown in Figure 5. Application of a very narrow band 13 C inversion pulse pair, as illustrated in Figure 5A, affords the possibility of quite high 13 C digital resolution with relatively few t1 increments precisely because only the portion of the 13 C frequency space refocused by the narrow band pulse is detected. For simultaneous dual-region selection, it is necessary to interleave data acquisition using the two pulses graphically presented in Figure 5B and C. After data acquisition, 2DFT sum of the two interleaved data sets would result in full signal to noise for the upfield region of the 13 C frequency space, whereas 2DFT difference would give full signal to noise for the downfield 13 C frequency space. It is important to understand that the upfield and downfield regions are the result of two separate data processing steps that are graphically represented in Figure 5D and E. The method presented here is both robust and simple to implement. It is the author’s feeling that the full Hadamard-encoded techniques outlined in Refs. [7–10] may well represent the beginnings of a major revolution in the general approach to multi-dimensional NMR spectroscopy typically used in support of the elucidation of chemical structure in the pharmaceutical industry. It is hoped that the presentation of this simple two-step procedure serves to illustrate the general concepts.

Simultaneous Dual-Region gHMBC

Simultaneous Dual-Region gHMBC 1301

Part II

Fig. 2. Plot of the full single-region gHMBC acquired with a 2 mg sample of cyclosporine-A using an INOVA 600 with HCN Cold Probe. Data were acquired with four scans in each of 128 t1 increments in 22 min. A normal 1 H spectrum is on top and a projection showing the detected 13 C resonances is shown on the side. The boxed areas denote the regions to be selected in subsequent dual Hadamard-encoded experiments.

1302 Part II

Pharmaceutical Sciences

Part II F1 (ppm)

F1 (ppm)

20

20

30

30

40

40

50

50

60

60

70

8

B

70

A 7

6

5

4

3

2

1

7

8

6

F2 (ppm)

5

4

3

2

1

3

2

1

F2 (ppm)

F1 (ppm) 169.5

F1 (ppm) 169.5

170.0

170.0 170.5

170.5

171.0

171.0

171.5

171.5

172.0

172.0

172.5

172.5

173.0

173.0

173.5

173.5

174.0

174.0

C

174.5

D

174.5 175.0

8

7

6

5

4

F2 (ppm)

3

2

1

8

7

6

5

4

F2 (ppm)

Fig. 3. Panels A–D compare regions from both traditional gHMBC and simultaneous dual-region Hadamard-encoding using the pulse sequence described in Figure 1. Panels A and C are simply expansions from the normal gHMBC depicted in Figure 2, whereas panels B and D are the result of 2DFT sum or difference of Hadamard-encoded data acquired in a single experiment in exactly the same experimental time as the gHMBC data shown in panels A and C. The data in panels A and C (gHMBC) were acquired with four scans/t1 increment. The Hadamard-encoded data in panels B and D were acquired with two scans per increment for both of the region-selected encoding 13 C pulses.

Simultaneous Dual-Region gHMBC

Simultaneous Dual-Region gHMBC 1303

F1 (ppm) 170.0 170.5 171.0 171.5 172.0 172.5 173.0 173.5 174.0 174.5 175.0 8

7

6

5

4

3

2

1

F2 (ppm)

A

B

C

D B+C

E B-C

Fig. 5. Graphical representation of band-selection pulses required to acquire data in a two-step Hadamard matrix to generate the results shown in Figure 3. (A) For the single-region band-selection of carbonyl 13 C resonances described in Figure 4 only a very narrow refocusing of the 13 C carbonyl resonances is required. Figure 5B and C illustrate the pair of band-selected inversion pulses required for Hadamard-encoding. The anti-echo refocusing shown in the inverted portion of Figure 5C is achieved by application of a 90◦ phase shift for the band-selection of that π pulse portion of the 13 C frequency window. In Figure 5D and E the effect of processing the two encoded data sets, respectively, as sum and difference is illustrated. Figure 5D depicts 5B plus 5C, whereas Figure 5E depicts 5B minus 5C.

Part II

Fig. 4. Single C=O region-selected data acquired with the 2 mg sample of cyclosporine-A in 20 min. Data were acquired with two scans for each of the 256 t1 increments. The data were acquired and processed as a single phase with sinebell window functions.

1304 Part II

Pharmaceutical Sciences

Part II

References 1. Bax A, Summers MF. J. Am. Chem. Soc. 1986;108:2093. 2. Crouch RC, Martin GE. J. Magn. Reson. 1991;92:189. 3. Crouch RC, Spitzer TD, Martin GE. Magn. Reson. Chem. 1992;30:595. 4. Stott K, Stonehouse J, Keeler J, Hwang T-L, Shaka AJ. J. Am. Chem. Soc. 1995;117:4199. 5. Gaillet G, Lequart C, Debeire P, Nuzillard J-M. J. Magn. Reson. 1999;139:454. 6. Krishnamurthy K. J. Magn. Reson. 2001;153:144.

7. Kupce E, Nishida T, Freeman R. Prog. NMR Spectrosc. 2003;42:95. 8. Kupce E, Freeman R. J. Magn. Reson. 2003;162:158. 9. Kupce E, Freeman R. J. Magn. Reson. 2003;163:56. 10. Kupce E, Freeman R. J. Magn. Reson. 2003;163:158. 11. Crouch RC, Boyer R, Johnson R, Krishnamurthy K. Magn. Reson. Chem. 2003;42:301. 12. Kupce E, Wagner G. J. Magn. Reson. 1995;109B:329. 13. Kupce E, Freeman R. J. Magn. Reson. 1996;118A:299. 14. Meissner A, Moskay D, Nielson NC, Sorensen OW. J. Magn. Reson. 1997;124:245.

1305

Mark W. Maciejewski1 , Alan S. Stern2 , Glenn F. King1 , and Jeffrey C. Hoch1 1 Department

of Molecular, Microbial and Structural Biology, University of Connecticut Health Center, CT, USA 2 Rowland Institute at Harvard, Cambridge, MA, USA

Abstract The twin challenges of biomolecular NMR spectroscopy are sensitivity and resolution. Nonuniform sampling, in which data collection is tailored to meet the requirements of sensitivity or resolution, enables new approaches to meet these challenges. We describe the use of nonuniform sampling in biomolecular NMR and the use of maximum entropy reconstruction to process the data. Data collected using nonuniform sampling cannot be processed using conventional methods such as the discrete Fourier transform or linear prediction, because these methods require data sampled at uniform intervals. A two- to three-fold reduction in acquisition time can routinely be realized for each indirect time dimension that employs nonuniform sampling, making this approach a valuable complement to cryogenic probes and high-field magnets for the most demanding biomolecular NMR applications. The approach is especially well suited for high-throughput applications of NMR spectroscopy in drug discovery. The development of Fourier transform (FT) NMR spectroscopy by Ernst and Anderson [1] in the late 1960s set the stage for the subsequent development of multidimensional experiments [2–4]. By resolving individual nuclear coherences, these experiments enable systematic determination of the structure of biological macromolecules in solution. However, the need to avoid spectral aliasing when using the discrete FT (DFT) to compute frequency spectra from the time domain data imposes substantial data collection requirements; it is not unusual for each additional dimension to increase the amount of data that must be collected, and the time required to complete the experiment, by two orders of magnitude. The Nyquist theorem places a lower bound on the sampling rate, or time between samples (the dwell time), needed to avoid aliasing. This rate is inversely proportional to the width of the spectrum. Frequency resolution comparable to the natural line widths of the resonances is needed to exploit the dispersion of resonances at a given magnetic field, and high-resolution requires data collected at long acquisition times, in essence because the closer two resonances are in frequency, the longer it takes for a detectable difference in their signals to evolve. Since the DFT requires samples

Graham A. Webb (ed.), Modern Magnetic Resonance, 1305–1311.
C 2008 Springer.

collected at uniformly spaced intervals, acquiring data at long acquisition times requires collecting samples at all of the intervening multiples of the dwell time. The data collection requirements increase with increasing magnetic field, because the increased dispersion of resonances necessitates shorter dwell times. In the 1980s, linear prediction (LP) extrapolation of NMR data beyond the sampled interval was introduced [5,6], and it remains in common use today. LP extrapolation dramatically reduces the need for aggressive apodization when working with short data records, providing improvements in both sensitivity and resolution in favorable situations. However, LP has drawbacks; it has recently been shown that LP extrapolation generates falsepositive peaks when applied to noisy data or when used to extrapolate too far beyond the measured interval [7]. A more subtle but insidious defect is that LP extrapolation can cause slight frequency shifts of real signals in these situations. These shifts are not apparent on casual inspection but have potentially serious consequences, such as improper resonance assignments. Furthermore, the extent to which NMR data can be extrapolated using LP depends mainly on the signal-to-noise (S/N ) ratio, and consequently additional LP extrapolation is not a general solution to increased data requirements imposed by the increased spectral dispersion at higher magnetic fields. Conventional approaches employing the DFT, including LP extrapolation, require acquisition times that can be prohibitively long, especially for biomolecules that are fleetingly stable or for data collected at very high magnetic field. Thus, there is considerable impetus for developing alternative methods that can produce highresolution multidimensional spectra in less time. A wide variety of different approaches are actively being developed. One approach involves multiplexing, resulting in coherence frequencies that encode more than one resonance frequency in a single dimension. These include “reduced dimensionality” experiments such as G-matrix FT (GFT) [8,9] and Hadamard encoding [10]. A drawback of these approaches is that they require special radiofrequency pulse sequences as well as additional post-processing steps to disentangle the multiplexed

Part II

Nonuniform Sampling in Biomolecular NMR

1306 Part II

Pharmaceutical Sciences

Part II

frequencies. A different approach to multiplexing has been developed by Frydman and colleagues [11], which utilizes orthogonal magnetic field gradients to partition the sample into “subsamples,” each reporting a different part of the spin response. This approach is not practical when sensitivity is already a limiting factor for measuring the response of the entire sample, but as advances in probe technology continue to improve sensitivity and gradient strength (which determines the resolution), it will likely find increasing application in biomolecular NMR. Parametric methods that model the NMR signal as a sum of exponentially decaying sinusoids, such as LP singular value decomposition [12], Hankel SVD [13], filter diagonalization method (FDM) [14], maximum likelihood [15], and Bayesian [16] methods are in principle capable of producing high-resolution spectra from very short data records. These methods often exhibit “spontaneous splitting,” resulting in multiple sinusoids for a single resonance when applied to noisy or nonideal (non-Lorentzian) data, complicating the parametric interpretation. Indeed the regularized resolvent transform [17], in which an FDM-like approach is used to extrapolate the signal, rather than compute a parametric decomposition, was developed to avoid this problem. Another class of methods utilizes symmetries or redundancy in multidimensional spectra to avoid acquiring all combinations of the evolution times for the different dimensions. An approach inspired by the methods used in medicine to reconstruct images in computerized axial tomography exploits the relationship between line projections in the frequency domain and linear cross sections in the time domain. Called back-projection reconstruction [18], this approach can, in favorable circumstances recover a complete two-dimensional spectrum from a small number of cross sections through the full time domain data matrix. However, it has been pointed out that in the general case n cross sections are needed to unambiguously resolve n peaks in multiple dimensions [19], and for complex biomolecules, this may not result in significant time savings. Another approach, called three-way decomposition [20], exploits the fact that line shapes in multiple dimensional spectra can be decomposed into products of one-dimensional line shapes. Back-projection reconstruction and three-way decomposition are special cases of nonuniform sampling in the time domain. In principle, any method that approaches spectrum reconstruction as an inverse problem, that is, computes spectra and inverts the spectrum to obtain “mock” data that are then compared with the empirical data for consistency, can be used with nonuniformly sampled time domain data. A very general method that makes no assumptions regarding the nature of the signals, and imposes no particular constraints on the subset of the time domain data matrix that must be collected, is maximum entropy (MaxEnt) reconstruction [21]. It has

been shown recently to be particularly robust, yielding fewer false positives and more accurate frequencies than LP extrapolation [7]. It is also versatile: it can be used to construct spectra from the same data utilized by backprojection reconstruction, or can be used in combination with other methods, such as GFT (Jim Sun and Gerhard Wagner, personal communication). In principle, it can be used with any pulse sequence, and the computed spectra are compatible with conventional analysis tools developed for Fourier spectra. In this chapter, we describe the use of nonuniform sampling and MaxEnt reconstruction in multidimensional NMR experiments and its application to biomolecules. In practice, factors of two to three savings in data acquisition time per indirect time dimension are readily achieved. The robustness and versatility of MaxEnt reconstruction make it a powerful tool for improving sensitivity and resolution, and saving time in biomolecular NMR experiments.

MaxEnt Reconstruction The MaxEnt reconstruction of the spectrum of a complexvalued time series d is the spectrum f which maximizes the entropy S(f), subject to the constraint that the mock data m, given by the inverse DFT of the spectrum, is consistent with the time series d. Consistency is defined by the condition C(f, d) ≤ C0

(1)

where C(f, d) is the unweighted chi-squared statistic, C(f, d) =

M−1 

|m i − di |2 =

i=0

M−1 

|IDFT(f )i − di |2 ,

(2)

i=0

and C0 is an estimate of the noise level; IDFT is the inverse DFT. A definition of the entropy S(f) applicable to complex-valued spectra can be derived from consideration of the entropy of an ensemble of spin-1/2 particles [22] or by extending the Shannon information formula to the complex domain along with suitable differentiability conditions [23]. The two approaches yield equivalent formulations for the entropy of a complex spectrum  ⎛ ⎞   N −1 | | de f + 4 + | f n |2 de f 2 f  n | fn | ⎠ S(f) = − log ⎝ de f 2 n=0   − 4 + | f n |2 de f 2 ,

(3)

where def is a scale factor. In principle, the quantummechanical derivation prescribes the value of def (it depends on the sensitivity of the spectrometer and the

Nonuniform Sampling

O = S(f ) − λC (f, d) ,

(4)

where the value of the Lagrange multiplier λ is adjusted to obtain C = C0 . The value of λ depends on the values of the parameters def and C0 , and on the data. Practical guidelines for choosing the values of def and C0 are described elsewhere [24]. The formulas for entropy and constraint readily generalize to multidimensional data. However, because MaxEnt reconstruction is nonlinear, MaxEnt must be the last timeto-frequency transformation applied. So if MaxEnt is to be used in t1 , then the t2 dimension must be transformed to the frequency domain first. And if MaxEnt is to be used for both t1 and t2 , then it must be applied to both dimensions simultaneously, rather than one at a time. The IDFT operation appearing in Equation (2), and the corresponding sums in Equations (2) and (3), can be applied along as many dimensions as one likes. Applying MaxEnt to reconstruct separately each t1 row, or each t1 –t2 plane, of a three-dimensional spectrum raises another problem. The extent of the nonlinearity of the reconstructions can vary from row to row (or from plane to plane), resulting in distorted peak shapes. One way to avoid this is to process the entire data set as a single unit, extending the sums in Equations (2) and (3) to cover the entire spectrum, rather than working on a single row or plane at a time. Of course, such an approach will entail drastically increased data storage requirements. A simpler solution is to use a fixed value of λ rather than iterating to attain C = C0 [25]. The correct value for λ can be determined by choosing a representative row, then computing the normal MaxEnt reconstruction with an appropriate value for C0 . In quantitative applications, it is important to use one of these methods—full data set MaxEnt reconstruction or constant λ reconstruction— to assure that the nonlinearity is uniform across the spectrum, and thus it can be calibrated in a reliable way. The nature of the nonlinearity of MaxEnt reconstructions is fairly well understood—in general, MaxEnt reconstruction tends to scale intensities down, compared to the DFT, with small amplitudes scaled down more than large amplitudes [26]. In addition to requiring calibration when used for quantitative applications, such as measuring nuclear Overhauser effects, the nonlinearity of MaxEnt reconstruction heightens the distinction between sensitivity, which is the ability to distinguish peaks from noise, and signal-to-noise ratio (S/N ). For MaxEnt reconstruction, S/N is not a reliable indicator of sensitivity.

Nonuniform Sampling The matched filter is a function that follows the decay envelope of the FID; multiplying the FID by the matched filter results in optimal S/N in the DFT spectrum [27]. In essence, the matched filter places more emphasis on the portion of the data where the S/N (in the time domain) is highest. The same idea can be applied to sampling in the time domain: instead of sampling at uniform intervals, the signal can be sampled more frequently when it is strong, and less frequently when it is weak [28–32]; we call this nonuniform sampling. Nonuniform sampling prevents the use of LP or the DFT, since they require data sampled at uniform intervals, but MaxEnt handles such data because the chi-squared statistic in Equation (2) can be computed from just the collected samples. Nonuniform sampling provides additional flexibility in balancing the trade-offs between acquisition time, resolution, and sensitivity. The potential for time saving accrues in the indirect time dimensions of multidimensional experiments, since the relaxation delay between transients renders the total acquisition time insensitive to the number of samples acquired in the direct time dimension. In contrast, the total experiment time increases in direct proportion to the number of indirect time samples. A list of evolution times for which data are acquired is called a sampling schedule. Examples of nonuniform sampling schedules are depicted in Figure 1. A uniform grid is used to signify evenly spaced points in the various time dimensions. A large dot indicates a sampled time interval and small dots indicate times not sampled. Panel A illustrates an exponential sampling schedule in one indirect dimension (t1 ), and panel B illustrates an exponential sampling schedule in two indirect dimensions. Panel C also illustrates a schedule employing randomly selected points chosen from a two-dimensional exponential distribution; this helps to suppress coherent artifacts that can be introduced by schedules that use exactly the same time values for successive rows or columns (such as the schedule in panel B). Panel D illustrates a schedule appropriate for reduced sampling of constant-time dimensions, where there is no decay. For COSY-type spectra, an optimal sampling schedule can be constructed from a sine-modulated exponential distribution. In principle, it is possible to construct more complex sampling schedules that make explicit consideration of the dimensionality of the experiment or the distribution of expected frequencies, for example, along the lines of the samples used for back-projection reconstruction.

Example Applications Comparisons of spectra obtained using conventional linear sampling and nonuniform sampling applied to two

Part II

number of spins in the sample), but it is more convenient to treat def as an adjustable parameter. Essentially, it determines the scale at which the nonlinearity of MaxEnt becomes pronounced. The MaxEnt solution is found by maximizing the objective function

Example Applications 1307

1308 Part II

Pharmaceutical Sciences

Part II Fig. 1. Nonuniform sampling schedules in one and two indirect time dimensions. Large and small dots indicate times for which samples are collected and not collected, respectively. The schedules depicted are (A) exponential in t1 and linear in t2 , (B) exponential in both t1 and t2 , (C) randomly sampled from an exponential distribution in t1 and t2 , and (D) random in t1 and t2 .

experiments important for pharmaceutical applications of protein NMR are shown in Figures 2 and 3. Figure 2 shows 15 N–1 H HSQC spectra of a 42-residue atracotoxin from the Australian funnel-web spider Hadronyche versuta. Panel A is the spectrum computed using LP extrapolation, shifted sine-bell windowing, and DFT to produce a 1024-point spectrum in f 1 from 512 uniformly spaced samples in t1 . Panel B was computed in the same fashion and to the same size in f 1 , using 64 samples in t1 . The spectrum in panel C was computed using MaxEnt

reconstruction, using 64 t1 samples with exponentially increasing time between samples. Panels D, E, and F show an expanded region of panels A, B, and C, respectively. The resolution obtained from 64 t1 samples using MaxEnt reconstruction and nonuniform sampling is comparable to that obtained using uniform sampling and conventional processing; however, data acquisition required a factor of eight less time. This time saving can be important for highthroughput applications such as screening compound libraries for molecules that bind to a protein target.

Nonuniform Sampling

Example Applications 1309

Part II

Fig. 2. 15 N–1 H HSQC spectra for a 42-residue atracotoxin. (A) Computed using LP extrapolation and DFT from 256 linearly spaced samples in t1 , (B) using LP extrapolation and DFT from 64 linearly spaced samples in t1 , and (C) using MaxEnt reconstruction from 64 exponentially spaced samples in t1 . Panels D, E, and F depict an enlarged region from panels A, B, and C, respectively.

Extension of nonuniform sampling and MaxEnt reconstruction to two indirect time dimensions is straightforward, enabling the acquisition of triple resonance spectra for sequential assignment of proteins in 1 day [33]. Experiments in which the signals of interest span a wide range of intensities (such as those involving nuclear Overhauser effects) remain a challenge. The very nature of the nonlinearity of MaxEnt reconstruction, which is to scale down weak signals more than strong signals, makes it more difficult to choose parameters of the reconstruction (def and C0 ) that enable weak signals to be distinguished from noise. Nonuniform sampling can introduce additional nonlinearities. An example is shown in Figure 3, in which exponential nonuniform sampling and MaxEnt reconstruction have been applied to the two indirect dimensions of a 3D 15 N–1 H HSQC-NOESY experiment. Panels A and B shows f 1 – f 3 cross sections at the f 2 frequency 120.27 ppm, close to the amide 15 N chemical shift of residue Cys 21 (120.24 ppm); A is from the linearly sampled DFT spectrum and B is computed using

MaxEnt reconstruction from data containing one-eighth fewer samples. Remarkably, the MaxEnt spectrum exhibits resolution superior to the spectrum computed from the larger linearly sampled data set. This is evidenced by the disappearance of resonances at the f 3 frequencies of 8.69 and 8.22 ppm, which correspond to residues with amide 15 N chemical shifts of 120.53 and 120.30 ppm, respectively, which have maxima in adjacent f 2 planes. However, the nonlinearity (which depends on the values chosen for def and C0 , as well as the sampling schedule) in this instance makes it difficult to observe several weak peaks. Panels C and D are linear cross sections from panels A and B at the f 3 frequency of the amide proton of Cys 21, scaled vertically to the height of the maximum in each panel. While the noise level is greatly reduced, so are the weak peaks, making clear the difficulty caused by high dynamic range. While careful choice of the nonuniform sampling schedule and parameters of the MaxEnt reconstruction can in principle permit observation of the missing weak

1310 Part II

Pharmaceutical Sciences

Part II Fig. 3. Two-dimensional cross sections at f 2 = 120.27 ppm from 3D 15 N–1 H HSQC-NOESY spectra of a 42-residue atracotoxin. (A) Computed using LP extrapolation and DFT, (B) using nonuniform sampling in t1 and t2 and MaxEnt reconstruction. (C) and (D) are linear cross sections from panels A and B at the f 3 frequency of the amide proton of Cys 21 (9.34 ppm).

Nonuniform Sampling

Concluding Remarks MaxEnt reconstruction is a particularly versatile method of spectrum analysis, capable of providing highresolution spectral estimates from short data records. Its ability to determine spectra from data sampled at arbitrary times enables the use of nonuniform sampling to improve resolution or reduce data acquisition time. Model-based methods, such as Bayesian analysis or FDM, offer many of the same capabilities but are not suited to the computation of spectra containing non-Lorentzian lines. The availability of efficient algorithms and powerful computers has removed what were once perceived as significant barriers to the widespread application of MaxEnt reconstruction in NMR. The ability to collect high-resolution spectra in substantially less time not only enables more efficient use of expensive spectrometers, but also enables investigation of biomolecular systems that are fleetingly stable. These capabilities are particularly well suited to the demands of applications of NMR in drug discovery, including high-throughput screening and rapid assignment of protein spectra.

Acknowledgments We thank Drs. Peter Schmieder, David Rovnyak, and Gerhard Wagner for sage advice. Financial support was provided by the US National Institutes of Health (GM47467), the US National Science Foundation (MCB 9316938 and MCB 0234638), and The Rowland Institute at Harvard.

References 1. Ernst RR, Anderson WA. Rev. Sci. Instrum. 1966;37:93– 102. 2. Jeener J. Oral Presentation, Ampere International Summer School: Yugoslavia, 1971. 3. Aue WP, Bartholdi E, Ernst RR. J. Chem. Phys. 1976; 64: 2229–46.

4. Oschkinat H, Griesinger C, Kraulis PJ, Sorensen OW, Ernst RR, Gronenborn AM, Clore GM. Nature. 1996;332:374–6. 5. Zeng Y, Tang J, Bush CA, Norris JA. J. Magn. Reson. 1989;83:473–83. 6. Ni F, Scheraga HA. J. Magn. Reson. 1986;70:506–11. 7. Stern AS, Li K-B, Hoch JC. J. Am. Chem. Soc. 2002;124: 1982–93. 8. Kim S, Szyperski T. J. Am. Chem. Soc. 2003;125:1385– 93. 9. Ding K, Gronenborn AM. J. Magn. Reson. 2002;156:262–8. 10. Kupce E, Freeman R. J. Magn. Reson. 2003;162:300–10. 11. Frydman L, Scherf T, Lupulescu A. Proc. Natl. Acad. Sci. USA 2002;99:15858–62. 12. Barkhuijsen H, de Beer R, Bovee WM, Creyghton JH, van Ormondt D. Magn. Reson. Med. 1985;2:86–9. 13. de Beer R, van den Boogaart A, van Ormondt D, Pijnappel WW, den Hollander JA, Marien AJ, Luyten PR. NMR Biomed. 1992;5:171–8. 14. Mandelshtam VA, Taylor HS, Shaka AJ. J. Magn. Reson. 1998;133:304–12. 15. Chylla RA, Markley JL. J. Biomol. NMR 1995;5:245–58. 16. Kotyk JJ, Hoffman NG, Hutton WC, Bretthorst GL, Ackerman JJH. J. Magn. Reson. 1995;A116:1–9. 17. Chen J, Shaka AJ, Mandelshtam VA. J. Magn. Reson. 2000; 147:129–37. 18. Kupce E, Freeman R. J. Am. Chem. Soc. 2003;125:13958–9. 19. Coggins BE, Venters RA, Zhou P. J. Am. Chem. Soc. 2004; 126:1000–1. 20. Orekhov VY, Ibraghimov I, Billeter M. J. Biomol. NMR 2003; 27:165–73. 21. Wernecke SJ, D’ Addario LR. IEEE Trans. Comp. 1977;26: 351–64. 22. Daniell GJ, Hore PJ. J. Magn. Reson. 1989;84:515–36. 23. Hoch JC, Stern AS, Donoho DL, Johnstone IM. J. Magn. Reson. 1990;86:236–46. 24. Hoch JC, Stern AS. Methods Enzym. 2001;338:159–78. 25. Schmieder P, Stern AS, Wagner G, Hoch JC. J. Magn. Reson. 1997;125:332–9. 26. Donoho DL, Johnstone IM, Stern AS, Hoch JC. Proc. Natl. Acad. Sci. USA 1990;87:5066–8. 27. Ernst RR. Adv. Magn. Reson. 1966;2:1–135. 28. Barna JCJ, Laue ED. J. Magn. Reson. 1987;75:384–9. 29. Barna JCJ, Laue ED, Mayger MR, Skilling J, Worrall SJP. J. Magn. Reson. 1987;73:69–77. 30. Schmieder P, Stern AS, Wagner G, Hoch JC. J. Biomol. NMR 1993;3:569–76. 31. Schmieder P, Stern AS, Wagner G, Hoch JC. J. Biomol. NMR 1994;4:483–90. 32. Robin M, Delsuc MA, Guittet E, Lallemand JY. J. Magn. Reson. 1991;92: 645–50. 33. Rovnyak D, Frueh DP, Sastry M, Sun Z-Y, Stern AS, Hoch JC, Wagner G. J. Magn. Reson., 2004;170:15–21. 34. Rovnyak D, Hoch JC, Stern AS, Wagner G. J. Biomol. NMR. 2004;30:1–10.

Part II

peaks (for example, using a small value for C0 decreases the nonlinearity of the reconstruction), determination of the appropriate values is by no means straightforward, and remains the subject of active research [34]. Regardless of the application, the nonlinearities make error analysis an important adjunct to nonuniform sampling and MaxEnt reconstruction.

References 1311

Part II

Applications

1315

Leonard T. Nguyen, Elmar J. Prenner, and Hans J. Vogel Structural Biology Research Group, Department of Biological Sciences, University of Calgary Alberta, Canada T2N 1N4

Introduction

Solution Structures of Antimicrobial Peptides

Antimicrobial peptides are ubiquitous in nature and they aid in host defenses against bacterial infections. While initial discoveries of bacterial peptides go as far back as the 1950s, the 1990s saw an explosion in research to isolate novel peptides from all kingdoms of life. Out of the more than 800 eukaryotic peptides reported in the Antimicrobial Sequences Database (found at www.bbcm.units.it/∼tossi/pag1.htm), more than 700 entries were reported since 1990. The rapid growth of this field is fueled by the potential of these compounds to be used in a clinical setting [1,2] and replace commonly used antibiotics. The latter are losing potency against diseasecausing bacterial strains that are increasingly becoming resistant. The unique qualities of the various antimicrobial peptides make them an incredibly diverse family in terms of sequence, structure and by extension, mechanism of action [3]. Typically shorter than 50 amino acid residues, antimicrobial peptides are prime candidates for structural study by NMR spectroscopy. Cyclic and disulfide cross-linked peptides usually have a stable structure in aqueous solution. However, many of the linear peptides do not form defined structures in aqueous solution and do so only in a membrane mimetic environment [4]; this finding supports the idea that peptide-lipid interactions are critical for antimicrobial action and/or insertion [5]. In addition to solution structures calculated mostly from standard twodimensional TOCSY and NOESY 1 H experiments [6], much progress has been made to characterize the insertion and orientation of the peptides in membrane bilayers through solid-state NMR measurements [7]. This chapter will examine the progress in structural studies of antimicrobial peptides by NMR. Recent reviews have described the solution structures of antimicrobial peptides in an attempt to elucidate general themes and mechanisms [8,9]. Therefore to avoid overlap in this chapter we will highlight the versatility of NMR as a tool to characterize the solution structures of antimicrobial peptides which possess some unique features. In addition, we will describe some solid-state NMR studies which focus on several well known peptides.

To date, more than 100 unique three-dimensional structures of antimicrobial peptides have been solved by NMR. The complete list consists of more than 80 parent peptides and roughly 20 synthetic peptide derivatives, resulting in great variability. Table 1 summarizes the information for selected solution structures of peptides that have been studied because of their potential as bactericidal agents. In order to be useful, such peptides cannot have high toxicity toward mammalian cells. Some representative structures are shown in Figures 1 and 2. Well-characterized antimicrobial peptides include magainin 2 from the skin of the African clawed frog (Figure 1A) [10], protegrin-1 from mice (Figure 1B) [11], and several members of the defensin family (Figure 2). Some rather distinct antimicrobial peptides have been included in Table 1 as well: These are human hepcidin (Figure 1C) [12], which is better known as a peptide involved in the regulation of iron absorption [13], three gramicidin S derivatives [14], which are listed in spite of the fact that the parent gramicidin S compound, which is secreted from a Bacillus species [15], possesses significant hemolytic activity, and a completely synthetic peptide, combi-1 [16], that was obtained by screening from a library of hexa-peptides produced by combinatorial chemistry. Various structural classification schemes have been used in the past ([8] for example). However, given the wide diversity of the peptides, a satisfactory organization scheme can misrepresent the properties of certain peptides. Nevertheless, the eukaryotic cationic peptides are often classified into broad groups according to the secondary structure of the parent peptide. As such researchers often talk about helical, β-sheet, turn or extended peptides. This nomenclature can be somewhat misleading at times as some of the larger peptides should really be considered as mini-proteins with various types of secondary structure. For example, insect defensin A [17], as a defensin, is grouped as a β-sheet peptide even though there is an α-helix present in the structure in addition to several β-strands. It should also be noted that sometimes the structure of a synthetic derivative deviates from that of the parent peptide. This is seen for a shortened derivative of

Graham A. Webb (ed.), Modern Magnetic Resonance, 1315–1323.
C 2008 Springer.

Part II

Structural Characterization of Antimicrobial Peptides by NMR Spectroscopy

1316 Part II

Pharmaceutical Sciences

Part II

Table 1: Examples of NMR derived high-resolution solution structures of antimicrobial peptides Membrane Mimetica

Peptide

Class

Source

Cecropin A Cecropin A-Magainin 2 hybrid (CA-MA) (4)b Cecropin A-Melittin hybrid (CA-ME) Magainin 2 (1) Ovispirin-1 (2) PGLa Sheep myeloid antimicrobial peptide (SMAP-29) Bovine lactoferricin (1) Drosomysin γ -1-H thionin Hepcidin-20 and -25 Human β defensin-3 (HBD-3) Human neutrophil protein-1 (HNP-1) Insect defensin A/Phormicin NaD1 floral defensin Protegrin-1 (2) Rabbit kidney defensin 1 (RK-1) Rhesus θ defensin-1 (RTD-1) Tachyplesin I (3) Tachystatin A Ac-RRWWRF-NH2 (Combi-1) Indolicidin (1) Tritrpticin NK-lysin Thanatin

α-helix α-helix

HFIP DPC, None

Hyalophora cecropia Synthetic

[87]

Ref.

α-helix

DPC

Synthetic

[88]

α-helix α-helix α-helix α-helix

DPC, SDS,TFE TFE DPC TFE

Xenopus laevis Synthetic (from Ovis aries) Xenopus laevis Ovis aries

[10]([90])

β-sheet β-sheet β-sheet β-sheet β-sheet β-sheet

None None None None None None

Bos taurus Drosophila melanogaster Hordeum vulgare Homo sapiens Homo sapiens Homo sapiens

[19]([18])

β-sheet β-sheet β-sheet β-sheet β-sheet β-sheet β-sheet Extended Extended Turn Loop Loop

None None None, DMSO None Acetonitrile None None SDS SDS, DPC SDS None None

Protophormia terraenovae Nicotiana alata Sus scrofa Oryctolagus cuniculus Macaca mulatta Tachypleus tridentatus Tachypleus tridentatus Synthetic Bos taurus Synthetic (from Sus scrofa) Homo sapiens Podisus maculiventris

[17]

Bacteriocins Carnobacteriocin B2 Gramicidin S derivatives (3) Microcin J25 Nisin A

α-helix Loop Loop Turn

TFE None None DPC, SDS

Carnobacterium piscicola Synthetic (from Bacillus brevis) Escherichia coli Lactococcus lactis

[103]

[88]([89])c

[91]([91]) [92] [93]

[41] [42] [12] [44] [39]

[94] [11]([95,96]) [97] [47] [98]([98]) [99] [16] [21]([20]) [100] [101] [102]

([14]) [36,61,62] [30]

a

Abbreviations used for membrane mimetic present are as follows: TFE, trifluoroethanol; DPC, dodecylphosphocholine micelles; HFIP, hexafluoroisopropanol; SDS, sodium dodecyl sulfate micelles, and DMSO, dimethyl sulfoxide. b Number in parentheses denotes the number of derivatives from a parent peptide for which structures are available. c Number in parentheses in the reference column refers to the structures of the derivatives.

bovine lactoferricin [18] which forms an extended peptide structure even though intact bovine lactoferricin itself [19] adopts a β-sheet conformation. Also, a cyclized derivative [20] of indolicidin [21] has quite a different structure than its parent compound. The prevalent theme among all eukaryotic cationic peptides is a structure with an amphipathic character. Thus the vast majority of the peptides have a preponderance of basic and hydrophobic amino acids, yet they have no

sequence homology. The cationic side of the structure is responsible for electrostatic interactions with the bacterial membranes, which have a high abundance of negatively charged phosphatidylglycerol phospholipids head groups. The charge-charge interactions are thought to be an important factor in providing long-range attraction between the peptide and bacterial membranes [3]. This mechanism allows for broad specificity in target cells while retaining low toxicity to eukaryotic membranes, which mostly have

NMR of Antimicrobial Peptides

N

N C Cys9

Cys8

Glu8 C N N

C

Fig. 1. Representative solution structures of antimicrobial peptides presented as ribbon diagrams: (A) magainin 2 (PDB ID 2MAG) in TFE solution, (B) protegrin-1 (PDB ID 1PG1), (C) hepcidin-20 showing the vicinal disulfide link between Cys8 and Cys9 (PDB ID 1M4E), and (D) microcin J25 showing the amide link between the sidechain of Glu8 and the N-terminal amino group (PDB ID 1Q71). These figures were prepared using MOLMOL [104]. (See also Plate 100 on page XX in the Color Plate Section.)

zwitterionic phosphatidylcholine head groups. However, a few anionic peptides from eukaryotic sources have been identified, for example in human lungs where Zn2+ chelation is required for activity [22], or from the toad Bombina maxima where the activity does not require the presence of cations [23]. The hydrophobic portion of the antimicrobial peptides allows them to interact with the acyl chain C

A

B C

N

N

C N C

C

N

Fig. 2. Solution structures of some defensin peptides in ribbon diagram: (A) bovine neutrophil β-defensin-12 (BNBD-12) (PDB ID 1BNB), (B) insect defensin A (PDB ID 1ICA), and (C) rhesus θ-defensin (RTD-1) (PDB ID 1HVZ). These figures were prepared using MOLMOL [104]. (See also Plate 101 on page XX in the Color Plate Section.)

region of membrane bilayers. The amphipathicity that balances both facets of the peptides plays a key role in target selectivity, as peptides that have greater hydrophobicity usually act as general toxins, such as the bee venom melittin [24]. Almost all α-helical linear peptide structures were solved using either a micelle suspension or an organic co-solvent environment because the vast majority does not form a defined structure in aqueous solution. Highly represented in this category are the peptides obtained from the skins of frog or toad species. β-sheet peptides meanwhile, which encompass the mini-protein defensin family, come from a variety of plants and animals and are known for their characteristic disulfide linkages between cysteine residues [25,26]. Highly disulfide bridged peptides quite frequently act as antimicrobial peptides because the crosslinking would reduce their flexibility and promote stability against proteolytic enzymes. In contrast to the αhelical peptides, structures of the β-sheet peptides have mostly been solved in an aqueous environment and in fact, binding of protegrin-1 to lipid bilayers resulted in aggregation, which prevented detailed structural analysis [27,28]. Highly oligomerized peptides on membranemimetic surfaces, while possibly more representative of in vivo behavior, unfortunately pose difficulties in structure elucidation. Also included in Table 1 are a few antimicrobial peptides from bacterial origin. Bacteria use these to protect themselves from unrelated microbes in their ecological niche. These “biological warfare” peptides contain many unusual amino acids which are either the result of extensive post-translational modifications of ribosomally synthesized peptides (e.g. lantibiotics and microcins) or they are synthesized non-ribosomally on large peptide synthesis complexes (e.g. peptaibols). These peptides, often referred to as bacteriocins [29], have been the focus of many studies and nisin A [30] from Lactococcus lactis particularly has received a lot of attention to date, because this peptide is currently widely used as a food preservative [32]. Moreover, it is active at lower concentrations than the cationic antimicrobial peptides. The more potent activity could be explained when it was discovered that nisin, and related lantibiotics, form a specific complex with phospholipid II, a precursor for bacterial cell wall biosynthesis [33,34]. Very recently, an NMR paper was published showing the complex structure of nisin with lipid II (Figure 3) [35]. This is a unique contribution as lipid/peptide interactions are usually too transient to experimentally capture their complex structures. The reader is encouraged to read this intriguing paper describing the “pyrophosphate cage” structure for the nisin/lipid II complex. The solution structure of microcin J25 (Figure 1D) [36], another bacterial peptide with uncommon biochemical properties, will be discussed later in this chapter.

Part II

C

Solution Structures of Antimicrobial Peptides 1317

1318 Part II

Pharmaceutical Sciences

Part II Fig. 3. The contact region between nisin and 3LII, a lipid II variant, as determined from the NMR solution structure of the complex in dimethyl sulfoxide. (A) The N-terminal part of nisin (shown in van der Waals surface) encages the two phosphate groups (PA and PB ) of 3LII, which also contains the amino sugar N-acetylmuramic acid (MurNAc) and an isoprene tail. The side chains of nisin are labeled, with the dehydrated residues dehydroalanine, Dha, and dehydrobutyrine, Dhb. (B) The intermolecular hydrogen bond network between the nisin backbone and the 3LII pyrophosphate group (in spheres). Hydrogen bonds with high occurrence in the ensemble of structures are indicated by dashed lines, with corresponding residues labeled (with Abu representing α-aminobutyric acid). An arrow indicates the position of Ala3-Cβ, at which bioactivity is disrupted by the addition of a methyl group. The figures were taken with permission from Hsu et al. [35]. (See also Plate 102 on page XXI in the Color Plate Section.)

Defensins; Structure Determination and Dimerization Properties The defensins (Figure 2) are cationic antimicrobial miniproteins that play a key role in the innate immune defense of organisms. Major families are found in plants, insects, and vertebrates including humans. They are distinguished by the position and cross-linking patterns of six disulfide-bonded cysteine residues. A three-stranded β-sheet structure is found in most of the defensins. Plant and insect defensins have their own distinct disulfide connectivity patterns while the disulfides of the mammalian defensins [26] are separated into three major classes: the α- and β-defensins and the more recently discovered θ-defensins. Along with the cathelicidins [37], the defensins make up the major family of antimicrobial peptides produced in humans. They are of great biological importance as they can also play a key role in stimulating the adaptive immune response, substituting for chemokines [28]. They are found on the skin and other epithelial linings and in phagocytic neutrophils [26]. Defensins are of a small enough size (30–50 amino acids) that their proton NMR spectra can be fully assigned by regular TOCSY and NOESY NMR spectroscopy [6]. The first structures solved were those of rabbit neutrophil defensin-5 (NP-5) [38], NP-2 [39], and human neutrophil protein-1 (HNP-1) [38]. These early studies revealed a typical monomeric defensin structure with an amphipathic triple stranded antiparallel β-sheet. The structure was further characterized by hairpin formation and several other tight turns which appear to be stabilized by the disulfide bonds. The structure of the bovine neutrophil β-defensin-12 (Figure 2A) was also shown to be similar

to the previously studied α-defensins [40]. While the structure of insect defensin A (Figure 2B) has similar features to the mammalian defensins, it contains an additional “cysteine-stabilized α-helix” (CSH) motif where the α-helix is stabilized by disulfide bridging to the nearby antiparallel β-sheet [17]. This CSH motif is also found in other insect defensins such as in drosomycin from Drosophila melanogaster [41] as well as in a few plant defensins such as the γ-1-H and γ-1-P thionins from Hordeum vulgare and Triticum turgidum, respectively [42]. While the majority of the structures of the defensins show monomeric molecules, a few human peptides are thought to dimerize at low concentrations in aqueous solution. The rationale for this is unclear but it may be a firststep toward oligomerization which is required for pore formation in membranes. For example, the crystal structure of human neutrophil protein-3 (HNP-3) [43] showed an amphipathic symmetrical dimer of peptides. Diffusion NMR and dynamic light scattering, which were used to estimate the approximate size of human β-defensin3 (HBD-3) in solution, also suggest the presence of a dimeric complex for this peptide, but not for other βdefensins [44]. Distinct chemical shifts for a region within the first β-strand of HBD-3 indicate that it is likely involved in intermolecular β-sheet formation. However, unambiguous NOE crosspeaks in the spectra could not be found to identify the exact dimerization contacts between the two β-defensin molecules in the dimer. Diffusion NMR, using the PG-SLED approach [45], is a facile technique to study the dimerization and multimerization properties of peptides and small proteins in solution [12,44,46].

NMR of Antimicrobial Peptides

Hepcidin; Identification of Disulfide Cross-Linking The structure determination of the liver-synthesized peptide hepcidin presented a challenge in establishing the correct disulfide bond connectivities. Hepcidin was originally discovered in human urine [49] and plasma ultrafiltrate [50] in two forms of 20 and 25 amino acid residues long. The peptide plays an important role in the control and regulation of iron uptake; in addition, hepcidin has also been found to have antimicrobial and antifungal properties [50]. As observed with several antimicrobial peptides, there is a high proportion of Cys in hepcidin. All eight of its Cys residues are involved in disulfide bonds [50], but the specific connectivities between the residues could not be determined by conventional proteolysis methods [51]. Therefore, the solution structure determination of the two peptides by 2D 1 H NMR was used to establish the connectivities [12]. As done previously for the 51-residue pheromone Er-23, where the disulfide bond pattern could also not be determined by chemical methods [52], initial structure calculations incorporated only NOE restraints and bond angle constraints, without the disulfide bonds or the hydrogen bonds being included. The initial NOE patterns of the hepcidins themselves only clarified two out of the four unambiguous disulfide links. Hydrogen exchange experiments in D2 O helped to establish the position of some hydrogen bonds, and once these were introduced all the disulfide cross-links could be assigned in further rounds of the structure calculations. This approach resulted in a highly unusual vicinal cysteine bridge between the adjacent Cys residues 8 and 9. Support for this unexpected connectivity was obtained from other NMR parameters, such as a lack of amide proton correlations in the fingerprint regions for these two residues in the TOCSY and NOESY spectra, as well as broad peaks for their α and β-protons, indicating a conformational exchange process on the NMR timescale. The role of this unique structural feature in the hepcidins is still unclear.

Vicinal disulfides have been observed in a few enzymes where they were found to be critical for enzyme function [53]. They have not yet been found in any other antimicrobial peptide. The overall structure of the hepcidins is a distorted two-stranded β-sheet where the vicinal disulfide bridge is found at the turn of the hairpin. In a similar manner, disulfide connectivities have been determined on the basis of NMR experiments alone for other peptides such as the antifungal Alo-3 [54] and the insecticidal pea albumin 1, subunit b [55]. These authors used ambiguous disulfide restraints and reduced charges on the sulfur groups in the early stages of the structure calculation, with the ARIA software [56] to establish the correct connectivities.

Microcin J25; Characterization of Unusual Linkages The bacteriocin microcin J25 is secreted by certain strains of Escherichia coli; its antibacterial mechanism of action has been suggested to involve cell membrane disruption [57]. However, several bacterial strains with specific mutations in the β-subunit of RNA polymerase have been found to be resistant to the peptide [58], suggesting another mode of action. The original solution structure determined for the peptide [59] revealed a distorted, highly compact β-sheet structure with the main chain N- and C-termini joined together by an amide bond.Unexpectedly, microcins with their circular backbone cleaved by thermolysin retained their antimicrobial activity as well as the core structure [60]. Subsequent NMR analysis of chemically synthesized linear and headto-tail cyclic peptides revealed that the spectra of these synthetic peptides were different from those of the natural compound [36]. A reanalysis of the original NMR data for native microcin J25 revealed that the expected sequential NOEs linking the glycines of both termini were absent. Nevertheless, mass spectrometry of the native peptide showed that there was one extra link present in the peptide, which had been formed through a dehydration reaction. A careful NMR analysis of microcin J25 revealed that the compact peptide did not have a head-to-tail cyclized backbone, but rather a novel side chain-to-backbone ring structure (Figure 1D). This linkage involves an amide link between the amino terminus of the peptide and the carboxyl side chain of Glu 8, with an overall knot structure where the C-terminal end is threaded through the ring in a rigid conformation. In fact, this noose-shaped structure was simultaneously reported in two other independent NMR studies [61,62]. The internal cross-link found in microcin J25 has been found in other bacterial proteins, such as small peptide inhibitors [63], where it is believed to contribute to a greater resistance to proteolytic degradation.

Part II

A recently solved structure of interest is that of the cyclic θ defensin-1 (RTD-1) [47] from rhesus macaque leukocytes. The biosynthesis of RTD-1 involves the translation of two α-defensin-related nonapeptides which are combined by head-to-tail backbone ligation of the peptides and the formation of three disulfide bridges between these two molecules [48]. The resulting NMR-derived structure (Figure 2C), determined in 10% acetonitrile, shows an antiparallel β-sheet with very well defined turns but an overall flexible structure as judged from the poor superposition of the final 10 structures. The lack of a welldefined structure was attributed to bending motions at the center of the cyclic peptide [47].

Solution Structures of Antimicrobial Peptides 1319

1320 Part II

Pharmaceutical Sciences

Part II

Solid-State NMR Experiments: Peptide Orientation in Bilayers Magainin 2 As a classical example of amphipathic α-helical antimicrobial peptides, magainin, and its natural and synthetic analogues have been intensively studied [64]. Membrane pore formation is thought to be at the center of its mechanism of action, given that magainin has been shown to disrupt the electrochemical gradient across cell membranes [65]. Hence, recent NMR studies have focused on characterizing its interactions with membrane bilayers. Residuespecific 15 N labeling of a series of synthetic magainin 2 peptides has been used in solid-state NMR measurements of oriented membrane samples. This work conclusively showed that the helix formed by magainin, and the related PGLa peptide, lie in an orientation that is parallel to the membrane surface of oriented bilayer samples [66], in agreement with fluorescence quenching experiments [67]. The cationic side of the magainin helix is exposed to the aqueous environment and interacts with the surrounding polar lipid head groups, while the hydrophobic half of the peptide is embedded into the membrane and interacts with the fatty acyl chains in the bilayer. Additionally, a recent hydrogen/deuterium exchange NMR experiment confirms this parallel arrangement in SDS micelles [68], with lower exchange factors associated with the hydrophobic residues and higher exchange rates seen for the polar residues. These findings have helped in explaining the mechanism of pore formation for these peptides by excluding the “barrel-stave” model of pore formation in favor of the “toroidal” pore model [5] (shown in a schematic in Figure 4). In the barrel-stave model, lytic peptides oligomerize at very low peptide:lipid ratios on the membrane and then form peptide-lined pores by orienting

Fig. 4. The toroidal pore model as originally proposed for magainin 2. After a threshold concentration is reached, the peptides change their conformation from a parallel orientation on the bilayer plane. The peptides transiently cross the membrane to adopt a conformation perpendicular to the bilayer plane. Contacts with the phospholipid head groups remain because the pore is lined with peptides intercalated with lipids.

perpendicularly to the membrane plane. A good example of such a barrel-stave pore is provided by alamethicin [9], a peptide that is inhibitory of prokaryotic as well as eukaryotic cells. In the toroidal pore model, after reaching a high peptide:lipid threshold, the peptides associate loosely with each other to form transmembrane pores that are lined with peptides separated by lipid head groups. This lipid-peptide rearrangement has been supported in the case of magainin 2 through fluorescence experiments showing rapid flip-flop of the lipids from one side of the bilayer leaflet to the other upon addition of the peptide [70]. Solid-state 31 P NMR studies of lipid bilayers showed that upon addition of MSI-78, a synthetic analogue of magainin, the temperature-dependent transition from the fluid lamellar to the hexagonal phase was inhibited [71]. This means that the peptide induced a positive curvature strain in the membrane which is consistant with toroidal pore formation. Recently, TRNOE experiments [72] of magainin 2 transiently bound to dilaurylphosphatidylcholine (DLPC) vesicles suggested that the peptide formed an antiparallel dimer structure; this interpretation was based on numerous long-range NOE crosspeaks that were not compatible with the peptide being in a monomeric state [73]. This study contests that solid-state NMR studies of magainin 2 missed this dimer formation because the 15 N labeling only provides information for the main chain conformation. Meanwhile, side chain conformations and therefore oligomerization contacts are harder to determine. The impact of this dimer formation on the currently accepted model of pore formation by magainin is presently unclear.

Protegrin-1 Solid-state NMR measurements with protegrin-1, a β-sheet peptide, showed different results compared to magainin 2. Chemical shifts of 13 CO and 15 N labels on Val-16 of the peptide were measured, and the two-stranded hairpin was calculated to be rotated by 48 ± 5◦ from the bilayer normal of the DLPC bilayers [74]. This orientation and the length of the peptide still allow for the positively charged arginine side chains to protrude out of the hydrophobic core of the membrane and interact with the lipid head groups. Three of the six Arg’s are near the termini of the peptide and the other three are adjacent to each other near the hairpin turn, while the residues on the strands are hydrophobic to allow for favorable interactions with the acyl portion of the membrane. In a related solid-state NMR experiment, relatively low concentrations of paramagnetic Mn2+ were used to induce distance-dependent dipolar relaxation to investigate the depth of insertion of specific residues of protegrin-1 bound to DLPC bilayers [75]. This was accomplished through specific labeling of four selected residues with

NMR of Antimicrobial Peptides

C, and the order of penetration of the residues was in agreement with the tilted model for protegrin-1 insertion. While these experiments shed an interesting light on the arrangement of protegrin-1 in bilayers, it should be recognized that the DLPC lipids used have shorter acyl chains by 4–6 carbons than physiologically relevant phospholipids. Solid-state NMR studies making use of the magicangle-spinning (MAS) technique to obtain higher resolution have also been reported. These allowed the study of protegrin-1 bound to palmitoyloleoylphosphatidylcholine (POPC) bilayers, a more relevant membrane system [27]. However, the 13 C signals of the isotopically labeled protegrin-1 indicated a high level of aggregation in these bilayers; therefore proper characterization of the bound peptides could not be achieved. Be that as it may, information about the lipid arrangement could be obtained by studying 2 H NMR spectra of deuterium-labeled POPC with increasing concentrations of protegrin-1. The results clearly showed a greater proportion of unordered lipids in the lamellar phase [74]. These findings support the toroidal pore mechanism in the case of the protegrin-1 peptide where the membrane is under positive curvature stress and the lipids can move between the inner and outer leaflets.

Conclusions and Future Directions Standard 2D proton NMR spectroscopy remains the primary method to obtain reliable and detailed structural information about antimicrobial peptides in solution or bound to micelles. As we have discussed, when performing structure calculations based on NMR data alone, one must carefully consider at what stage of the process it is justified to include hydrogen bonding, disulfide bridging or special features such as the knot motif in microcin J25 [36,61,62] as a constraint in the calculations. As the population of easily solvable peptide structures decreases, new NMR strategies will be necessary to fully characterize antimicrobial peptides in a more realistic membrane environment, where many are thought to exert their antimicrobial effects [3]. To this end solid-state NMR has already started to provide a snapshot of such peptides in a membrane bilayer. In the future, it should be possible to use fully isotope-labeled and deuterated peptides and TROSY NMR approaches to study these antimicrobial peptide in small vesicles, as is already being done for some membrane proteins [76,77,78]. Vesicles are a better membrane mimetic than the frequently used micelles. The development of lipid bicelles, bilayers composed of lipids with different chain lengths forming small discoidal shapes [79], could also prove useful in determining more accurate membrane-bound solution structures. Multimerization of the peptides presents difficulties in data

analysis and yet, the oligomerization states could be more reflective of biologically relevant conditions. Also, the structural data coming from solution NMR studies should be supplemented by dynamic information to gain further insights. Novel approaches using homonuclear NMR are now available [80]. Additionally, a multitude of other aspects of the mechanism of action of antimicrobial peptides remain to be characterized. In particular, studies of their interactions with the lipopolysaccharides of the outer membrane of Gram negative bacteria need to be done. Although many peptides exert their cytotoxicity through membrane disruption, in the case of peptides that do not act on the membrane, interaction studies with their intracellular targets will need to be pursued. In addition to being bactericidal, other useful properties are starting to emerge for many of these peptides [81]. A fair number of them also selectively exhibit toxicity toward tumor cells [82], fungi [83], and viruses [84]. Interaction studies with relevant receptors can lead to an understanding of a peptide’s role in the anti-inflammatory response [85], cell signaling pathways [86], or its capacity to stimulate the adaptive immune response [28]. These alternate properties present further directions in which research in this field can go. Great strides have been made in the characterization of the solution structures of antimicrobial peptides by highresolution NMR. In addition, solid-state NMR is beginning to provide valuable insights into the positioning of these peptides in membrane environments. Such information will be crucial for the design and development of novel classes of powerful pharmaceuticals to combat bacterial infections.

Acknowledgments The authors would like to thank Drs. David Schibli, Howard Hunter, and Weiguo Jing for many insightful discussions. This research is supported by the Canadian Institute for Health Research and the Alberta Heritage Foundation for Medical Research.

References 1. Hancock RE. Expert. Opin. Investig. Drugs. 2000;9:1723. 2. Boman HG. J. Intern. Med. 2003;254:197. 3. Epand RM, Vogel HJ. Biochim. Biophys. Acta. 1999;1462: 11. 4. Marassi FM, Opella SJ. Curr. Opin. Struct. Biol. 1998;8: 640. 5. Matsuzaki K. Biochim. Biophys. Acta. 1999;1462:1. 6. Wuthrich K (Ed). NMR of Proteins and Nucleic Acids, John Wiley and Sons: New York, 1986. 7. Bechinger B. Biochim. Biophys. Acta. 1999;1462:157. 8. Hwang PM, Vogel HJ. Biochem. Cell Biol. 1998;76:235.

Part II

13

References 1321

1322 Part II

Pharmaceutical Sciences

Part II

9. Powers JP, Hancock RE. Peptides. 2003;24:1681. 10. Gesell J, Zasloff M, Opella SJ. J. Biomol. NMR 1997;9:127. 11. Fahrner RL, Dieckmann T, Harwig SS, Lehrer RI, Eisenberg D, Feigon J. Chem. Biol. 1996;3:543. 12. Hunter HN, Fulton DB, Ganz T, Vogel HJ. J. Biol. Chem. 2002;277:37597. 13. Leong WI, Lonnerdal B. J. Nutr. 2004;134:1. 14. Lee DL, Hodges RS. Biopolymers. 2003;71:28. 15. Ando S, Nishikawa H, Takiguchi H, Lee S, Sugihara G. Biochim. Biophys. Acta. 1993;1147:42. 16. Jing W, Hunter HN, Hagel J, Vogel HJ. J. Pept. Res. 2003;61:219. 17. Cornet B, Bonmatin JM, Hetru C, Hoffmann JA, Ptak M, Vovelle F. Structure. 1995;3:435. 18. Schibli DJ, Hwang PM, Vogel HJ. FEBS Lett. 1999;446:213. 19. Hwang PM, Zhou N, Shan X, Arrowsmith CH, Vogel HJ. Biochemistry. 1998;37:4288. 20. Rozek A, Powers JP, Friedrich CL, Hancock RE. Biochemistry. 2003;42:14130. 21. Rozek A, Friedrich CL, Hancock RE. Biochemistry. 2000;39:15765. 22. Fales-Williams AJ, Gallup JM, Ramirez-Romero R, Brogden KA, Ackermann MR. Clin. Diagn. Lab. Immunol. 2002; 9:28. 23. Lai R, Liu H, Hui LW, Zhang Y. Biochem. Biophys. Res. Commun. 2002;295:796. 24. Matsuzaki K, Yoneyama S, Miyajima K. Biophys. J. 1997;73:831. 25. White SH, Wimley WC, Selsted ME. Curr. Opin. Struct. Biol. 1995;5:521. 26. Ganz T. Science 2002;298:977. 27. Buffy JJ, Waring AJ, Lehrer RI, Hong M. Biochemistry 2003;42:13725. 28. Roumestand C, Louis V, Aumelas A, Grassy G, Calas B, Chavanieu A. FEBS Lett. 1998;421:263. 29. Riley MA, Wertz JE. Annu. Rev. Microbiol. 2002;56:117. 30. Van Den Hooven HW, Doeland CC, Van De KM, Konings RN, Hilbers CW, Van De Ven FJ. Eur. J. Biochem. 1996;235: 382. 31. Breukink E, de Kruijff B. Biochim. Biophys. Acta. 1999; 1462:223. 32. Cleveland J, Montville TJ, Nes IF, Chikindas ML. Int. J. Food Microbiol. 2001;71:1. 33. Twomey D, Ross RP, Ryan M, Meaney B, Hill C, Antonie Van Leewenhoek 2002;82:165. 34. Breukink E, Wiedemann I, van Kraaij C, Kuipers OP, Sahl H, Kruijff B. Science 1999;286:2361. 35. Hsu ST, Breukink E, Tischenko E, Lutters MA, de Kruijff B, Kaptein R, Bonvin AM, van Nuland NM. Nat. Struct. Mol. Biol. 2004;11:963. 36. Rosengren KJ, Clark RJ, Daly NL, Goransson U, Jones A, Craik DJ. J. Am. Chem. Soc. 2003;125:12464. 37. Zanetti M, Gennaro R, Scocchi M, Skerlavaj B. Adv. Exp. Med. Biol. 2000;479:203. 38. Pardi A, Zhang XL, Selsted ME, Skalicky JJ, Yip PF. Biochemistry 1992;31:11357. 39. Pardi A, Hare DR, Selsted ME, Morrison RD, Bassolino DA, Bach AC. J. Mol. Biol. 1988;201:625. 40. Zimmermann GR, Legault P, Selsted ME, Pardi A. Biochemistry 1995;34:13663.

41. Landon C, Sodano P, Hetru C, Hoffmann J, Ptak M. Protein Sci. 1997;6:1878. 42. Bruix M, Gonzalez C, Santoro J, Soriano F, Rocher A, Mendez E, Rico M. Biopolymers. 1995;36:751. 43. Hill CP, Yee J, Selsted ME, Eisenberg D. Science 1991;251:1481. 44. Schibli DJ, Hunter HN, Aseyev V, Starner TD, Wiencek JM, McCray PB Jr, Tack BF, Vogel HJ. J. Biol. Chem. 2002;277:8279. 45. Wilkins DK, Grimshaw SB, Receveur V, Dobson CM, Jones JA, Smith LJ. Biochemistry. 1999;38:16424. 46. Weljie AM, Yamniuk AP, Yoshino H, Izumi Y, Vogel HJ. Prot. Sci. 2003;12:228. 47. Trabi M, Schirra HJ, Craik DJ. Biochemistry. 2001;40:4211. 48. Tang YQ, Yuan J, Osapay G, Osapay K, Tran D, Miller CJ, Ouellette AJ, Selsted ME. Science. 1999;286:498. 49. Park CH, Valore EV, Waring AJ, Ganz T. J. Biol. Chem. 2001;276:7806. 50. Krause A, Neitz S, Magert HJ, Schulz A, Forssmann WG, Schulz-Knappe P, Adermann K. FEBS Lett. 2000;480:147. 51. Creighton TE. In: TE Creighton (Ed). Protein Structure, A Practical Approach. IRL Press Ensyham: Oxford, 1989, p 155. 52. Zahn R, Damberger F, Ortenzi C, Luporini P, Wuthrich K. J. Mol. Biol. 2001;313:923–931. 53. Wang X, Connor M, Smith R, Maciejewski MW, Howden ME, Nicholson GM, Christie MJ, King GF. Nat. Struct. Biol. 2000;7:505. 54. Barbault F, Landon C, Guenneugues M, Meyer JP, Schott V, Dimarcq JL, Vovelle F. Biochemistry. 2003;42:14434. 55. Jouvensal L, Quillien L, Ferrasson E, Rahbe Y, Gueguen J, Vovelle F. Biochemistry. 2003;42:11915. 56. Nilges M, O’Donoghue SI. Prog. NMR Spec. 1992;32:107. 57. Rintoul MR, de Arcuri BF, Salomon RA, Farias RN, Morero RD. FEMS Microbiol. Lett. 2001;204:265. 58. Delgado MA, Rintoul MR, Farias RN, Salomon RA. J. Bacteriol. 2001;183:4543. 59. Blond A, Cheminant M, Segalas-Milazzo I, Peduzzi J, Barthelemy M, Goulard C, Salomon R, Moreno F, Farias R, Rebuffat S. Eur. J. Biochem. 2001;268:2124. 60. Blond A, Cheminant M, Destoumieux-Garzon D, SegalasMilazzo I, Peduzzi J, Goulard C, Rebuffat S. Eur. J. Biochem. 2002;269:6212. 61. Wilson KA, Kalkum M, Ottesen J, Yuzenkova J, Chait BT, Landick R, Muir T, Severinov K, Darst SA. J. Am. Chem. Soc. 2003;125:12475. 62. Bayro MJ, Mukhopadhyay J, Swapna GV, Huang JY, Ma LC, Sineva E, Dawson PE, Montelione GT, Ebright RH. J. Am. Chem. Soc. 2003;125:12382. 63. Katahira R, Uosaki Y, Ogawa H, Yamashita Y, Nakano H, Yoshida M, J. Antibiot. (Tokyo) 1998;51:267. 64. Matsuzaki K. Biochim. Biophys. Acta. 1998;1376:391. 65. Westerhoff HV, Juretic D, Hendler RW, Zasloff M. Proc. Natl. Acad. Sci. U.S.A. 1989;86:6597. 66. Bechinger B, Zasloff M, Opella SJ. Protein Sci. 1993;2: 2077. 67. Matsuzaki K, Murase O, Tokuda H, Funakoshi S, Fujii N, Miyajima K. Biochemistry 1994;33:3342. 68. Veglia G, Zeri AC, Ma C, Opella SJ. Biophys. J. 2002;82:2176.

NMR of Antimicrobial Peptides

88. Shin SY, Kang JH, Hahm KS. J. Pept. Res. 1999;53: 82. 89. Oh D, Shin SY, Kang JH, Hahm KS, Kim KL, Kim Y. J. Pept. Res. 1999;53:578. 90. Hicks RP, Mones E, Kim H, Koser BW, Nichols DA, Bhattacharjee AK. Biopolymers 2003;68:459. 91. Sawai MV, Waring AJ, Kearney WR, McCray PB Jr, Forsyth WR, Lehrer RI, Tack BF. Prot. Eng. 2002;15:225. 92. Bechinger B, Zasloff M, Opella SJ. Biophys. J. 1998; 74:981. 93. Tack BF, Sawai MV, Kearney WR, Robertson AD, Sherman MA, Wang W, Hong T, Boo LM, Wu H, Waring AJ, Lehrer RI. Eur. J. Biochem. 2002;269:1181. 94. Lay FT, Schirra HJ, Scanlon MJ, Anderson MA, Craik DJ. J. Mol. Biol. 2003;325:175. 95. Lai JR, Huck BR, Weisblum B, Gellman SH. Biochemistry 2002;41:12835. 96. Shankaramma SC, Athanassiou Z, Zerbe O, Moehle K, Mouton C, Bernardini F, Vrijbloed JW, Obrecht D, Robinson JA. Chembiochem. 2002;3:1126. 97. McManus AM, Dawson NF, Wade JD, Carrington LE, Winzor DJ, Craik DJ. Biochemistry. 2000;39:15757. 98. Laederach A, Andreotti AH, Fulton DB. Biochemistry. 2002;41:12359. 99. Fujitani N, Kawabata S, Osaki T, Kumaki Y, Demura M, Nitta K, Kawano K. J. Biol. Chem. 2002;277:23651. 100. Schibli DJ, Hwang PM, Vogel HJ. Biochemistry. 1999; 38:16749. 101. Andersson M, Gunne H, Agerberth B, Boman A, Bergman T, Olsson B, Dagerlind A, Wigzell H, Boman HG, Gudmundsson GH. Vet. Immunol. Immunopathol. 1996;54:123. 102. Mandard N, Sodano P, Labbe H, Bonmatin JM, Bulet P, Hetru C, Ptak M, Vovelle F. Eur. J. Biochem. 1998;256: 404. 103. Wang Y, Henz ME, Gallagher NL, Chai S, Gibbs AC, Yan LZ, Stiles ME, Wishart DS, Vederas JC. Biochemistry. 1999;38:15438. 104. Koradi R, Billeter M, Wuthrich K. J. Mol. Graph. 1996; 14:29.

Part II

69. Laver DR. Biophys. J. 1994;66:355. 70. Matsuzaki K, Murase O, Fujii N, Miyajima K. Biochemistry. 1996;35:11361. 71. Hallock KJ, Lee DK, Ramamoorthy A. Biophys. J. 2003; 84:3052. 72. Post CB. Curr. Opin. Struct. Biol. 2003;13:581. 73. Wakamatsu K, Takeda A, Tachi T, Matsuzaki K. Biopolymers 2002;64:314. 74. Yamaguchi S, Hong T, Waring A, Lehrer RI, Hong M. Biochemistry 2002;41:9852. 75. Buffy JJ, Hong T, Yamaguchi S, Waring AJ, Lehrer RI, Hong M. Biophys. J. 2003;85:2363. 76. Fernandez C, Hilty C, Wider H, Guntert P, Wuthrich K. J. Mol. Biol. 2004;336:1211. 77. Hwang PM, Choy WY, Lo EI, Chen L, Forman-Kay JD, Raetz CR, Prive GG, Bishop RE, Kay LE. Proc. Natl. Acad. Sci. U.S.A. 2002;99:13560. 78. Arora A, Abildgaard J, Bushweller JH, Tamm LK. Nat. Struct. Biol. 2001;8:334. 79. Glover KJ, Whiles JA, Wu G, Yu N, Deems R, Struppe JO, Stark RE, Komives EA, Vold RR. Biophys. J. 2001;81:2163. 80. Schleucher J, Wijmenga SS. J. Am. Chem. Soc. 2002;124: 5881. 81. Kamysz W, Okroj M, Lukasiak J. Acta. Biochim. Pol. 2003;50:461. 82. Lichtenstein A, Ganz T, Selsted ME, Lehrer RI, Blood. 1986;68:1407. 83. Lupetti A, Danesi R, van ’t Wout JW, van Dissel JT, Senesi S, Nibbering PH. Expert. Opin. Investig. Drugs. 2002; 11:309. 84. Cole AM, Lehrer RI. Curr. Pharm. Des. 2003;9:1463. 85. Murphy CJ, Foster BA, Mannis MJ, Selsted ME, Reid TW. J. Cell Physiol. 1993;155:408. 86. Charp PA, Rice WG, Raynor RL, Reimund E, Kinkade JM Jr, Ganz T, Selsted ME, Lehrer RI, Kuo JF. Biochem. Pharmacol. 1988;37:951. 87. Holak TA, Engstrom A, Kraulis PJ, Lindeberg G, Bennich H, Jones TA, Gronenborn AM, Clore GM. Biochemistry 1988;27:7620.

References 1323

1325

Muriel Delepierre1 and Lourival D. Possani2 1 Unit´ e

de R´esonance Magn´etique Nucl´eaire des Biomol´ecules, CNRS URA 2185, 75 724 Paris Cedex 15, France of Molecular Medicine, Institute of Biotechnology, National Autonomous University of Mexico, Cuernavaca 62210, Mexico

2 Department

Introduction Animal venoms, such as those from snakes, scorpions, cone snails, sea anemones, spiders and honeybees, are rich sources of different classes of ligands that affect the function of eukaryotic cells and tissues by blocking ion channels [1]. These ligands recognize membrane bound proteins with different specificities and affinities, thus providing scientists with unique tools to investigate diverse areas of neuroscience, protein chemistry and evolution, including the ionic channels of excitable membranes, the phylogeny of proteins and structure–function relationships of proteins [2,3]. It is unclear why toxins with similar structures display such different functions, for example why do they show specific affinities for ion channels from a particular animal? Specific recognition of a wide range of ion channel types allows the use of their pharmacological properties. A given toxin can act on several different channels, of which only one may be of therapeutic interest [4], rendering its use as a therapeutic agent dangerous [5]. Nevertheless, successful approaches have been developed for the production of immunosuppressive agents from toxins and structural analogs [4] for the treatment of chronic pain [6] and for the treatment of various human disorders such as multiple sclerosis [7], cancer, diabetes, cardiovascular, and neurological diseases [1,8]. The fact that animal toxins are built around limited disulfide bonded frameworks, makes them resistant to degradation and permissive to mutations, meaning that they can be used as templates for protein engineering [9,10]. Before the detailed structure of a potassium channel was solved by X-ray crystallography [11], animal toxins were mainly used to map the channel-binding site and to identify the structural attributes that make a channel unique [12]. With present knowledge, the use of these toxins can be switched to the design or identification of new compounds that compete or interfere with their action upon ion channels. Unfortunately, for Na+ channels no high-resolution Graham A. Webb (ed.), Modern Magnetic Resonance, 1325–1330.
C 2008 Springer.

structural model is yet available [13]. Therefore, further structure–activity relationship studies are required on animal toxins to determine the molecular basis of the toxin–ion channel interaction in order to engineer molecules with novel biological functions and/or therapeutic potential [1]. This chapter will focus on toxins extracted from scorpion venoms that are active on sodium and potassium channels. Sodium channel toxins are much more abundant in scorpion venom than potassium channel toxins, which account for less than 0.1% of total venom. In addition, toxins affecting potassium channels usually make only a minor contribution to the effects of envenomation on humans. It is therefore not surprising that the first toxin structure to be solved was that of a sodium channel-specific toxin, the Centruroides sculpturatus variant-3 toxin [14].

Use of NMR to Determine the Structure of Rare Components It is essential to solve the three-dimensional structure of a toxin to determine how it interacts with its receptors. NMR and X-ray crystallography are the two most powerful tools available for structural determination and have been extensively used to obtain structural information on toxins. X-ray crystallography is limited by the necessity to obtain crystals, whereas NMR is limited by the size of the molecule and the low sensitivity of the method, making these two techniques complementary. NMR, which is particularly useful for determining the structure of small molecules, can also be used to obtain information on the dynamics of macromolecules over a wide timescale range and to study molecular interactions [15,16]. However, NMR is inherently less sensitive than almost all other analytical methods as a result of the small energy gap between ground and excited states. Indeed, the minimal quantity that can be analyzed by NMR is usually in the

Part II

Pharmaceutical Applications of Ion Channel Blockers: Use of NMR to Determine the Structure of Scorpion Toxins

1326 Part II

Pharmaceutical Sciences

Part II

nanomolar range or higher, precluding its routine use in trace analysis. Nevertheless, the sensitivity of NMR has been improved considerably in the last decade by increasing the field strength—900 MHz is now available—and by increasing probe performance [17]. The use of cryogenic probes increases the coil quality factor, improving sensitivity by four- or fivefold compared to conventional 5 mm probes [18]. Sensitivity can be also improved by optimizing the sample volume as a function of the specific availability or solubility of a given sample. Three types of small sample volume probes have been developed: (i) Microprobes that use small diameter (3 mm) vertical sample tubes with volumes ranging from 120 to 150 µl. These probes, designed to couple HPLC with NMR, use saddle-type radio frequency (RF) coils. They were initially used in the pharmaceutical industry and are now becoming more widely used in academic laboratories for the characterization of small volumes of natural products and other rare samples [19]. The sample volume can be further reduced to 70–80 µl by using 3 mm Shigemi tubes. In this case, the bottom of the tube that is not in the receiver coil is filled with glass with the same dielectric properties as the solvent used. The amount of material necessary and/or the experimental time can be decreased further by using a 3-mm cryogenic probe [20]. Finally, 1.7 mm submicro inverse-detection gradient NMR probes first described in 1998 are compatible with sample volumes of just 20–30 µl [21]. (ii) The nano-NMR probe was first developed to characterize molecules covalently attached to solid-phase synthesis beads [22]. It uses a 40 µl observable volume that does not need to be filled without compromising the line widths [23] and identical concentrations to standard probes. High detection efficiency is achieved by placing the entire sample in the receiver coil. Magnetic susceptibility contributions to the line width around or within the sample are removed by spinning samples very rapidly (1–4 kHz) at the magic angle. This produces a very narrow water lineshape even at the base of the water signal (Figure 1). However, the spinning of samples in the presence of both RF and magnetic field inhomogeneities modulates the effective field, dramatically reducing the performance of isotropic mixing in TOCSY experiments when conventional composite-pulse mixing sequences are used. Adiabatic mixing sequences are less susceptible to such modulations [24] and perform considerably better in TOCSY experiments at the magic angle [25]. (iii) Microcoil NMR probes were first described in 1994 [26]. Their capillaries can contain nanoliter to micro-

8

6

4

2

δ 1H (ppm) Fig. 1. One-dimensional spectra of the Pi4 toxin [56]. The lower spectrum corresponds to natural (CONH2 ) Pi4 toxin at 380 µM 15% D2 O obtained at 500 MHz with a nano-NMR probe (Varian INNOVA) while top spectrum corresponds to synthetic (COO− ) Pi4 toxin at 100 µM 15% D2 O, 3 mm Shigemi tube obtained at 600 MHz with a cryoprobe (Varian INNOVA). In both cases 80 µg of protein was dissolved in 5 mM pH 4.0 CD3 COONa buffer. The water was suppressed with presaturation (courtesy of Inaki Guijarro).

liter volumes and the probes are built with a solenoid coil of about 1 mm. They were developed to couple capillary HPLC or electrophoresis to NMR [27]. Although impressive results have been obtained [28], most of these microprobes are still in the prototype phase. The small volume probes make it possible to collect spectra with excellent resolution, sensitivity, and lineshape with small amounts of product, making this technique a promising option when sample size is a real limitation. The combined use of small volume probes—3 mm and below—cryogenic technology, and Shigemi tubes makes it possible to determine chemical structures at unprecedented low concentrations. The use of these technological improvements has allowed the reduction of sample amounts to less than 2% of the initial quantity required. This is evidenced by the fact that for the first scorpion toxin solved by NMR, the sodium channel-specific insect toxin AaHIT, about 12 mg was needed [29], whereas for the first potassium channel-specific scorpion toxin, charybdotoxin (ChTX) 8 mg was required [30]. Nowadays, as little as 200 µg is necessary to obtain structure of potassium channel-specific toxin with a reasonable precision, i.e. with an average of 8–14 constraints per residue [31,32].

Structure of Scorpion Toxins by NMR

Sodium channel toxins typically contain 60–70 amino acid residues. Two classes of toxin were initially described based on their pharmacological properties and their ability to bind to two different sites on the extracellular surface of the sodium channel [33,34]. Alpha scorpion toxins (αScTxs) slow down sodium channel inactivation by binding to receptor site 3 in a voltage-dependent manner. These have been mostly found in the venom of Asian and African scorpions. Beta scorpion toxins (β-ScTxs) bind to receptor site 4 of sodium channels and alter the mechanism of activation of the channel. They were initially found in the New World scorpion (North and South America). This rule is not absolute, because α-ScTxs have been found in the New World and conversely β-ScTxs in the Old World. Displacement experiments on excitable tissues have shown that the two binding sites are not related. In addition to channel specificity, the sodium channel toxins also display species specificity, for example some α-like toxins are active on insects and mammals [33]. Based on toxicity tests, binding experiments and electrophysiological experiments, sodium channel toxins can be divided into peptides that are specifically toxic to either one or combinations of groups of animals. In this context, the term specific does not mean “exclusively effective” but refers to low toxicity or affinity in one group and/or higher toxicity or affinity in another. The three-dimensional structures of several sodium channel toxins have been determined by NMR or X-ray crystallography. All toxins (α, β, and α-like) have the same overall structure based on a conserved scaffold with different insertions, deletions, and mutations. Inserted fragments are named loops J, M, B, and F and correspond, respectively, to loop 1 between β strand 1 and the α-helix, loop 2 between the α-helix and the β strand 2, loop 3 between β strands 2 and 3 and finally the C-terminal end. The α-ScTxs have a short J loop and a long B loop, while the β-ScTxs have a long J loop and a short B loop. Excitatory toxins have short J and B loops. An example of such insertion is the excitatory insect toxin from Buthotus judaicus (BjxtrIT) in which a five-amino acid insertion occurs just before the α-helix in loop 1 and a short αhelix is inserted in the C-terminal part of the molecule [35]. The α-ScTxs contain an insertion at around residue 42 of the B loop that is believed to play a key role in the modulation of toxic activity [36]. Based on experimental evidence, three criteria appear to be important for sodium channel specificity, namely: (i) the presence of a positively charged residue at the N-terminus (usually lysine and sometimes arginine), immediately followed by a negatively charged residue such as glutamic acid or aspartic acid at position 2; (ii) a con-

served aromatic cluster; and (iii) a positively charged group (e.g. lysine or arginine) at position 13 in β-type toxins or position 58 in α-type toxins. Most of the scorpion toxin structures that have been reported so far have been assigned to the α-type group or are weakly toxic to mammals. In North and South America, two of the most dangerous scorpions for mammals are Centruroides noxius and Tityus serrulatus, respectively. In each case, the most toxic components for mammals are β-type toxins active on sodium channels, respectively Ts1 and Cn2. Ts1 is a β-ScTxs and is the major component of the venom of T. serrulatus. The structure of Ts1, also known as Ts-γ or toxin VII, has been determined ˚ [37] while by X-ray diffraction to a resolution of 1.7 A the structure of Cn2 has been determined by NMR [38]. Cn2 is composed of one α-helix and three β strands. It contains 10 aromatic residues, several of which are located on the same face of the molecule, face A, as is generally observed for sodium channel toxins. This face was thought to be the determinant for the sodium channel specificity. The orientation of the C-terminal region differs between α- and β-toxins as do the sizes of the J loop and B loop regions. The unique orientation of the Cn2 C-terminal region stems from a proline in a cis conformation. Interestingly Ts1 C-terminal orientation is similar to that of Cn2 but here the cis-proline is replaced by a helix turn. Whereas binding of β toxins to a common site on the sodium channel may be explained by the same recognition pattern on face A (i.e. the face containing the hydrophobic cluster), the difference in their toxicities may be related to the distribution of residues on the opposite face (face B). In Cn2 and Ts1, this face, formed by loop 1 and the C-terminal end, displays a large number of basic amino acid residues, whereas in weakly active toxins these residues are replaced by neutral amino acids. The fact that the first crustacean-specific toxin for which a structure was solved came from a scorpion of the same genus as Cn2 (Centruroides limpidus limpidus) revealed structural features that might be responsible for the species specificity: mammals versus crustaceans. Whether a particular toxin of the genus Centruroides is specific towards mammals or arthropods depends on the nature of a few residues clustered in two well-defined regions (mainly loops between secondary structure elements). Mutational analyses in which the two regions are exchanged or point mutations introduced at these residues are required to corroborate these results.

NMR Structures of Toxins Active on Potassium Channels Potassium channel toxins also vary considerably in terms of sequence and specificity, and can act on a wide range of potassium channels. Although potassium channel toxins

Part II

NMR Structures of Toxins Active on Sodium Channels

NMR Structures of Toxins Active on Potassium Channels 1327

1328 Part II

Pharmaceutical Sciences

Part II

from scorpions of the family Buthidae are mostly present as minor components of the venom, the increasing interest in venom constituents has led to efforts to improve the separation and identification of all venom compounds. So far, more than 120 different K+ channel toxins have been found in scorpion venoms [39,40]. They have been divided into three subfamilies based on their primary structures and functions: α-, β- and γ-KTxs. The α-KTx family is the largest one, containing more than 80 members in 18 subfamilies active on Kv, BKCa , or SKCa channels. They are short-chain toxins, consisting of 23–41 amino acids and three or four disulfide bridges [41]. The four members of the β-Ktxs family identified so far are longchain toxins composed of 60–64 amino acids and containing only three disulfide bridges [42]. Twenty-six members of the γ-Ktxs family have been identified. Members of this family are HERG channel blockers consisting of 36–47 amino acids and three or four disulfide bridges [43]. With the exception of the κ-Hefutoxins [44], all the K+ channel-specific toxins isolated from scorpion venoms are structurally related. They are characterized by the cysteine-stabilized α/β motif (CS-αβ) in which Ci–C j and Ci+4–C j+2 disulfide bridges link the α-helix to the second strand of the β-sheet. κ-Hefutoxin 1 adopts a unique three-dimensional fold, consisting of two parallel helices linked by two disulfide bridges and without any β-sheets [44]. The first potassium channel toxin to be identified was noxiustoxin in 1982 [45]. Its three-dimensional structure was solved in 1995 [46]. The first potassium channel for which the structure was solved was charybdotoxin [47] discovered in 1985 [48]. The structures of several potassium channel toxins have been elucidated in the last 15 years either by X-ray or by NMR. This was made possible by following developments including: (i) chemical synthesis associated with efficient refolding, (ii) recombinant methods allowing uniform 15 N and 13 C labeling of these toxins, (iii) advances in methods for the extraction and purification, and (iv) recent developments in highly sensitive NMR methodology allowing structure determination with just a few hundred micrograms of compound [49, and references therein]. Among the 120 K+ channel toxins the α-Ktx6 subfamily, which now comprises five different members, is characterized by the presence of an additional disulfide bridge that, in most members, fixes the C-terminus to the turn following the α-helix. Two different arrangements of disulfide bonds have been described for members of this subfamily. Pi1 [50], Pi4 [51], Pi7 [32,51] and HsTX1 [52] show half-cystine pairing of C1–C5, C2–C6, C3– C7, C4–C8, whereas MTX displays a different S–S bond topology C1–C5, C2–C6, C3–C4, C7–C8 [53]. The different S–S bonds in MTX link contiguous regions in the sequence; however MTX displays the characteristic structure of a small scorpion toxin [54]. Interestingly, no target

has yet been found for toxin Pi7 despite it having a similar overall fold to all potassium channel toxins [32]. Another toxin containing four disulfide bridges has also been purified from T. serrulatus, TXT α-KTx17, but the additional disulfide is on the N-terminal end and the structure of this toxin has not yet been solved [55]. The three-dimensional structures of the three Pandinus imperator toxins have been determined using nanomolar amounts of the natural compounds and the nano-NMR probe approach and/or 3 mm Shigemi tubes [31–32,56]. The structures of HsTX1 [57] and MTX [54] were determined using synthetic molecules. The structure of Pi4 was determined using both the natural and synthetic toxins, showing that both have the same fold and disulfide pairing [56]. Disulfide bridges can be identified from the spatial interactions observed between Hα and Hβ and between Hβ and Hβ of cysteines involved in the disulfide bridge. However, in cysteine-rich proteins, such as scorpion venom toxins, it is not always possible to obtain conclusive results, as connections between Hβ protons can involve more than two cysteines. S–S bonds have been successfully assigned by NMR and modeled using different methodologies for several proteins [57,58]. Structures are calculated using NMR data only, with no disulfide bridges, to propose topologies compatible with NMR restraints for all possible combinations of disulfide bridges. Given that the bond between the sulfur atoms can only be formed if ˚ the the distance between the two sulfur atoms is 2.03 A, structures with lowest experimental energies are analyzed in terms of the statistical distribution of distances between cystine sulfur atoms [32]. Even when other experimental data are available, NMR and structure calculations offer the possibility of testing the disulfide bridge pattern determined by other techniques and can be routinely used to locate the S–S bonds of proteins of unknown structures. The structures of only two members of the γ-KTx family have been solved: BeKm-1 [59] and CnErg1 [60]. They have different disulfide bridging patterns and do not share any significant homology in primary structure, but both block the turret region of the HERG channel [61,62]. Their structure was solved using the recombinant toxin [59], the synthetic toxin [63] and the native toxin in the case of CnErg1 [64]. They adopt the α-KTx CS-αβmotif; however the turn of helix observed in the N-terminal part of CnErg1 is unique to the family of potassium channel toxins [63,64]. BeKm-1 and CnErg1 share similar interaction surfaces with the HERG channel. Evolutionary [65] and structural analyses demonstrated the presence of two important functional residues that are mainly located in two patches—one hydrophobic and the other hydrophilic— located at the two opposite heads of the toxin molecules [40,54,61,62]. Therefore, scorpion γ-KTx toxins seem to share the same interaction mode regardless of their primary structures.

Structure of Scorpion Toxins by NMR

The structures determined so far for scorpion toxin acting on ion channels seem to converge to a unique global fold CS-αβmotif even though they exert a wide range of biological functions. It is therefore reasonable to expect that molecular modeling will be enough to establish reasonable structural models. However, it is probably the unique fine details of structure (i.e. subtle changes in secondary structure elements and/or the relative orientation of side chains and/or the dynamic properties of the molecule) that are important for channel recognition and species specificity. Two recent findings emphasize how fine structure can help to elucidate the structure–function relationships of scorpion toxins. The first one concerns the finding that a α-KTx toxin, BmTx3, isolated from the venom of the scorpion Buthus martensi Karsch, can display dual activity, blocking HERG activity and A-type potassium currents [66]. The second is the finding of the inhibitor cystine knot (ICK) [67] fold among scorpion toxins [68]. Furthermore, preliminary results on venom proteomics [39] seem to suggest that a large number of toxins from the venom of various animal species remain to be structurally characterized, thus it is reasonable to expect that new folds and new potent molecules will be discovered.

References 1. Lewis RJ, Garcia ML. Nat. Rev. Drug Discov. 2003:2;790. 2. Menez A. Toxicon. 1998:36;1557. 3. Mouhat S, Jouirou B, Mosbah A, De Waard M, Sabatier J-M. Biochem. J. 2004:378;717. 4. Kem W, Pennington MW, Norton RS. Drug Discov. Des. 1999:15/16;111. 5. Kalman K, Pennington MW, Lanigan MD, Nguyen A, Rauer H, Mahnir V, Paschetto K, Kem WK, Grissmer S, Gutrman GA, Christian EP, Cahalan MD, Norton RS, Chandy KG. J. Biol. Chem. 1998:273;32697. 6. Nielsen K, Schroeder T, Lewis RJ. J. Mol. Recognit. 2000:13;55. 7. Beeton C, Wulff H, Barbaria J, Clot-Faybesse O, Pennington M, Bernard D, Cahalan MD, Chandy KG, B´eraud E. Proc. Natl. Acad. Sci. U.S.A. 2001:98;13942. 8. Chandy KG, Wulff H, Beeton C, Pennington M, Gutman GA, Chalan MD. Trends Pharmacol. Sci. 2004:25;280. 9. Vita C, Roumestand C, Toma F, Menez A. Proc. Natl. Acad. Sci. U.S.A. 1995:92;6404. 10. Mer G, Kellenberger E, Lefevre JF. J. Mol. Biol. 1998:281;235. 11. Doyle DA, Morais Cabral JM, Pfuetzner RA, Kuo A, Gulbis JM, Cohen SL, Chait BT, MacKinnon J. Science. 1998:280;69. 12. Aiyar J, Withka JM, Rizzi JP, Singleton DH, Andrews GC, Lin W, Boyd J, Hanson D, Simon M, Dethlefs B, Lee CL, Hall JE, Gutman GA, Chandy G. Neuron. 1995:15;1169.

13. Blumenthal KM, Seibert AL. Cell Biochem. Biophys. 2003:38;1. 14. Fontecilla Camp JC, Almassy RJ, Suddath FL, Watt DD, Bugg CE. Proc. Natl. Acad. Sci. U.S.A. 1980:77;6496. 15. Wolff N, Guenneugues M, Gilquin B, Drakopoulou E, Vita C, M´enez A, Zinn-Justin S. Eur. J. Biochem. 2000:267;6519. 16. Cui M, Shen J, Briggs JM, Fei W, Wu J, Zhang Y, Luo X, Chi Z, Ji R, Jiang H, Chen K. J. Mol. Biol. 2002:318;417. 17. Hoult D, Richards RE. J. Magn. Reson. 1976:24;71. 18. Styles P, Soffe NF, Scott CA, Cragg DA, Row F, White DJ, White PCJ. J. Magn. Reson. 1984:60;397. 19. Crouch RC, Martin GE. J. Nat. Prod. 1992:55;1343. 20. Russel DJ, Hadden CE, Martin GE, Gibson AA, Zen AP, Cavolan JL. J. Nat. Prod. 2000:63;1047. 21. Subramanian R, Lam MM, Web AG. J. Magn. Reson. 1998:133;227. 22. Keifer PA. Drug Discov. Today 1997:2;468. 23. Delepierre M, Roux P, Chaffotte A-F, Goldberg ME. Magn. Reson. Chem. 1998:36;645. 24. Kupce E, Freeman R. J. Magn. Reson. A. 1995:115;273. 25. Kupce E, Keifer PA, Delepierre M. J. Magn. Reson. 2001:148;115. 26. Wu N, Peck TL, Webb AG, Magin RL, Sweedler JV. Anal. Chem. 1994:66;3849. 27. Olson DL, Peck TL, Webb AG, Magin RL, Sweedler JV. Science. 1995:270;1967. 28. Martin GE, Guido JE, Robins RH, Sharaf MHM, Tackie AN, Schiff PL Jr. J. Nat. Prod. 1998:61;555. 29. Darbon H, Weber C, Braun W. Biochemistry. 1991:30;1836. 30. Bontems F, Roumestand C, Boyot P, Gilquin B, Doljansky Y, Menez A, Toma F. Eur. J. Biochem. 1991:196;19. 31. Delepierre M, Prochnicka-Chalufour A, Possani LD. Biochemistry. 1997:36;2649. 32. Delepierre M, Prochnicka-Chalufour A, Boisbouvier J, Possani LD. Biochemistry. 1999:38;16756. 33. Gordon D, Savarin P, Gurevitz M, Zinn-Justin S. J. Toxicol. Toxin Rev. 1998:17;131. 34. Possani LD, Becerril B, Delepierre M, Tytgat J. Eur. J. Biochem. 1999:264;287. 35. Oren DA, Froy O, Amit E, Kleinberger-Doron V, Gurevitz M, Shaanan B. Structure. 1998:6;10995. 36. Fontecilla-Camps JC, Habersetzer-Rochat C, Rochat H. Proc. Natl. Acad. Sci. U.S.A 1988:85;7443. 37. Polikarpov I, Matilde MS Jr, Marangoni S, Toyama MH, Teplyakov A. J. Mol. Biol. 1999:290;175. 38. Pintar A, Possani LD, Delepierre M. J. Mol. Biol. 1999:287;359. 39. Tytgat J, Chandy KG, Garcia ML, Gutman GA, MartinEauclaire M-F, van der Walt JJ, Possani LD. Trends Pharmacol. Sci. 1999:20;444. 40. Rodriguez de la Vega RC, Merino E, Becerril B, Posan LD. Trends Pharmacol. Sci. 2003:24;222. 41. Batista CVF, del Pozo L, Zamudio FZ, Contreras S, Becerril B, Wanke E, Possani LD. J. Chromatogr. B. 2004:803; 55. 42. Legros C, Ceard B, Bougis PE, Martin-Eauclaire MF. FEBS Lett. 1998:431;375. 43. Corona M, Zurita M, Possani LD, Becerril B. Toxicon. 1996:34;251. 44. Srinivasan KN, Sivaraja V, Huys I, Sasaki T, Cheng B, Kumar TKS, Sato K, Tytgat J, Yu C, Cheng San BC, Ranganathan

Part II

Conclusion

References 1329

1330 Part II

Part II

45. 46. 47. 48. 49.

50. 51. 52. 53.

54. 55.

Pharmaceutical Sciences

S, Bowie HJ, Kini RM, Gopalakrishnakone P. J. Biol. Chem. 2002:277;30040. Possani LD, Martin BM, Svendsen I. Carlsberg Res. Commun. 1982:47;285. Dauplais M, Gilquin B, Possani LD, Gurrola-Briones G, Roumestand C, Menez A. Biochemistry. 1995:34; 16563. Bontems F, Roumestand C, Gilquin B, Menez A, Toma F. Science. 1991:254;1521. Miller C, Moczydlowski E, Latorre R, Philipps M. Nature. 1985:313;316. Possani LD, Becerrill B, Tytgat J, Delepierre M. In: Lopatin A, Nichols CG (Eds). Ion Channel Localization Methods and Protocols. Humana Press Inc., Totowa, NJ, 2000, pp 145– 66. Olamendi-Portugal T, Gomez-Lagunas F, Gurrola GB, Possani LD. Biochem. J. 1996:315;977. Olamendi-Portugal T, Gomez-Lagunas F, Gurrola GB, Possani LD. Toxicon. 1998:36;759. Lebrun B, Romi-Lebrun R, Martin-Eauclaire M-F, Yasuda A, Ishiguro M, Oyama Y, Pongs O, Nakajima T. Biochem. J. 1997:328;321. Kharrat R, Mabrouk KP, Crest M, Darbon H, Oughideni R, Martin-Eauclaire M-F, Jacquet G, El Ayeb M, Van Rietschoten J, Rochat H, Sampieri F. Eur. J. Biochem. 1996:242;491. Blanc E, Sabatier JM, Kharrat R, Meunier S, El Ayeb M, Van Rietschoten J, Darbon H. Proteins Struct. Funct. Genet. 1997:29;321. Becerril B, Marangoni S, Possani LD. Toxicon. 1997:35; 821.

56. Guijarro IJ, M’Barrek S, Olamendi-Portugal T, GomezLagunas F, Garnier D, Rochat H, Sabatier J-M, Possani LD, Delepierre M. Protein Sci. 2003:12;1844. 57. Savarin P, Romi-Lebrun R, Zinn-Justin S, Lebrun B, Nakajima T, Gilquin B, M´enez A. Protein Sci. 1999:8;2672. 58. Nilges M, Macias MJ, O’Donoghue SI, Oschkinat H. J. Mol. Biol. 1997:269;408. 59. Boisbouvier J, Albrand J-P, Blackledge M, Jaquinod M, Schweitz H, Lazdunski M, Marion D. J. Mol. Biol. 1998:283;205. 60. Korolkova YV, Bocharov EV, Angelo K, Maslennikov IV, Grinenko OV, Lipkin AV, Nosyreva ED, Pluzhnikov KA, Olesen SP, Arseniev AS, Grishin EV. J. Biol. Chem. 2002:277;43104. 61. Gurrola GB, Rosati B, Rocchetti M, Pimienta G, Zaza A, Arcangeli A, Olivotto M, Possani LD, Wanke E. FASEB J. 1999:13;953. 62. Zhang M, Korolkova YV, Liu J, Jiang M, Grishin EV, Tseng GN. Biophys. J. 2003:84;3022. 63. Pardo-Lopez L, Zhang M, Liu J, Jiang M, Possani LD, Tseng GN. J. Biol. Chem. 2002:277;16403. 64. Torres AM, Bansal P, Alewood PF, Bursill JA, Kuchel PW, Vandenberg JI. FEBS Lett. 2003:539;138. 65. Frenal K, Xu C-Q, Wolff N, Wecker K, Gurrola G, Zhu S-Y, Chi C-W, Possani LD, Tytgat J, Delepierre M. Proteins Struct. Funct. Bioinform. 2004:56(2);367–75. 66. Huys I, Xu C-Q, Wang C-Z, Vacher H, Martin-Eauclaire M-F. Biochem. J. 2004:378;745. 67. Craik DJ. Toxicon. 2001:39;1809. 68. Zhu S, Darbon H, Dyason K, Verdonk F, Tytgat J. Faseb J. 2003:17;1765.

1331

Christian Damblon and Gordon C.K. Roberts Biological NMR Centre, Department of Biochemistry, University of Leicester, Leicester LE1 7RH, UK

Introduction The β-lactam antibiotics are among the most useful antibacterial chemotherapeutic agents, for both human and animal use, but their efficiency is continuously being challenged by the emergence of resistant strains of pathogenic bacteria. Production of β-lactamases, which inactivate these antibiotics by hydrolyzing their endocyclic amide bond, is the most important resistance mechanism [1]. β-Lactamases have been divided into four classes on the basis of their amino acid sequences and catalytic mechanisms [2]. Class B enzymes, the metallo-β-lactamases (MBLs), are ∼30 kDa metallo-proteins which require one or two zinc ion(s) for their activity [3]. In the last 20 years, MBL-mediated resistance has appeared in several pathogenic strains including Bacteroides fragilis, Aeromonas hydrophila, Stenotrophomonas maltophilia, and Serratia marcescens [4], and some MBL genes are being disseminated rapidly by horizontal transfer, involving both plasmid and integron-borne genetic elements [5–9]. The spread of MBL-mediated bacterial resistance to βlactams is thus a matter of concern, particularly in view of the very wide substrate profile displayed by many MBLs, which can hydrolyze almost all the known β-lactam antibiotics, including the carbapenems such as imipenem [10]. The MBLs share a small number of conserved motifs, but otherwise show significant sequence diversity. Crystal structures have been determined for the MBLs from B. cereus [11,12], B. fragilis [13–15], Chryseobacterium meningosepticum [16] and S. maltophilia [17], and for the IMP-1 enzyme from Pseudomonas aeruginosa [18]. These structures reveal a unique “αββα. sandwich” characteristic of this family of enzymes and distinct from other zinc-dependent amide hydrolases such as thermolysin and carboxypeptidase A. It has subsequently become apparent that the metallo-β-lactamase fold is a widespread one [19–21]. Two zinc cations are present in the active site of the MBLs for which crystal structures are available. For example, in site I of the B. cereus enzyme, the zinc is co-ordinated by the imidazole rings of three histidine residues, H86, H88, and H149 (H116, H118, and H196 according to the standard numbering of MBLs [22]) and Graham A. Webb (ed.), Modern Magnetic Resonance, 1331–1337.
C 2008 Springer.

one water molecule. This water (or hydroxide) bridges to the zinc in site II, which is also co-ordinated by a histidine (H210 in B. cereus), a cysteine, an aspartate and a second water (or a carbonate ion). The zinc ligands are not absolutely conserved among MBLs, some enzymes having an Asn in place of a His in position 116 (site I ligand) and others having an Asp in place of the Cys in site II. These differences may contribute to some of the observed differences in substrate profiles and zinc affinities between different MBLs. The precise role of the two metals in catalysis is still unclear (e.g. [3,23–26]); mechanisms have been proposed in which only the zinc in site I is involved in catalysis [23], and in which both zinc cations play essential catalytic roles [3,26]. In fact, it remains unclear whether the mononuclear or binuclear enzyme is the more physiologically relevant; recent measurements suggest that at physiological metal ion concentrations most MBLs may exist as the apoenzyme, and that the presence of substrate leads to enhanced zinc binding in the mononuclear form of the enzyme [27]. It is notable that, while the presence of the second zinc does seem to be necessary for obtaining the maximum catalytic efficiency in MBLs, most of them can catalyze substrate hydrolysis with only one zinc cation bound [28–30]. Indeed, catalytic activity of the A. hydrophila MBL is actually inhibited by the binding of a second zinc cation [30]. It has been proposed that the observed conservation of the ligating residues at site II may be explained if a translocation of the metal ion between the two metal sites is involved in catalysis, and that this might explain the observation of catalytic activity with only one zinc per molecule of enzyme. The range of active site architectures for the MBLs makes the discovery of useful broad-spectrum inhibitors a challenging task, although compounds of several different chemical classes which inhibit the enzyme in vitro have been described [15,18,23,31–44]. Structural information, from crystallography or NMR, is available for the binding of a biphenyl tetrazole, which binds to the zinc in site II [31], and for a number of mercaptocarboxylates [18,41,44,45]. We recently demonstrated that a simple mercaptocarboxylate, thiomandelic acid

Part II

Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β-Lactamases

1332 Part II

Pharmaceutical Sciences

Part II

(β-mercapto phenylacetic acid), is a reasonably potent (sub-micromolar) inhibitor for most MBLs, except for the enzyme from A. hydrophila [45]. The kinetic data for thiomandelic acid fitted a competitive pattern of inhibition, and no evidence was obtained for irreversible binding of thiomandelic acid, at least to the B. cereus enzyme. Such a wide spectrum of activity against MBLs is unprecedented in previously published data concerning MBL inhibitors. For instance, thiomandelic acid is only about 25-fold less potent as an inhibitor of the enzyme from B. fragilis than of the IMP-1 enzyme from P. aeruginosa. This is in marked contrast to the reported thioester inhibitors, which are very much poorer inhibitors of the B. fragilis enzyme [37,38,43]. To obtain structural information to guide the design of improved broad-spectrum MBL inhibitors, we have used NMR to study the interaction of thiomandelic acid with the B. cereus MBL, BcII, particularly with its metal ions. The approaches we have used are described below.

Effect of Inhibitor Binding on the Backbone Amide Resonances A two-dimensional proton–nitrogen correlation spectrum (1 H-15 N HSQC) of 15 N-labelled BcII allows the rapid observation of most of the backbone amide 1 H and 15 N resonances. We have assigned these resonances to individual residues in the free protein [45], and monitoring changes in these resonances thus provides a convenient method for identifying the regions of the enzyme affected by inhibitor binding—the “chemical shift mapping” approach has become a widely used method for the initial characterization of protein–ligand and protein–protein interactions. Figure 1 shows a comparison of the 1 H-15 N HSQC spectra of the free enzyme and of the enzyme in the presence of R-thiomandelic acid (R-TM). Most of the spectrum remains essentially unaffected, but some significant changes are observed, generally involving residues which are close to the zinc binding sites. Only very few residues in the secondary structure elements of the core of the protein are affected by the inhibitor; some 90 residues in this part of the structure which give well-resolved resonance signals are unaffected by the addition of inhibitors. While the resonance assignment for the free enzyme is complete, the resonances in the complex are not yet assigned, so that the magnitude of the shifts of individual resonances cannot be accurately determined. We have thus used the “minimum chemical shift” approach [46–48], in which the chemical shift difference from a given crosspeak in the free protein to the closest cross-peak in the complex is calculated, to identify the residues most affected by inhibitor binding; these are indicated on the structure in Figure 1. As would be expected for a competitive inhibitor, many of the residues showing significant

changes in chemical shift are in the active-site region of the structure, near the zinc ions. Only two or three residues in the structural core of the enzyme show significant changes. Substantial amide chemical shift perturbations are, however, observed for residues in the “flexible-flap” region (the β3 –β4 loop, residues 30–40), which in the absence of inhibitor is some distance from the active site (Figure 1). Shift changes are seen for E30, L31, F34, and V39 in this loop. (A significant shift perturbation is also observed for W59, which is located at the beginning of a loop on the top of the cavity; since the side-chain of W59 extends towards the β3 –β4 loop, this could result from a movement of the loop.) There is good evidence that this loop plays a role in ligand binding and catalytic activity [18,41,44,45,49,50]. Loop deletion and loop swapping experiments with the B. fragilis, B. cereus and IMP-1 MBLs [49,50] have shown that the loop influences not only the K i of inhibitors but also both the K M and kcat of substrates, suggesting that it plays a role in the catalytic mechanism. In the structure of the complex of IMP-1 MBL with a large mercaptocarboxylate inhibitor [18], there are clear hydrophobic contacts between the inhibitor and residues of the β3–β4 loop. These interactions might stabilize the loop, which is flexible in the absence of the inhibitor [44], in a “closed” position. However, our own studies with the B. cereus enzyme have shown that the residues of this loop are affected by the binding of much smaller inhibitors, not only thiomandelic acid [45] but even the simple mercaptoacetic acid [Kumar, S. unpublished work]. This clearly shows that hydrophobic interactions with the bound ligand cannot be the origin of the loop movement; there must be a “trigger” mechanism related to the coordination of the metal ions.

Effect of Inhibitor Binding on the Imidazole Resonances of the Metal Ligands In examining the effect of inhibitor binding on the active site of the B. cereus enzyme, the imidazole resonances of the metal-binding histidine residues (H86, H88, and H149 in site I, H210 in site II) provide valuable probes. The imidazole NH resonances appear between 12 and 15 ppm (Figure 2). They have been assigned to individual residues by using 1 H-15 N HMQC spectra optimized for observation of the long-range 1 H-15 N couplings in the imidazole ring [29], which allow one to connect the NH resonances to those of the two nitrogens and the two CH protons in each imidazole. At pH 6.4, 298K, in the absence of inhibitors, relatively sharp resonances are observed for H86 and H88, and a broad resonance for H210, but no signal is apparent for H149. At lower pH or temperature, the H210 resonance sharpens and a signal for H149 appears. These differences in linewidth reflect the

Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β

Part II

101

Effect of Inhibitor Binding on the Imidazole Resonances 1333

T150 G36

G148 G167

106

D90

111

S190 G179

S57 D152 G211

G214

G209

W59

V170

V155 S41 L178

116

R91

H86

I188

N118

L169 S58

N

L220 A184

D56

121 W189

E30 L31

126

C168

V207

L21 F34 V39 I84

131

K171

H149

11.1

10.6

10.1

9.6

9.1

8.6

8.1

7.6

7.1

6.6

6.1

5.6

H

Fig. 1. 1 H-15 N HSQC spectra of B. cereus MBL BcII alone in black, and in its complex with R-TM in red. Assignments of residues affected by ligand binding are labelled on the HSQC spectra and are displayed on the X-ray structure of the free BcII. Residues affected in the flexible β3-β4 loop are labelled in blue on the HSQC and on the structure. Other residues affected by the presence of the inhibitor are coloured in green on the structure. (See also Plate 103 on page XXI in the Color Plate Section.)

1334 Part II

Pharmaceutical Sciences

Part II

Fig. 2. Part of the 1 H NMR spectrum of B. cereus MBL BcII in the presence and absence of R-TM, showing the imidazole NH resonances of the metal-binding histidine residues.

differences in accessibility of the histidine imidazole NHs for exchange with the solvent. Gradual addition of Rthiomandelic acid to the enzyme results in a progressive decrease in the intensity of the imidazole NH signals from the free enzyme, and a progressive increase in a new set of signals attributable to the enzyme-inhibitor complex. The changes are complete at a 1:1 ratio of inhibitor to enzyme, and the spectrum of the fully formed complex is shown in Figure 2. It is clear that the imidazoles in both metal binding sites are markedly affected by inhibitor binding. In site I, both H86 and H88 show large (>0.5 ppm) changes in NH chemical shift on inhibitor binding. In addition, inhibitor binding clearly tends to decrease the rate of exchange with water of the imidazole NHs of the metalbinding histidines, thus sharpening their resonances, although the magnitude of this effect varies significantly from one residue to another. Thus, the resonances of H86 and H88 are reasonably sharp in the spectrum of the free enzyme, and their linewidths are little affected by inhibitor binding, while the resonance of H210 is broad in the spectrum of the free enzyme and is markedly sharpened by the binding of R-thiomandelic acid so that in the complex it has approximately the same linewidth as H86 and H88. The NH resonance of H149, on the other hand, is too broad to see in the spectrum of the free enzyme; only the binding of the inhibitor decreases the imidazole NH exchange rate of this residue sufficiently to yield an observable resonance—in fact one as sharp as those of the other three active site histidines in this complex. The binding of R-thiomandelic acid thus changes not only the timeaverage environment of the zinc-ligating residues, as reflected in their chemical shifts, but also their dynamics, as reflected in the rate of exchange of their NH protons with water.

Direct Observation of the Active-Site Metals In the majority of MBLs, zinc can be exchanged with cadmium to yield catalytically active enzymes [51], and in the case of CcrA, the structure of the cadmium-substituted enzyme has been shown to be essentially identical to that of the zinc enzyme [52]. Isotopes of cadmium provide very convenient spectroscopic probes, allowing direct studies of the co-ordination and dynamics of the metal ion. We have used NMR and, in collaboration with Prof. R. Bauer (Copenhagen), perturbed angular correlation of γ rays (PAC) spectroscopy [29,45,53,54] to make detailed studies of metal binding to the B. cereus MBL. The enzyme with two cadmium ions bound gives a 113 Cd NMR spectrum containing two distinct resonances at ∼140 and ∼260 ppm [29]. We have assigned these two signals to the two metal binding sites of the enzyme by using a variety of multinuclear NMR experiments, particularly 113 Cd-edited 1 H and 1 H-15 N HSQC experiments, which allow us to connect the 113 Cd signals to assigned 1 H and 15 N resonances from the metal ligands (see Figure 3; [29,45,53,54]). Analysis of the 1 H-15 N HSQC spectra of the B. cereus MBL enzyme with only one cadmium ion bound demonstrates the existence of a mononuclear cadmium enzyme, which is distinct from the apoenzyme, and from the species with two cadmiums bound. This mononuclear enzyme shows one single broad 113 Cd NMR signal at ∼175 ppm, suggesting at first sight that cadmium occupies only one of the two sites under these conditions; however, two well defined cadmium PAC signals are observed [53]. The only explanation for these observations comes from considering the dynamics of metal binding and the different time regimes monitored by the two methods.

Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β

Effects of Thiomandelate Binding on the 113 Cd Spectrum 1335

The timescales to which the two methods are sensitive are quite different: 111m Cd-PAC experiments can monitor dynamics in a time regime from about 0.1 ns to 100 ns, whereas chemical exchange effects in 113 Cd NMR, with large chemical shift differences arising from differences in coordination, can typically monitor dynamics from 0.01 to 10 ms. The most straightforward explanation consistent with both the NMR data and the PAC data at [Cd(II)]/[E] ratios ≤1, is that a single cadmium bound to the enzyme jumps between the two binding sites on a time scale between 100 ns and 0.01 ms. Thus, the long-held assumption that the two different macroscopic dissociation constants for metal ion binding to the enzyme, reflecting a higher affinity of the metal for site I was shown to be wrong; they in fact reflect negative co-operativity in metal binding [53]. Recent studies of the wild-type and mutant enzyme by optical spectroscopy and EXAFS experiments

indicated that there is a similar negative co-operativity in the binding of zinc and cobalt [55].

Effects of Thiomandelate Binding on the 113 Cd Spectrum The BcII gives a 113 Cd NMR spectrum containing two distinct resonances at ∼140 and ∼260 ppm [29,53]. Addition of one molar equivalent of R-thiomandelic acid to the B. cereus MBL with two cadmium ions bound leads to a substantial downfield shift of both the 113 Cd resonances, to ∼343 and ∼372 ppm (Figure 3). Addition of 0.5 molar equivalent of thiomandelate leads to a spectrum containing both resonances corresponding to the free enzyme, and those corresponding to the complex, with approximately equal intensities, indicating that thiomandelate binding is

Part II

Fig. 3. NMR spectra of di-cadmium BcII in the presence of one equivalent of R-TM. (A) 1 H-113 Cd HMQC spectrum; (B) 1 H-15 N HMQC spectrum of the imidazole resonances, allowing assignment of the imidazole cross-peaks in A. Signal splittings in the nitrogen dimension due to the N-Cd coupling are observed for the four active site histidines. (C), Model of the binding of R-TM to the active site of BcII [49].

1336 Part II

Pharmaceutical Sciences

Part II

in slow exchange on the 113 Cd NMR timescale. Further addition of the inhibitor from one to two molar equivalents does not change the NMR spectra. The direction and magnitude of the inhibitor-induced change in chemical shift is consistent with that expected for the co-ordination of a sulphur atom to each of the cadmium ions, and hence with the idea that the inhibitor binds directly to both metal sites. Direct evidence for this, and the assignment of the two 113 Cd resonances in the inhibitor complexes, can be obtained from the combination of 1 H-113 Cd HMQC spectra and 1 H-15 N HSQC spectrum optimized for observation of imidazole resonances [29], shown in Figure 3. In the 1 H113 Cd HMQC spectrum, cross-peaks are observed from the 343 ppm 113 Cd resonance for three imidazoles, His88, His86, and His149. The His88 can easily be assigned because it is the only metal-bound histidine which is in the NεH tautomeric form, and thus readily recognized in the 1 H-15 N HSQC spectrum1 . The 343 ppm 113 Cd resonance can thus be unambiguously assigned to the cadmium in site I. There is also a strong cross-peak to a signal at 5.36 ppm, which can be assigned to the CαH (benzylic proton) of the bound R-thiomandelate (R-TM). This inhibitor resonance also gives a strong cross-peak to the low-field 113 Cd resonance (Figure 3), which in turn shows cross-peaks to the imidazole protons of His210 and the βproton(s) of Cys168, and can thus be assigned to the metal in site II. The observation that, in a complex containing one molar equivalent of inhibitor, the R-TM resonance at 5.36 ppm gives a cross-peak in the 1 H-113 Cd HMQC spectrum to both 113 Cd resonances, is clear evidence that R-TM co-ordinates through its sulphur atom to both cadmium ions. As described above, the 113 Cd spectrum of the enzyme having one cadmium equivalent bound contains a single resonance, reflecting a rapid intramolecular exchange of the single metal between the two sites. Addition of one molar equivalent of R-TM to the mono-cadmium enzyme led to a 113 Cd spectrum having two resonances at the same chemical shifts as those observed for the complex of the di-cadmium enzyme. Furthermore, the 113 Cd-edited 1 H spectra of the R-TM complex demonstrated that the thiomandelate CαH resonance at 5.35 ppm showed 1 H113 Cd scalar coupling to two cadmium nuclei, just as observed for the di-cadmium enzyme, demonstrating that the thiomandelate-enzyme complex contains two cadmium ions. These observations clearly demonstrate that addition of thiomandelate to a sample of enzyme containing cadmium at a [Cd]/[E] ratio of one, leads to the formation 1

The other two histidines have not been individually assigned in the cadmium-substituted enzyme, but can be tentatively assigned by comparison with the assigned resonances in the zinc enzyme [45].

of the R-TM complex of the di-cadmium enzyme together with a corresponding amount of apoenzyme [54]. This was confirmed by analysis of the backbone amide 1 H-15 N HSQC spectrum. The presence of the inhibitor, which binds to both metal ions, thus induces positive cooperativity in metal binding, in marked contrast to the negative co-operativity in cadmium binding observed in the absence of inhibitor [29].

Conclusion The power of NMR in determining the three-dimensional structure of proteins and their complexes in solution is well-established. Full structure determination of course provides the most detailed picture of a drug–protein complex, and in the present case the determination of the structure of the complex of R-thiomandelic acid with the B. cereus MBL is in progress. However the experiments described here demonstrate some of the range of NMR experiments which can provide very valuable structural and dynamic information on ligand binding short of a full structure; others are reviewed elsewhere [56,57]. Experiments of this kind, which can be carried out rapidly, have considerable value in understanding drug–protein interactions and in comparing the binding of related compounds to a target, and promise to have a continuing place in structure-based drug design.

References 1. Frere JM, Mol. Microbiol. 1995;16(3):385. 2. Matagne A, Dubus A, Galleni M, Fr`ere JM. Nat. Prod. Rep. 16,1 (1999). 3. Wang Z, Fast W, Valentine AM, Benkovic SJ. Curr. Opin. Chem. Biol. 1999;3(5):614. 4. Payne DJ. Med J. Microbiol. 1993;39(2):93. 5. Laraki N, Franceschini N, Rossolini GM, Santucci P, Meunier C, de Pauw E, Amicosante G, Frere JM, Galleni M. Antimicrob. Agents Chemother. 1999;43(4):902. 6. Riccio ML, Franceschini N, Boschi L, Caravelli B, Cornaglia G, Fontana R, Amicosante G, Rossolini GM. Antimicrob. Agents Chemother. 2000;44(5):1229. 7. Iyobe S, Kusadokoro H, Ozaki J, Matsumura N, Minami S, Haruta S, Sawai T, O’Hara K. Antimicrob. Agents Chemother. 2000;44(8):2023. 8. Chu YW, Afzal-Shah M, Houang ET, Palepou MF, Lyon DJ, Woodford N, Livermore DM. Antimicrob. Agents Chemother. 2001;45(3):710. 9. Yano H, Kuga A, Okamoto R, Kitasato H, Kobayashi T, Inoue M. Antimicrobial. Agents and Chemother. 2001;45(5): 1343. 10. Livermore DM, Woodford N. Curr. Opin. Microbiol. 2000; 3(5):489. 11. Fabiane SM, Sohi MK, Wan T, Payne DJ, Bateson JH, Mitchell T, Sutton BJ, Biochemistry 1998;37(36):12404.

Structure and Dynamics of Inhibitor and Metal Binding to Metallo-β

35. Toney JH, Hammond GG, Fitzgerald PMD, Sharma N, Balkovec JM, Rouen GP, Olson SH, Hammond ML, Greenlee ML, Gao YD, J. Biol. Chem. 2001;276(34):31913. 36. Goto M, Takahashi T, Yamashita F, Koreeda A, Mori H, Ohta M, Arakawa Y. Biol. Pharm. Bull. 1997;20(11):1136. 37. Payne DJ, Bateson JH, Gasson BC, Proctor D, Khushi T, Farmer TH, Tolson DA, Bell D, Skett PW, Marshall AC, Reid R, Ghosez L, Combret Y, Marchand-Brynaert J. Antimicrob. Agents Chemother. 1997;41(1):135. 38. Payne DJ, Bateson JH, Gasson BC, Khushi T, Proctor D, Pearson SC, Reid R, Fems. Microbiol. Lett. 1997;157: 171. 39. Payne DJ, Bateson JH, Gasson BC, Khushi T, Proctor D, Pearson SC, Reid R. FEMS Microbiol. Lett. 1997;157(1):171. 40. Page MI, Laws AP. Chem. Commun. 1998;16:1609. 41. Scrofani SDB, Chung J, Huntley JJA, Benkovic SJ, Wright PE, Dyson HJ. Biochemistry. 1999;38(44):14507. 42. Greenlee ML, Laub JB, Balkovec JM, Hammond ML, Hammond GG, Pompliano DL, Epstein Toney JH. Bioorgan. Med. Chem. Lett. 1999;9(17):2549. 43. Hammond GG, Huber JL, Greenlee ML, Laub JB, Young K, Silver LL, Balkovec JM, Pryor KD, Wu JK, Leiting B, Pompliano DL, Toney JH. FEMS Microbiol. Lett. 1999;179(2):289. 44. Huntley JJA, Scrofani SDB, Osborne MJ, Wright PE, Dyson HJ. Biochemistry. 2000;39(44):13356. 45. Mollard C, Moali C, Papamicael C, Damblon C, Vessilier S, Amicosante G, Schofield CJ, Galleni M, Frere JM, Roberts GCK. J. Biol. Chem. 2001;276(48):45015. 46. Lian LY, Barsukov I, Golovanov AP, Hawkins DI, Badii R, Sze KH, Keep NH, Bokoch GM, Roberts GCK. Structure 2000;8(1):47. 47. Farmer BT. Nat. Struct. Biol. 1996;3(12):995. 48. Williamson RA, Carr MD, Frenkiel TA, Feeney J, Freedman RB. Biochemistry. 1997;36(45):13882. 49. Yang Y, Keeney D, Tang X, Canfield N, Rasmussen BA, J. Biol. Chem. 1999;274(22):15706. 50. Moali C, Anne C, Lamotte-Brasseur J, Groslambert S, Devreese B, Van Beeumen J, Galleni M, Frere JM. Chem. & Biol. 2003;10(4):319. 51. Paul-Soto R, Bauer R, Frere JM, Galleni M, Meyer-Klaucke W, Nolting H, Rossolini GM, de Seny D, HernandezValladares M, Zeppezauer M, Adolph HW. Biol. J. Chem. 1999;274(19):13242. 52. Concha NO, Rasmussen BA, Bush K, Herzberg O, Protein Sci. 1997;6(12):2671. 53. Hemmingsen L, Damblon C, Antony J, Jensen M, Adolph HW, Wommer S, Roberts GCK, Bauer R, Amer. J. Chem. Soc. 2001;123:10329. 54. Damblon C, Jensen M, Ababou A, Barsukov I, Papamicael C, Schofield CJ, Olsen L, Bauer R, Roberts GCK, J. Biol. Chem. 2003;278(31):29240. 55. de Seny D, Heinz U, Wommer S, Kiefer M, Meyer-Klaucke W, Galleni M, Frere JM, Bauer R, Adolph HW. Biol. J. Chem. 2001;276(48):45065. 56. Roberts GCK, Curr. Opin. Biotech. 1999;10(1):42. 57. Roberts GCK. Drug Discov. Today. 2000;5(6):230.

Part II

12. Carfi A, Du E´ee, Galleni M, Fr`ere JM and Dideberg O. Acta Cryst. 1998;D54:313. 13. Carfi A, Duee E, Paul-Soto R, Galleni M, Frere JM, Dideberg O, Acta Crystallogr. D. Biol. Crystallogr. 1998;54(Pt 1):45. 14. Concha NO, Rasmussen BA, Bush K, Herzberg O. Structure 1996;4(7):823. 15. Fitzgerald PM, Wu JK, Toney JH. Biochemistry 1998; 37(19):6791. 16. Garcia-Saez I, Hopkins J, Papamicael C, Franceschini N, Amicosante G, Rossolini GM, Galleni M, Frere JM, Dideberg O. J. Biol. Chem. 2003;278(26):23868. 17. Ullah JH, Walsh TR, Taylor IA, Emery DC, Verma CS, Gamblin SJ, Spencer J. Mol. J. Biol. 1998;284(1):125. 18. Concha NO, Janson CA, Rowling P, Pearson S, Cheever CA, Clarke BP, Lewis C, Galleni M, Frere JM, Payne DJ, Bateson JH, Abdel-Meguid SS, Biochemistry. 2000;39(15):4288. 19. Aravind L. In Silico Biol. 1999;1(2):69. 20. Daiyasu H, Osaka K, Ishino Y, Toh H. FEBS. Lett. 2001; 503(1):1. 21. Melino S, Capo C, Dragani B, Aceto A, Petruzzelli R. Trends Biochem. Sci. 1998;23(10):381. 22. Galleni M, Lamotte-Brasseur J, Rossolini GM, Spencer J, Dideberg O, Frere JM. Antimicrob. Agents Chemother. 2001; 45(3):660. 23. Bounaga S, Laws AP, Galleni M, Page MI. Biochem. J. 1998; 331(Pt 3):703. 24. Rasia RM, Vila AJ. Biochemistry. 2002;41(6):1853. 25. Wang ZG, Fast W, Benkovic SJ. Biochemistry. 1999;38(31): 10013. 26. Cricco JA, Orellano EG, Rasia RM, Ceccarelli EA, Vila AJ. Coordination Chem. Rev. 192,519 (1999). 27. Wommer S, Rival S, Heinz U, Galleni M, Frere JM, Franceschini N, Amicosante G, Rasmussen B, Bauer R, Adolph HW. J. Biol. Chem. 2002;277(27):24142. 28. Paul-Soto R, Hernadez-Valladares M, Galleni M, Bauer R, Zeppezauer M, Fr`ere JM, Adolph HW. FEBS Lett. 1998;438: 137. 29. Damblon C, Prosperi C, Lian LY, Barsukov I, Paul-Soto R, Galleni M, Fr`ere JM, Roberts GCK. J. Am. Chem. Soc. 1999;121:11575. 30. Hernandez Valladares M, Felici A, Weber G, Adolph HW, Zeppezauer M, Rossolini GM, Amicosante G, Frere JM, Galleni M. Biochemistry. 1997;36(38):11534. 31. Toney JH, Fitzgerald PM, Grover-Sharma N, Olson SH, May WJ, Sundelof JG, Vanderwall DE, Cleary KA, Grant SK, Wu JK, Kozarich JW, Pompliano DL, Hammond GG. Chem. Biol. 1998;5(4):185. 32. Nagano R, Adachi Y, Imamura H, Yamada K, Hashizume T, Morishima H, Antimicrob. Agents Chemother. 1999;43(10): 2497. 33. Walter MW, Felici A, Galleni M, Paul-Soto R, Adlington RM, Baldwin JE, Fr`ere JM, Gololobov M, Schofield CJ. Bioorg. Med. Chem. Lett. 1996;6(20):2455. 34. Walter MW, Hernandez Valladres M, Adlington RM, Amicosante G, Baldwin JE, Fr`ere JM, Galleni M, Rossolini GM, Schofield CJ. Bioorg. Chem. 1999;27:35.

References 1337

1339

David A. Gell and Joel P. Mackay School of Molecular and Microbial Biosciences, University of Sydney, Sydney, NSW 2006, Australia

Introduction Specific interactions between pairs or larger sets of proteins are central to all processes that take place within an organism, and many of these interactions may prove to be suitable targets for therapeutic intervention. In order to target such complexes effectively, detailed structural information is highly desirable, and NMR spectroscopy is one of the methods of choice for providing this information. Recent methodological advances have enabled full structure determination for increasingly large proteins and protein complexes. These methods include TROSY pulse sequences, sample deuteration, and the use of additional long-range structural restraints derived from residual dipolar couplings (RDCs) and paramagnetic effects to complement traditional NOE-based methods. For protein–protein complexes, filtered/edited NOE experiments with differentially labeled samples greatly reduce spectral complexity through selecting only intra- or intermolecular NOEs. While NMR cannot currently compete with X-ray crystallography in the determination of the structures of huge macromolecular machines such as the ribosome, it has many advantages for the study of more transient protein complexes, for which crystals are less likely to form. Relatively weak or transient interactions (high µM–mM dissociation constants) allow cells to respond dynamically to rapidly changing conditions; for example, many cell surface receptor interactions [1] and electron transport processes [2] rely on transient interactions. Likewise, transcription factor complexes are often formed from combinations of low-affinity interactions [3], facilitating exchange of factors that alter the transcriptional activity of the complex [4]. For weak protein complexes full structure determination by NMR may be prevented by exchange processes, but protein–protein interfaces and association constants may still be determined using chemical shift perturbation experiments. In such cases, shift perturbation data or long-range conformational restraints can be used to drive the docking of a complex, based on known structures of subunit components.

Graham A. Webb (ed.), Modern Magnetic Resonance, 1339–1346.
C 2008 Springer.

This chapter will outline the approaches mentioned above, together with experimental difficulties that are likely to be encountered. Examples both from our own work and from that of others will be used for illustration.

Tackling the Size Issue for Larger Protein Complexes In general, the study of protein complexes involves dealing with larger species. Unfortunately, the slower tumbling (measured by τ c , the molecular correlation time) of larger species leads to increased transverse relaxation rates. Faster transverse relaxation reduces the efficiency with which magnetization can be transferred through scalar couplings. As a consequence, the experiments used to make resonance assignments perform more poorly as τ c increases. This has effectively placed a rough upper limit of ∼30 kDa on the size of a system that can be solved using standard triple resonance methods [5]. However, several recently developed approaches have opened the door to the analysis of much larger proteins and protein complexes.

Deuteration The first strategy involves the preparation of protein samples in which some percentage of protons is substituted by deuterons. This can be achieved by protein over-expression in cells grown on deuterated media [6]. Two advantages are conferred by the incorporation of 2 H. Firstly, the relaxation of 13 C nuclei is brought about primarily through the dipolar interaction with their attached proton(s). Because the magnitude of the dipolar interaction is directly proportional to the gyromagnetic ratios of the nuclei involved, substitution of 1 H for 2 H (γ H /γ D = 6.5) reduces the magnitude of this interaction significantly. With the addition of 2 H decoupling that removes the line broadening effects that arise from scalar coupling to a quadrupolar nucleus, 13 C line widths are dramatically reduced. Secondly, the general reduction in

Part II

NMR Spectroscopy in the Analysis of Protein–Protein Interactions

1340 Part II

Pharmaceutical Sciences

Part II

proton density reduces spin diffusion significantly, resulting in narrower 1 H line widths. Growth medium for protein deuteration is made up in 2 H2 O, and cells often require acclimatization on this media to grow well [7]. To achieve complete deuteration (perdeuteration), labeled carbon sources are required in addition to 2 H2 O (e.g. 2 H, 13 C-glucose or acetate). The extent of deuteration aimed for will depend on the goal of the study. Perdeuteration, followed by back exchange of labile backbone and side chain amide deuterons for protons, as will generally occur during protein purification in 1 H2 O, will yield the greatest improvements in sensitivity and resolution for backbone triple resonance experiments and chemical shift perturbation experiments (see below). However, lower levels of deuteration, down to ∼50% [8], are required if side chain NOEs are to be detected. Alternatively, several groups have pioneered the use of biosynthetically directed labeling strategies that can be used to produce, for example, 2 H-, 13 C-, and 15 N-labeled proteins that bear protonated methyl groups on Leu, Val, and the δ1 methyl of Ile [9]. Such protein samples allow NOEs to be detected between methyl groups, which often lie in the protein core and provide a substantial number of structural constraints. Indeed, HN –HN , HN –methyl, and methyl–methyl NOEs alone may be sufficient to define the global fold of a large protein [6].

TROSY The other recent breakthrough that is revolutionizing the analysis of large systems is transverse relaxation optimized spectroscopy (TROSY) [10]. Transverse relaxation of amide 1 H and 15 N spins at high field strengths is dominated by two main mechanisms: dipole–dipole (DD) interactions and chemical shift anisotropy (CSA). Amide protons additionally experience significant relaxation (∼40% of T2 [11]) from remote 1 H spins, but this latter term can be largely eliminated by perdeuteration, as described above. Interference between DD and CSA relaxation (cross-correlation) has the effect of enhancing the relaxation of one component of the HN doublet while slowing the relaxation of the other; the two doublet components thus display different line widths. The same is true for the 15 N doublet, and an undecoupled 2 × 2 multiplet in an HSQC has one component that is considerably narrower in both dimensions. The degree of relaxation interference changes with field strength, because CSA but not DD relaxation is field dependent, and at ∼1 GHz the two relaxation mechanisms essentially cancel each other in the narrow component. The line width of this component consequently becomes relatively insensitive to the rotational correlation time of the protein, and the TROSY pulse sequence element (a simple modification

of the HSQC pulse sequence) selects this signal. TROSY promises a great deal for the analysis of large systems and high-quality TROSY-HSQC spectra have already been recorded of systems up to ∼900 kDa (a GroEL–GroES complex [12]). The TROSY element has been incorporated into a number of triple resonance experiments, and in combination with deuteration these sequences have been used to make massive strides in the size of systems that are amenable to solution NMR analysis. Backbone assignments of malate synthase, a 723-residue monomer [13], and aldolase, a 110-kDa octamer [14], have been reported recently. Perhaps even more excitingly, the structures of several integral membrane proteins solubilized in lipid micelles have been determined [15–17]. These latter studies hold substantial promise for the analysis of interactions between receptors and effector proteins. The benefits of TROSY and deuteration have already been applied to map the interaction surfaces of large protein complexes using chemical shift perturbation experiments. Complexes formed between Ras and Byr2 [18], calreticulin and Erp57 [19], p53 and Hsp90 [20], and FimC and FimH [21] have all been characterized in this fashion.

Reducing Complexity: Differential Isotope Labeling An additional issue that arises in NMR studies of protein complexes is the increase in the number of resonances. A simple method to alleviate this problem involves preparing complexes in which individual subunits carry different isotope labeling patterns. Most commonly, one protein will be 15 N- and 13 C-labeled, while its partner will be unlabeled. This arrangement has two advantages. First, HSQC spectra show signals from only the labeled partner, simplifying chemical shift titration experiments. Second, heteronuclear filters can be incorporated into pulse programs to either select or filter out magnetization from protons attached to specific nuclides, and hence select specific subsets of inter- or intramolecular resonances. The double half-filtered NOESY spectrum comprises four sub-spectra that are created by the linear combination of four data sets recorded with different phase cycles [22]. The most useful of these sub-spectra contain only NOEs between protons attached to different nuclides. For example, the 13 C(ω1 )-filtered/13 C(ω2 )-selected sub-spectrum of a 13 C(ω1 , ω2 ) double half-filtered NOESY will display NOEs between protons not bound to 13 C in ω1 and protons exclusively bound to 13 C in ω2 . Figure 1 shows such a sub-spectrum from a data set recorded on a complex of two zinc finger domains [23] in our laboratory, and many other studies have utilized this technique to gather NOE constraints for structure determination.

I215 QG2

I215 HG13

A235 QB/ T227 QG2

A235 QB

1

H chemical shift, ω1 (ppm) 3.50 G208 HA1 G208 HA2 4.00

V205 HA 1.00 1

0.50

Obtaining Long-Range Structural Information 1341

Fig. 1. Heteronuclear filtered/selected NOESY data. A portion of a two13 C-filtered, 13 C-edited dimensional NOESY spectrum of a 1:1 complex formed from 15 N and 13 C-labeled GATA-1 N-finger and unlabeled U-shaped finger 1, highlighting a number of intermolecular NOEs (Liew et al. (2005) PNAS 102, 583–8). This sub-spectrum contains only NOEs between protons attached to 12 C (ω2) and protons attached to 13 C (ω1).

0.00

H chemical shift, ω2 (ppm)

Obtaining Long-Range Structural Information A recent innovation has been the use of long-range structural information to supplement the conventional NOE data. Two methods for deriving long-range structure calculation restraints are described below; these methods have particular significance for protein complexes as they provide information about the relative orientation of subunits.

Line ‘em Up—Residual Dipolar Couplings In NMR spectra of species that do not undergo rotational motion, resonances are split not only by scalar couplings (J ), but also by dipolar couplings (D) involving nuclei that are nearby in space. These dipolar couplings can be extremely large (thousands of Hz), but average to zero if the molecule tumbles rapidly in solution. Because the magnitude of a dipolar coupling is dependent on the orientation of each internuclear vector relative to the static field, couplings between different pairs of nuclei reveal the orientation of those vectors relative to a common frame of reference (and therefore relative to each other). This represents valuable structural information, not least because it reports on the relative orientations of distant parts of a molecule or complex, in contrast to the more usual NOEs and scalar coupling constants [24]. The trick in creating a useful structural tool from this concept has been to create a physical environment where a protein exhibits a small degree of preferential alignment. Too much alignment results in the observation of many dipolar couplings that render spectra un-interpretable, but a small amount yields RDCs of up to ∼10–20 Hz that are observed only for spin pairs that are strongly coupled: generally one-bond interactions. Solutions containing low

concentrations of a highly aligned and asymmetric component provide suitable environments for these experiments. Bacteriophage, disc-shaped lipid bicelles (both of which align completely in a magnetic field) and compressed polyacrylamide gels have all been used [24], and in such media, the reorientation of a protein becomes slightly non-isotropic. The magnitudes of the RDCs can be tuned by altering the concentration of the alignment medium. These splittings (e.g. 1 DHNN , 1 DHαCα) can be measured directly in simple 2D and 3D NMR experiments such as undecoupled HSQC and HNCO spectra, and methodologies have been developed to incorporate these data into standard structure calculation programs [25]. It should be noted that there is some degeneracy in the derived data, with more than one possible relative orientation of a vector leading to the same RDC. This fact, together with issues with the energy landscapes generated by dipolar couplings in simulated annealing protocols, have somewhat hindered the widespread use of such data to date [24]. Dipolar coupling data have been demonstrated to be beneficial in a number of scenarios, including validation of NMR structures [26] and DNA structure determination [27]. One of their most powerful uses may prove to be in defining the relative orientations of domains in a multi-domain proteins [28–30] or of protein subunits in a complex [31,32].

Paramagnetic Nuclei Paramagnetic species cause a multitude of effects on the relaxation and chemical shifts of neighboring nuclei [33]. Paramagnetics induce both T1 and T2 relaxation in nearby nuclei through a number of mechanisms, and traditionally

Part II

I215 HG12/ L223 QD1

Analysis of Protein–Protein Interactions

1342 Part II

Pharmaceutical Sciences

G60

Mapping Protein–Protein Interfaces Chemical Shift Perturbation Experiments The chemical shift of a nuclear spin is extremely sensitive to the physical and chemical environment of the spin, and has been used extensively to map the amino acids in a protein that directly contact a binding partner. Figure 2A shows 1 H, 15 N HSQC data from the titration of a solution of unlabeled α-hemoglobin into uniformly 15 N-labeled alpha hemoglobin stabilizing protein (AHSP [42]). 1 H, 15 N HSQC spectra are generally used in such titrations because of their high sensitivity and excellent resonance dispersion. It can be seen that a subset of the signals shift significantly upon the addition of αhemoglobin, and it can be inferred that the amino acids that correspond to these signals (Figure 2B) either reside on the α-hemoglobin binding surface of AHSP, or are involved in a conformational change upon α-hemoglobin binding Feng et al. (2004) CELL 119, 629–40. Because of this latter possibility, care must be taken in interpreting the results of chemical shift titrations; complementary data from mutagenesis studies are often extremely useful [23,43]. Identification of binding surfaces is potentially an

B

106

108

G16

110

112 S19

N chemical shift (ppm)

A

nitroxide radical [41]), can be attached to proteins by chemical cross-linking.

15

Part II

cause problems in the structure determination of metalloproteins such as copper(II) plastocyanin. However, these effects show predictable distance and angle dependence, and can therefore provide useful structural information [33,34]. Pseudocontact shifts (PCSs) are chemical shift changes induced through the interaction of nuclei with the anisotropic susceptibility tensor χ of the paramagnetic center. These changes are dependent on the position of the nucleus relative to the χ tensor and fall off with the third power of distance, thereby providing useful long-range structural information [35]. Furthermore, interaction between the χ tensor and the strong static magnetic field can lead to preferential orientation of the protein, generating RDCs. Because PCSs are long-range effects, chemical shift changes can be induced across protein–protein interfaces and the information can therefore be used to orient protein complexes. This approach has been successful for complexes containing cytochromes [36–38] as well as for a DNA polymerase III complex [34]. Many proteins contain metal sites that can accommodate paramagnetic ions. For example many Ca2+ binding sites can bind paramagnetic lanthanide ions, which due to their highly anisotropic paramagnetic susceptibilities are good choices for PCS and RDC studies [39]. In the absence of natural metal binding sites, spin-labels, such as gadolinium-EDTA [40], cobalt-EDTA [35], and TEMPO (which contains a stable

Q61

9.0

8.8 1

114

S12

T57 8.6

8.4

8.2

8.0

H chemical shift (ppm)

Fig. 2. Chemical shift perturbation experiment. (A) Sections of 1 H, 15 N HSQC spectra of 15 N-labeled AHSP are shown in the absence (solid lines) and the presence (dashed line) of 1.2 molar equivalents of unlabeled α-hemoglobin. Resonance assignments are indicated for several residues that undergo significant chemical shift changes (indicated by arrows), as well as several that remain unperturbed. (B) A ribbon representation of the AHSP NMR structure, highlighting the residues that undergo significant chemical shift changes upon titration with α-hemoglobin (space-fill). These residues constitute the major portion of the α-hemoglobin binding surface of AHSP (Feng et al. (2004) CELL 119, 629–40).

Analysis of Protein–Protein Interactions

Recently, a new method for mapping protein interfaces, called cross-saturation [45], has been employed to map interaction interfaces of a number of protein complexes [46– 48]. In this approach, the protein whose interaction surface is to be mapped is 2 H- and 15 N-labeled and mixed with its unlabeled partner. Saturation of high-density 1 H nuclei of the unlabeled protein spreads to the back-exchanged amide protons on the interaction surface of the doublelabeled protein by spin diffusion and can be detected through quenching of signals in a 1 H, 15 N TROSY-HSQC. Because this technique relies on through-space proximity of the interaction surfaces, conformational changes would not be expected to affect the results as they do with chemical shift mapping studies.

Protein–Protein Interactions and Chemical Exchange The phenomenon of chemical exchange is observed in many different guises in NMR spectra. In the examination of protein–protein interactions, the rate of exchange between free and bound protein forms can lead to poor NMR spectra if the timescale of the binding process compared to the chemical shift timescale is inappropriate. As is well known, slow and fast exchange regimes exist [49]. Consider a single proton at a protein–protein interface. In slow exchange, the lifetimes of both the bound and free states are sufficiently long (>∼1 s) that a separate signal is observed for each. In fast exchange (∼µs or shorter lifetimes), a single signal at an averaged chemical shift (determined by the populations of the bound and free states) is observed. At the extremes of these regimes, no signal broadening occurs. However, in between lies the intermediate exchange regime, where substantial line broadening and non-Lorentzian lineshapes can occur. In the situation where two species with equal populations interconvert in a unimolecular reaction (A ↔ B), the maximum amount of line broadening occurs when the rate of exchange (kex ) is given by: πν kex = √ 2

(1)

1.0 0.8 ∆δ (ppm)

Cross-saturation

where ν is the frequency difference for a spin between two states. For bimolecular reactions (A + B ↔ C), the mathematics is more complex, but analogous expressions, albeit approximate, can be derived. In cases of intermediate exchange, the degree of line broadening depends on several parameters, namely the off-rate for the complex (kd ), the equilibrium constant (K A ), the line widths of the free and bound species, and the concentrations of each component. For typical frequency changes that might be observed in the formation of a protein complex (∼10–600 Hz for 1 H), and assuming diffusion controlled on-rates for complex formation, equilibrium constants in the range 103 –106 M−1 are most likely to give rise to NMR spectra exhibiting intermediate exchange. As noted above, regulatory interactions of this strength are common in cells. Chemical shift titration data can be used to determine interaction affinities. For systems in fast exchange, non-linear least squares fitting of the population averaged chemical shift to a simple hyperbolic binding function can be carried out. For slow exchange, disappearance and/or reappearance of signals from the free or bound protein states can be used to generate a binding curve. In the case of intermediate exchange, where significant line broadening occurs, the analysis is more complex, but binding constants may still be calculated [50] (Figure 3). Line widths often increase substantially during the course of such a titration as a result of exchange broadening, but then decrease toward the end as the fraction of labeled protein in the complexed state nears one. If the

0.6 0.4 0.2 0.0 0

1

2

3

4

5

[Pnr-NF]/[Ush-F9] Fig. 3. Deriving binding affinity from NMR data. Plot of the change of 15 N chemical shift (δ) for residue V21 of U-Shaped finger 9 (Ush-F9) upon binding to the N-finger of Pannier (Pnr-NF). The analysis [62] was based using a model of intermediate exchange [50] and reveals an association constant of 2.5 × 104 M−1 .

Part II

important early step in the design of targeted therapeutics, and chemical shift titrations have been used extensively in the screening of small molecule libraries to find lead compounds that bind a target protein [44]. Chemical shift titration data can also be used to determine equilibrium constants (for K A ∼ 103 –106 M−1 ).

Protein–Protein Interactions and Chemical Exchange 1343

Pharmaceutical Sciences

B

Q29

Q29

N34 N19 Y31

T23

T23

Y20

S2

S2

H32

Q29

K22 S17

N19 L25

K28

D15

E5 V6

F18

K9 A4

T13 K8

K33

Y10 T13

K9

K35

A4

M7

M7

A26

120

K8 K35

S12

Y24

A3

V21

C11

9.00

V6

F18

Y10

A3

V21

115

D15 E5

A26

S12

I16 H32

C14 N34

H27

Q29

Y31

K22

C14

N19

I16

F30 N34

110

N34

chemical shift (ppm)

Part II

A

15N

1344 Part II

125

C11

8.00

7.00 1H

9.00

8.00

7.00

chemical shift (ppm)

Fig. 4. The effect of intermediate chemical exchange on NMR spectra. 1 H, 15 N HSQC spectra of a point mutant (C32H) of zinc finger 9 from U-shaped, in the absence (left) and presence (right) of 5 molar equivalents of its binding partner, the N-terminal zinc finger from GATA-1 [62].

binding is relatively weak, however, it may be impractical to reach this point, preventing assignment of the complex and determination of a structure. In the titration shown in Figure 4, the signals that disappear correspond to residues that are known to lie at the protein–protein interface from mutagenesis studies [23], but it can be seen that they do not reappear even following the addition of 5 molar equivalents of unlabeled binding partner. In principle, manipulation of solution conditions might allow one to move from intermediate to fast or slow exchange; however, in practice, the narrow range of conditions under which protein structural studies can be carried out often precludes this possibility. Protein complexes that exhibit unfavorable chemical exchange properties therefore often remain difficult to characterize by NMR.

Stitching Up Proteins for Improved Stability In many cases, the preparation of protein complexes for structural studies is hindered by the instability of one or both components. In such cases, it may prove advantageous to either co-express the two partners [51] or to express engineered fusions of the interacting proteins. For example, the LIM domain proteins LMO2 and LMO4 are extremely unstable in isolation, which hampered studies in our laboratory of their structures and their interac-

tion with the co-regulator ldb1. In contrast, single-chain LMO2-ldb1 and LMO4-ldb1 fusion proteins [52] could be readily purified and their structures determined [53]. Volkman et al. used an identical strategy in their analysis of the N-WASP:WIP complex [11]: the EVH1 domain from N-WASP was poorly expressed and insoluble, but these problems were alleviated in a N-WASP:WIP fusion protein. A similar approach may hold promise for the analysis of weak protein complexes in which chemical exchange prevents the acquisition of high-quality data. By constraining diffusion of the two components, it may be possible to shift the system to a more favorable exchange regime. Indeed, covalent end-to-end ligation of two individually expressed recombinant proteins could potentially be achieved using intein technology (self-splicing protein modules [54]), thus permitting differential labeling of the two components as described above.

Docking Protein Complexes An interesting new development is the use of NMR data to guide docking of protein–protein complexes, given the known 3D structures of the subunits. The in silico docking of two proteins in the absence of direct experimental data has generally been unsuccessful because of the large

Analysis of Protein–Protein Interactions

References 1345

Summary In this chapter, we have outlined the major approaches to characterizing protein–protein complexes using solution NMR spectroscopy, with a particular emphasis on recently developed experimental techniques. Essentially all of these approaches are applicable to other macromolecular complexes, and it is likely that we will see a substantial increase in both the number and the complexity of the systems analyzed over the next 5 years. Fig. 5. HADDOCK derived structure of a complex formed between TACC3 and the third zinc finger of FOG. The TACC3 dimeric coiled-coil (dark) and the FOG zinc finger (light). The model was generated using the known structure of FOG finger 3, a homology model of the TACC3 coiled-coil, and a combination of alanine scanning mutagenesis and chemical shift perturbation data (Simpson et al. (2004) J Biol Chem 279, 39789–97).

number of degrees of freedom presented by two protein surfaces. However, long-range structural information derived from RDCs or paramagnetic effects, along with limited intermolecular NOEs, can be used to orient protein subunits in docking routines. For example, the N-terminal domain of enzyme I (EIN) and the histidine phosphocarrier, HPr were successfully docked using 231 1 DHNN RDC restraints and only nine intermolecular NOEs. The final ˚ of the previously solved docked complex was within 1.3 A

References 1. van der Merwe PA, Barclay AN, Trends Biochem. Sci. 1994;19(9):354. 2. Crowley PB, Ubbink M. Acc. Chem. Res. 2003;36(10):723. 3. Ptashne M, Gann A. Curr. Biol. 1998;8(22):R812. 4. Naar AM, Lemon BD, Tjian R. Annu. Rev. Biochem. 2001; 70:475. 5. Sattler M, Schleucher J, Griesinger C. Prog. NMR Spectrosc. 1999;34:93. 6. Gardner K, Lay L. Annu. Rev. Biophys. Biomol. Struct. 1998; 27:357. 7. Venters RA, Huang CC, Farmer BT 2nd, Trolard R, Spicer LD, Fierke CA. J. Biomol. NMR. 1995;5(4):339. 8. Nietlispach D, Clowes RT, Broadhurst RW, Ito Y, Keeler J, Kelly M, Ashurst J, Oschkinat H, Domaille PJ, Laue ED. J. Am. Chem. Soc. 1996;118(2):407. 9. Gardner K, Kay L. J. Am. Chem. Soc. 1997;119:7599.

Part II

NMR structure of the complex [55]. Such approaches can equally be employed to determine the domain packing of multi-domain proteins [56]. The identification of intermolecular NOEs requires the time-consuming assignment of side chain resonances. A number of approaches have been recently described for docking protocols that do not rely on NOE data [57–59]. These methods utilize chemical shift perturbation, mutagenesis and RDC data, which require only backbone assignments. HADDOCK [58] docks two proteins of known structure in the CNS-based ARIA protocol [60,61]. The use of ARIA allows the introduction of ambiguous interaction restraints (AIRs) derived from a wide range of NMR and other biophysical data. Chemical shift titration data are introduced by defining AIRs between each amino acid on the identified interaction surface of one protein and all of the residues on the surface of the partner protein. Mutagenesis data can be included in a similar fashion. We recently used this procedure to generate a model of the complex formed between a zinc finger from the transcriptional regulator FOG and a coiled-coil domain from TACC3 (Figure 5). One limitation of docking approaches is that they rely on the assumption that the structures of the bound and complexed proteins remain unchanged.

1346 Part II

Pharmaceutical Sciences

Part II

10. Pervushin K, Riek R, Wider G, W¨uthrich K. Proc. Natl. Acad. Sci. U.S.A. 1998;94:12366. 11. Volkman BF, Prehoda KE, Scott JA, Peterson FC, Lim WA. Cell. 2002;111(4):565. 12. Fiaux J, Bertelsen EB, Horwich AL, Wuthrich K. Nature. 2002;418(6894):207. 13. Tugarinov V, Muhandiram R, Ayed A, Kay LE. J. Am. Chem. Soc. 2002;124(34):10025. 14. Salzmann M, Pervushin K, Wider G, Senn H, Wuthrich K. J. Am. Chem. Soc. 2000;122:7543. 15. Fernandez C, Adeishvili K, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 2001;98(5):2358. 16. Arora A, Abildgaard F, Bushweller JH, Tamm LK. Nat. Struct. Biol. 2001;8(4):334. 17. Hwang PM, Choy WY, Lo EI, Chen L, Forman-Kay JD, Raetz CR, Prive GG, Bishop RE, Kay LE. Proc. Natl. Acad. Sci. U.S.A. 2002;99(21):13560. 18. Gronwald W, Huber F, Grunewald P, Sporner M, Wohlgemuth S, Herrmann C, Kalbitzer HR. Structure. 2001;9(11):1029. 19. Frickel EM, Riek R, Jelesarov I, Helenius A, Wuthrich K, Ellgaard L. Proc. Natl. Acad. Sci. U.S.A. 2002;99(4):1954. 20. Rudiger S, Freund SM, Veprintsev DB, Fersht AR. Proc. Natl. Acad. Sci. U.S.A. 2002;99(17):11085. 21. Pellecchia M, Sebbel P, Hermanns U, Wuthrich K, Glockshuber R. Nat. Struct. Biol. 1999;6(4):336. 22. Otting G, W¨uthrich K. J. Magn. Reson. 1989;85:586. 23. Liew CK, Kowalski K, Fox AH, Newton A, Sharpe BK, Crossley M, Mackay JP. Structure. 2000;8:1157. 24. Bax A, Kontaxis G, Tjandra N. Methods Enzymol. 2001; 339:127. 25. Tjandra N, Omichinski JG, Gronenborn AM, Clore GM, Bax A. Nat. Struct. Biol. 1997;4(9):732. 26. Clore GM, Garrett D. J. Am. Chem. Soc. 1999;121:9008. 27. Ramos A, Grunert S, Adams J, Micklem DR, Proctor MR, Freund S, Bycroft M, St Johnston D, Varani G. EMBO J. 2000;19(5):997. 28. Fisher M, Losonczi J, Prestegard J. Biochemistry. 1999; 38:9013. 29. Walters KJ, Lech PJ, Goh AM, Wang Q, Howley PM. Proc. Natl. Acad. Sci. U.S.A. 2003;100(22):12694. 30. Tsui V, Zhu L, Huang TH, Wright PE, Case DA. J. Biomol. NMR. 2000;16(1):9. 31. Garrett DS, Seok YJ, Peterkofsky A, Gronenborn AM, Clore GM. Nat. Struct. Biol. 1999;6(2):166. 32. Wang G, Louis JM, Sondej M, Seok YJ, Peterkofsky A, Clore GM. EMBO J. 2000;19(21):5635. 33. Ubbink M, Worrall JA, Canters GW, Groenen EJ, Huber M. Annu. Rev. Biophys. Biomol. Struct. 2002;31:393. 34. Pintacuda G, Keniry MA, Huber T, Park AY, Dixon NE, Otting G. J. Am. Chem. Soc. 2004;126(9):2963. 35. Gaponenko V, Sarma SP, Altieri AS, Horita DA, Li J, Byrd RA. J. Biomol. NMR. 2004;28(3):205. 36. Ubbink M, Ejdeback M, Karlsson BG, Bendall DS. Structure. 1998;6(3):323.

37. Guiles RD, Sarma S, DiGate RJ, Banville D, Basus VJ, Kuntz ID, Waskell L. Nat. Struct. Biol. 1996;3(4):333. 38. Worrall JA, Liu Y, Crowley PB, Nocek JM, Hoffman BM, Ubbink M. Biochemistry. 2002;41(39):11721. 39. Barbieri R, Bertini I, Cavallaro G, Lee YM, Luchinat C, Rosato A. J. Am. Chem. Soc. 2002;124(19):5581. 40. Arumugam S, Hemme CL, Yoshida N, Suzuki K, Nagase H, Berjanskii M, Wu B, Van Doren SR. Biochemistry. 1998;37(27):9650. 41. Jain NU, Venot A, Umemoto K, Leffler H, Prestegard JH. Protein Sci. 2001;10(11):2393. 42. Kihm AJ, Kong Y, Hong W, Russell JE, Rouda S, Adachi K, Simon MC, Blobel GA, Weiss MJ. Nature. 2002; 417(6890):758. 43. Schmiedeskamp M, Rajagopal P, Klevit RE. Protein Sci. 1997;6(9):1835. 44. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Science. 1996;274(5292):1531. 45. Takahashi H, Nakanishi T, Kami K, Arata Y, Shimada I. Nat. Struct. Biol. 2000;7(3):220. 46. Lane AN, Kelly G, Ramos A, Frenkiel TA. J. Biomol. NMR. 2001;21(2):127. 47. Nishida N, Sumikawa H, Sakakura M, Shimba N, Takahashi H, Terasawa H, Suzuki EI, Shimada I. Nat. Struct. Biol. 2003;10(1):53. 48. Shao W, Im SC, Zuiderweg ER, Waskell L. Biochemistry. 2003;42(50):14774. 49. Sanders J, Hunter B. Modern NMR Spectroscopy. Oxford University Press: Oxford, 1987. 50. Feeney J, Batchelor J, Albrand J, Roberts G. J. Magn. Reson. 1979;33:519. 51. Demarest SJ, Martinez-Yamout M, Chung J, Chen H, Xu W, Dyson HJ, Evans RM, Wright PE. Nature. 2002;415(6871): 549. 52. Deane JE, Sum E, Mackay JP, Lindeman GJ, Visvader JE, Matthews JM. Protein Eng. 2001;14:493. 53. Deane J, Mackay J, Kwan A, Sum E, Visvader J, Matthews J. EMBO J. 2003;22:2224. 54. Hofmann RM, Muir TW. Curr. Opin. Biotechnol. 2002;13(4): 297. 55. Clore GM. Proc. Natl. Acad. Sci. U.S.A. 2000;97(16):9021. 56. Dosset P, Hus JC, Marion D, Blackledge M. J. Biomol. NMR. 2001;20(3):223. 57. Dobrodumov A, Gronenborn AM. Proteins. 2003;53(1):18. 58. Dominguez C, Boelens R, Bonvin AM. J. Am. Chem. Soc. 2003;125(7):1731. 59. Fahmy A, Wagner G. J. Am. Chem. Soc. 2002;124(7):1241. 60. Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, Read RJ, Rice LM, Simonson T, Warren GL. Acta. Crystallogr. D. Biol. Crystallogr. 1998;54(Pt 5):905. 61. Nilges M. J. Mol. Biol. 1995;245(5):645. 62. Kowalski K, Liew CK, Matthews JM, Gell DA, Crossley M, Mackay JP. J. Biol. Chem. 2002;277:35720.

1347

Robert E. London Laboratory of Structural Biology, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709, USA

Introduction The complex macromolecules that have evolved to serve a broad range of biological functions present surfaces with highly variable topological, electrostatic, and hydrogen bonding properties. These surfaces provide the basis for interactions that play critical roles in catalysis and signal transduction, and offer a limitless landscape of potential targets for chemical agents designed to interfere with these functions. A broad range of analytical techniques has been developed to exploit this topological information when it is available, but it is clear that flexible ligands can interact in unanticipated ways with their theoretical targets, and often with unintended targets as well. Substitution of an amino group for a single oxygen substituent to convert folate to aminopterin results in a potent antifolate drug which, unexpectedly, adopts a dramatically different conformation in the complex with the target enzyme, dihydrofolate reductase [1,2]. Studies of the inhibition of alpha-chymotrypsin by a series of 1-acetamidoboronic acids unexpectedly revealed that when the side chain of the inhibitor was altered from a p-chlorophenyl to a 1-naphthyl group, the enzyme reversed its usual preference for l-enantiomers, which more closely resemble the natural substrates, and bound more tightly to the d-enantiomer [3]. The sensitivity of NMR spectroscopy to ligand– macromolecule interactions makes it of considerable value for the process of ligand identification and modification, and a broad range of approaches have been described that involve direct observations of either the target macromolecule [4] or the ligand [5]. The use of any particular approach is generally strongly dependent on the particular problem under study, however direct NMR studies of biologically important macromolecules are limited by the size and complexity of the macromolecule, its solubility, the general requirement for isotopic labeling, and the more extensive analysis which may be required. Alternatively, observations of ligands have been demonstrated to be useful for selecting lead compounds from complex mixtures [6] and for obtaining insight into the conformation of bound ligands [7], and are often more straightforward to interpret. Graham A. Webb (ed.), Modern Magnetic Resonance, 1347–1356.
C 2008 Springer.

One approach for the analysis of ligand– macromolecule interactions involves the identification of ternary complexes formed from pairs of ligands in the presence of a macromolecule. In some cases, biologically interesting complexes have formed from adventitious inorganic ions such as aluminum or beryllium fluorides [8]. The presence of adventitious zinc ions led to the identification of a trypsin-Zn2+ -bis(5-amidino-2benzimidazolyl)methane ternary complex, which formed a basis for the further development of serine protease inhibitors [9]. Arsenate and borate complexes have been identified due to the use of cacodylate and borate buffers [10,11]. Vanadate complexes have been identified in toxicological and bioinorganic studies [12–14]. These complexes have been detected based on their biochemical effects, or by structural methods including X-ray crystallography and NMR spectroscopy. Arsenate complexes often mimic phosphate complexes, both structural and biochemically, while vanadate complexes often form pentavalent structures that mimic the transition state involved in nuclease and phosphatase biochemistry [12–15]. Vanadate, and aluminum and beryllium fluorides are all capable of mimicking the terminal phosphate group of nucleotides, e.g. turning ADP into the ADP-VO4 analog of ATP [16–18]. Boronate and borate complexes most frequently mimic the tetrahedral transition state of serine proteases and mechanistically related enzymes [19–21]. Alternatively, enzymatic catalysis frequently involves the formation of ternary complexes that include two, and occasionally more, organic ligands. For example, many enzyme-catalyzed redox reactions involve ternary enzyme-substrate–cofactor complexes. Formation of transient complexes involving macromolecules and organic ligands can be studied using transferred interligand Overhauser effects (ILOEs) [22–27]. A brief outline of NMR approaches used to study ternary complexes is described in the present article.

Borate Complexes and Their Study by NMR Spectroscopy Studies of ternary and higher order complexes formed from inorganic ions have provided insights into the

Part II

Identification and Characterization of Ternary Complexes Using NMR Spectroscopy

1348 Part II

Pharmaceutical Sciences

Part II

toxicological properties of these ions, valuable structural information, insight into catalytic mechanisms, as well as fundamental information of use for the design of inhibitors. In some cases, these ions interact with active site residues of enzymes in ways that mimic transition state structures. A comprehensive review of this area is beyond the scope of the present article, but our recent studies of borate complexes illustrate many of the features of these studies. Boric acid (pK = 9.0) is a convenient buffer for high pH assays of enzymes or other biochemical phenomena. The use of borate buffers has led to a number of unanticipated observations that resulted from the presence of borate–ligand complexes. For example, borate has been found to inhibit pyridine nucleotide and flavin-requiring enzymes as a result of complex formation with these cofactors [28,29]. A more dramatic effect is observed in studies of the membrane protein γ-glutamyl transpeptidase (γ-GT), an enzyme important for glutathione metabolism that transfers γ-linked glutamyl groups between peptides, amino acids, or other acceptors. For this enzyme, serine and borate, which individually exert little effect on the enzyme, exhibit a dramatic inhibitory synergy [11]. This inhibitory synergy apparently results from the stabilization of a ternary γ-GT–serine–borate complex, in which the borate adopts a tetrahedral, anionic form that is esterified to both an enzyme threonine hydroxyl group, as well as to the hydroxyl group of the serine ligand [20,21]. This ternary complex thus mimics a putative transition state in which the threonine hydroxyl group of the active site binds to the γ-carbon of the glutamyl group in glutathione or another substrate, creating a transition state with tetrahedral geometry. Although direct NMR observations of membrane associated enzymes such as γ-GT are difficult, the general features of such ternary complexes have been observed for the serine protease trypsin [30–32]. In these studies, several cationic alcohols with presumed affinity for the trypsin S1 specificity binding pocket were studied. The molecules were selected so that the alcohol function would be appropriately positioned to interact with a borate bound to the active site serine-195. Boron-11 spectra of samples containing borate, trypsin, and 4-aminobutanol (4AB) are characterized by a resonance with a chemical shift that corresponds to the boric acid–borate equilibrium at the pH used for the study, as well as an upfield shifted resonance corresponding to a tetrahedral borate which is now in slow exchange with the free boric acid/borate species (Figure 1). Binary mixtures of borate–trypsin or borate–4AB, or a ternary mixture of borate, trypsin, and 3aminobutanol do not show the upfield shifted resonance. The resonance observed at −17.3 ppm can therefore be assigned to a ternary trypsin–borate–4AB complex. Analogous complexes have now been observed in the crystalline state [31,32]. Interestingly, the

crystallographic studies demonstrate that the nature of the borate complex formed at the active site is significantly influenced by lattice contacts, so that a ternary trypsin–borate–4AB complex analogous to that observed in solution could be observed in P31 21 crystals [32], while only quaternary trypsin–borate–4AB–ethylene glycol complexes were observed in P21 21 21 crystals [31]. In the P21 21 21 crystals, formation of the active site borate complex was found to be correlated with a 4% expansion of the crystal along the b-axis. These results were interpreted to indicate that the additional stability of the quaternary complex was required to overcome crystal packing effects. This study therefore illustrates the significant effect that lattice contacts can have on crystallographic screening for enzyme ligands. Most of the inorganic complexes which have been studied by NMR, involve nuclei with spin >1/2, and hence are subject to quadrupolar relaxation. For quadrupolar nuclei bound to very large molecules such that ωτ  1, where ω is the Larmor frequency of the nucleus and τ its rotational correlation time, the relaxation behavior is multiexponential. For nuclei with half-integral spin in this limit, the central 1/2 ↔ −1/2 transition can give rise to a relatively narrow signal, while the remaining transitions are very broad and generally unobserved [33– 36]. Subject to the above assumptions, the linewidth of the central transition is then inversely related to ω2 τ , i.e. the linewidth actually narrows as the rotational correlation time of the complex τ increases: ν1/2 =

1 2s + 3  Q 2J1 (ωS ) + ((s + 3/2) = π T2 πs 2 (2s − 1)   × (s − 1/2) − 1) J2Q (2ωS (1)

where J Q (ω) =

3χ 2 160



τ 1 + ω2 τ 2

 (2)

χ is the quadrupole coupling constant, nucleus S has spin s and Larmor frequency ωS , τ is the rotational correlation time, and isotropic motion is assumed. In the limit ωτ  1, we obtain:   1 2(2s + 3) 1 ≈ 1 + ((s + 3/2) ν1/2 = π T2 πs 2 (2s − 1) 8  2 3χ 1 (3) × (s − 1/2) − 1) 160 ω2 τ This inverse dependence on τ contrasts with the linewidth predictions based on other relaxation mechanisms such as the dipolar interaction and chemical shift anisotropy, which predict ν 1/2 ∼ J (0) ∼ τ c . Exchange

NMR Studies of Ternary Complexes

Borate Complexes and Their Study by NMR Spectroscopy 1349

Part II

A)

4

0

–4

–8

–12

–16

–20

–24

ppm

B)

Fig. 1. (A) Boron-11 NMR spectra obtained at 160.6 MHz, 25◦ C on samples containing (top to bottom) 5 mM borate and 2 mM porcine pancreatic trypsin; 25 mM borate and 250 mM 4-aminobutanol; 5.0 mM borate, 50 mM 3-aminopropanol, and 2 mM trypsin; 5.0 mM borate, 50 mM 4-aminobutanol, and 2 mM trypsin. Insets show a 10-fold vertical expansion. All samples contained 20 mM HEPES, pH 8.0 (based on Figure 1 of Ref. [30]). (B) Active site of trypsin showing complex formed with borate, 4-aminobutanol, and ethylene glycol (EG) (based on Figure 1F of Ref. [31]). A similar ternary complex without the EG has also been observed for crystals in the P31 21 space group [32].

broadening due to the chemical exchange between uncomplexed and variously complexed forms of the ion also can make significant contributions. The observation of narrow resonances for quadrupolar nuclei in

macromolecules is illustrated by the 11 B NMR spectrum of 4-fluorobenzeneboronic acid (FBBA) in the presence of a small ligand, sorbitol, and the enzyme subtilisin Carlsberg (MW = 27.3 kD), shown in Figure 2. As

1350 Part II

Pharmaceutical Sciences

Part II

FBBA-subtilisin

OH F

B OH

FBBA-sorbitol

FBBA

10

5

0

–5

–10

–15

–20

–25

–30

–35

ppm

Fig. 2. Boron-11 NMR spectrum of 5 mM 4-fluorobenzeneboronic acid, 1.5 mM sorbitol, 2 mM subtilisin Carlsberg (MW = 27.3 kD) in 50 mM HEPES, pH 7.5 obtained at 11.75 T (160.6 MHz 11 B frequency), 5◦ C.

predicted above, the resonance arising from the boronate– subtilisin complex is significantly narrower than the resonances for either free boronate or for the boronate–sorbitol complex. The 11 B linewidth of the uncomplexed FBBA is determined by quadrupolar relaxation of the 11 B nucleus (spin = 3/2) and by boronic acid–boronate exchange, while the linewidths of the FBBA–sorbitol and FBBA– subtilisin complexes are probably determined primarily by quadrupolar 11 B relaxation. In addition to this unintuitive dependence of linewidth on rotational correlation time, the quadrupolar nuclei are also subject to a secondorder dynamic frequency shift which is also inversely correlated with ω (i.e. B0 ) [35,36]: δ=

 8s(s + 1) − 6  Q Q L (ω ) − L (2ω ) S S 1 2 s 2 (2s − 1)2

(4)

Ternary Complexes Involving Organic Molecules

where L Q (ω) =

2

3χ 160



ωτ 1 + ω2 τ 2 2

 (5)

so that δ≈

4s(s + 1) − 3 s 2 (2s − 1)2



3χ 2 160

  1 ω

Expression of the shift in ppm results in a ω−2 dependence. Observations on the relaxation behavior of various metal ion complexes with transferrin have verified the predicted linewidth and shift behavior [37–39]. In summary, 11 B NMR studies provide a useful way of characterizing transient complexes that may form involving borate esters. The exchange behavior among various complexes is often sufficiently slow on the NMR time scale to allow observation of separate resonances. Similar complexes can be observed in the crystalline state, however in this case, lattice contacts can significantly alter the observation of these weak complexes. The observation of such complexes can provide useful insight into the development of boronated as well as non-boronated ligands.

for ωτ  1 (6)

Although ternary complexes formed from inorganic ions can provide a useful basis for the development of high affinity ligands, it is desirable to generalize this approach by removing the restriction to inorganic ions. The observation of ILOEs provides one basis for obtaining information on the spatial proximity and relative orientation of two (or potentially more) ligands that form a relatively weak complex with a macromolecular binding site. Nuclear Overhauser effect (NOE) interactions involving

NMR Studies of Ternary Complexes

Ternary Complexes Involving Organic Molecules 1351

Part II

Interligand Overhauser Effect

NOE H H L1

L2

L2

L1 H H

Receptor

Receptor

Fig. 3. Schematic illustration of the interaction of two ligands, L1 and L2 , with a macromolecule. NOE interactions between the protons on the two different ligands can be transferred to the free ligands due to dissociation of the complex.

protons on the two ligands are transferred to the pair of uncomplexed ligands much as the intramolecular NOE information is transferred to the free ligand in a transferred NOE experiment (Figure 3). As in the transferred NOE experiment, these experiments are typically performed using a significant excess of ligand concentration, i.e. [L1 ], [L2 ]  [E], where L1 and L2 are the two ligands and E is the enzyme or other macromolecule forming the complex. Since this approach involves direct observation of the ligand resonances, but not the resonances of the macromolecule, it does not require complex resonance assignment strategies. Further, as with the transferred NOE experiment, it works well on macromolecules that are too large for direct structural determination using currently available NMR methodology. However, in contrast with the transferred NOE experiment, this experiment is considerably less subject to misinterpretation resulting from non-specific binding, ligand aggregation, or ligand misbinding. The latter effect arises because ligand binding is essentially an exploratory process in which a ligand may become transiently lodged in an enzyme cleft in a non-productive conformation such that the structural constraints will not allow it to readily adopt a productive conformation. It then becomes most probable for the ligand to dissociate, and subsequently it may reassociate in a conformation which can lead to a productive complex. If the misbound ligand remains associated with the

macromolecule for a time that is on the order of the rotational correlation time, it is then able to develop an NOE characteristic of the bound state, and this NOE information characteristic of the misbound complex will be transferred to the free ligand, along with the NOE information from the correctly bound ligand. For both transferred NOE and interligand NOE studies to provide useful information, the exchange rates for both ligands must be reasonably fast compared with the T1 time scale of the protons that are under observation. This allows NOE interactions characteristic of the bound ligand(s) to be observed regardless of whether the exchange is fast or slow on the chemical shift time scale, although in general, slow exchange on the shift time scale is less likely to satisfy the T1 time scale constraint noted above, and leads to non-linear NOE build-up curves. The long T1 values for the uncomplexed ligand allow it to be used as a “magnetic storage device” for NOE information that is formed in the macromolecular complex. The kinetic behavior of such a system is described using eight rate constants, seven of which are independent, as shown in Scheme 1. The fractional populations of the uncomplexed enzyme/macromolecule, pE, the two binary complexes pEL1 and pEL2 , and the ternary complex pEL1 L2 , can be expressed as a function of the concentrations and rate constants, as described previously [24]. From a practical standpoint, we have generally found that

1352 Part II

Pharmaceutical Sciences

Part II

L1 k1 E

EL1

k-1 k-2

k4

k-4

k2 L2

L2

L1 k3

EL2

k-3

EL1L2

Scheme 1

association rate constants of ∼108 /M/s and dissociation rate constants of 103 /s that correspond to equilibrium dissociation constants of ∼10 µM produce strong effects that are readily observed under typical sample conditions. An illustrative 2D NOESY spectrum for a system with two spins is shown in Figure 4. The 1D spectrum for this system consists of two sets of three resonances: for spin 1 on L1 , the shift will generally be dependent on whether the ligand is free, in a binary complex with E, or in a ternary complex with E and L2 . The diagonal peaks are shown as open circles. The exchange cross peaks which connect specific spins interconverting among the free, binary, and ternary complexed forms are in black. The NOE cross peaks connecting spins 1 and 2 on the ternary complex are shown in light gray. Of course, since the spins are postulated to be on different ligands, the values for the binary complexes and the free ligands will be zero in the absence of exchange. Finally, the dark gray cross peaks are exchange-mediated NOE interactions which become non-zero when both chemical exchange and NOE interactions are present. For example, such cross peaks can connect spin 1 in a ternary complex with spin 2 in the free state. When the exchange is sufficiently rapid compared with the relevant spin lattice relaxation rates, the sets of nine cross peaks in each of the four boxes shown in the figure exhibit the same time dependence, differing only in a magnitude. When this condition is not met, the elements in each box can exhibit different dependences on the mixing time. In a typical experiment of this type, one observes only a single resonance for each spin due to the fact that the concentrations of the binary and ternary complexes are much lower than the free ligand concentrations, or, more typically, due to fast exchange on the chemical shift time scale. If the latter condition is met, it is appropriate

to sum the NOE build-up curves for the nine peaks in each of the indicated boxes. This represents the optimal condition for the transferred NOE and interligand NOE observations. In this case, the sum of the nine NOE cross peaks, Ai j , can be represented by another matrix, Ci j [24]. Simulations of the dependence of the interligand NOE (ILOE) on mixing time have been performed for a range of geometric structures and kinetic rate constants [24]. In general, the ILOE curves exhibit a dependence on mixing time which is similar to the transferred NOE curves, characterized by an initial build up and a subsequent decay as relaxation effects become dominant. Calculations corresponding to the intensities of the NOESY cross peaks, Ai j , and to the sums of elements, Ci j , have been performed for a model system that contains two ligands with three spins ˚ A ternary comarranged linearly and separated by 2.5 A. plex was modeled in which all six nuclei of both ligands are also arranged in a line, with nucleus 3 on ligand 1 and ˚ as shown nucleus 4 on ligand 2 also separated by 2.5 A, below:

Using parameters: [E] = 0.4 mM, [L1 ] = [L2 ] = 5 mM, τ 1F = τ 2F = 10−10 s; τB = 10−7 s, k1 = k2 = k3 = k4 = 108 /M/s, k−1 = k−2 = k−3 = k−4 = 10−3 /s, and other parameters as given in Ref. [24], curves showing

NMR Studies of Ternary Complexes

2Free 2Bin 2Ter

C12

C11

1Free

A11

1Bin 1Ter

1Ter 1Bin 1Free

Fig. 4. Schematic representation of a 2D NOESY spectrum for a system containing a macromolecule, E, and two ligands, L1 and L2 , which can form binary or ternary complexes with E. Nucleus 1 is assumed to be positioned on L1 , and nucleus 2 on L2 . The observation of separate resonances for each spin in the free, binary, and ternary complexes would only be possible under slow exchange conditions on the chemical shift time scale, but is useful for illustrative purposes. The exchange peaks which interconvert nucleus 1 (or 2) between free, binary, and ternary complexes are shown in black. Cross peaks arising from NOE interactions between nuclei 1 and 2 in the ternary complex are shown in light gray. Note that if spins 1 and 2 are on different ligands, there will be no cross peaks for the uncomplexed ligands and for either binary complex. Cross peaks that arise due to a combination of exchange and NOE interactions are shown in dark gray. Under conditions of fast exchange, only the sums of the peaks in each square, i.e. the Ci j matrix elements, are observed (based on Figure 1 of Ref. [24]).

2Ter 2Bin 2Free

the dependence of C12 , C56 , and C34 on the mixing time are shown in Figure 5A. Note that the initial slopes of the C12 , C56 , and C34 curves are identical, consistent with the equal r12 , r56 , and r34 internuclear distances in the model ternary complex which, based on the above parameters, corresponds to 0.996 of the total enzyme concentration available. Most of the difference that can be observed between C12 and C34 arises due to the faster decay of the C34 NOESY interaction, which arises primarily from the presence of two nearest neighbor spins in the ternary complex, compared with one nearest neighbor spin for nucleus 1. Thus, a comparison of the C23 with C34 curves shows greater similarity (not shown). In a typical transferred NOE experiment, the internuclear distances of the complexed ligand can be calibrated by comparing the initial slope of the NOE build-up curves with that obtained for a pair of nuclei that have a fixed internuclear distance. For example, in a study in-

volving NADPH, the calibration curve could correspond to the H5–H6 protons on the nicotinamide ring. For the interligand NOE, the analysis becomes more complex if the two ligands have significantly different kinetic parameters. If the fraction of the ternary complex pEL1 L2 becomes significantly less than one and the fraction of one binary complex, e.g. pEL1 becomes significant, ligand 1 will be subject to greater transferred NOE effect than ligand 2. In general, the ILOE will scale with the weaker binding ligand. As shown in Figure 5B, increasing the off rates for L2 to 105 /s results in a relatively modest increase in pEL1 and decreased initial slopes for the C56 and C34 build-up curves. Further increasing the off rate of L2 to 106 /s lowers the fraction of E in the ternary complex to 1/3, resulting in a significantly greater initial slope for C12 than for C56 or C34 , which exhibit similar values (Figure 5C). Increasing k−2 = k−4 to 107 /s largely eliminates the binding of ligand 2, giving pEL1 = 0.95 (Figure 5D). Interestingly, the initial slope calculated

Part II

A12

Ternary Complexes Involving Organic Molecules 1353

1354 Part II

Pharmaceutical Sciences

Part II Fig. 5. Sensitivity of interligand Overhauser effects to weaker binding. The set of simulations shown above illustrates the effect of increasing the dissociation rate constant of L2 , while leaving that of L1 fixed, with k−1 = k−3 = 103 /s. All association rate constants are set at 105 /M/s. The calculated fractions of E involved in the ternary (pEL1 L2 ) and binary (pEL1 ) complexes are shown for the corresponding dissociation rate constants above each figure. (A) k−1 = k−2 = k−3 = k−4 = 103 /s; (B) k−2 = k−4 = 105 /s; (C) k−2 = k−4 = 106 /s; (D) k−2 = k−4 = 107 /s. Other parameters as given in Ref. [24] (based on Figure 6 in Ref. [24]).

for C34 is greater than that calculated for C56 , and no significant transferred NOEs are predicted for ligand 2. This results because for these parameters, L2 is largely free, and free L2 experiences an NOE of opposite sign to the bound ligand. Hence, the presence of free L2 reduces the observed build-up. However, no such subtraction effect arises for the interaction between nuclei 3 and 4, since these do not interact when L2 is not bound to E.

ILOE Observations—Type II Dihydrofolate Reductase Interligand Overhauser effects have been observed in several systems of interest [22,23,25–27]. In general, such effects can be observed in non-productive ternary enzyme complexes that involve inhibitors or other molecules that are not substrates for the enzyme under study. Ternary complexes involving enzyme substrates can be observed if the enzyme has been inactivated by mutagenesis. Alternatively, enzymes involved in redox chemistry can be studied if the pair of ligands are both in the oxidized or

reduced state, rather than the reduced–oxidized combination that functions as a substrate–cofactor pair. Of course, binding to other types of macromolecules can also be studied, and in this case, there are no concerns about ligand tur over. Dihydrofolate reductase catalyzes the NADPHdependent reduction of dihydrofolate to tetrahydrofolate. Type II dihydrofolate reductase is a plasmid-encoded enzyme that is structurally unrelated to the extensively studied type I enzyme, and confers antifolate drug resistance on bacteria containing the plasmid. The enzyme is a symmetric tetramer that binds the substrate and cofactor in a central pore [27]. Figure 6 illustrates portions of a NOESY experiment for a sample containing 0.1 mM R67 DHFR, 5 mM folate, and 5 mM NADP [25]. The region of the spectrum including the CH2 protons of folate (at the F9 position) are shown in Figure 6. Transferred NOE cross peaks to the folate F7, proton, and to the folate F12/16 and F13/15 protons on the p-aminobenzoyl group are readily observed. ILOE peaks connecting the folate F9 protons to the nicotinamide NH-4 and NH-5 protons on the NADP also can be observed in the spectra. All of these peaks exhibit time-dependent build-up curves expected for the

NMR Studies of Ternary Complexes

HF7

HN5

HF 12/16

Part II

HN4

Ternary Complexes Involving Organic Molecules 1355

HF 13/15

900 ms

HF9

4.2

HA8/HA2′

4.2

4.4 4.3 F1 (ppm)

700 ms

4.2

4.4

4.3

500 ms

4.4

4.3

300 ms

8.0

7.0 F2 (ppm)

Fig. 6. A portion of the NOESY spectrum illustrating the cross peaks connecting the folate H-9 protons with the aromatic protons of folate and NADP+ , obtained on a sample containing 0.1 mM R67 DHFR, 5 mM folate, and 5 mM NADP+ in 100 mM Tris-d11, pH 8.0 in D2 O. The spectra correspond to mixing times of 300, 500, 700, and 900 ms, as indicated (based on Figure 2 of Ref. [25]).

NOE experiment. The results obtained in this study are intuitive since the ILOE peaks are observed near the positions (NH-4 on NADPH, C-6 on folate) at which redox chemistry occurs. The very weak ILOE peaks involving the folate F-7 were also useful for determining the relative positions of the two ligands. The potential application of ILOE studies for ligand discovery has been demonstrated in experiments on alcohol dehydrogenase, in which a sample contained NADH and an inhibitor, m-methylbenzamide, as well as a reduc-

ing agent, sodium cyanoborohydride, included in order to maintain the NADH in the reduced state [26]. In this study, unanticipated cross peaks were observed between the BH3 protons on the cyanoborohydride and the methylbenzamide protons. It was concluded that the methylbenzamide binds primarily to a hydrophobic pocket in the active site that is involved in binding larger substrates, such as benzyl alcohol, while the cyanoborohydride formed a transient complex with the active site Zn2+ ion [26].

1356 Part II

Pharmaceutical Sciences

Part II

Summary The identification and characterization of ternary complexes involving organic molecules and inorganic ions should provide a useful basis for the further development of targeted ligands. Although the example of borate discussed above involved the serine protease trypsin, recent structural data indicate that borate complexes can form outside of the active site, supporting the possibility of targeting protein–protein interactions in a more general way. Comparisons with crystallographic analyses of the same complexes indicate that crystal screening approaches, while generally very useful, are subject to potentially significant influences that result from lattice contacts. ILOEs can provide useful insight into the relative binding and orientation of ligands that reversibly form ternary complexes. The presence of unanticipated complexes has been observed in some studies, supporting the potential application of this approach to the design of novel ligands.

References 1. Charlton PQ, Young DW, Birdsall B, Feeney J, Roberts GCK. J. Chem. Soc. Chem. Commun. 1979;922. 2. Bystroff C, Oatley SJ, Kraut J. Biochemistry 1990;29: 3263. 3. Stoll VS, Eger BT, Hynes RC, Martichonok V, Jones JB, Pai EF. Biochemistry 1998;37:451. 4. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Science 1996;274:1531. 5. Meyer B, Peters T. Angew. Chem. Int. Ed. 2003;42:864. 6. Mayer M, Meyer B. Angew. Chem. Int. Ed. 1999;38:1784. 7. Post CB. Curr. Opin. Struct. Biol. 2003;13:581. 8. Sternweis PC, Gilman AG. Proc. Natl. Acad. Sci. U.S.A. 1982;79:4888. 9. Katz BA, Clark JM, Finer-Moore JS, Jenkins TE, Johnson CR, Ross MJ, Luong C, Moore WR, Stroud RM. Nature 1998;391:608–12. 10. Boyle FA, Cook ND, Peters TJ. Clin. Chim. Acta 1988;172:291–6. 11. Revel JP, Ball EG. J. Biol. Chem. 1959;254:577.

12. Lindquist RN, Lynn JL Jr, Lienhard GE. J. Am. Chem. Soc. 1973;95:8762. 13. Borah B, Chen C-W, Egan W, Miller M, Wlodawer A, Cohen JS. Biochemistry 1985;24:2058. 14. Crans DC, Smee JJ, Gaidamauskas E, Yang L. Chem. Rev. 2004;104:849. 15. Davies DR, Interthal H, Champoux JJ, Hol WGJ. J. Med. Chem. 2004;47:829. 16. Goodno CC. Proc. Natl. Acad. Sci. U.S.A. 1979;76:2620. 17. Maruta S, Henry GD, Sykes BD, Ikebe M. J. Biol. Chem. 1993;268:7093. 18. Fisher AJ, Smith CA, Thoden JB, Smith R, Sutoh K, Holden HM, Rayment I. Biochemistry 1995;34:8960. 19. Walker B, Lynas JF. Cell. Mol. Life Sci. 2001;58:596. 20. Tate SS, Meiser A. Proc. Natl. Acad. Sci. U.S.A. 1978;75: 4806. 21. London RE, Gabel SA. Arch. Biochem. Biophys. 2001;385: 250. 22. Barsukov IL, Lian LY, Ellis J, Sze KH, Shaw WV, Roberts GCK. J. Mol. Biol. 1996;262:543. 23. Li D, DeRose EF, London RE. J. Biomol. NMR 1999;15:71. 24. London RE. J. Magn. Reson. 1999;141:301. 25. Li D, Levy LA, Gabel SA, Lebetkin MS, DeRose EF, Wall MJ, Howell EE, London RE. Biochemistry 2001;40:4242–52. 26. Li D, London RE. Biotechnol. Lett. 2002;24:623–9. 27. Pitcher WH III, DeRose EF, Mueller GA, Howell EE, London RE. Biochemistry 2003;42:11150. 28. Roush A, Norris ER. Arch. Biochem. 1950;29:345. 29. Smith KW, Johnson SL. Biochemistry 1976;15:560. 30. London RE, Gabel SA. Biochemistry 2002;41:5963. 31. Transue TR, Krahn JM, Gabel SA, DeRose EF, London RE. Biochemistry 2004;43:2829. 32. Babine RE, Rynkiewicz MJ, Jin L, Abdel-Maguid SS. Lett. Drug Des. Discov. 2004;1:35. 33. Bull TE, Forsen S, Turner DL. J. Chem. Phys. 1979;70:3106. 34. Werbelow LG. J. Chem. Phys. 1979;70:5381. 35. Werbelow LG, Pouzard G. J. Phys. Chem. 1981;85:3887. 36. Werbelow LG. In: DM Grant (Ed). Encyclopedia of Nuclear Magnetic Resonance, Vol. 6. John Wiley & Sons: New York, 1995, p 4092. 37. Aramini JM, Vogel HJ. J. Am. Chem. Soc. 1993;115:245. 38. Aramini JM, Germann MW, Vogel HJ. J. Am. Chem. Soc. 1993;115:9750. 39. Aramini JM, McIntyre DD, Vogel HJ. J. Am. Chem. Soc. 1994;116:11506.

1357

Mike P Williamson Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK

Abstract The transferred NOE allows NOE information to be obtained on a bound ligand, but appearing on the resonances of the free ligand. It is thus a very useful method for gaining information on the bound ligand. Its major limitation is that it only works well in a range of dissociation constants between 1 mM and 10 nM. A key negative control experiment is to demonstrate the absence of NOEs arising from nonspecific binding, by addition of competitor ligands. A second is to measure NOEs in the absence of protein. For quantitative work, it is also necessary to minimize spin diffusion, by using relatively short mixing times. Recent work has shown that spin diffusion in TRNOE is no worse than spin diffusion in standard NOE spectra of proteins. With care, TRNOE can thus provide powerful quantitative information on bound conformation.

Affinities and Timescales The appearance and information content of an NMR spectrum of a protein/ligand mixture depend strongly on the rates of binding and tumbling of the components. This is why it is important to think about these rates, and understand their implications. The binding of a ligand to a protein can be written L+P

kon

 L·P k

(1)

off

The overall on-rate is kon [L][P], and the off-rate is koff [L.P]. At equilibrium, these are by definition equal. The dissociation constant K d is [L][P]/[LP], which is therefore also equal to koff /kon . The association constant K a is simply 1/K d . For most ligand/protein interactions, binding happens as fast as ligand and protein can come together: in other words, the on-rate is diffusion controlled. The on-rate (kon ) is usually assumed to have a value of around 108 M−1 s−1 , and will only be significantly slower if the protein binding site is somehow buried or hidden (which is unusual, particularly for inhibitors as compared to substrates). We can therefore get a good estimate of the off-rate if we know the dissociation constant, since koff ≈ 108 K d . A good drug Graham A. Webb (ed.), Modern Magnetic Resonance, 1357–1362.
C 2008 Springer.

often has a K d in the low nM range, and therefore has a koff of around 1 s−1 or less. By contrast, initial targets derived from screening programes will typically have K d values in the high µM or low mM range, and will therefore have off-rates of around 105 s−1 . This leads to very different NMR behavior. We also need to consider that the concentrations of samples in NMR are comparable to some of these numbers, and therefore the fraction of ligand bound is strongly dependent on K d . It is easily shown that for the simple binding equilibrium shown above, the concentration of free ligand is given by [L] = 0.5([L0 ] − [P0 ] − K d + {([L0 ] − [P0 ] − K d )2 + 4[L0 ]K d }1/2 )

(2)

where [L0 ] and [P0 ] are the total concentrations of ligand and protein respectively. So for example, if the total protein concentration is 50 µM, K d is 10 nM, and there is a 1:1 ratio of ligand to protein, then almost all the ligand is bound (only ca. 1.4% is free); if there is a 10-fold excess of ligand over protein, then to a good approximation one equivalent is bound and the other 9 are free. By contrast, with the same protein concentration, a 10-fold ligand excess and a much weaker K d of 1 mM, the concentration of bound ligand is only 16 µM (i.e. only one-third of the protein is bound), and 97% of the ligand is free. Thus, the amount of bound complex depends strongly on K d . This will be important when we come to think about what the NMR spectrum looks like. The off-rate is important for another reason, namely because it defines whether exchange processes are in “fast” or “slow” exchange [1]. Confusingly, there are different definitions of what is meant by fast and slow, depending on what we are interested in. The most obvious is the chemical shift timescale: whether we see one signal or two. This depends on ν, the difference in resonance frequency between the two signals, expressed in Hz. Fast exchange is when koff  ν. Thus if for example we are considering a 1 H spectrum at 500 MHz, in which free and bound signals are 0.3 ppm different, ν = 150 Hz, so koff must be 1000 s−1 or greater for fast exchange (corresponding to a K d of 10 µM or weaker). Slow exchange requires koff 1.12, it is negative (solid line).

The Transferred NOE

HA

This is a process analogous to heat diffusion, whereby an NOE from one proton to another can subsequently diffuse to other protons throughout the protein. There are thus several pathways for NOE transfer, and therefore any spin diffusion immediately renders distance information less useful. This can be illustrated with a simple three-spin example (Figure 2), which shows that a third proton that bridges between the two protons being measured can reduce the apparent distance between them if it is significantly closer to both of them than they are to each other. Clearly, this is much more of a problem when trying to measure long internuclear distances. It is hard to detect the presence of spin diffusion (except using ROESY experiments, as described below), and therefore spin diffusion is normally handled by using rather wide ranges for NOE distance calibration and short mixing times. This clearly makes structure calculation less precise.

0.25

rIX

X 0.2

rISapp 2.82

rIX °

2.2

3.0 A

S

I

NOE

0.15

2.97 2.5

0.1

3.00

4.0

0.05 0.0 0.0

Part II

Spin Diffusion

The Transferred NOE 1359

0.05

0.1

0.15

0.2

time (s) ˚ apart, and there Fig. 2. Spin diffusion. Protons S and I are 3.0 A is a third spin X equidistant between S and I . The figure shows ˚ results for three positions of X with distances rIX = rXS of 2.2 A ˚ (dotted) and 4.0 A ˚ (solid line). The time (dashed line), 2.5 A course of IS NOESY crosspeak buildup is shown, calculated for a smallish protein with a correlation time of 5 ns on a 500 MHz ˚ the IS spectrometer. It can be seen that whereas for rIX = 4 A ˚ there is a “lag time” of NOE builds up smoothly, for rIX = 2.2 A about 30 ms, during which spin X has little effect on the NOE, but then the IS NOE buildup rate increases because of spin diffusion via X . The initial buildup over 200 ms was fitted by hand and converted to an apparent distance by taking the sixth root of the buildup rate. Spin diffusion reduces the apparent distance rIS ˚ when proton X is close to both I and S. In from 3.0 to 2.82 A most cases, such a reduction in apparent distance would have little effect on the calculated structure.

NOE HB

ring flip HB′ NOE (spin diffusion) HC Fig. 3. The effect of ring flipping on NOEs. Proton A generates an NOE at B. If the phenylalanine ring flips 180◦ , then B ends up at B  , and can pass an NOE on to C by spin diffusion. This will be significant as long as the flip rate is greater than the NOE crossrelaxation rate, which is almost always the case. The buildup rate of NOE at proton C suggests it is close to A, when in fact it can ˚ away. be easily 8–10 A

Internal Mobility Internal mobility can affect the NOE in several ways. The main result, as for spin diffusion, is to make the relationship between NOE and distance less precise. Any mobility will cause the protons to have a range of distance separations. Because of the r −6 averaging, this almost always results in an apparent shortening of the distance. Very close protons cannot be brought any closer together, and so the distance averaging effect is more significant with more distant protons. In practice, this often makes the r −6 effect end up looking more like a r −5 or even r −4 effect. Potentially the most severe effect of motion is when there is an NOE to an aromatic proton. Flipping of the ring can lead to the NOE then being passed on by spin diffusion to protons on the other side of the ring (Figure 3). This will have a marked effect on apparent distances in the vicinity of the ring.

The Transferred NOE The transferred NOE makes use of a particularly useful phenomenon, namely that the rate of buildup of the NOE is approximately proportional to the correlation time (Figure 1). The average cross-relaxation rate in a ligand exchanging between free and bound states is given by = NF σF + NB σB

(4)

where NF and NB are the fractions of free and bound ligand. The cross-relaxation rate of bound ligand, σ B , is

1360 Part II

Pharmaceutical Sciences

Part II

much greater than the cross-relaxation rate of free ligand, σ F . It is therefore easy to produce a situation where the observed NOE is dominated by cross-relaxation in the bound state. Depending on the exact values of σ B and σ F (which depend on the tumbling rate of bound and free ligand), ligand:protein ratios of 10–40 can be and are often used, and give useful quantitative results. A further helpful aspect of this high ligand:protein ratio is that one can use protein concentrations in the low µM range and still measure NOEs. An NOE measurement (almost always run as a NOESY experiment) then produces NOEs characteristic of the bound state. In the usual situation in which the off-rate is fast on the chemical shift timescale, the only signals observed are sharp and virtually identical to those of free ligand, and consequently are easily assigned. We therefore observe NOEs characteristic of the bound state, but seen on the signals of the free state. The experiment has a characteristic unusual for NMR spectra of proteins, that the experiment actually gets better as the protein gets bigger, because σ B increases, and therefore the NOE is dominated more and more by the cross-relaxation rate in the bound state. For many ligands, σ F is so small as to be insignificant. It can (and should) always be checked by running a control NOE experiment in the absence of protein. For intermediate sized ligands, such as peptides, there is a particularly advantageous situation, in that the cross-relaxation rate goes through zero at ωτ c ≈ 1; as it happens, this is where many peptides come. This greatly reduces the need to worry about NOEs in the free peptide. There have been a number of detailed simulations carried out on the TRNOE [4–7]. Figure 4 shows simulation results of a closely similar experiment, the STD experiment (discussed later), which show a complex dependence on K d [8]. At K d values weaker than 1 mM (righthand side of Figure 4), the TRNOE falls off to zero because there is not enough ligand bound to give rise to an NOE. Conversely, at K d values stronger than about 10 nM (left-hand side of Figure 4), the off-rate is so slow that bound ligand hardly ever dissociates from the complex, and the TRNOE again goes to zero. There is thus in this simulation a “window” of K d of between 10 nM and 100 µM where the TRNOEs are not greatly affected by exchange rates and can be related simply to distance (see however the discussion in the figure legend). For really quantitative results (i.e. an NOE that is related only to the distance in the complex and is not affected by exchange), it is necessary to have koff much greater than the cross-relaxation rate. An approximate guide to estimating cross-relaxation rates in proteins has been suggested: [3] ˚ Protons cannot get much σ ≈ −24 × MWt (kDa)/r6 (A). ˚ apart, which means that the largest possicloser than 2 A ble cross-relaxation rate is approximately 0.375 × MWt, or 19 s−1 for a “typical” 50 kDa protein. This implies that koff should be at least 100 s−1 , or alternatively K d should

Fig. 4. The effect of irradiating a protein resonance (open circle) on ligand signal intensity, shown for two ligand signals, A (diamonds, solid line) and B (squares, dashed). The effect calculated is STD, which is closely similar to what would be measured ˚ except for the in a TRNOE experiment. All distances are 3 A distance from ligand proton B to its nearest neighbor which is ˚ This near neighbor causes a rapid loss of NOE from B 1.8 A. by spin diffusion: this loss (and the consequent reduction in the NOE measured at B) is not as severe in the case of TRNOE. A 10:1 excess of ligand over protein is used in the calculation, with kon = 109 s−1 M−1 , τ c = 10−7 s, a 600 MHz spectrometer and a uniform leakage relaxation rate of 0.3 s−1 on all protons. The free ligand is assumed to have zero NOE. Using data from Jayalakshmi and Krishna [8].

be 1 µM or weaker. Of course, for protons at more realistic distances, this limit can be relaxed somewhat. As the molecular weight of the protein increases, the limitations on K d become more stringent, such that for example for a 500 kDa protein, K d should be 10 µM or weaker. This implies that for properly quantitative results, the window of K d is very small, being only between 1 µM and 100 µM. Fortunately this is a range that includes a good proportion of ligands of interest, and one can go at least a factor of 10 to either side without losing much quantitative information. It should also be added that TRNOEs have been observed and used even with much more tightly bound ligands. Useful TRNOEs were for example observed for a ligand with a K d of 30 nM [9]. Some of the early work on TRNOEs was overly precise and optimistic in the power of TRNOE to produce structural information. The pendulum therefore later swung in the opposite direction, with dire warnings about the dangers of over-interpreting TRNOE spectra [1]. More recent work has brought the pendulum back towards its proper position, and has shown that with proper controls, TRNOEs can give distance information that is very nearly as accurate as from conventional structure calculations. Apart from the problems discussed below, the main limitation is simply that ligands often bind in fairly extended

The Transferred NOE

1. Spin Diffusion. This has already been described above, but in the TRNOE there is a further problem, that there could be spin diffusion from one ligand proton to another, going via a protein proton [10]. Such a transfer would be very hard to detect, and would seriously affect measured distances. Recent simulations have shown that in practice, this very seldom happens, basically because most ligands bind on the exposed surfaces of proteins and therefore are not close enough to the protein for this to happen (cf. Figure 2) [11,12]. In particular, if the ligand has a reasonably high density of protons (for example, a peptide) it would be uncommon for ligand-protein-ligand spin diffusion to affect a distance measurement. The spin diffusion problem is less severe if a short presaturation time is used [12]. However, the large excess of free ligand effectively dilutes the buildup of NOE on the bound ligand by exchanging with it, with the consequence that one can use longer mixing time for TRNOE experiments than for conventional NOE experiments on proteins: mixing times of 100–200 ms are common, even with very large proteins. (It should also be added that the large number of TRNOEs that can be measured also means that one inaccurate TRNOE is less likely to perturb the structure.) This does not remove the problem of spin diffusion within the ligand, which is as bad as it is in a conventional NOE measurement, and still implies a need for caution in interpreting NOEs as distances. It has been noted that spin diffusion (of any kind) can be detected using a transferred ROESY experiment, in which spin diffusion gives rise to smaller NOEs of opposite sign to direct NOEs. Bax and co-workers have provided an elegant example of a TRNOE that disappeared on using a transferred ROESY experiment, and was thereby shown to be due to spin diffusion [13]. However, this experiment typically requires a low peptide:protein ratio (and more careful control experiments in the absence of protein), because the ratio of σ B to σ F is not as great as for NOESY. It is also more prone to artifacts, and has not proved as popular as one might expect. Where used, it tends to be used as a check that measured NOEs are genuine, rather than as a stand-alone method. 2. Nonspecific Binding. Usually, one uses something like a 30-fold excess of ligand over protein. Under these circumstances, a ligand that binds specifically, but in addition binds nonspecifically but 100-fold more weakly, will be fully bound at the tight binding site and almost fully bound at the weaker site, Equation (2). It will thus give equally

strong TRNOEs from its strongly and weakly bound conformations (indeed, if it binds very strongly at the more specific site, these TRNOEs will be small for reasons shown in Figure 4, and thus almost all the TRNOE will in fact come from nonspecific binding!). This gives rise to averaged TRNOEs, which, if not spotted, will produce misleading distance information. This is best detected using a control experiment, in which a strong competitive ligand is added [7]. This will displace the ligand from the binding site, and any remaining TRNOEs must therefore arise from nonspecific binding. Nonspecific binding appears to be very common, and therefore a control experiment is extremely valuable: indeed, for quantitative results such an experiment is crucial.

Practical Implementation One of the beauties of TRNOE is that it is simple to do: one just needs to carry out a conventional NOESY experiment (in conjunction with a control using a tight inhibitor, to rule out nonspecific binding). It is common

Cl N N

N

N

O

Cl

1 MeO2S

Cl N

Me

N

N

N

O

2

N N

N

CN

N O O

3

CN

Part II

geometries on protein surfaces, which means that most of the distances measurable by NOE are already close in terms of the covalent structure, and thus there are relatively few informative NOEs that can be measured. It is worth looking at the two most important pitfalls, as listed below:

The Transferred NOE 1361

1362 Part II

Pharmaceutical Sciences

Part II

to add a relaxation filter to the experiment, often in the form of a T2 or T1ρ filter, to reduce broad signals from the protein and therefore make the baseline flatter and reduce integration problems. As an example, compound 1 is a lead compound as an inhibitor of farnesyltransferase (FTase), and therefore a potential antitumor agent. It binds to FTase with an IC50 of 2 nM, and its off-rate is therefore too slow to give TRNOEs. However, an analogue 2 has an IC50 of only 475 nM and is therefore suitable for TRNOE experiments. Measurements were carried out using 64 µM enzyme and 2 mM 2 (ca. 1:30), a typical concentration for such experiments. TRNOE experiments were carried out using several mixing times, going up to 300 ms. Secondary binding was identified by addition of an excess of a competitive ligand. Measurement of TRNOEs revealed an NOE between the cyanophenyl ring and the piperazinone ring, which prompted the authors to design a cyclised ligand 3 that resembled the bound conformation and has an IC50 of 0.1 nM.

Related Experiments The TRNOE is closely related to two other experiments described in this volume, namely saturation transfer difference (STD) and transferred cross-correlated relaxation. The only difference between TRNOE and STD is that in STD one saturates the protein and transfers magnetization from protein to bound ligand, whereas in TRNOE one uses intramolecular NOEs in the bound ligand. (It is also true that STD is normally run as a steady-state NOE experiment, with irradiation of the protein resonance for several seconds, while TRNOE is normally run as a 2D transient experiment.) The dependence on off-rate and molecular size is the same, and it is therefore applicable in the same circumstances. This of course means that STD does not produce a signal for very tightly bound ligands. A neat method has been devised to get around this

limitation, using competition experiments—for example, an STD seen to a relatively weakly binding ligand disappears on addition of a much tighter binding competitor, thus allowing the tight binding competitors to be identified in high-throughput screens. The same would hold with the TRNOE also, but it would of course be a less useful result. A combination of STD and TRNOE experiment is particularly valuable in screening strategies [14,15].

References 1. Neuhaus D, Williamson MP. The Nuclear Overhauser Effect in Structural and Conformational Analysis, 2nd ed. WileyVCH: New York, 2000. 2. Sandstr¨om J. Dynamic NMR Spectroscopy. Academic Press: Paris, 1982. 3. Post CB. Curr. Op. Struct. Biol. 2003;13:581–588. 4. Clore GM, Gronenborn AM. J. Magn. Reson. 1982;48:402– 417. 5. Campbell AP, Sykes BD. Annu. Rev. Biophys. Biomolec. Struct. 1993;22:99–122. 6. Lippens GM. Cerf C, Hallenga K. J. Magn. Reson. 1992;99:268–281. 7. Ni F. Prog. Nucl. Magn. Reson. Spectrosc. 1994;26:517– 606. 8. Jayalakshmi V, Krishna NR. J. Magn. Reson. 2002;155:106– 118. 9. Weimar T, Petersen BO, Svensson B, Pinto BM. Carbohydr. Res. 2000;326:50–55. 10. Jackson PL. Moseley HNB, Krishna NR. J. Magn. Reson. Ser. B 1995;107:289–292. 11. Eisenmesser EZ. Zabell APR, Post CB. J. Biomol. NMR, 2000;17:17–32. 12. Zabell APR, Post CB. J. Biomol. NMR, 2002;22:303–315. 13. Arepalli SR. Glaudemans CPJ. Daves GD. Kovac P, Bax A. J. Magn. Reson. Ser. B 1995;106:195–198. 14. H. Kogelberg, Solis D, Jimenez-Barbero J. Curr. Op. Struct. Biol. 2003;13:646–653. 15. Stockman BJ, Dalvit C. Prog. Nucl. Magn. Reson. Spectrosc. 2002;41:187–231.

1363

Mark S. Searle1 , Graham Balkwill1 , Huw E.L. Williams1 , and Evripidis Gavathiotis2 1 Centre

for Biomolecular Sciences, School of Chemistry, University Park, Nottingham NG7 2RD, UK; and 2 Laboratory of Molecular Biophysics, The Rockefeller University, New York, NY 10021, USA

Drug–Quadruplex Interactions Studied by NMR DNA quadruplex structures have been of considerable interest as supramolecular self-assembling systems [1] and because of their potential biological importance in the regulation of a variety of processes within the cell cycle including replication, transcription, and recombination [2]. The observation that the telomeric repeats at the ends of chromosomes (5
-TTAGGG in humans) are able to assemble in the presence of monovalent cations into folded structures containing stacked G-tetrads has bought them into focus as a drug target, notably to interfere with telomere maintenance in immortalized human tumorderived cell lines by inhibiting the enzyme telomerase [1–6]. The novel fluorinated polycyclic quinoacridinium cation RHPS4 (Figure 1a) shows enhanced binding to higher-ordered DNA structures (triplex/quadruplex) and is a potent inhibitor of telomerase function [7,8]. To investigate the drug–quadruplex interaction, as part of a rational ligand design approach, we have studied by NMR the structure (Figure 1b) and dynamics of the RHPS4 complex with the intermolecular parallel-stranded quadruplex d(TTAGGGT)4 , formed from the human telomeric repeat [9,10].

Exchange Rates for Drug Binding to Quadruplex DNA Fluorescence quenching studies have established that RHPS4 interacts with d(TTAGGGT)4 with a binding affinity of 2.2 × 105 M [7]. NMR lineshape analysis during drug titration studies shows that the drug is in fast exchange between free and bound states with the quadruplex accommodating drug molecules at both the 5
-AG and 5
-GT steps at the ends of the core G-quadruplex structure [10]. Drug complexation was monitored using the downfield guanine imino proton resonances between 10 and 12 ppm. These initially shift upfield due to ring current effects from the drug, broadening significantly at a bound drug:DNA ratio of ∼0.3, but subsequently sharpen as the fully bound state is reached.

Graham A. Webb (ed.), Modern Magnetic Resonance, 1363–1367.  C 2008 Springer.

For a ligand-binding interaction where free and bound states are in fast exchange [11,12], the observed chemical shift δ obs is given by, δobs = δf pf + δb pb

(1)

The parameters δ f and δ b are the chemical shifts for the free DNA and the fully bound DNA:ligand complex, and pf and pb are the fractional populations of these species ( pf + pb = 1), enabling the chemical shift change to be used to determine pb as a function of drug concentration. The observed line width (LWobs ) is related to the populations of free and bound states with an additional exchange contribution given by the expression, LWobs = pf. LWf + pb LWb + [ pb (1− pb )2 4π(ν)2 ]/koff (2) where koff is the off-rate (per second) and ν the chemical shift difference (Hz) between the free and fully bound states, and LWf and LWb the line widths in the free and bound states. The exchange contribution, and hence the off-rate, can be determined by monitoring LWobs as a function of the fraction bound ( pb ) at each site. Selected 1D NMR spectra from the RHPS4 titration study at 303 K are shown in Figure 2a, with line width changes as a function of the pb shown in Figure 2b. Data for the guanine imino proton resonances G5 and G6 were fitted to Equation (2) using values of ν of 205 and 415 Hz, giving estimated values for koff of 1814 (±170)/s and 3812 (±250)/s. Given the uncertainty in line width measurements at values of pb close to ∼0.3 (Figure 2), the factor of ∼2 difference in exchange rates measured using the two different probes probably represents a more realistic estimate of possible errors, at least in this context, and hence we report a mean value for koff of 2800 (±1000)/s. Thus, despite the fact that RHPS4 is a potent inhibitor of telomerase its interaction with quadruplex DNA is highly dynamic. Structural analysis reveals that the drug is not able to intercalate between the G-tetrads since the core structure is highly stabilized by K+ ions which are

Part II

NMR Kinetic Measurements in DNA Folding and Drug Binding

1364 Part II

Pharmaceutical Sciences

Part II Fig. 1. (a) Chemical structure of the telomerase inhibitor RHPS4; (b) NMR structure of the 2:1 complex of RHPS4 bound at the 5 -AG and 5 -GT sites in the intermolecular parallel-stranded quadruplex formed from the human telomeric repeat d(TTAGGGT)4 . Only the core AGGGT structure is shown (5 -terminal thymines are relatively disordered). From Ref. [10]. (See also Plate 104 on page XXII in the Color Plate Section.)

octahedrally co-ordinated to the O6 carbonyl of guanine bases in adjacent tetrads. Displacement of these ions is energetically costly. Consequently, quadruplex-binding ligands appear to associate with the terminal G-tetrads, binding at the 5 -AG and 5 -GT steps within the sequence 5 -TTAGGGT (Figure 1b), where the adjacent nucleotides are less well structured [9,10]. As we have shown, small energy barriers to binding and dissociation permit fast kinetics at these sites.

DNA Hairpin Folding and Slow Exchange Eq uilibria The mechanisms and pathways by which biomolecules fold to their native structures is well developed in the protein folding arena; however, the folding of DNA and large RNA assemblies is still poorly understood [13]. The process is likely to be just as complex, involving multiple pathways and kinetic traps originating from misfolded structures and non-native interactions [13]. DNA and RNA hairpins, consisting of a double stranded stem region and connecting loop, occur naturally in single stranded nucleotide sequences and serve a range of important biological functions. A number of DNA mini-hairpin sequences have been identified with extraordinary stability containing GNNA and GNA loop sequences (N = any nucleotide) where the purine bases fold back to form a G–A wobble base pair. In the most stable hairpin structures, the G–A pair preferentially stacks on an adjacent C–G base pair in the adjoining double stranded stem region [14,15]. We have exploited the 5 -CGNAG minihairpin motif (with a 5 -GAA loop) in the design of hairpins with mismatched or extrahelical bases for studies of drug recognition and binding [16]. We recently identified a 13-mer sequence (5 -GCTACGTAGTCGC) that folds in

solution to form a hairpin with C–T mismatch (Figure 3a). In the presence of 100 mM NaCl (10 mM sodium phosphate, pH 7.0) the folded state is in slow exchange with the unstructured single strand with two sets of resonances assignable in 2D NMR spectra. The observation of two distinct species in slow exchange is unusual and indicative of a large activation energy barrier between folded and unfolded conformations. Significant chemical shift differences are observed between resonances in the two conformers (up to ∼0.3 ppm), making this an ideal system for characterizing the folding kinetics via magnetization transfer experiments.

Slow Exchange Between Two Conformers In the general case, two conformers A and B are in slow exchange when the first order rate constants kA and kB for interconversion are νAB , where ν AB is the chemical shift difference in Hz between a nucleus X in the two structural environments A and B. Selective perturbation of one signal can result in magnetization transfer to the other due to the exchange between environments. Assuming that the longitudinal relaxation rates in the two conformers are not considerably faster than the exchange rate, the time-dependent change in intensity after the perturbation can enable the rate of exchange to be measured [11,17]. Selective inversion of nucleus X in conformation A results in the recovery of the magnetization of A via a double exponential process, MA (t) = C1 eλ1t + C2 eλ2t + MA∞

(3)

Similarly, inversion of A leads to transfer of magnetization to B with recovery of the magnetization of B also

DNA Folding and Binding Kinetics by NMR

Slow Exchange Between Two Conformers 1365

Part II

(a)

(a)

(b)

1 0.8

intensity

0.6 0.4 0.2 0 -0.2 -0.4

(b)

-0.6

140

0

line width (Hz)

120

G6 G5

100 80 60 40 20 0 0

0.2

0.4

0.6

0.8

1

Fraction bound

Fig. 2. (a) Selected 600 MHz 1D NMR spectra at 303 K showing the guanine imino proton resonances of the G-quadruplex structure d(TTAGGGT)4 as a function of the fraction of drug bound ( pb ); resonances are shifted upfield by ring current effects from the bound drug and initially broaden (maximum line width at pb ∼ 0.3) before again sharpening as the binding sites are filled. From Ref. [10]. (b) Measured line width (LWobs ) as a function of the fraction of drug bound showing the best fit to Equation (2).

0.2

0.6 0.4 time (s)

0.8

1

Fig. 3. (a) Proposed equilibrium between the hairpin conformation of the 13-mer d(GCTACGTAGTCGC) and the disordered single strand with interconversion rates kA and kB . (b) Magnetization transfer data showing the recovery of magnetization after selective inversion of the A4 H8 signal for the single stranded conformation at 8.23 ppm (open squares), and the change in intensity of A4 H8 (hairpin conformation) due to magnetization transfer (black dots); 600 MHz data collected at 308 K. Both data sets were analyzed using a double exponential fit [see Equations (3) and (4)]. The experimental parameters obtained are as follows (open squares): C1 = 0.458, λ1 = 2.787, C2 = 0.765, λ2 = 9.892, MA∞ = 0.677, and MA0 = 0.521; (black dots) C3 = 0.895, λ1 = 2.572, C4 = 0.455, λ2 = 10.014, MB∞ = 1.022, and MB0 = 0.582.

following a double exponential decay, MB (t) = C3 eλ1t + C4 eλ2t + MB∞

(4)

where MA∞ and MB∞ are the equilibrium longitudinal magnetization of X in the two conformations A and B. The various parameters have the following relationships as

1366 Part II

Pharmaceutical Sciences

Part II

solutions of McConnell’s equations [18]: λ1 = {−(k1A + k1B ) + [(k1A − k1B )2 + 4kA kB ]1/2 }/2 (5) λ2 = {−(k1A + k1B ) − [(k1A − k1B ) + 4kA kB ] 2

1/2

}/2 (6)

C1 = [(λ2 +

k1A )(MA∞

−

MA0 )

−

kB (MB∞

−

MB0 )]/

(λ1 − λ2 ) C2 = [−(λ1 +

(7)

k1A )(MA∞

−

MA0 )

kB (MB∞

−

MB0 )]/

k1A )(MB∞

−

MB0 )]/

+

(λ1 − λ2 ) C3 =

[−kA (MA∞

(8) −

MA0 )

− (λ1 +

(λ1 − λ2 ) C4 =

[kA (MA∞

−

(9) MA0 )

+ (λ2 +

k1A )(MB∞

−

MB0 )]/

(λ1 − λ2 )

(10)

where k1A and kA , and k1B and kB are related by k1A = kA + R1A

(11)

k1B = kB + R1B

(12)

where kA and kB are the required rate constants (Figure 3), R1A and R1B the longitudinal relaxation rates, and MA0 and MB0 are proportional to the initial intensities immediately after the perturbation to the magnetization of A. Since conformers A and B are in equilibrium, the following relationship must also hold: kA MA∞ = kB MB∞

(13)

The double exponential nature of the time-dependence of the magnetization means that λ1 and λ2 can only be determined independently with any certainty if they differ by more than a factor of 3, otherwise only the combination C1 λ1 + C2 λ2 can be determined, as discussed in detail elsewhere [11,17]. This is remedied by repeating the experiment by selectively inverting signal B to provide another pair of independent combinations.

DNA Hairpin Folding K inetics by Magnetization Transfer Inversion transfer experiments were used to characterize the hairpin to single strand equilibrium at 308 K where the two conformers are significantly populated (MA∞ /MB∞ = kB /kA = 0.662). Selective inversion was achieved using a 180◦ Gaussian-shaped pulse with the excitation bandwidth optimized to give the desired selectivity. Perturba-

tion of the A4 H8 signal at 8.23 ppm in the single strand conformation results in transfer of magnetization to the resonance of the folded hairpin at 8.28 ppm. The recovery curves for both resonances are illustrated in Figure 3b. A double exponential fit resolves two rate constants which differ by a factor >3. Making use of Equation (13) to substitute into Equations (11) and (12), and subsequent substitutions into Equations (5)–(10) using experimentally determined values for λ1 , λ2 , and C1 –C4 , enables all of the unknown rate constants in Equations (11) and (12) to be determined, from which we deduce that kA = 3.3/s and kB = 2.1/s at 308 K. Rate constants were similarly determined from the reverse experiment (inversion of the hairpin signal at 8.28 ppm), and from inversion of other resonance pairs, with all the data in very close agreement. We have chosen experimental conditions under which the two conformations are of similar stability (at 308 K, K eq = 1.6) with the slow rate of interconversion indicating that the two states are separated by a large activation barrier. Recent thermodynamic investigations involving mutation of residues in various mini-loop sequences, including 5 -CGNAG, suggest that stacking interactions are highly co-operative with base substitutions or functional group mutations leading to large changes in stability [19,20]. Thus, a significant kinetic barrier may originate from the requirement for a highly structured loop region in the transition state which may have many of the interactions of the fully folded state. Our investigations have shown that the introduction of the destabilizing C–T mismatch has enabled the equilibrium to be sufficiently perturbed to be able to detect both folded and disordered species simultaneously in slow exchange. Magnetization transfer methods are ideally suited to measuring the kinetics of such systems, and have similarly been employed in characterizing coiled-coil transitions in the folding of GCN4 leucine zipper peptides [21].

Acknowledgments We acknowledge our collaborators Professor Malcolm Stevens and Dr. Robert Heald on the drug–quadruplex studies, and the support from the EPSRC of the UK and the School of Chemistry, University of Nottingham.

References 1. Neidle S, Parkinson G. Nat. Rev. Drug Discov. 2002;1:383. 2. Wheelhouse RT, Sun D, Han H, Han FX, Hurley LH. J. Am. Chem. Soc. 1998;120:3261. 3. Mergny J-L, Mailliet P, Lavelle F, Riou J-F, Laoui A, Helene C. Anticancer Drug Des. 1999;14:327. 4. Davies JT. Angew. Chem. Int. Ed. 2004;43:668.

DNA Folding and Binding Kinetics by NMR

13. Pan J, Thirumalai D, Woodson SA. J. Mol. Biol. 1997;273:7. 14. Hirao I, Kawai G, Yoshizawa S, Nishimura Y, Ishido Y, Watanabe K, Miura K. Nucleic Acids Res. 1994;22: 576. 15. Yoshizawa S, Kawai G, Watanabe K, Miura K, Hirao I. Biochemistry. 1997;36:4761. 16. Colgrave ML, Williams HEL, Searle MS. Angew. Chem. Int. Ed. 2002;41:4754. 17. Led JJ, Gesmar H, Abildgaard F. Methods Enzymol. 1989; 176:311. 18. McConnell HM. J. Chem. Phys. 1958;28:430. 19. Moody EM, Bevilacqua PC. J. Am. Chem. Soc. 2003;125: 2032. 20. Moody EM, Bevilacqua PC. J. Am. Chem. Soc. 2003;125: 16285. 21. d’ Avignon DA, Bretthorst GL, Holtzer ME, Holtzer A. Biophys. J. 1998;74:3190.

Part II

5. Haider SM, Parkinson GN, Neidle S. J. Mol. Biol. 2003; 326:117. 6. Schouten JA, Ladame S, Mason SJ, Cooper MA, Balasubramanian S. J. Am. Chem. Soc. 2003;125:5594. 7. Heald RA, Modi C, Cookson JC, Hutchinson I, Laughton CA, Gowan SM, Kelland LR, Stevens MFG. J. Med. Chem. 2002;45:590. 8. Gowan SM, Heald RA, Stevens MFG, Kelland LR. Mol. Pharmacol. 2001;60:981. 9. Gavathiotis E, Heald RA, Stevens MFG, Searle MS. Angew. Chem. Int. Ed. 2001;40:4749. 10. Gavathiotis E, Heald RA, Stevens MFG, Searle MS. J. Mol. Biol. 2003;334:25. 11. Lian L-Y, Roberts GCK. In: GCK Roberts (Ed). NMR of Macromolecules. IRL Press, Oxford, 1993, p 152. 12. Craik DJ, Pavlopoulos S, Wickham G. NMR in Drug Design. CRC Press Inc., New York, 1996, p 423.

References 1367

1369

Søren M. Kristensen1 , Marina R. Kasimova2 , and Jens J. Led1 1 Department 2 Optics

of Chemistry, University of Copenhagen, DK-2100 Copenhagen, Denmark; and and Fluid Dynamics Department, Risø National Laboratory, DK-4000 Roskilde, Denmark

Introduction The power of NMR as a method for characterizing the solution state of proteins stems from the highly specific way that NMR reports on protein structure and dynamics. This is due to the fact that NMR parameters, such as chemical shifts, relaxation rates and exchange-rate parameters report specifically about the local environment around the investigated nuclei. This obviously requires that the relevant NMR resonances have been assigned to the individual nuclei of the protein structures, which with present day techniques for isotope labeling and multidimensional NMR methods is possible for proteins with molecular weights higher than 30 kDa. After the resonances have been assigned, the overall three-dimensional structure of a protein can be determined from distance and angular constraints derived from additional spectral information. However, a detailed NMR determination of the threedimensional solution structure of proteins may be hampered by the dynamics of the protein molecules. At the same time, a detailed characterization of the flexible regions in the protein is highly interesting due to the functional importance of these regions, as shown by several NMR studies in recent years [1–4]. To solve this immediate conflict, that is, to alleviate the impact of the dynamics on the NMR structure determination of proteins, while at the same time obtaining important information on their dynamics and flexibility, several approaches can be used. Here we illustrate some of the approaches that have been used in the studies of the structure and function of two flexible insulin mutants and a flexible state of human growth hormone (hGH). One of the approaches concerns the determination of the relatively fast exchanging amide protons involved in weak intra- and intermolecular hydrogen bonds. This approach is useful in cases where the exchange rates of the amide protons are relatively fast while separate amide proton signals can still be observed, that is, the exchange takes place within an hour or less. This approach is illustrated using a monomeric insulin mutant that,

Graham A. Webb (ed.), Modern Magnetic Resonance, 1369–1376.
C 2008 Springer.

although flexible, still forms a well-defined structure in water. Another approach focuses on the use of the helixpromoting solvent 2,2,2-trifluoroethanol (TFE) to stabilize helices inherently encoded in the amino acid residue sequence of the protein. This approach is illustrated in the study of a highly flexible insulin mutant without a well-defined structure in water. TFE strengthens helical structures in peptides and proteins and induces helical structures in segments which have the propensity for forming helices [5,6]. It has also been found that TFE stabilizes β-turns and β-hairpins in proteins [7,8]. On the other hand, TFE disrupts the quaternary structure of proteins and reduces interactions between non-polar residues [9,10]. However, a complete assignment of the resonances may not always be possible. In particular, resonances in highly flexible regions in proteins may be broadened beyond detection and thereby preventing an assignment of these resonances. Still, NMR can provide valuable information about the structure and dynamics of such highly flexible proteins, despite an incomplete assignment. Here, this is exemplified by the study of hGH. A partial assignment can be used to track structural changes caused by variations in the solution conditions. Changes in pH, for instance, may result in alterations in the charge distribution on the protein surface, which may modulate both intra- and intermolecular coulombic interactions and hydrogen bond interactions. This, in turn, may give rise to changes in NMR parameters such as chemical shifts, relaxation rates and exchange rates of labile protons. Similarly, changes in the salt concentration and the buffer composition can modulate structure and stability of specific regions in proteins, as well as intermolecular interactions, while changes in temperature are likely to influence the stability of the structure of a protein and to affect exchange phenomena. Hence, by measuring the NMR parameters for a protein at different solution conditions, it is possible to obtain information about non-native and destabilized structural states, even if only a partial assignment is available.

Part II

The Use of NMR in the Studies of Highly Flexible States of Proteins: Relation to Protein Function and Stability

1370 Part II

Pharmaceutical Sciences

Part II

Insulin Flexibility and Activity Insulin is a 5.8-kDa peptide hormone with two chains, a 30-residue long A-chain and a 21-residue long B-chain, and with a three-dimensional structure as shown in Figure 1. Native insulin forms well-defined dimers and higher aggregates thereof in water solution, and in the presence of Zn2+ it forms a well-defined hexamer complexed with two zinc ions. However, the active form of insulin is the monomer. Therefore, the structure of several monomeric insulin mutants have been determined by NMR, and their in vivo activities have been compared, in order to elucidate the function of insulin and the mechanism of its interaction with the receptor, through the correlation between its structure and flexibility and its activity. However, active monomeric mutants are often highly flexible, making it necessary to use special NMR techniques to extract the dynamical and structural information. Here we describe the NMR investigations of two monomeric mutants, the des-[Phe(B25)] mutant, that is, an insulin mutant in which Phe in position B25 has been removed, and the T(B27)P, P(B28)T mutant (PT insulin), where Thr(B27) and Pro(B28) of the native insulin have been interchanged. In both cases, a hydrophobic patch that is surface

Fig. 1. The native insulin monomer consists of a 21-residue long A-chain (yellow) and a 30-residue long B-chain (red). Two disulfide bridges (SS bridge), CysA7-CysB7 and CysA20-CysB19, tie the two chains together, while a third disulfide bridge connects CysA6 and CysA11. The N- and C-terminal ends of the A-chain form α-helices (AI : A2–A8 and AII : A13–A20) connected by a loop. The central part of the B-chain forms a third helix (B10–B19). The C-terminal end of the B-chain forms an extended structure, while the N-terminal end is unstructured. (See also Plate 105 on page XXII in the Color Plate Section.)

exposed in native insulin is turned toward the inside of the molecule, resulting in intramolecular hydrophobic interactions that eliminate the aggregation propensity. Both mutants have a biological activity that is higher than that of native insulin. However, they also have flexible structures. In the case of des-[Phe(B25)]insulin this results in increased exchange rates of the H-bonded amide protons involved in the stabilization of the α-helices, while in the case of PT insulin the structure is too unstable in water solution to be determined by NMR.

Des-[Phe(B25)] Insulin: Quantitative Exchange Rates of Weakly H-bonded Amide Protons in Proteins from a Single 2D Experiment To get more detailed information about the structure of the des-[Phe(B25)] mutant, the relatively fast exchange rates of the H-bonded amide protons were determined quantitatively using the two-dimensional (2D) NMR method for measuring exchange rates of the order of reciprocal hours (h−1 ). This method is based on a single two-dimensional NOESY or TOCSY spectrum of the protein. It exploits the line broadening of the amide cross-peaks in the indirect dimension, caused by the hydrogen/deuterium exchange of the amide proton during the indirect evolution period when the protein is dissolved in D2 O. The method is described in detail in Ref. [11]. It allows the determination of amide proton exchange rates that are considerably faster than those, which can be measured by the conventional 2D approach based on series of NOESY spectra [12]. Furthermore, it is far less time consuming than the conventional 2D approach, since it is based on a single 2D experiment. At the same time, the accuracy of the rates determined by the method is high because data points from all the slices in the indirect dimension of the 2D experiment contribute to the determination of the rates. The accuracy can be further improved if the method is combined with a spectral analysis base on linear prediction. Moreover, the range of exchange rates covered by the method can be further expanded using the linear prediction model method described previously [13]. Using this method together with standard methodologies for NMR structure determination of proteins, it was found [14] that, although the des-[Phe(B25)] mutant has the normal insulin structure in water (Figure 1), the exchange rates of the H-bonded amide protons are relatively fast. Thus, the rates are about an order of magnitude faster than those of an insulin mutant where the hydrophobic patch is surface exposed, namely the B9(Asp) or S(B10)D mutant where Ser in position B9 has been substituted by Asp [15,16]. This clearly indicates that the structure of the des-[Phe(B25)] mutant is considerably less stable than that of the B9(Asp) mutant. Most importantly, the absence in the des-[Phe(B25)] insulin of

NMR Studies of Highly Flexible Proteins

b

0.40 B24 A19

0.60

0.80

1.00

B24

7.20

7.00 1

(ppm)

B24

A19

1H

A19

A19

Part II

a

Insulin Flexibility and Activity 1371

6.80

H (ppm)

7.20

7.00 1

6.80

H (ppm)

Fig. 2. The amide/aromatic region of an 800 MHz 1 H–1 H NOESY spectra of PT insulin showing the correlations between the aromatic side chain protons and aliphatic side chain protons (a) in pure water and (b) in 35% TFE-d3 . The broadening of the signals in the spectrum recorded in water reflects the high flexibility of the side chains. (Reproduced with permission from Ref. [17].)

slow exchanging amide protons, corresponding to the intermolecular hydrogen bonds associated with the dimer formation, show unambiguously that the mutant does not form a dimer in water solution. This is in contrast to the B9(Asp) mutant that forms a well-defined dimer in water, as reflected by the slow exchange of amide protons involved in intermolecular hydrogen bonds that stabilize the dimer.

PT Insulin: Stabilizing the Structure of an Unstructured Protein by 2,2,2-trifluoroethanol In the case of proteins that are sufficiently flexible in water to escape any NMR structure determination, secondary or tertiary, the helix-promoting solvent TFE may stabilize the structure [10], allowing the determination of a sufficient number of NOE constraints for a regular NMR structure determination. This was demonstrated in a study of the structure and folding propensity of the highly flexible PT insulin mutant [17]. PT insulin is biologically active with an activity that is 50% higher than that of native insulin [18]. It can, therefore, adapt the conformation necessary for its binding to the insulin receptor (IR). However, it is also unusually flexible with a poorly defined structure in water. This is indicated by the broad NOESY cross-peaks between the side chain protons observed for the mutant dissolved in water (Figure 2a). Moreover, only a few slowly exchanging amide protons can be detected in the protein dissolved in water, and only at low temperature (15 ◦ C). The slowly exchanging amide protons are all involved in weak hydrogen bonds associated with very loose secondary or tertiary structure elements. The strength of a hydrogen bond can be measured by the protection factor, P, of the amide proton involved in the H bond [19]. A more detailed de-

scription of the protection factor is given below together with the studies of hGH. Suffice it to mention here that a normal hydrogen bond corresponds to a protection factor of ≥25 [20], while the protection of the observed slow exchanging amide protons in PT insulin in water solution are in the range from 2.1 to 18.7. When dissolved in a mixture of 35% TFE and water, PT insulin assumes a more stable structure, as shown by the relatively sharp and well-defined NOESY cross-peaks between the side chain protons observed for the mutant dissolved in this solvent (Figure 2b). Thus, a sufficient number of NOE-derived distance constraints and slowly exchanging amide protons could be identified to allow a regular NMR structure determination using the program X-PLOR [21]. Moreover, the structure that PT insulin assumes is identical to the structure of native insulin shown in Figure 1. That is, even though PT insulin does not have a well-defined structure in water, it has the propensity to form the native fold.

Model for the Insulin–Receptor Interaction It is well established that the flexibility of the insulin molecule is important for its function. Thus, it was found [22] that an insulin mutant, where the N-terminal end of the A-chain was tied to the C-terminal end of the B-chain by a short peptide link, was biologically inactive even though it has a three-dimensional structure similar to that of native insulin (Figure 1). Also NMR studies of insulin mutants in which the C-terminal end of the B-chain is turned away from the rest of the molecule [23,24] indicate that such a structural change is required for insulin to interact with its receptor. In the light of these findings, the concomitant high flexibility and biological activity of PT insulin is interesting.

1372 Part II

Pharmaceutical Sciences

Part II

The Acid State of Human Growth Hormone

Fig. 3. Model of the interaction of native insulin with the insulin receptor (IR). I, free insulin in solution; II, detachment of the C-terminal part of the B-chain (vertical helix) from the rest of the molecule; III, unfolding of the AI -helix; IV, refolding of the AI -helix after navigating into the binding site of receptor (gray) and binding to the receptor. The enhanced biological activity of the PT insulin mutant could result from the fact that the AI -helix of PT insulin is very loose or non-existing in water, yet it has the propensity to form the helix (see text). Therefore, PT insulin does not need to go through a genuine unfolding before navigating into the active site of the receptor. (Reproduced with permission from Ref. [17].) (See also Plate 106 on page XXII in the Color Plate Section.)

Together with the observation that the mutant has the propensity to form the native fold, these characteristics led to the suggestion of the mechanism for the binding of insulin to its receptor (IR) shown in Figure 3. This mechanism also includes a detachment of the C-terminal end of the B-chain from the rest of the molecule before the interaction with the receptor can take place, as suggested in previous studies [23,24].

In order to rationalize a rather puzzling study of hGH, the so-called acid state that occurs at pH 2.7, we shall compare this state with the native state of hGH at pH 7.0. hGH is a 191-residue four-helix bundle protein with a classical upup-down-down cytokine topology [25]. The native state is characterized by a high thermal and chemical stability, similar to most native states of globular proteins. The acid state of hGH is more difficult to categorize, because it has characteristics of both the molten globule and the native state. Similar to the native state, the acid state has a high stability toward both chemical [26] and thermal denaturation and shows a cooperative unfolding behavior [27]. These features are contradictory to a classic definition of molten globules as defined by Ptitsyn [28]. They also differ from the behavior of other four-helix bundle proteins at similar solution conditions. Thus, the two cytokines, interleukin-2 and interleukin-4, were reported to adopt a classical molten globule state [29] and a “highly ordered molten globule” state [30], respectively, at low pH. In accordance with the stability data, far-UV CD measurements indicate that the structural properties of the acid state of hGH are very similar to those of its native state, suggesting that the helical content of the two states are similar. However, the near-UV CD spectrum of the acid state deviates from the spectrum of the native state and is more similar to the spectrum of denatured hGH [26]. This indicates that the aromatic side chains in the acid state are disordered, just like in the denatured state or a molten globule state. It is puzzling how it is possible to have a structural state with both a high stability toward denaturation and, at the same time, a disordered and flexible protein core. Using standard three-dimensional triple-resonance techniques combined with uniform 13 C- and 15 N-labeling, only partial assignments of the hGH molecule were obtained [31]. In total, 60% and 75% of the backbone resonances were assigned for the native and acid states, respectively. Complete resonance assignment was not possible because of severe resonance overlap and line broadening effects [31]. However, even without access to the solution structure of hGH, NMR investigations of the protein allow a detailed comparison of the native and the acid states. The investigation includes analyses of the backbone chemical shifts, the amide and indole hydrogen/deuterium exchange rates, and backbone amide 15 N relaxation rates. Backbone Cα and Hα chemical shifts contain information about the secondary structure, and provide means for accessing the structural integrity of a protein under a particular set of experimental conditions. This information is derived by calculation of secondary chemical shifts and a subsequent comparison with reference values for helix,

NMR Studies of Highly Flexible Proteins

Part II

ph 7.0

The Acid State of Human Growth Hormone 1373

ph 2.7

Fig. 4. Distribution of measured protection factors (P) for the native state (pH 7) and the acid state (pH 2.7) of human growth hormone. The protection factors are color mapped onto the three-dimensional structure as indicated in the figure. Structural parts colored in gray correspond to regions where the amide proton exchange rates could not be quantified. The figure was created with UCSF chimera [40]. (See also Plate 107 on page XXIII in the Color Plate Section.)

sheet, and extended structure [32]. Exchange-rate constants of labile protons report directly on solvent accessibility and in particular on the protection of labile protons by hydrogen bonds [33]. Sequence-related differences in intrinsic exchange-rate constants, kintr , and their dependencies upon pH, are taken into account by calculation

of protection factors P = kex /kintr [19], where kex is the experimentally derived exchange-rate constant. Measurement of 15 N longitudinal (R1 ) and transverse (R2 ) relaxation rates, and [1 H]–15 N heteronuclear NOE for each assigned amide 1 H–15 N pair, and analysis within the framework of the Lipari–Szabo model-free formalism [34,35]

S2 1.0

pH 7.0

0.5 0.0

0

50

100

150

S2 1.0

pH 2.7

0.5 0.0 0

50

100 Sequence

150

Fig. 5. Squared generalized order parameters (S 2 ) for the native state (pH 7.0) and the acid state (pH 2.7) of human growth hormone mapped as a function of the amino acid sequence. The black horizontal bars indicate the position of the four central helices A, B, C, and D forming the helix bundle (residues 9–34, 72–97, 106–128, and 155–184, respectively) and the gray horizontal bars indicate the position of the minor helices observed in the crystal structure of the hormone–receptor complex (PDB entry 3HHR).

1374 Part II

Pharmaceutical Sciences

Part II

provide a structural mapping of the so-called generalized order parameters, which reflect the restriction of the N–H bond vector on the pico- to nanosecond dynamics on a scale from zero to one. The 15 N relaxation measurements also provide information about exchange processes on the micro- to millisecond timescale as reflected in line broadening effects, and finally, information about the overall molecular tumbling rate and rotational diffusion anisotropy [36,37]. Backbone Cα and Hα assignments for the two states form the basis for a chemical-shift-based prediction of secondary structure [32]. This provides secondary structural data for the protein at the level of the individual residues without a priori knowledge of the overall structure. A comparison with the crystal structures of the free hormone (PDB entry 1HGU) and of the hormone bound to the extracellular domain of the receptor (PDB entry

Fig. 6. Mapping of the aromatic residues in ball-and-stick representation on the ribbon diagram of human growth hormone shows a high density of aromatic groups in the north end of the molecule and a low density of aromatic groups in the south end of the molecule. The figure was created with UCSF chimera [40]. (See also Plate 108 on page XXIII in the Color Plate Section.)

3HHR) shows a high degree of similarity between the solution-state predictions at high and low pH and the two crystal structures [31]. The general conclusion that can be drawn from the NMR data is that the helix bundle scaffold is preserved at both high and low pH. However, subtle differences exist, including evidence for a helical propensity from residue 102 to 107 in the native solution state at neutral pH, which is not seen in the acid state. Also, the secondary chemical shifts indicate that the mini-helix from residue 94 to 100, observed in the crystal structure of the hGH–receptor complex but not in the crystal structure of free hGH, is present in the solution states. The partial assignment of hGH gives access not only to structural information. NH exchange and backbone 15 N relaxation report on the flexibility of a molecule, and hence can help in the understanding of molecular stability.

NMR Studies of Highly Flexible Proteins

structure and stability of proteins can be obtained from a multitude of available NMR data, despite an incomplete assignment of the NMR spectra.

Acknowledgement The 800 MHz spectra were acquired at The Danish Instrument Center for NMR Spectroscopy of Biological Macromolecules.

References 1. Feher VA, Cavanagh J. Nature. 1999;400:289. 2. Eisenmesser EZ, Bosco DA, Akke M, Kern D. Science. 2002; 295:1520. 3. Ma LX, Hass MAS, Vierick N, Kristensen SM, Ulstrup J, Led JJ. Biochemistry. 2003;42:320. 4. Hansen DF, Hass MAS, Christensen HM, Ulstrup J, Led JJ. J. Am. Chem. Soc. 2003;125:6858. 5. Buck M, Radford SE, Dobson CM. Biochemistry. 1993;32: 669. 6. Hamada D, Kuroda Y, Tanaka T, Goto Y. J. Mol. Biol. 1995; 254:737. 7. Blanco FJ, Ortiz AR, Serrano L. Fold. Des. 1997;2:123. 8. Mabrouk K, Vanrietschoten J, Rochat H, Loret EP. Biochemistry. 1995;34:8294. 9. Albert JS, Hamilton AD. Biochemistry. 1995;34:984. 10. Buck M. Q. Rev. Biophys. 1998;31:297. 11. Olsen HB, Gesmar H, Led JJ. J. Am. Chem. Soc. 1993;115: 1456. 12. Wagner G, W¨uthrich K. J. Mol. Biol. 1982;160:343. 13. Moss R, Gesmar H, Led JJ. J. Am. Chem. Soc. 1994;116: 747. 14. Jørgensen AMM, Olsen HB, Balschmidt P, Led JJ. J. Mol. Biol. 1996;257:684. 15. Kristensen SM, Jørgensen AMM, Led JJ, Balschmidt P, Hansen FB. J. Mol. Biol. 1991;218:221. 16. Jørgensen AMM, Kristensen SM, Led JJ, Balschmidt P. J. Mol. Biol. 1992;227:1146. 17. Keller D, Clausen R, Josefsen K, Led JJ. Biochemistry. 2001;40:10732. 18. Clausen R, Jørgensen TG, Jørgensen KH, Johnsen AH, Led JJ, Josefsen K. Eur. J. Endocrinol. 2002;147:227. 19. Bai YW, Milne JS, Mayne L, Englander SW, Proteins Struct. Funct. Genet. 1993;17:75. 20. Mori S, Abeygunawardana C, Berg JM, Vanzijl PCM. J. Am. Chem. Soc. 1997;119:6844. 21. Brunger AT. X-PLOR: A system for X-ray Crystallography and NMR. Yale University Press: New Haven, Connecticut, 1992. 22. Derewenda U, Derewenda Z, Dodson EJ, Dodson GG, Bing X, Markussen J. J. Mol. Biol. 1991;220:425. 23. Hua QX, Shoelson SE, Kochoyan M, Weiss MA. Nature. 1991; 354:238. 24. Ludvigsen S, Olsen HB, Kaarsholm NC. J. Mol. Biol. 1998;279:1. 25. de Vos AM, Ultsch M, Kossiakoff AA. Science. 1992;255: 306.

Part II

For the native state of hGH, protection factors were determined for 22 backbone amide proton and despite the scarcity of the data, it is apparent that the protection factors are high throughout the helix bundle with values in the range 6 × 103 to 1 × 106 . Furthermore, the indole NH of tryptophan 86 shows a high degree of protection (P ∼ 1 × 106 ) consistent with the hydrogen bond to Oδ1 of Asp169 first suggested by Bewley and Li [38] and later verified in the crystal structures. In the acid state, the backbone protection factors are orders of magnitude smaller than in the native state, and are in the range from 9 × 103 to 2 × 103 , with the highest protection factors in the south end of the molecule. Figure 4 graphically illustrates the distribution of protection factors in the two solution states. Also, the Trp86 indole NH is exhibiting fast exchange with the solvent consistent with absence of the hydrogen bond to Oδ1 of Asp169 observed in the native state. The backbone 15 N order parameters, S 2 , for the native state are uniformly high throughout the sequence, corresponding to a highly rigid globular protein. This is in agreement with the high protection factors observed for this state. The acid state, on the other hand, displays increased flexibility throughout the sequence, but most pronounced in the long loop regions between the four main helices, as shown in Figure 5. In particular, the loop between helices C and D display order parameters as low as ∼0.2, which is comparable to what is expected for an unfolded state [39]. Thus, the acid state of hGH is characterized by a preserved helix bundle structure with slightly increased overall flexibility as compared to the native state, but with highly mobile inter-helical loop regions. The notion of the acid state as a well-defined, though more flexible, helix bundle with flexible loop regions is consistent with the high stability of the acid state toward denaturation. Also the native-like far-UV CD spectrum is indicative of a helix bundle structure with largely preserved inter-helical stabilizing interactions. On the other hand, the similarity of the near-UV CD spectrum with that of the denatured state of the protein suggests a certain degree of disorder of the aromatic groups, and seems to contradict the picture of a native-like helix bundle with preserved inter-helix stabilizing interactions. As shown in Figure 6, the majority of the aromatic side chains are located in the north end of the molecule. Thus, a unified picture is emerging, where the north end of the helix bundle is opening up resulting in a disorder of aromatic groups and an increased solvent access to amide groups— in agreement with near-UV CD and NH exchange data. At the same time, inter-helix interactions in the south end of the molecule are preserved, which can account for the high stability toward denaturation. All taken together, the NMR investigation of hGH shows that important and detailed information about the

References 1375

1376 Part II

Pharmaceutical Sciences

Part II

26. DeFelippis MR, Kilcomons MA, Lents MP, Youngman KM, Havel HA. Biochim. Biophys. Acta-Prot. Struct. Mol. Enzymol. 1995;1247:35. 27. Kasimova MR, Milstein SJ, Freire E. J. Mol. Biol. 1998; 277:409. 28. Ptitsyn OB. J. Protein Chem. 1987;6:273. 29. Dryden D, Weir MP. Biochim. Biophys. Acta. 1991;1078:94. 30. Redfield C, Smith RAG, Dobson CM, Nat. Struct. Biol. 1994;1:23. 31. Kasimova MR, Kristensen SM, Howe PWA, Christensen T, Matthiesen F, Petersen J, Sørensen HH, Led JJ. J. Mol. Biol. 2002;318:679. 32. Wishart DS, Sykes BD, Methods Enzymol. 1994;239:363.

33. Englander WS, Downer NW, Teitelba H. Annu. Rev. Biochem. 1972;41:903. 34. Lipari G, Szabo A. J. Am. Chem. Soc. 1982;104:4546. 35. Lipari G, Szabo A. J. Am. Chem. Soc. 1982;104:4559. 36. Tjandra N, Feller SE, Pastor RW, Bax A. J. Am. Chem. Soc. 1995;117:12562. 37. Tjandra N, Wingfield P, Stahl S, Bax A. J. Biomol. NMR. 1996;8:273. 38. Bewley TA, Li CH. Arch. Biochem. Biophys. 1984;233:219. 39. Farrow NA, Zhang OW, Formankay JD, Kay LE. Biochemistry. 1997;36:2390. 40. Huang CC, Couch GS, Pettersen EF, Ferrin TE, Pac. Symp. Biocomput. 1996;1:724.

1377

John C. Lindon, Elaine Holmes, and Jeremy K. Nicholson Biomedical Sciences Division, Imperial College London, Sir Alexander Fleming Building, South Kensington, London SW7 2AZ, UK

Introduction Since the decoding of the human genome, there has been much interest in using changes in gene expression to discover the basis of disease and for the identification of new drug targets. However to date, the promise of the approach is yet to be comprehensively realized and there remains a difficulty in relating such changes to real conventional end points used in diagnosis and pharmaceutical development. The simultaneous measurement of many gene expression changes is termed transriptomics, and is usually carried out in an automatic fashion using so-called gene microarrays. The relevance of such gene changes is not always clear, which has led subsequently to efforts focused on the consequent protein level changes (a subject termed proteomics), but again it is not always possible to relate such changes directly to pathological events. On the other hand, metabonomics, a whole systems approach for evaluating metabolite changes and pathway analysis through study of biofluids and tissues [1] offers a process whereby real end points can be obtained. In complex organisms, these three levels of biomolecular organization and control are highly interdependent, but they can have very different time scales of change. All of the technologies, which rely on analytical chemistry methods, result in complex multivariate data sets that require a variety of chemometric and bioinformatic tools for interpretation. The aim of such procedures is to extract biochemical information that is of diagnostic or prognostic value, and that reflects actual biological events. The areas where metabonomics impacts pharmaceutical R&D are: r validation of animal models of disease, including genetically modified animals; r preclinical evaluation of drug safety and ranking of compounds; r assessment of safety in clinical trials and after product launch; r quantitation, or ranking, of the beneficial effects of pharmaceuticals both in development and clinically; r improved understanding of idiosyncratic toxicity; r improved, differential diagnosis and prognosis of clinical diseases; Graham A. Webb (ed.), Modern Magnetic Resonance, 1377–1385.
C 2008 Springer.

r better understanding of environmental population effects through epidemiological studies; r patient stratification (pharmaco-metabonomics); r nutrition, interactions between drugs, and between drug and diet; r studies in environmental science. Importantly, metabonomics also allows time-dependent patterns of change in response to stimuli to be measured. In multi-cellular organisms, there are a broad variety of time scales, varying widely according to gene, protein, pathway, and tissue. One important potential role for metabonomics therefore is to direct the timing of proteomic and genomic analyses in order to maximize the probability of observing “omic” biological changes that are relevant to functional outcomes. Metabonomics, formally defined as the quantitative measurement of the dynamic multi-parametric metabolic response of living systems to pathophysiological stimuli or genetic modification, thus provides an approach which leads to real world end points. Metabolites can be identified and quantified, and changes can be related to health and disease, and are changeable by therapeutic intervention. The subject of metabonomics has been reviewed recently [2–4].

Metabonomics Analytical Technologies The two principal methods that can produce metabolic profiles of biomaterials, comprise 1 H NMR spectroscopy [2] and mass spectrometry (MS) [5], the latter usually including a separation stage such as GC or LC. NMR has the advantages of being non-destructive, also applicable to intact tissues using magic angle spinning (MAS) methods [6], and provides detailed information on molecular structure, especially in complex mixture analysis. In addition, NMR can also be used to probe molecular dynamics as well as concentrations. MS is inherently considerably more sensitive than NMR, but it is necessary generally to employ different separation techniques for different classes of substances. Quantitation in MS can also be impaired by variable ionization effects, such that derivatization may be necessary. Most published mammalian studies have used NMR spectroscopy, but LC–MS is increasing in usage.

Part II

NMR-based Metabonomics Techniques and Applications

1378 Part II

Pharmaceutical Sciences

Part II

Typically metabonomics is carried out on biofluids that provide an integrated view of the whole systems biology. The biochemical profiles of the main diagnostic fluids, blood plasma, cerebrospinal fluid (CSF), and urine, reflect both normal variation and the impact of drug effects or disease on single or multiple organ systems [7]. Urine and plasma are obtained in a non- or minimally invasive fashion, and hence are appropriate for clinical trials monitoring and disease diagnosis. Different biofluids have characteristically different metabolic profiles (Figure 1). Although urine and plasma are the main diagnostic fluids, others have also been used in special cases. These include CSF, seminal fluids, digestive fluids, pathological fluids

such as cyst fluid, and administered fluids such as dialysis fluids. A standard 1 H NMR spectrum of urine typically contains thousands of sharp lines from predominantly low molecular weight metabolites. Plasma contains low and high molecular weight components, which give a wide range of signal line widths. Protein and lipoprotein signals dominate 1 H NMR spectra of plasma, with small molecule fingerprints superimposed on them. Standard editing experiments based on T1 , T1ρ , T2 , or diffusion coefficients can be used to select only the contributions from proteins, and other macromolecules and micelles, or alternatively to select only the signals from the small molecule metabolites. Each biofluid yields a characteristic 1 H NMR

plasma

bile

urine

Fig. 1. 800 MHz 1 H NMR spectra with water peak suppression of control human biofluids, urine, gall bladder bile, and blood plasma showing characteristically different biochemical profiles.

4.0

3.5

3.0

2.5

2.0

1.5

1.0

ppm

Metabonomics using NMR

The most powerful of these “hyphenated” approaches being HPLC–NMR–MS, [11] which can provide the full array of NMR and MS-based molecular identification tools. These include MS–MS for fragment ions and Fourier transform (FT)-MS or time-of-flight (TOF)-MS for accurate mass measurement and hence derivation of molecular empirical formulae. Within the last few years, the development of highresolution 1 H MAS NMR spectroscopy has had a substantial impact on the ability to analyze intact tissues [12]. Rapid spinning of the sample (∼4–6 kHz typically) at an angle of 54.7◦ relative to the applied magnetic field serves to reduce line broadening effects caused by sample heterogeneity, residual dipolar couplings, and residual chemical shift anisotropy. Thus, it is possible to obtain very high quality NMR spectra of whole tissue samples with no sample pretreatment. Such experiments indicate that diseased or toxin-affected tissues have substantially different metabolic profiles to those taken from healthy organs [13]. In addition, MAS NMR spectroscopy can be used to access information regarding the compartmentalization of metabolites within cellular environments. Such MAS NMR-based metabonomics can also be applied to in vitro systems such as tissue extracts [14], Caco-2 cells [15], or spheroids [16]. The NMR spectrum of a sample can be thought of as an object in a multi-dimensional set of metabolic coordinates, the values of which could be the spectral intensity at every data point. Similarity or differences between samples can then be evaluated using multivariate statistical methods or other pattern recognition approaches. One simple approach that has been widely used is to effect a dimension reduction from the typically 32K or 64K NMR intensity points describing each spectrum, down to a few dimensions to visualize similarities and differences between samples. This can be accomplished using principal components analysis (PCA). This constructs latent variables from linear combinations of the original descriptors to describe as much variation as possible in the data set, with first principal component (PC) explaining the maximum variation and successive components explaining decreasing amounts of variance with the constraint that all PCs are orthogonal to each other. The approach gives rise to two matrices based on the data, a scores and a loadings matrix. A plot of the PC scores shows the relationships between the samples (each point on the plot comprises one sample) and the PC loadings provide information on those variables, which contribute to the position of the samples in the scores plot. This is illustrated in Figure 2 for rat urine NMR spectra. The scores plot shows three clusters of samples, those of NMR spectra of control urine and those of NMR spectra of urine from animals dosed with two different liver toxins, hydrazine and a substance known as ANIT. These cause different biochemical changes and produce the separate clusters (Figure 2a). Examination of

Part II

spectroscopic fingerprint in which the spectral intensity distribution is determined by the relative concentrations of solutes, and in some cases by their intermolecular interactions. Standard spectra typically take only a few minutes to acquire using robotic flow injection methods. These entail robotic sample preparation involving buffering, addition of a chemical shift and quantitation standard such as TSP, and addition of D2 O as a magnetic field lock signal for the spectrometer. The large interfering NMR signal arising from water in all biofluids is easily eliminated by use of appropriate NMR solvent suppression methods such as NOESY presaturation or WET. Using NMR flow probes, the capacity for NMR analysis has increased enormously and now up to 200–300 samples per day can be measured. Identification of the biomarkers (i.e. metabolites that change in level as a consequence of the pathology) detected in a biofluid NMR spectrum can involve the application of a number of other NMR techniques including two-dimensional (2D) experiments [2]. Although 1 H NMR spectra of urine and other biofluids are very complex, many resonances can be assigned directly based on their chemical shifts, signal multiplicities, and by adding authentic material and remeasuring the spectrum. 2D NMR spectroscopy can also be useful for spreading the signals out and for working out the connectivities between signals, thereby enhancing the information content and helping to identify biochemical substances. These include the 1 H–1 H J -resolved experiment, which reduces the contribution of macromolecules and yields information on the multiplicity and coupling patterns of resonances. Other 2D experiments such as COSY and TOCSY provide 1 H–1 H spin–spin coupling connectivities. Use of other nuclei can be important to help assign NMR peaks and here inverse-detected heteronuclear correlations, usually 1 H–13 C, can also be obtained by use of sequences such as HSQC or HMBC. A major improvement in the scope of metabonomics has been made possible by the commercialization of miniaturized NMR probes. Now it is possible to study metabolic profiles by NMR using as little as 2–20 µl of sample, and examples have been published using CSF and blood plasma [8,9]. Cryogenic NMR probe technology, whereby the NMR detector coil and preamplifier are cooled to about 20 K, is now commercially available and provides an improvement in spectral signal–noise ratios of up to 500%. This improvement permits the routine use of natural abundance 13 C NMR spectroscopy of biofluids, such as urine or plasma, with acquisition times that enable a high throughput of samples. Information-rich 13 C NMR spectra of urine can be obtained using appropriately short acquisition times suitable for biochemical samples when using a cryogenic probe [10]. For metabolite identification, directly coupled chromatography–NMR spectroscopy methods can be used.

Metabonomics Analytical Technologies 1379

1380 Part II

Pharmaceutical Sciences

Part II

Fig. 2. Classification of rat urine samples based on their 1 H NMR spectra using principal components (PC) analysis. Animals were dosed with hydrazine or ANIT, two model liver toxins operating by different biochemical mechanisms, or with dosing vehicle alone (controls). (a) PC scores plot where each spectrum is reduced to a single point and samples with similar metabolic profiles appear close to each other. (b) PC loadings plot where each point corresponds to the signal intensity a single NMR spectral region of width 0.04 ppm and the position of a spectral region point determines which regions are responsible for the corresponding clustering in the scores plot [17].

t[2] ANIT

(a)

Controls t[1]

Hydrazine

w∗c[2]

(b)

bile acids

glucose 2-OG, succinate 2.44 3

3.92 2.54

hippurate 3.04

citrate

creatine

3.28 2.24

w∗c[1]

2./

taurine

2-AA, citrulline

the loadings plot, Figure 2b, gives information on which NMR spectral regions are responsible for the clustering in the scores plot, i.e. a cluster with a low PC1 and a low PC2 score (such as hydrazine) will appear in the bottom left hand quadrant and the corresponding loadings or altered NMR spectral regions will be in the same quadrant. In the real world, biochemical changes caused by disease, nutrition, lifestyle, drug therapy, or drug toxicity develop and recover in real time and there can be complex, time-related changes in NMR spectra of biofluids. Hence, in order to develop automatic classification methods, it has proved efficient to use advanced chemometrics and bioinformatics approaches, known as supervised methods, that also take time into account. In general, these

methods allow the quantitative description of the multivariate boundaries that characterize and separate each class of sample in terms of their metabolic profiles, some providing a level of probability for the classification. By using a variety of chemometric methods, it is possible to use such models to provide classification probabilities or even quantitative response factors for a wide range of sample types. It is important to build and test such chemometric models using independent training data and validation data sets. Subtle biochemical changes in 1 H NMR spectroscopic profiles of biofluids can be obscured in such analyses by interfering factors such as variations in pH, which can cause changes in NMR chemical shifts because of differences in ionization state of some molecules. In NMR

Metabonomics using NMR

Selected Applications of Metabonomics Preclinical Drug Candidate Safety Assessment There is evidence through the recurring need to withdraw drugs from the market place that drug safety assessment approaches used in the pharmaceutical industry can still fail. There is a need for methodologies that can pick up potential problems earlier, faster, more cheaply, and more reliably. A recent survey of market withdrawals during the period 1960–1999 has identified hepatotoxicity as the most common reason for withdrawal [20]. Minimizing attrition caused by drug adverse effects is therefore one of the most important aims of pharmaceutical R&D, and metabonomics has been used extensively in evaluating the adverse effects of candidate drugs. In this application, NMR-based metabonomics can be used for (i) definition of the metabolic hyperspace occupied by normal samples, (ii) rapid and simple classification of the sample as normal or abnormal (this enables spectrometer automation for data acquisition), (iii) classification of target organ toxicity, (iv) site and mechanism of action within the organ, (v) identification of biomarkers of toxic effect, and (vi) evaluation of the time course of effect, e.g. the onset, evolution, and regression of toxicity. The role that metabonomics has in the evaluation of xenobiotic toxicity has been comprehensively defined by the Consortium for Metabonomic Toxicology (COMET) formed between five pharmaceutical

companies and Imperial College, London, UK [21]. The aim of this project is to define methodologies and to apply metabonomic data generated using 1 H NMR spectroscopy of urine and blood serum for preclinical toxicological screening of candidate drugs. This has been achieved by generating databases of spectral and conventional results for a wide range of model toxins (147 in total) that serve as the raw material for computer-based expert systems for toxicity prediction. The project goals of the generation of comprehensive metabonomic databases (now around 35,000 NMR spectra) and multivariate statistical models for prediction of toxicity (initially for liver and kidney toxicity in the rat and mouse) have completely been achieved. A study operated to the same detailed protocol and using the same model toxin was carried out over seven sites in the companies and contract research organizations. This was used to probe both the analytical and biological variation that could both arise through the use of metabonomics. The inter-site NMR analytical reproducibility revealed a high degree of robustness where split samples were analyzed both at Imperial College and at various company sites, giving a coefficient of variation of about 1.6%. The biological variability was evaluated by a detailed comparison of the ability of the companies to provide consistent urine and serum samples with all samples measured at Imperial College. There was a high degree of consistency between samples from the various companies, and the differences between samples were small compared to the biochemical effects of the toxin, where dose-related effects could be distinguished [22]. Following this successful start, metabonomic models have been constructed for urine from control rats and mice, enabling identification of outlier samples and the metabolic reasons for the deviation. Building on this, and with the completion of all planned studies, and the development of new chemometrics methodology to meet the challenges thrown up by this project, workable expert systems for prediction of liver and kidney toxicity have been generated. To achieve this goal, a new approach to the classification of a large data set of COMET samples has been developed, termed—Classification Of Unknowns by Density Superposition (CLOUDS)—a novel non-neural implementation of a classification technique developed from probabilistic neural networks [23]. Modeling the urinary NMR data according to organ of effect (control, liver, kidney, or other organ), using a model training set of 50% of the samples and predicting the other 50%, over 90% of the test samples were classified as belonging to the correct group with only a 2% misclassification rate between the classes. This work showed that it is possible to construct predictive and informative models of metabonomic data, delineating the whole time course of toxicity—the ultimate goal of the COMET project.

Part II

spectroscopy of urine, one means of limiting the effects of pH on the chemical shift of sensitive moieties is to add a standard amount of buffer to the sample prior to NMR spectroscopic analysis. Alternatively, mathematical algorithms can be used to realign the chemical shifts of resonances from protons near ionizable groups displaced by pH effects [18]. Some spectral regions, such as those containing water or urea, are very variable due to water NMR peak suppression effects. In addition, many drug compounds or their metabolites are excreted in biofluids and these can obscure significant changes in the concentration of endogenous components. Therefore, it is usual to remove these redundant spectral regions prior to chemometric analysis. One example of a robust automatic data reduction method that has been widely used is the division of the NMR spectrum into regions of equal chemical shift ranges followed by signal integration within those ranges [19]. Automatic data reduction of 2D NMR spectra can be performed using a similar procedure, in which the spectrum is divided by a grid containing squares or rectangles of equal size, and the spectral integral in each volume element is calculated.

Selected Applications of Metabonomics 1381

1382 Part II

Pharmaceutical Sciences

Part II

Genetic Differences and other Physiological Effects by Metabonomics In order to determine therapeutic effects, it is necessary to understand any underlying physiological variation and to this end, metabonomics can be used to separate classes of experimental animals such as mice and rats according to a number of inherent and external factors based on the endogenous metabolite patterns in their biofluids [24]. Such differences may help explain differential toxicity of drugs between strains and inter-animal variation within a study. Metabonomics is also being used for the phenotyping of mutant or transgenic animals and the investigation of the consequences of transgenesis such as the transfection process itself [25]. This suggests that the method may be appropriate for following treatment regimes such as gene therapy. It is important to differentiate often-seen unintended consequences of the genetic engineering process from the intended result, because pharmaceutical companies are developing genetically engineered animal models of disease using transfection procedures. Metabonomic approaches can give insight into the metabolic similarities or differences between mutant or transgenic animals and the human disease processes that they are intended to simulate and their appropriateness for monitoring the efficacy of novel therapeutic agents. The importance of gut microfloral populations on urine composition has been highlighted by a study in which axenic (germ-free) rats were allowed to acclimatize in normal laboratory conditions and their urine biochemical composition was monitored for 21 days [26]. Many other effects can be distinguished using metabonomics, including male/female differences, age-related changes, estrus cycle effects in females, diet, diurnal effects, and interspecies differences and similarities [24].

Integrated Metabonomic Studies The value of obtaining multiple NMR data sets from various biofluid samples and tissues of the same animals collected at different time points has been demonstrated. This procedure has been termed “integrated metabonomics,” [13,14] and can be used to describe the changes in metabolic chemistry in different body compartments affected by exposure to toxic drugs. Such timed profiles in multiple compartments are themselves characteristic of particular types and mechanisms of pathology and can be used to give a more complete description of the biochemical consequences than can be obtained from one fluid or tissue alone. An example is given in Figure 3 that shows 1 H NMR spectra from intact liver tissue, tissue extracts, and blood plasma from a mouse after administration of a toxic dose of paracetamol. Building on this, it has also

been possible to integrate data from transcriptomics and metabonomics to find common metabolic pathways implicated by both gene expression changes and changes in metabolism [27].

Disease Diagnosis Many examples exist in the literature on the use of NMRbased metabolic profiling to aid human disease diagnosis, including the use of plasma to study diabetes, CSF for investigating Alzheimer’s disease, synovial fluid for osteoarthritis, seminal fluid for male infertility and urine in the investigation of drug overdose, renal transplantation, and various renal diseases. Most of the earlier studies have been reviewed [7]. Some studies have been undertaken in the area of cancer diagnosis using perchloric acid extracts of various types of human brain tumor tissue.[28] The spectra were classified using neural network software giving ∼85% correct classification. Tissues themselves can be studied by metabonomics using the MAS technique and published examples include prostate cancer [29] and renal cell carcinoma [30]. Other recent studies include an NMR-based urinary metabonomic study of multiple sclerosis in humans and non-human primates [31]. Currently, the only reliable diagnostic method for coronary heart disease (CHD) is the injection of X-ray opaque dye into the blood stream and visualization of the coronary arteries using X-ray angiography. This is both expensive and invasive with an associated 0.1% mortality and 1–3% of patients experiencing adverse effects. Recently metabonomics has been applied to provide a method for diagnosis of CHD non-invasively through analysis of a blood serum sample using NMR spectroscopy [32]. Patients were classified into two groups, those with normal coronary arteries and those with triple coronary vessel disease, as based on an angiographic examination. Around 80% of the NMR spectra were used as a training set to provide a two-class model after appropriate data filtering techniques had been applied and the samples from the two classes were easily distinguished. The remaining 20% of the samples were used as test set and their class was then predicted based on the derived model with a sensitivity of 92% and a specificity of 93% based on a 99% confidence limit for class membership. It was also possible to diagnose the severity of the CHD that was present by employing serum samples from patients with stenosis of one, two, or three of the coronary arteries. Although this is a simplistic indicator of disease severity, separation of the three sample classes was evident even though none of the wide range of conventional clinical risk factors that had been measured was significantly different between the classes.

Selected Applications of Metabonomics 1383

Metabonomics using NMR

glucose and glycogen

choline

–CH–CH2–CH– acetate

3.0

Mono-UFA Gly-C1(DG/TG/PE/PI) Gly-C3(DG) Gly-C3(PC/PE)

-N+(Me3)3

Gly-C2(PC/PE(PI/TG)

2.5

2.0

1.5

1.0 -(CH2)n

ppm R-CH3

CH-CH-CH2-CH-CH

18.1 18.2 20.4

Fβ :R-CH2-CH2-CO

4.0 3.5 CHO2OD

HOD

Fα :R-CH2-CH2-CO

4.5

5.0

CH=CHCH2 CH2CH2CO

GPC/PC

β–Gle

HOD

CH2CH2CO

-CH-CH=(CH2-CH-CH)3 (204. 22 6) -CH=CH-CH2-CH=CH(18:2)

Gly–C2

TMAQ betomene

CH–CH–CH2–CH3

CH2OCOR (Glyceryl)

(CH2)n

–CH–CH–

(b)

Part II

(a)

Chol-19

Chol-C18

Gly-C2(DG)

5.0

(c)

4.5

4.0

3.5

3.0

2.5

2.0

1.5

Cho/Pcho Ala

1.0

ppm

lactate

sugars and amino acids lle/Leu Val

Leu lactate

(d)

2400

HOD β-Gle

α-Gle

5.0

4.5

4.0

3.5

3.0

2.5

EDTA(E2)

EDTA(E1) HOD

2.0

1.5

Ca-EDTA2 (E2)

1.0

ppm

lactate

Ca-EDTA3(E1)

VLDL/LDL(CH2)3 Mg-EDTA1(E2)

β-Gle

TMA

VLDL/LDL-CH3

α-Gle

Paracetamol N-acetyl lactate pyruvate

5.0

4.5

4.0

3.5

3.0

2.5

val/les 3-HB

acetate

2.0

1.5

1.0

ppm

Fig. 3. (a) 600 MHz 1 H MAS NMR CPMG spin-echo spectrum of intact liver tissue; (b) 600 MHz standard 1 H NMR spectrum of a lipid-soluble liver tissue extract; (c) 600 MHz solvent suppressed 1 H NMR spectrum of an aqueous-soluble liver tissue extract; (d) 500 MHz 1 H NMR CPMG spectrum of blood plasma. All spectra are from animals treated with a toxic dose of paracetamol (500 mg/kg) and sacrificed at 240 min after dosing. Key: 3HB, 3-D-hydroxybutyrate; Cho, choline; Chol, cholesterol; Glu, glucose; GPC, glycerophosphorylcholine; Gly, glycerol; LDL, low-density lipoprotein; P Cho, phosphocholine; TMAO, trimethylamine-N oxide; VLDL, very low-density lipoprotein.

1384 Part II

Pharmaceutical Sciences

Part II

Conclusions Although there continues to be a need for advances in metabonomic analytical technologies both in NMR and MS, NMR is likely to remain the method of choice for a broad impartial survey of metabolic profiles, especially given recent gains in sensitivity through the use of cryoprobe detectors. MS coupled to a separation stage is always likely to yield better detection limits for specific classes of metabolite, but is inherently less general. NMR-based metabonomics is now recognized as an independent and widely used technique for evaluating the toxicity of drug candidate compounds, and it has been

adopted by a number of pharmaceutical companies into their drug development protocols. For drug safety studies, it is possible to identify the target organ of toxicity, derive the biochemical mechanism of the toxicity, and determine the combination of biochemical biomarkers for the onset, progression and regression of the lesion. Additionally, the technique has been shown to be able to provide a metabolic fingerprint of an organism (metabotyping) as an adjunct to functional genomics, and hence has applications in the design of drug clinical trials and for evaluation of genetically modified animals as disease models. The potential and real impact of metabonomics on all stages of pharmaceutical R&D is encapsulated in Figure 4.

NDA

IND

Compound libraries

Clinical trials Phase 1

HTS hits

Launch

Pharmacology & efficacy

Toxicity

Target characterisation

Efficacy optimization

Safety assessment and compound ranking

Clinical trials Phase 2

Clinical trials Phase 3

Patientselection (pharmacometabonomics) Doselevel setting in man Clinical safety Clinical efficacy (absolute) Clinical efficacy (relative to other drugs)

Clinical trials Phase 4

Patient monitoring Epidemiology Mechanisms of idiosyncratic toxicity

Fig. 4. The role of metabonomics in the various stages of pharmaceutical R&D. IND—Investigational new drug application— required before administration to man is permitted. NDA—new drug application—required before a new product launch.

Metabonomics using NMR

References 1. Nicholson JK, Lindon JC, Holmes E. Xenobiota. 1999;29: 1181. 2. Lindon JC, Nicholson JK, Holmes E, JR Everett. Concepts Magn. Reson. 2000;12:289. 3. Nicholson JK, Connelly J, Lindon JC Holmes E. Nat. Rev. Drug Discov. 2002;1:153. 4. Lindon JC, Holmes E Nicholson JK. Anal. Chem. 2003;75: 384A. 5. Plumb RS, Stumpf CL, Gorenstein MV, Castro-Perez JM, Dear GJ, Anthony M, Sweatman BC, Connor SC Haselden JN. Rapid Commun. Mass Spectrom. 2002;16:1991. 6. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall PJD, Nicholson JK Lindon JC. Anal. Commun. 1997;34:107. 7. Lindon JC, Nicholson JK Everett JR. Annu. Rep. NMR Spectrosc. 1999;38:1. 8. Khandelwal P, Beyer CE, Lin Q, McGonigle P, Schechter LE, Bach C. J. Neurosci. Methods. 2004;133:181. 9. Griffin JL, Nicholls AW, Keun HC, Mortishire-Smith RJ, Nicholson JK, Kuehn T. Analyst. 2002;127:582. 10. Keun HC, Beckonert O, Griffin JL, Richter C, Moskau D, Lindon JC, Nicholson JK. Anal. Chem. 2002;74:4588. 11. Lindon JC, Nicholson JK, Wilson ID. J. Chromatogr. B. 2000;748:233. 12. Garrod SL, Humpfer E, Spraul M, Connor SC, Polley S, Connelly J, Lindon JC, Nicholson JK, Holmes E. Magn. Reson. Med. 1999;41:1108. 13. Waters NJ, Holmes E, Williams A, Waterfield CJ, Farrant RD, Nicholson JK. Chem. Res. Toxicol. 2001;14:1401.

14. Coen M, Lenz EM, Nicholson JK, Wilson ID, Pognan F, Lindon JC. Chem. Res. Toxicol. 2003;16:295. 15. Lamers RJAN, Wessels ECHH, van der Sandt JJM, Venema K, Schaafsma G, van der Greef J, van Nesselrooij JHJ. J. Nutr. 2003;133:3080. 16. Bollard ME, Xu JS, Purcell W, Griffin JL, Quirk C, Holmes E Nicholson JK. Chem. Res. Toxicol. 2002;15:1351. 17. Lindon JC, Holmes E, Nicholson JK. Prog. NMR Spectrosc. 2001;39:1. 18. Brown TR, Stoyanova R. J. Magn. Reson. 1996;112:32. 19. Farrant RD, Lindon JC, Rahr E, Sweatman BC. J. Pharm. Biomed. Anal. 1992;10:141. 20. Fung M, Thornton A, Mybeck K, Wu JH-H, Hornbuckle K, Muniz E. Drug Inf. J. 2001;35:293. 21. Lindon JC, Nicholson JK, Holmes E, Antti H, Bollard ME, Keun H, Beckonert O, Ebbels TM, Reily MD, Robertson D, Stevens GJ, Luke P, Breau AP, Cantor GH, Bible RH, Niederhauser U, Senn H, Schlotterbeck G, Sidelmann UG, Laursen SM, Tymiak A, Car BD, Lehman-McKeeman L, Colet JM, Loukaci A, Thomas C. Toxicol. Appl. Pharmacol. 2003;187:137. 22. Keun HC, Ebbels TMD, Antti H, Bollard M, Beckonert O, Schlotterbeck G, Senn H, Niederhauser U, Holme E, Lindon JC, Nicholson JK. Chem. Res. Toxicol. 2002;15:1380. 23. Ebbels T, Keun H, Beckonert O, Antti H, Bollard M, Holmes E, Lindon J, Nicholson J. Anal. Chim. Acta. 2003;490:109. 24. Lindon JC, Holmes E, Bollard ME, Stanley EG, Nicholson JK. Biomarkers. 2004;9:1–31. 25. Griffin JL, Sang E, Evens T, Davies K, Clarke K. FEBS Lett. 2002;530:109. 26. Nicholls AW, Mortishire-Smith RJ, Nicholson JK. Chem. Res. Toxicol. 2003;16:1395. 27. Coen M, Nicholson JK, Lindon JC, Lenz EM, Wilson ID, Ruepp SU, Pognan F. J. Pharm. Biomed. Anal. 2004;35:93. 28. Maxwell RJ, Martinez-Perez I, Cerdan S, Cabanas ME, Arus C, Moreno A, Capdevila A, Ferrer E, Bartomeus F, Aparicio A, Conesa G, Roda JM, Carceller F, Pascual JM, Howells SL, Mazucco R, Griffiths JR. Magn. Reson. Med. 1998;39:869. 29. Tomlins A, Foxall PJD, Lindon JC, Lynch MJ, Spraul M, Everett JR, Nicholson JK. Anal. Commun. 1998;35:113. 30. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall PJD, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1998;17:125. 31. ‘t Hart BA, Vogels JTWE, Spijksma G, Brok HPM, Polman C, van der Greef J. J. Neuro. Sci. 2003;212:21. 32. Brindle JT, Antti H, Holmes E, Tranter G, Nicholson JK, Bethell HWL, Clarke S, Schofield PM, McKilligin E, Mosedale DE, Grainger DJ. Nat. Med. 2002;8:1439.

Part II

Using metabonomics, it has proved possible to derive new biochemically based assays for disease diagnosis and to identify combination biomarkers for disease, which can then be used to monitor the efficacy of drugs in clinical trials. Thus, based on differences observed in metabonomic databases from control animals and from animal models of disease, diagnostic methods and biomarker combinations might be derivable in a preclinical setting. Similarly, the use of databases to derive predictive expert systems for human disease diagnosis and the effects of therapy, require compilations from both normal human populations and patients before, during, and after therapy. In human studies, metabonomics also has the potential for disease diagnosis and possibly even prognosis.

References 1385

1387

Harald Schwalbe and Julia Wirmer Institute for Organic Chemistry and Chemical Biology, Center for Biomolecular Magnetic Resonance, Johann Wolfgang Goethe-Universit¨at, Marie-Curie-Strasse 11, D-60439 Frankfurt/M, Germany

Protein Misfolding Diseases The failure of proteins to fold into their functional forms can occasionally lead to “protein misfolding” or “protein conformational” diseases. Many among the most common and debilitating of these diseases are associated with the formation of protein amyloid, an insoluble material that is deposited as fibrils or plaques in different tissues and organs of the body. Amyloid formation is known to be accelerated by a variety of cellular factors, including metal ions, such as copper and zinc, and interactions with other species, such as lipids and RNA. It is implicated in many medical conditions including Alzheimer’s disease and the transmissible prion disorders. It is becoming increasingly recognized that the switch from a normal to a diseased state of the cell in protein misfolding diseases is induced by a shift in the equilibrium between different conformational and aggregation states of a polypeptide chain that are present under normal conditions. The native state [N] of a protein can be investigated in great detail using X-ray crystallography and NMR-spectroscopy. Our understanding of states other then the native one, however, is only now emerging. Liquid and solid-state high-resolution NMR spectroscopy are the major structural techniques to determine structure and dynamics of the polypeptide chain under a variety of different conditions. They describe proteins in their native state [N], in unfolded states [U], in transient intermediates [I], and in their fibrillar and prefibrillar states (Figure 1).

Natively Unfolded Proteins Involved in Protein Misfolding Diseases Non-native and unfolded states of proteins have also come into interest recently from the observation that an axiomatic linking of the function of a protein to a persistent fold might not be general, because a number of proteins have been identified that lack intrinsic globular structure in their normal functional form [1–3]. The expressions “intrinsically unstructured” and “natively unfolded” are being used synonymously, the latter being coined by Schweers et al. in 1994 in the context of structural Graham A. Webb (ed.), Modern Magnetic Resonance, 1387–1391.
C 2008 Springer.

studies of the protein tau [4]. Intrinsically unstructured proteins are extremely flexible, non-compact, and reveal little if any secondary structure under physiological conditions. In 2000, the list of natively unfolded protein comprised 100 entries [5]. Natively unfolded proteins are implied in the development of a number of neurodegenerative diseases including Alzheimer’s disease (deposition of amyloid-β, τ-protein, α-synuclein), Down’s syndrome and Parkinson disease to name a few [5]. They are predicted to be ubiquituous in the proteome [6,7] and algorithms available as a web-program (http://dis.embl.de/) have been developed to predict protein disorder [8]. According to the predictions, 35–51% of eucaryotic proteins have at least one long (>50 residues) disordered region and 11% of proteins in Swiss-Prot and between 6 and 17% of proteins encoded by various genomes are probably fully disordered [6]. Proteins predicted to be intrinsically unstructured show low compositional complexity. These regions sometimes correspond to repetitive structural units in fibrillar proteins. Therefore, it does not seem unlikely that lack of structure of the polypeptide chains in some states of a protein plays an important role in the development of fibrillar states and this further supports the importance for detailed structural and dynamic investigations of non-native states of proteins. It has also been noted by Gerstein that the average genomially encoded protein is significantly different in terms of size and amino acid composition from folded proteins in the PDB [9]. This difference would indicate that the structures deposited in the PDB are not random and in turn that they cannot be taken as representative for the entire structural diversity of polypeptide chains.

Brief Background in NMR Parameters NMR is able to provide both dynamic and structural information about proteins in a variety of different states at atomic resolution. NMR has the potential for probing residual structure, the size of aggregating molecules, and variation in the internal dynamical properties on the basis of diffusion-weighted NMR spectroscopy, heteronuclear relaxation measurements, paramagnetic enhancement of

Part II

Protein Misfolding Disease: Overview of Liquid and Solid-State High Resolution NMR Studies

1388 Part II

Pharmaceutical Sciences

Part II Fig. 1. The characteristics of states accessible to a protein differ widely in their structural and kinetic properties. The different conformations adopted by a polypeptide chain can range from the native, often monomeric state (c), in which a single conformation exists and which is built up from secondary structure elements and their specific arrangements, to the ensemble of conformers representing the random coil state of a protein (a). The individual members of this ensemble have widely different compaction, dynamics, local, and non-local conformations. Protein folding preceeds via formation of folding intermediates (b) whose structure and dynamics may be modulated by protein–protein interactions with molecular chaperones such as GroEL (scaled down by a factor of 2) (e) shown in the figure. At the other extreme of conformational states, proteins can aggregate and form oligomeric states called fibrils (d). Liquid and solid-state NMR spectroscopy can provide detailed information on structure and dynamics in all of these states.

relaxation induced by paramagnetic spin labels, and residual dipolar couplings (summarized in Table 1). More recently, NMR techniques have been developed to characterize the structural transitions from unfolded states of protein, via low molecular oligomers to fibrils. Such studies include both static and time-resolved experiments.

Proteins Involved in Misfolding Diseases Studied by NMR Two proteins intimately linked to neurodegenerative illnesses seem to have a possible function in copper

homeostasis: the amyloid precursor protein (APP) of Alzheimer’s disease, and the prion protein (PrP), which causes transmissible spongiform encephalopathies (TSE). APP gives rise to the amyloid Aβ peptide and the prion protein PrPc can convert to the variant isoform PrPSc of Creutzfeldt-Jakob’s disease (CJD). The protein Cu,Zn dismutase (SOD) is a metalloprotein for which familial mutants are linked to familial amyotrophic lateral sclerosis (FALS). The three proteins are known to interact: The prion protein specifically binds copper and native PrP charged with copper has superoxide dismutase (SOD1) activity. In turn, overexpression of APP alters brain copper content and attenuates SOD1. On the other hand,

Table 1: Overview of NMR parameters and their conformational dependence NMR parameter

Conformational dependence

Chemical shift δ (ppm) Scalar couplings n J (Hz) Homonuclear NOEs (a.u.)

Multiple torsion angles: φ, ψ, ω, χ 1 Single torsion angles via Karplus equations Distances, dependence on correlation time, and motional properties Motional properties, dependence on τc , S 2 , τe a Overall shape, dynamics, Sb Exchangeable HN Radius of hydration (Rh ) Accessible Trp, Tyr, His

Heteronuclear relaxation (Hz) Residual dipolar couplings RDC (Hz) H/D exchange Diffusion Photo CIDNP aτ c

correlation time for overall rotational tumbling, S 2 order parameter of local dynamics, τe correlation time for local dynamics b S order parameter for local dynamics

Protein Misfolding Disease

Amyloid Precursor Protein Cleavage of APP by β- and γ-secretases results in amyloid forming peptides containing 39–43 residues referred to as the Amyloid β-peptide (Aβ). The Aβ is the major constituent of the plaques in the brains of Alzheimer patients. The presence of copper inhibits Aβ production and stimulates the non-amyloidogenic pathway of APP in cells and in transgenic animals overexpressing APP. The soluble and the fibrillar form of various peptides of Aβ have been studied using liquid and solid-state NMR spectroscopy: conformations in the soluble form range from α-helical conformations via random coil to β-sheet conformations, depending on solvent condititions, pH values and peptide concentration. In membrane mimicking environments such as SDS, the N-terminus is unstructured (residues 1–14) while α-helical conformation is found from residue 15–36 with a kink from residues 25–27 [10– 13]. A high propensity for aggregation is found in solution and the conformation in solution is often found to be in an equilibrium between random coil conformation [14] and β-sheet conformers [15] with increased β-sheet propensity at high concentration, high temperature and high salt conditions. However, at very low concentrations and at 0 ◦ C, a conformational equilibrium between a left handed 3(10) helix and random coil conformations is found [16]. Solid-state NMR studies revealed the structure of the amyloid fibrils formed by Aβ (1−40): approximately the first 10 residues of the peptide are disordered in the fibril, residues 12–24 and 30–40 adopt β-strand conformations and form parallel β-sheets through intermolecular hydrogen bonding. The two β-sheets within a single peptide unit form an antiparallel sheet [17]. The structural model found for the Aβ(1–40) is very similar to structural models of smaller (11–25, 16–22) and larger (1–42) peptides [18,19].

Prion Protein Prion protein (PrPc ) is a glycosyl-phosphatidyl-inositol (GPI) anchored membrane protein of 35 kD. Its conformational altered scrapie form (PrPSc ) is associated with neurodegenerative diseases. Cellular PrPc is synthesized in three topologic forms, secreted sec PrP and the single-spanning membrane proteins of opposite orientations Ntm PrP, and Ctm PrP. Increases in Ctm PrP are associated with the development of neurodegenerative disease. The conformational properties of a protein are generally thought to be determined by specific elements in the sequence; for the prion protein these are the N-terminal

sequence, a hydrophobic stretch around residue 100, a putative transmembrane domain, and the C-terminus for GPI anchor addition. The signal sequence plays a decisive role in determining the membrane orientation of the PrP. Ctm PrP contains an uncleaved N-terminal signal peptide. NMR solution structures have been reported for PrPc from several species [20–31]. Interestingly, it was shown that only few of the disease related mutations lead to reduced stability of N [26]. Residues 23–121 comprise a flexibly disordered tail in contrast to the well-ordered 125–228 core part. Cu2+ binding to histidines located in the flexible part of the protein (high affinity octarepeat regions are found between residues 51–90 and low affinity binding sites are formed by His96 and His111) induces a shift towards β-sheet conformation as shown by liquid-state NMR [32–36]. In particular, the conformational shift induced by binding of His96 and His111 is remarkable, as these residues are part of the infectious region (residues 90–231) [24]. β-sheet conformation has also been observed by liquid and solid-state NMR studies on fibrillar states: a hydrogen exchange study of the fibrillar form of the 106–121 peptide revealed 50% β-sheet structure, which is located in the center portion of the peptide [37]. Complete β-sheet structure was found by solid-state NMR investigation on the fibrillar state of a peptide corresponding to the infectious P101L mutant of residues 89–143 [38].

α-Synuclein α-synuclein is associated with Parkinson’s disease, a neurodegenerative disorder affecting some five million people worldwide. This disease is characterized by the formation of proteinaceous inclusions called Lewy bodies that are known to be composed primarily of aggregated αsynuclein. There is mounting evidence that the oligomeric species that are formed during the process of aggregation of the monomeric α-synuclein are at the origin of the pathology. In solution, α-synuclein can be either found in a monomeric, substantially unfolded conformation, or it can form several morphologically different types of aggregates, including oligomers, amorphous aggregates, and amyloid-like fibrils. Its normal function is poorly understood, but it is believed that it includes transient or reversible lipid interactions [39]. NMR studies of the lipid bound form of α-synuclein revealed that the C-terminus of the protein remains free in solution, while the N-terminus of the protein binds to the lipids in a helical conformation [40–43]. This helical conformation in the lipid bound state remains nearly unchanged in the early onset mutations (A30P, A53T) [43]. More interesting results were found by NMR studies on the natively unfolded monomeric form free in solution: the protein is more compact than expected for a highly denatured protein [44] and residual helical

Part II

for the other systems discussed here, α-synuclein and transthyretin, no metal induced misfolding is currently being discussed in the literature.

α-Synuclein 1389

1390 Part II

Pharmaceutical Sciences

Part II

structure is present in the N-terminal part of α-synuclein (residues 18–31) [40,45,46]. This residual helical structure is diminished in one of the early onset mutants A30P, while enhanced β-sheet propensity around the mutation site in the other early onset mutant A53T is observed [45]. Changes in these parts of the sequence (residues 22–93) were also observed upon binding of aggregation promoting polyamines at the C-terminal part (residues 109–140) [47]. Taking all NMR investigations on the monomeric form in solution together, clearly the N-terminal part of the protein is the aggregation nucleus.

Cu-Zn-Superoxide Dismutase Mutations in the gene for cytosolic SOD are linked with familial FALS. The active site of each monomer of SOD (153 residues) contains one copper and one zinc ion. The enzyme catalyzes the dismutation of superoxide to dioxygen and hydrogen peroxide. FALS-related mutations are concentrated either in the metal-binding loop or in loops III (38–40) and V (90–93). In this protein, metal ions have a fundamental role for protein stabilization, correct folding, and in increasing protein rigidity. Binding of the metal ions is of fundamental importance to stabilize its secondary structure and to define the tertiary interactions relevant for the relative orientation of the two sheets forming as can be shown by a comparison of the NMR strucutures in the presence and the absence of metal ions [48,49]. A comparison of WT SOD and the FALS-related mutant G93A based on NMR dynamics revealed that indeed a in the β-barrel loops III and V are a lot more dynamic and thus less stable in the mutant than in the wild type proteins [50].

Transthyretin Transthyretin (TTR) is a homotetrameric protein that is involved in the transport of thyroid hormones and retinol in human serum. A large number of mutations (more than 80 different ones) have been found that lead to misfolded forms of the protein. These misfolded forms are implicated in amyloid diseases such as familial amyloidotic polyneuropathy and senile systemic amyloidosis. A combination of hydrogen exchange experiments coupled with NMR detection have compared WT TTR under native conditions and under conditions that are known to promote amyloid formation: These studies revealed that amyloidogenic behavior is linked to destabilization of one half of the β-sandwich structure of TTR [51]. These findings were further confirmed by additional NMR studies of some of the amyloidogenic mutants [52,53]. Liquid-state NMR studies of amyloidogenic peptides (TTR10–20 and TTR105–115) showed that the peptides are unstructured in solution [54,55]. In contrast, the TTR105–115 peptide

in its fibrillar state is in an extended β-sheet conformation, with its backbone and side chain torsion angles close to their optimal values for this secondary structure element. In addition, long-range order could be detected that is generally associated with crystalline materials by solidstate NMR [56,57].

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Dyson HJ, Wright PE. Curr. Opin. Struct. Biol. 2002;12:54. Wright PE. Dyson HJ. J. Mol. Biol. 1999;293:321. Uversky VN. Eur. J. Biochem. 2002;269:2. Schweers O, Schonbrunn-Hanebeck E, Marx A, Mandelkow E. J. Biol. Chem. 1994;269:24290. Uversky VN, Gillespie JR, Fink AL. Proteins. 2000;41: 415. Tompa P. Trends Biochem. Sci. 2002;27:527. Uversky VN. Protein Sci. 2002;11:739. Linding R, Jensen LJ, Diella F, Bork P, Gibson TJ, Russell RB. Structure (Camb.) 2003;11:1453. Gerstein M. Fold Des. 1998;3:497. Coles M, Bicknell W, Watson AA, Fairlie DP, Craik DJ. Biochemistry 1998;37:11064. Shao H, Jao S, Ma K, Zagorski MG. J. Mol. Biol. 1999;285: 755. D’Ursi AM, Armenante MR, Guerrini R, Salvadori S, Sorrentino G, Picone D. J. Med. Chem. 2004;47:4231. Crescenzi O, Tomaselli S, Guerrini R, Salvadori S, D’Ursi AM, Temussi PA, Picone D. Eur. J. Biochem. 2002;269: 5642. Riek R, Guntert P, Dobeli H, Wipf B, Wuthrich K. Eur. J. Biochem. 2001;268:5930. Jarvet J, Damberg P, Bodell K, Eriksson LEG, Graslund A. J. Amer. Chem. Soc. 2000;122:4261. Jarvet J, Damberg P, Danielsson J, Johansson I, Eriksson LE, Graslund A. FEBS Lett. 2003;555:371. Petkova AT, Ishii Y, Balbach JJ, Antzutkin ON, Leapman RD, Delaglio F, Tycko R. Proc. Natl. Acad. Sci. U.S.A. 2002; 99:16742. Petkova AT, Buntkowsky G, Dyda F, Leapman RD, Yau WM, Tycko R. J. Mol. Biol. 2004;335:247. Antzutkin ON, Leapman RD, Balbach JJ, Tycko R. Biochemistry. 2002;41:15436. Riek R, Hornemann S, Wider G, Billeter M, Glockshuber R, Wuthrich K. Nature, 1996;382:180. Glockshuber R, Hornemann S, Riek R, Wider G, Billeter M, Wuthrich K. Trends Biochem. Sci. 1997;22:241. Billeter M, Riek R, Wider G, Hornemann S, Glockshuber R, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 1997;94:7281. Riek R, Hornemann S, Wider G, Glockshuber R, Wuthrich K. FEBS Lett. 1997;413:282. James TL, Liu H, Ulyanov NB, Farr-Jones S, Zhang H, Donne DG, Kaneko K, Groth D, Mehlhorn I, Prusiner SB, Cohen FE. Proc. Natl. Acad. Sci. U.S.A. 1997;94:10086. Donne DG, Viles JH, Groth D, Mehlhorn I, James TL, Cohen FE, Prusiner SB, Wright PE, Dyson HJ. Proc. Natl. Acad. Sci. U.S.A. 1997;94:13452. Riek R, Wider G, Billeter M, Hornemann S, Glockshuber R, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 1998;95:11667.

Protein Misfolding Disease

41. Bussell R Jr, Eliezer D. J. Mol. Biol. 2003;329:763. 42. Chandra S, Chen X, Rizo J, Jahn R, Sudhof TC. J. Biol. Chem. 2003;278:15313. 43. Bussell R Jr, Eliezer D. Biochemistry. 2004;43:4810. 44. Morar AS, Olteanu A, Young GB, Pielak GJ. Protein Sci. 2001;10:2195. 45. Bussell R Jr, Eliezer D. J. Biol. Chem. 2001;276:45996. 46. Yao J, Chung J, Eliezer D, Wright PE, Dyson HJ. Biochemistry. 2001;40:3561. 47. Fernandez CO, Hoyer W, Zweckstetter M, Jares-Erijman EA, Subramaniam V, Griesinger C, Jovin TM. Embo. J. 2004; 23:2039. 48. Assfalg M, Banci L, Bertini I, Turano P, Vasos PR. J. Mol. Biol. 2003;330:145. 49. Banci L, Bertini I, Cramaro F, Del Conte R, Viezzoli MS. Biochemistry. 2003;42:9543. 50. Shipp EL, Cantini F, Bertini I, Valentine JS, Banci L. Biochemistry. 2003;42:1890. 51. Liu K, Cho HS, Lashuel HA, Kelly JW, Wemmer DE. Nat. Struct. Biol. 2000;7:754. 52. Liu K, Kelly JW, Wemmer DE. J. Mol. Biol. 2002;320: 821. 53. Niraula TN, Haraoka K, Ando Y, Li H, Yamada H, Akasaka K. J. Mol. Biol. 2002;320:333. 54. Jarvis JA, Craik DJ. J. Magn. Reson. B. 1995;107:95. 55. Jarvis JA, Kirkpatrick A, Craik DJ. Int. J. Pept. Protein. Res. 1994;44:388. 56. Jaroniec CP, MacPhee CE, Astrof NS, Dobson CM, Griffin RG. Proc. Natl. Acad. Sci. U.S.A. 2002;99:16748. 57. Jaroniec CP, MacPhee CE, Bajaj VS, McMahon MT, Dobson CM, Griffin RG. Proc. Natl. Acad. Sci. U.S.A. 2004;101: 711.

Part II

27. Liu H, Farr-Jones S, Ulyanov NB, Llinas M, Marqusee S, Groth D, Cohen FE, Prusiner SB, James TL. Biochemistry. 1999;38:5362. 28. Zahn R, Liu A, Luhrs T, Riek R, von Schroetter C, Lopez Garcia F, Billeter M, Calzolai L, Wider G, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 2000;97:145. 29. Lopez Garcia F, Zahn R, Riek R, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 2000;97:8334. 30. Calzolai L, Lysek DA, Guntert P, von Schroetter C, Riek R, Zahn R, Wuthrich K. Proc. Natl. Acad. Sci. U.S.A. 2000;97: 8340. 31. Zahn R, Guntert P, von Schroetter C, Wuthrich K. J. Mol. Biol. 2003;326:225. 32. Stockel J, Safar J, Wallace AC, Cohen FE, Prusiner SB. Biochemistry. 1998;37:7185. 33. Viles JH, Cohen FE, Prusiner SB, Goodin DB, Wright PE, Dyson HJ. Proc. Natl. Acad. Sci. U.S.A. 1999;96:2042. 34. Jackson GS, Murray I, Hosszu LL, Gibbs N, Waltho JP, Clarke AR, Collinge J. Proc. Natl. Acad. Sci. U.S.A. 2001;98:8531. 35. Jones CE, Abdelraheim SR, Brown DR, Viles JH. J. Biol. Chem. 2004;279:32018. 36. Belosi B, Gaggelli E, Guerrini R, Kozlowski H, Luczkowski M, Mancini FM, Remelli M, Valensin D, Valensin G. Chembiochem. 2004;5:349. 37. Kuwata K, Matumoto T, Cheng H, Nagayama K, James TL, Roder H. Proc. Natl. Acad. Sci. U.S.A. 2003;100:14790. 38. Laws DD, Bitter HM, Liu K, Ball HL, Kaneko K, Wille H, Cohen FE, Prusiner SB, Pines A, Wemmer DE. Proc. Natl. Acad. Sci. U.S.A. 2001;98:11686. 39. Clayton DF, George JM. Trends Neurosci. 1998;21:249. 40. Eliezer D, Kutluay E, Bussell R Jr, Browne G. J. Mol. Biol. 2001;307:1061.

References 1391

1393

NMR Spectroscopy for Functional and Binding High-Throughput Screening Marina Veronesi and Claudio Dalvit Chemistry Department, Nerviano Medical Sciences, 20014 Nerviano, Milan, Italy

NMR screening has emerged as a powerful and reliable approach for identification of potential drug candidates [1]. The technique is now recognized for its impact on the drug discovery process and has become an important tool for lead identification, lead validation, and lead optimization in many pharmaceutical companies and universities [2–17]. A plethora of different NMR experiments has been proposed in the literature for performing these tasks. Two of these approaches, recently introduced, use fluorine NMR spectroscopy. FAXS (fluorine chemical shift anisotropy and exchange for screening) [18,19] and 3-FABS (three fluorine atoms for biochemical screening) [20] allow one to perform binding and functional highthroughput screening (HTS), respectively, and determine the dissociation binding constant (K D ) and the 50% mean inhibition concentration (IC50 ) of the identified binders and inhibitors, respectively. This chapter provides an insight into the theory and practical aspects of these two experiments and presents applications to the screening of different biomolecular targets.

FAXS Competition ligand-based NMR screening experiments were introduced for overcoming all the limitations associated with ligand-based NMR screening experiments [21–23]. The screening of chemical mixtures or single compounds against the biomolecular target of interest is performed in the presence of a weak- to medium-affinity ligand of known binding constant referred to in its role as the spy or reporter. Changes in the transverse or selective longitudinal relaxation rates of the spy resonances are monitored. Signals from the molecules screened are not utilized. Often the screening is performed in the presence of an additional molecule that does not interact with the receptor and referred to in its role as the control molecule. This compound represents an internal reference. Competition NMR-based screening, originally proposed with proton detection experiments, was subsequently extended to fluorine detection experiments [18,19]. For these experiments it is sufficient to have a spy molecule containing either a CF or a CF3 moiety. This approach ofGraham A. Webb (ed.), Modern Magnetic Resonance, 1393–1399.
C 2008 Springer.

fers some unique advantages. (i) Absence of overlap permits the screening of large chemical mixtures and automated analysis of the spectra. (ii) Protonated solvents, buffers, or detergent do not interfere with the measurements thus allowing, for example, the screening against membrane proteins. (iii) The 19 F transverse relaxation rate R2 , given by Equation (1), is a sensitive parameter of binding events since it contains spectral densities calculated at zero frequency for both the heteronuclear 19 F–1 H dipolar interactions and the 19 F chemical shift anisotropy (CSA) interaction [24]: R2F =

 γF2 γH2h¯ 2 τc  1 1 4+ 6 20 1 + (ω − ωH )2 τc2 r F FH Hi i 3 6 + 2 2 1 + ω F τc 1 + ωH2 τc2  6 + 1 + (ωF + ωH )2 τc2   2 ηCSA 2 2 + σ 1 + B02 γF2 τc 15 3   1 2 + × 3 2(1 + ωF2 τc2 ) +

(1)

The Hi correspond to all the protons of the spy compound and of the protein close in space to the fluorine atom and rFHi is the internuclear distance between proton Hi and the fluorine atom of the spy molecule. σ is the CSA of the 19 F atom and is given by σ = σzz − (σx x + σ yy )/2 where the different σs are the components of the chemical shift tensor. The asymmetry parameter ηCSA is given by ηCSA = (3/2)(σx x − σ yy )/σ, B0 is the strength of the magnetic field, γ H and γ F are the proton and fluorine gyromagnetic ratios, respectively, ωH and ωF are the proton and fluorine Larmor frequencies, respectively, and τ c is the correlation time. Owing to the large CSA of 19 F (as much as few hundreds ppm) it will contribute significantly, according to Equation (1), to the transverse relaxation of the fraction of bound spy molecule [18,25–28]. CSA contribution to

Part II

19 F

1394 Part II

Pharmaceutical Sciences

Part II

Fig. 1. One-dimensional 19 F NMR spinecho (n2τ = 100 ms) spectra recorded for the identification of a spy and control molecule required for FAXS. The two molecules are the weak-affinity ligand PHA-739917 and the non-interacting trifluoroacetic acid (TFA) molecule and the protein is the p21 activated kinase. The chemical shifts are referenced to TFA. The concentration of PHA-739917 and TFA was 50 and 15 µM, respectively. The spectra were recorded in the absence (top) and the presence of 1.5 µM of the protein (middle). In the difference spectrum (bottom) only the signal of the spy molecule is visible. Reprinted with permission from Dalvit c 2002 Bentham Science Pubet al. [18].
lishers.

Control

Spy

w/o Protein

w Protein

Difference

15 14 13 12 11 10

9

8

7

6

5

4

3

2

1

ppm

δ 19F transverse relaxation is directly proportional to the square of the magnetic field thus resulting in a more pronounced effect at stronger magnetic fields. For a weak-affinity spy molecule the observed transverse relaxation R2,obs is given by [29,30]:

R2,obs

  [EL] [EL] = R2,bound + 1 − R2,free [LTOT ] [LTOT ]   [EL] [EL] 2 + 1− [LTOT ] [LTOT ] ×

4π 2 (δfree − δbound )2 K −1

(2)

where [EL]/[LTOT ] and (1 − [EL]/[LTOT ]) are the fraction of bound and free ligand, respectively, R2,bound and R2,free are the transverse relaxation rate constants for the ligand in the bound and free states, respectively. The last term in Equation (2) is the exchange term where δ bound and δ free are the isotropic chemical shifts of the fluorine resonance of the spy molecule in the bound and free states, respectively, and 1/K −1 is the residence time τ res of the spy molecule bound to the protein. Typically, a Carr–Purcell–Meibom–Gill (CPMG) spin-echo scheme [31,32] with a long 2τ interval between the train of 180◦ pulses (where 2τ > 5τres ) [33,34] is used in these experiments before the acquisition time. This is possible because the evolution with the heteronuclear 1 H– 19 F scalar couplings is refocused at the end of the scheme. However, the 2τ period should not be very long in order to

minimize signal attenuation originating from the spatial diffusion of the spy molecule. The steps required for the screening with FAXS are described below: 1. A library of 19 F (CF or CF3 ) containing molecules well characterized, chemically stable, and with high aqueous solubility is tested in mixtures against the receptor of interest for the identification of the potential spy and control molecules. STD [35], WaterLOGSY [36], or spin-echo 19 F experiments [18,19,37], as the example of Figure 1, are used for the identification of the two molecules. 2. The K D of the identified spy molecules are determined with either ITC or fluorescence spectroscopy. The spy molecule is then selected on the basis of its K D and the presence of only one binding site. 3. After the selection of the spy and control molecules, titration experiments as a function of the protein concentration are recorded and the intensity ratio of the two fluorine signals is plotted as a function of the fraction of protein-bound spy molecule [21], as shown in the example of Figure 2. The fraction of bound compound is calculated by using its K D value and the equation: [EL] [ETOT ] + [LTOT ] + K D = [LTOT ] 2[LTOT ]  ([ETOT ] + [LTOT ] + K D )2 − 4[ETOT ][LTOT ] − 2[LTOT ] (3)

19 F

NMR Spectroscopy

FAXS 1395

Part II

HSA (nM) 0

150

300

450

600

0.007 0.006

N N N

S

NH

0.005 [EL]/[LTOT]

F3C

(C) 15.6

ppm

O OH OH

0.004 0.003 0.002 0.001

F

0.000

(S) -64.0 ppm

0.3

δ 19F

0.4

0.5

0.6 0.7 I (s) / I (c)

0.8

0.9

Fig. 2. (Left) One-dimensional 19 F NMR spin-echo spectra (n2τ = 80 ms) recorded for the spy molecule 2-hydroxy 3-fluorobenzoic acid (s) and the control molecule 1-[5-(trifluoromethyl)1,3,4-thiadiazol-2-yl]piperazine(c) as a function of HSA concentration. The concentration of (s) and (c) was 50 and 25 µM, respectively. (Right) Plot of the signal intensity ratio (x-axis) of the two 19 F signals of (s) and (c) as a function of the fraction of bound spy molecule ([EL]/[LTOT ]) (y-axis). The last point on the right corresponds to the value in the absence of the protein. Two ratios ([EL]/[LTOT ]) were calculated using the limits of the ITC-derived K D value of 41 ± 3.3 µM for (s) and using Equation (3). Values indicated by open circles were calculated with a K D of 44.3 µM, values indicated by filled circles were calculated with a K D of 37.7 µM. The curves represent the best fits of the experimental points. Reprinted with c 2003 American Chemical Society. permission from Dalvit et al. [19].


where [ETOT ] and [LTOT ] are the total concentration of enzyme and ligand, respectively, and [EL] is the concentration of bound ligand. 4. The experimental conditions for the FAXS are then selected according to the graph of Figure 2. Compounds are tested in mixtures or as a single compound in the presence of the two molecules. Displacement of the spy molecule results in a shift of the R2 parameter toward the intrinsic of the free state. When mixtures are used, deconvolution is then performed for the identification of the active compound. The screening with FAXS against human serum albumin (HSA) is shown in Figure 3. For screening, a total spin-echo period (2nτ ) is selected for which the signal of the spy molecule is approaching zero. The presence in the mixture of 5-CH3 D, L Trp, and sucrose, known as non-binders, do not alter the spectrum of the spy molecule. In contrast, the presence in the mixture of the warfarin derivative 4-hydroxy-3-[1-( p-iodophenyl)-3oxobutyl] coumarin (PNU-24009) results in the reappearance of the signal of the spy molecule thus identifying the NMR-hit. 5. The extent of displacement of the spy molecule is then used to calculate the binding constant K I of the NMRhit. Since [LTOT ] in the NMR screening experiments

is known and fixed, the concentration of [EL] in the presence of the competing molecule can be calculated from the titration curve of Figure 2. The knowledge of [LTOT ], [EL], and [ETOT ] permits determination of the app apparent dissociation binding constant K D of the spy molecule in the presence of the competing molecule according to the equation: app

KD =

[ETOT ][LTOT ] − [ETOT ][EL] + [EL]2 − [LTOT ][EL] [EL] (4)

In the assumption of a simple competitive mechanism, the app K D is then used to extract the binding constant K I of the NMR-hit according to the equation: KI =

[I]K D app KD − KD

(5)

where [I] is the concentration of the NMR-hit. The NMRderived K I values obtained with a single experimental point compare favorably with the values derived from full titration fluorescence or ITC measurements [18,21,38]. A major advantage of NMR applied to these measurements is the direct determination of the concentration of the NMR-hits thus providing reliable K I values.

1396 Part II

Pharmaceutical Sciences

Part II

w HSA w/o HSA

N N F3C

N

S

NH

15.6

ppm

+ sucrose + sucrose + 5-CH3 W

+ 5-CH3 W + PNU-24009 O

O

O CH2

OH OH

OH

F

O I

-64.0

ppm

δ 19F

Fig. 3. FAXS performed with (c) (top) and (s) (bottom). The spectra were recorded with a total spin-echo period of 160 ms with an interval between the 180◦ pulses (2τ ) of 40 ms. The concentration of (s) and (c) was 50 and 25 µM, respectively. The spectra on the left were recorded in the absence of protein while all the other spectra were recorded in the presence of 600 nM HSA. Reprinted c 2003 American Chemical Society. with permission from Dalvit et al. [19].


3-FABS NMR has been extensively used for characterizing the product or products of an enzymatic reaction and for gaining insight into the kinetics of the reaction (see, for example, Refs. [39,40]). A high substrate concentration was necessary for these studies due to the low sensitivity of the NMR technique. The required high concentration represents a major hurdle for the utilization of this approach to functional screening purposes, since only the very strong inhibitors would be detected. The goal in a primary screening is the identification also of weak- and mediumaffinity inhibitors derived from a diversity of chemical classes. A way to overcome the sensitivity limitation is to tag the substrate with a CF3 moiety and use 19 F NMR (with proton decoupling) as the method of detection. The principle of this approach named 3-FABS [20] is described in Figure 4. The high receptivity of 19 F NMR spectroscopy, the 100% natural abundance of the isotope 19 F and the

presence of three fluorine atoms result in 19 F NMR signals of high intensity. Modification of the substrate through the enzymatic reaction results in changes of the fluorine chemical shift tensor components even when the CF3 moiety is distant from the reaction site. Therefore, distinct 19 F signals for the substrate and product (or products) are observed. The process involved with 3-FABS requires the following steps [20]: 1. Determination of the linear range (first-order region) of the enzymatic reaction. This is achieved by monitoring the course of the reaction within the NMR tube. 19 F spectra at different intervals are recorded and the 19 F signal integral of the substrate or product (products) is plotted as a function of the time elapsed from the beginning of the reaction. Screening and K M measurements are then performed in an end point format. The reaction is quenched with either a strong inhibitor, a denaturating or a chelating molecule after a delay for which the reaction is still linear and sufficient product is formed.

19 F

3-FABS 1397

SUBSTRATE

Enzymatic Reaction

PRODUCTS

δ

CF3 19F

2. Determination of the K M of the substrate and cosubstrate. This is achieved by performing the reaction at different substrate or cosubstrate concentrations. The K M is measured using the following equation [41]: V =

[S] Vmax KM + [S]

(6)

where V is the initial speed of the reaction and Vmax is the maximum speed that would be observed if all the enzyme [ETOT ] is complexed with the substrate (Vmax = K cat *[ETOT ], where K cat is the catalytic rate constant) and [S] is the substrate concentration. V is obtained experimentally by measuring the integral of the 19 F NMR signal of the product divided by the incubation time of the reaction. A plot of these values as a function of [S] permits the determination of K M and Vmax . 3. Screening, as shown in Figure 5, is performed in an end point format using either chemical mixtures or single compound at a substrate concentration in the range 2–5 K M . In every screening run a sample without test molecules is recorded. This sample represents the reference with 0% inhibition. Deconvolution of the active mixtures is then performed for the identification of the inhibitors.

4. The IC50 value of the detected inhibitor is obtained by simply recording experiments at different inhibitor concentration and by monitoring the integral of the product or substrate 19 F NMR signal as shown in Figure 5. A plot of these values as a function of the inhibitor concentration allows the determination of IC50 according to the equations: a) monitoring the substrate signal [Sw ] =

[Sw/o ] − [STOT ] + [STOT ] 1 + ([I]/IC50 )n

(7)

b) monitoring the product signal [Pw ] =

[Pw/o ] 1 + ([I]/IC50 )n

(8)

where [Sw ] and [Sw/o ] are given by the integrals of the substrate signal in the presence and absence of the inhibitor, respectively, [Pw ] and [Pw/o ] are given by the integrals of the product (or products) signal in the presence and absence of the inhibitor, respectively, and [STOT ] = [Pw/o ] + [Sw/o ] = [Pw ] + [Sw ]. [STOT ] is the substrate concentration used for the experiments and it represents an internal reference, [I] is the concentration of the inhibitor, IC50 the concentration of the inhibitor

Part II

Fig. 4. Schematic diagram of the 3FABS method. The enzyme in the example is a protease that cleaves the peptide bond with the formation of two products.

3-FABS (Three Fluorine Atoms for Biochemical Screening)

CF3

NMR Spectroscopy

Pharmaceutical Sciences

Part II

Fig. 5. Screening and deconvolution (top) and IC50 measurement (bottom) for compound H89 [43] performed with 3-FABS. The substrate is the N-terminal trifluoracetylated AKTide [44], the enzyme is the Ser/Thr kinase AKT1. The five compounds are (2-amino-6 methylquinazolin-4 ol, ethyl 2-quinoxalinecarboxylate, 5-methylbenzimidazole, methyl isoquinoline-3-carboxylate, and N-(2-{[(2E)-3-(4-bromophenyl)prop-2-enyl]amino}ethyl, known as H89). The activated enzyme AKT1, peptide, ATP, and compound concentrations were 25 nM, 30 µM, 131 µM (∼2 K M ), and 10 µM, respectively. S and P represent the CF3 signal of the peptide and the phosphorylated peptide, respectively. IC50 was measured with Equation (7), but the same value was obtained with Equation (8). The asterisks indicate the tiny amount of phosphorylated peptide in the presence of H89. Reprinted with permission c 2003 American from Dalvit et al. [20].
Chemical Society.

CF3-CO-ARKRERAYSFGHHA + 4 cpds + H89

P

+ 4 cpds

+ H89

S

* 0.120

* δ 19F

0.105 ppm

30 25

[Sw] in µM

1398 Part II

Br

20

IC50 = 0.72 ± 0.05 µM

15 H89

10

HN N NH SO2

5 0 0.1

1

10

[H89] in µM

at which 50% inhibition is observed and n is the cooperativity factor (known also as Hill slope). In the absence of allosteric effects (i.e. n = 1) a meaningful value for IC50 can be derived with a single experimental point. This is possible because the values for both plateaus are known. These are [Sw/o ] and [STOT ] if Equation (7) is used and [Pw/o ] and the 0 value if Equation (8) is used. Typically, enzyme concentration in the low nanomolar is used, but if the reaction is fast (substrates with large K cat /K M ) and the enzyme is stable at room temperature for few hours the concentration can be further reduced [42]. This can be appreciated in the example of Figure 6 where the screening against the protease trypsin is performed with only 50 pM enzyme concentration. With cryoprobe technology optimized to fluorine NMR detection it will be possible, in some fortunate cases, to perform 3-FABS at enzyme concentration in the femtomolar range.

Conclusion The fluorine NMR approaches FAXS and 3-FABS represent powerful tools for performing HTS and determining the K D and IC50 values of the identified hits. The low protein consumption required by these methodologies

P' P''

S reference S

+ leupeptin 13.48 13.46 13.44 13.42 13.40 13.38 13.36 13.34 13.32 13.30 13.28 13.26 13.24

ppm

δ 19F ARKRERAF(3-CF3)SFGHHA

Fig. 6. 3-FABS performed with the protease trypsin (Roche Molecular Biochemical Cat. No. 1418475) at an enzyme concentration of only 50 pM. The substrate concentration was 30 µM. The reaction was performed in 50 mM Tris pH 7.5 and 0.03% Triton X-100 (Sigma X-100) at 20 ◦ C and quenched after 6:40 hours with 0.5 mM phenylmethylsulfonyl fluoride (PMSF). S is the original peptide and P , P are two different products of the enzymic hydrolysis. In the presence of 20 µM leupeptin no product formation is observed. Reprinted with permission from c 2004 Elsevier Science Ltd. Dalvit et al. [42].


19 F

References 1. Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Science. 1996;274:1531. 2. Moore JM. Biopolymers (Peptide Science). 1999;51:221. 3. Hajduk PJ, Meadows RP, Fesik SW. Q. Rev. Biophys. 1999;32:211. 4. Roberts GCK. Drug Discov. Today. 2000;5:230. 5. Ross A, Senn H. Drug Discov. Today. 2001;11:583. 6. Peng JW, Lepre CA, Fejzo J, Abdul-Manan N, Moore JM. Methods Enzymol. 2001;338:202. 7. Diercks T, Coles M, Kessler H. Curr. Opin. Chem. Biol. 2001;5:285. 8. Pellecchia M, Sem DS, W¨uthrich K. Nat. Rev. Drug Discov. 2002;1:211. 9. Van Dongen M, Weigelt J, Uppenberg J, Schultz J, Wikstr¨om M. Drug Discov. Today. 2002;7:471. 10. Wyss D, McCoy MA, Senior MM. Curr. Opin. Drug Discov. Dev. 2002;5:630. 11. Stockman BJ, Dalvit C. Prog. NMR Spectrosc. 2002;41: 187. 12. Zartler ER, Yan J, Mo H, Kline AD, Shapiro MJ. Curr. Top. Med. Chem. 2003;3:25. 13. Fielding L. Curr. Top. Med. Chem. 2003;3:39. 14. Meyer B, Peters T. Angew. Chem. Int. Ed. 2003;42:864. 15. Coles M, Heller M, Kessler H. Drug Discov. Today. 2003;8:803. 16. Salvatella X, Giralt E. Chem. Soc. Rev. 2003;32:365. 17. Jahnke W, Widmer H. Cell. Mol. Life Sci. 2004;61: 580. 18. Dalvit C, Flocco M, Veronesi M, Stockman BJ. Comb. Chem. HTS. 2002;5:605.

References 1399

19. Dalvit C, Fagerness PE, Hadden DTA, Sarver RW, Stockman BJ. J. Am. Chem. Soc. 2003;125:7696. 20. Dalvit C, Ardini E, Flocco M, Fogliatto GP, Mongelli N, Veronesi M. J. Am. Chem. Soc. 2003;125:14620. 21. Dalvit C, Flocco M, Knapp S, Mostardini M, Perego R, Stockman BJ, Veronesi M, Varasi M. J. Am. Chem. Soc. 2002;124:7702. 22. Jahnke W, Floersheim P, Ostermeier C, Zhang X, Hemmig R, Hurth K, Uzunov DP. Angew. Chem. Int. Ed. 2002;41: 3420. 23. Siriwardena AH, Tian F, Noble S, Prestegard JH. Angew. Chem. Int. Ed. 2002;41:3454. 24. Canet D. Nuclear Magnetic Resonance Concepts and Methods. John Wiley & Sons: Chichester, 1996. 25. Hull WE, Sykes BD. J. Mol. Biol. 1975;98:121. 26. Gerig JT. Methods Enzymol. 1989;177:3. 27. Gerig JT. Prog. NMR Spectrosc. 1994;26:293. 28. London RE, Gabel SA. J. Am. Chem. Soc. 1994;116:2570. 29. Lian LY, Barsukov IL, Sutcliffe MJ, Sze KH, Roberts GCK. Methods Enzymol. 1994;239:657. 30. Craik DJ, Wilce JA. In: DG Reid (Ed). Protein NMR Techniques. Humana Press Inc.: New Jersey, 1997, pp 195–232. 31. Carr HY, Purcell EM. Phys. Rev. 1954;94:630. 32. Meiboom S, Gill D. Rev. Sci. Instrum. 1958;29:688. 33. Luz Z, Meiboom S. J. Chem. Phys. 1963;39:366. 34. Allerhand A, Gutowsky HS. J. Chem. Phys. 1964;41: 2115. 35. Mayer M, Meyer B. Angew. Chem. Int. Ed. 1999;38:1784. 36. Dalvit C, Pevarello P, Tat`o M, Veronesi M, Vulpetti A, Sundstr¨om M. J. Biomol. NMR. 2000;18:65. 37. Tengel T, Fex T, Emtenas H, Almqvist F, Sethson I, Kihlberg J. Org. Biomol. Chem. 2004;2:725. 38. Doerr AJ, Case MA, Pelczer I, McLendon GL. J. Am. Chem. Soc. 2004;126:4192. 39. Percival MD, Withers SG. Biochemistry. 1992;31:505. 40. Evans JNS. Biomolecular NMR Spectroscopy. Oxford University press, New York, USA, 1995, pp 237–340. 41. Segel IH. Biochemical Calculations. John Wiley & Sons, New York, USA, 1976. 42. Dalvit C, Ardini E, Fogliatto GP, Mongelli N, Veronesi M. Drug Discov. Today. 2004;9:595. 43. Reuveni H, Livnah N, Geiger T, Klein S, Ohne O, Cohen I, Benhar M, Gellerman G, Levitzki A. Biochemistry. 2002;41:10304. 44. Obata T, Yaffe MB, Leparc GG, Piro ET, Maegawa H, Kashiwagi A, Kikkawa R, Cantley LC. J. Biol. Chem. 2000;46:36108.

Part II

compares favorably with the concentration used with the other techniques used in HTS. In addition, the two NMR techniques have some important advantages. Their simplicity together with the possibility of directly monitoring the real concentration, purity, stability, and solubility of the screened compounds results in reliable lead molecule detection and precise quantification of their strength. It is envisioned that the speed and easy set-up of FAXS and 3-FABS together with their broad range of applications will play a major role in the drug discovery process for discovering potent, bioavailable, and safe clinical candidates.

NMR Spectroscopy

1401

Philip J. Hajduk Global Pharmaceutical Research and Development, Abbott Laboratories, Abbott Park, IL 60064, USA

Introduction High-throughput screening (HTS) of large corporate compound libraries has become the primary strategy for lead generation in the pharmaceutical industry. However, despite the advances in screening and chemistry technologies over the last decade, there has been little increase in the rate of discovery of quality drug leads [1,2]. One of the primary reasons for this problem is the nature of the hits that come from traditional HTS campaigns. Most hits tend to look “drug-like,” with molecular weights 2 nm/s

1.0 nm/s

0.3 nm/s

δ

α

0.04 nm/s 20

10

0 −10 [kHz]

ν

−20

10

30 20 20 −2 δ [10 Vm ]

Fig. 5. COF fits of 27 Al MAS of amorphous Alq3 obtained at different deposition rates. The insets show the distribution of the anisotropy parameter (δ) of the EFG tensor. The asymmetry parameter (η) is constant in all the fits.

indicating less disorder in the more slowly deposited samples. The bimodal distribution suggests the presence of two distinct conformations of the Alq3 molecule. It is tempting to assign the small peak at a lower δ to the facial isomer, which exhibits a smaller EFG anisotropy than the meridianal form also in crystalline forms of Alq3 . This would allow us to estimate the content of facial isomer in the amorphous samples at 8% ± 4%. However, a more sensitive NMR technique, such as variable angle correlation, is necessary to confirm this. Such experiments are currently underway in our laboratory. The observed dependence of the molecular order on the deposition rate correlates with changes in the charge transport properties. Figure 6 shows the variation of electroluminescence quantum efficiency with the current density of OLED devices prepared at different deposition rates. There is a significant difference in the performance

NMR of Organic Semiconductors

References 1545

10−4 Quant. Eff. (arb. units)

References 1. 2. 3. 4. 5. 10−5

10−5

10−4

0.04 nm/sec 0.3 nm/sec 1 nm/sec

6.

10−3

10.

J (A/cm2)

Fig. 6. Quantum Efficiency of the devices obtained by different thermal deposition rates(cf. text).

of the devices prepared at various evaporation rates. This must be due to structural changes at the molecular level, techniques such as XRD fails to pick up, in contrast to 27 Al NMR spectroscopy.

7. 8. 9.

11. 12. 13. 14. 15. 16.

Conclusions The kinetics of exchange between the inequivalent ligands in the meridianal isomer of Alq3 have been studied at different temperatures by 2D exchange proton NMR spectroscopy. The ligands were found to exchange on a time scale of about 5 s −1 at room temperature, and the exchange pattern can be explained using a simple firstorder mechanism based on 180◦ flips of the ligands. Using 27 Al solid-state NMR, it was found that the α polymorph consists of the meridianal isomer and the δ polymorph consists of the facial isomer. 1D 27 Al MAS NMR spectra of amorphous Alq3 films obtained at different deposition rates have been studied. Our results show that 27 Al solid state NMR is a powerful tool to study the structure and morphology of Alq3 -based devices, capable of quantifying the molecular disorder in amorphous Alq3 deposits.

17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Acknowledgments Parts of the present work have been supported by the National Science Foundation through CAREER awards to both MU (DMR-0094290) and FP (DMR-970220), as well as by a type G grant from the Petroleum Research Fund administered by the American Chemical Society to MU (PRF-36246-G7). FP also gratefully acknowledges

27. 28. 29. 30. 31.

Cacialli F. Phil. Trans. R. Soc. Lond. A. 2000; 358:173. Tang CW, Van Slyke SA. Appl. Phys. Lett. 1987; 51:913. Yannoni CS, Clarke TC. Phys. Rev. Lett. 1983; 51:1191. Mizoguchi K, Jpn. J. Appl. Phys., Part 1 1995; 34:1. Kaplan S, Conwell EM, Richter AF, MacDiarmid AG. Synth. Met. 1989; 29:E235. Stafstr¨om S, Sj¨ogren B, Wennerstr¨om O, Hjertberg T. Synth. Met. 1986; 16:31. Nechtschein M, Santier C. Physique J. 1986; 47:935. Bernier P, Stein PC, Lenoir C, Physica B. C 1986; 143:494. O’Gara J, Nazri G, MacArthur K. Solid State Ionics. 1991; 47:87. Simpson JH, Rice DM, Karasz, FE. Polym. J. Sci. B. Polym. Phys. 1992; 30:11. Cholli AL,Thakur M. Chem. J. Phys. 1989; 91:7912. Ago H, Tanaka K, Yamabe T, Miyoshi T, Takegoshi K, Terao T, Yata S, Hato Y, Nagura S, Ando N, Carbon. 1997; 35:1781. Tang CW, Van Slyke SA, Chen CH, Appl. J. Phys. 1989; 65:3610. Farchioni R, Grosso G, (Eds). Organic Electronic Materials, Springer: Berlin, Heidelberg, 2001. Curioni A, Boero M, Andreoni W. Chem. Phys. Lett. 1998; 294:263. CurioniA, Andreoni W, Treusch R, Himpsel FJ, Haskal E, Seidler P, Heske C, Kakar S, van Buuren T, Terminello LJ. Appl. Phys. Lett. 1998; 72:1575. Esposti A, Brinkmann M, Ruani G. Chem. J. Phys. 2002; 116:798. Amati M, Lelj F. Chem. Phys. Lett. 2002; 363:451. Utz, M Chen C Morton, Papadimitrakopoulos, M. F Am. J. Chem. Soc. 2003; 125:1371. Brinkmann M, Gadret G, Muccini, M Taliani C, Masciocchi N, Sironi A. Am. J. Chem. Soc. 2000; 122:5147. Braun M, Gmeiner J, Tzolov M, Coelle M, Meyer FD, Milius W, Hillebrecht H, Wendland O, von Sch¨utz JU, Br¨utting W. Chem. J. Phys. 2001; 114:9625. C¨olle M, Gmeiner J, Milius W, Hillebrecht H, Br¨utting W. Adv. Funct. Mater. 2003; 13:108. Schmidbaur H, Lettenbauer J, Wilkinson DL, M¨uller G, Kumberger O, Naturforsch Z. 1991; 46b:901. C¨olle M, Dinnebier RE, Br¨utting W. Chem. Comm. 2002; 2908. Chen BJ, Lai WY, Gao ZQ, Lee CS, Lee ST, Gambling WA. Appl. Phys. Lett. 1999; 75:4010. Cheng LF, Liao LS, Lai WY, Sun XH, Wong NB, Lee CS, Lee ST. Chem. Phys. Lett. 2000; 319:418. Kim SY, Ryu SY, Choi JM, Kang SJ, Park SP, Im S, Whang CN, Choi DS. Thin Solid Films 2001; 398–399:78. Qin DS, Li DC, Wang Y, Zhang JD, Xie ZY, Wang G, Wang LX. Appl. Phys. Lett. 2001; 78:437. Baker BC, Sawyer DT. Analyt. Chem. 1968; 40:1945. Jeener J, Meier BH, Bachmann T, Ernst RR. Chem. J. Phys. 1979; 71:4546. Utz M. Chem. J. Phys. 1998; 109:6110.

Part III

support of the Critical Technologies Program by Connecticut Innovations, Inc., and Trans-Lux Corp.

1547

J.L. Yarger,1 D.A. Buttry2 G.P. Holland1 1 Department

of Chemistry and Biochemistry, Arizona State University, Tempe, AZ 85287-1604, USA 2 Department of Chemistry, University of Wyoming, and Laramie, WY 82071, USA

Xerogels, aerogels and other nanoporous solids are an important class of materials in electrochemical and catalytic process design. The preparation of these materials by sol–gel processing has become an important field in materials science [1]. In hydrolytic processes, the texture and structure of sol–gel-based materials largely depend on synthetic parameters, such as (i) solvent, (ii) temperature, (iii) monomer to water ratio, etc. Considerable work has been done to identify the relevant synthetic parameters in order to control the surface and porosity of xerogels (and aerogels). NMR spectroscopy has been extensively used to investigate the effect of these parameters in the sols, prior to gelation, or in the xerogels, after aging and drying [2–5]. Xerogels (xero means dry) are derived from sol–gel precursors by evaporative drying, while aerogels are supercritically (or hypercritically) dried. Both of these sol– gel-based materials typically have high porosity, low density, and therefore, high surface areas. Intercalation into xerogels is attractive, not only because of the high surface area, but also because of the small distances in the solid that must be penetrated by the guest ions [6]. Again, NMR spectroscopy has played a crucial role in structural and surface elucidation of xerogel-based intercalation materials as well as elucidation of transport properties and local environments of the guest ions or material [7,8]. While much of the NMR characterization of xerogels has been devoted to silicate based materials [9–12], we will focus our examples on the vanadate xerogel system [13–17]. We will use our NMR studies of lithium intercalated V2 O5 xerogels as an extended example of several various solid-state NMR techniques that can be used to help elucidate the structure and dynamics of intercalated xerogel materials [18,19]. Because most of our NMR will focus on observation of the intercalating ions, the techniques presented are broadly applicable to other porous xerogel materials. Characterization of the host material is dictated strongly by the type of NMR nuclei present in specific xerogels. Hence, we will not focus on NMR characterization of the host xerogel, but rather the guest ions. Vanadium pentoxide xerogels have attracted considerable attention as a next generation cathode for rechargeable lithium batteries [20,21]. NMR studies of vanadium Graham A. Webb (ed.), Modern Magnetic Resonance, 1547–1550.  C 2008 Springer.

pentoxide xerogels serve as an excellent example of the need for high resolution magic angle spinning (MAS) techniques to properly study such materials. Figure 1 (left) shows static 7 Li NMR spectra a several V2 O5 samples prepared at different levels of lithiation. One can clearly see that as lithiation (i.e. reduction of vanadium (V) centers to vanadium (IV)) proceeds, the major resonance broadens and shifts. The peak asymmetry reveals that there may be more than one resonance, meaning that there may be more than one type of site for Li+ in this high surface area material. However, the static spectrum offers limited possibility of learning details about these various lithium environments. In contrast, Figure 1 (right) shows a 7 Li MAS NMR spectra the same sample. In these spectra, one clearly resolves two sites. The sharper resonance corresponds to Li+ at interfacial, ion-exchange sites and the broader resonance corresponds to Li+ inserted within the sheets of the V2 O5 layered lattice. In the context of battery materials, the availability of interfacial sites greatly increases the power density for a given material because of rapid and facile Li+ access to those sites. Hence, high-resolution 7 Li solid-state NMR provides an excellent probe for testing the potential of new intercalation materials for the next generation of lithium battery cathodes. In addition to identification and quantitation of Li+ binding sites in such materials, solid-state techniques also provide information on dynamics, and, indirectly, on more subtle structural questions. For example, the T1 (spinlattice) relaxation time provides information on loss of NMR excitation energy into the surrounding material. For the case of lithiated V2 O5 xerogels, we found that T1 for the two sites described above was dominated by interactions between the 7 Li nuclear spins and nearby unpaired electrons at V(IV) sites (polaron hopping). Further, it was determined that the T1 relaxation time for the Li+ ions inserted within the lattice was substantially shorter than for those at the interfacial sites. This type of information reveals both the extent to which Li+ interacts with electrons (polarons) within the material and provides information on the spatial relationship between Li+ and the electron (i.e. the site geometry). In preliminary work, we have modeled these data based on the through-space coupling between unpaired electrons at V(IV) sites and Li+ ions to determine the sites for the inserted Li+ .

Part III

Solid State NMR of Xerogels

1548 Part III

Materials Science

---------------------------------------------

Part III -80

-60

-40

-20

(E)

(E) (D)

(D) (C)

(C)

(B)

(B)

(A)

(A)

20

kHz

40

60

80

-10

-5

kHz

5

10

Fig. 1. Room temperature 7 Li static (left) and magic angle spinning (right). NMR spectra of lithium intercalated V2 O5 ∗ nH2 O xerogel. The electrochemically lithiated Lix V2 O5 ∗ nH2 O xerogel, where x is 0.03 (A), 0.17 (B), 0.55 (C), 0.84 (D) and 0.98 (E). Spectra were collected on a 400 MHz NMR system with 4 mm MAS probe (ωr = 10 kHz).

A second issue of critical importance for cathode material design optimization relates to the interplay between ion motion and electron motion and how these depend on their relative locations. Figure 2 shows an example of how electron–nuclear dipole coupling can influence the T1 relaxation time in lithiated vanadium pentoxide xerogels, Lix V2 O5 . It is widely thought that the motion of Li+ -2.8

τc (electron)

O

=

O =

-3.0

_ V _ O _ V4+_

ln(T1) - seconds

5+

-3.2 -3.4

Li+

ωNMR

-3.6 -3.8 -4.0 -4.2

τc ωNMR = 1 2.5

3

3.5

4

4.5

5

5.5

6

1000/T (1/K) Fig. 2. The natural logarithm of T1 versus inverse temperature for x = 0.03 () and x = 0.17 (•) Lix V2 O5 ·0.5H2 O xerogels. The solid lines represent the best spline fits to the experimental data and are included as a guide to the eye. The 7 Li static NMR T1 was determined from a standard population inversion experiment (π − π2 ). All measurements were performed on a 400 MHz NMR system (ω Li = 155 MHz).

through a cathode material is often the rate-limiting process that controls power density. The NMR results associated with Figure 2 demonstrate a type of charge pinning in Li+ insertion materials. This has also been observed in lithium vanadium phosphate cathode material, where it was shown that the extra unpaired electrons were ordered within the lattice and that the charge compensating Li+ ions were pinned to this ordered unpaired electron sublattice [22]. This suggests that electron-ion pinning may be important in terms of understanding charge transport in such systems. Ionic mobility and binding in xerogel materials can further be explored to monitor ion-exchange processes. An example of this is displayed in Figure 3. In the 7 Li MAS NMR spectra of electrochemically lithiated V2 O5 xerogels it was postulated that peak A represented interfacial, ion-exchangeable environments, while peak B corresponded to Li+ intercalation sites. This was verified by exposing Li0.2 V2 O5 xerogel to a NaClO4 electrolyte solution, and monitoring the 7 Li MAS NMR response as a function of exposure time. These results provide strong evidence that peak A originates from an intrinsic ionexchange site at the xerogel interface as it rapidly decays when the xerogel is exposed to Na+ counter ions. This demonstrates that simple solid-state NMR spectroscopy can be effectively employed to follow charge compensation at surfaces in nanoporous solids like xerogels and further, that these types of environments can be distinguished from interlamellar intercalation sites in these materials. Considerable effort has been made towards increasing the conductive properties in V2 O5 xerogel materials by incorporating a conductive polymer within the xerogel matrix. It was shown with solid-state NMR that similar

Solid State NMR of Xerogels

0.20 0.15 0.10 0.05 0.00 0

10

20

30 40 Time/mins.

50

60

Fig. 3. Time dependence of ion-exchange in Li0.2 V2 O5 xerogel exposed to 1 M NaClO4 in propylene carbonate plotted as the integrated peak ratio ( BA ) as a function of exposure time in minutes. Peaks A and B correspond to ion-exchangeable (interfacial) and intercalated (interlamellar) lithium sites, respectively.

ion-exchange processes are important in understanding how these V2 O5 nanocomposites form and function. Two types of lithiated V2 O5 nanocomposites containing polyaniline (PANI) and poly(aniline N -propane sulfonic acid) (PSPAN) have been studied with 7 Li MAS NMR [19]. In the 7 Li MAS spectra of the PANI/V2 O5 nanocomposite essentially no ion-exchange site was observed, while the PSPAN/V2 O5 composite displayed a significant

(A)

(B)

-800

-400

200

0

-200 -400 -600 -800

F1 (ppm)

-600

-400

0

F2 (ppm)

-800

-600

-200

800 600 400 200

ion-exchangeable component. PANI is cationic causing it to function as an anion exchange material. Therefore, the lack of Li+ present in ion-exchange sites is interpreted to result due to PANI interacting with these sites during the synthetic process. In contrast, a significant ion-exchange resonance was observed in the PSPAN/V2 O5 nanocomposite. This was attributed to the fact that PSPAN is a self-doped conducting polymer. It possesses pendent anionic groups that can inherently provide charge compensation during doping of the polyaniline-like groups in the main chains. The 7 Li NMR results of these composite materials not only provide information regarding the local Li environment in the material, but also indirectly provide some insight into how the conductive polymer and V2 O5 xerogel host interact during the sol–gel synthetic process. For quadrupolar nuclei (I > 12 ), the quadrupole coupling constant and the asymmetry parameter carry valuable information about the nature of anisotropic interactions [23]. However, in the static and MAS 7 Li NMR (I = 32 ) spectra of Lix V2 O5 xerogel the quadrupolar effects are masked by the paramagnetic broadening, especially at high V4+ concentrations (large Li+ loadings). Two-dimensional separation techniques have been employed in solid-state NMR spectroscopy to separate competing interactions, such as, dipolar and chemical shift anisotropy interactions [24]. In our case, we use a 2D spinecho experiment or a three-pulse sequence to separate out the contribution from quadrupolar and paramagnetic interactions of lithium-7 (I = 32 ) in Lix V2 O5 xerogel. The 2D contour plot, together with the F1 and F2 projections, is shown in Figure 4. The two frequency dimensions are

-200 0 200

400

400

600

600

800

800 800 600 400 200

0

F1 (ppm)

Integrated Peak Ratio (A/B)

B

-200 -400 -600 -800

F2 (ppm)

Fig. 4. 7 Li two-dimensional static spin-echo NMR spectrum of (A) LiV2 O5 xerogel (C Q = 90 kHz) and (B) γ-LiV2 O5 crystalline material (C Q = 121 kHz). F2 dimension displays static spectrum affected by both quadrupolar and paramagnetic interaction. This method eliminates the paramagnetic interaction in the F1 dimension allowing a model-free extraction of the 7 Li quadrupole coupling constant (C Q ) in this dimension.

Part III

A 0.25

Solid State NMR of Xerogels 1549

1550 Part III

Materials Science

Part III

asymmetrically correlated with respect to the diagonal. This is caused by the influence of the Li+ –V4+ paramagnetic dipolar interactions. From the resolvable satellite transitions (± 12 ⇔ ± 32 ) we can identify the paramagnetic frequency shift and the quadrupole coupling constant (C Q ). The influence of paramagnetic dipolar interaction on the quadrupolar broadened 7 Li NMR spectrum of LiV2 O5 materials is shown in Figure 4. The 7 Li quadrupole coupling constant was determined to be C Q = 90 and 121 kHz for LiV2 O5 xerogel and γ-LiV2 O5 crystalline material, respectively. It is surprising to find that C Q is smaller for the disordered xerogel, when compared to the ordered crystalline material. This indicates that though the host material is disordered, the lithium ions are in an ordered high-symmetry environment. In summary, we show several NMR experiments used to elucidate the structural and dynamic environment of lithium ions in a xerogel host material. By using 2D separation, relaxation, high-resolution MAS, and exchange NMR experiments, we are able to completely characterize the motion and structural environment of lithium ions in vanadate xerogels. All the NMR techniques discussed can be used on any lithium host material.

Acknowledgments Support provided from the National Science Foundation, the Department of Defense, the W.M. Keck Foundation and the Department of Energy.

References 1. Brinker C, Scherer G, Sol–Gel Science: The Physical and Chemical of Sol-Gel Processing. Academic Press; New York, 1990. 2. Cros F, Malier L, Korb J, Chaput F. J. Sol–Gel Sci. Technol. 1998;13:457.

3. Framery E, Mutin P. J. Sol–Gel Sci. Technol. 2002;24: 191. 4. Okada K, Tomita T, Kameshima Y, Yasumori A, MacKenzie K, J. Mater. Chem. 1999;9:1307. 5. Pickup D, Mountjoy G, Holland M, Wallidge Newport G, Smith M. J. Mater. Chem. 2000;10:1887. 6. Faceelli J, de Dias A. Acs Symposium Series. American Chemical Society; Washington, D.C., 1999. 7. Babu P, Tong Y, Kim H, Wieckowski A. J. Electronal. Chem. 2002;524:157. 8. Lee Y, Wang F, Grey C. J. Am. Chem. Soc. 1998;120: 12601. 9. Pickup D, Mountjoy G, Wallidge G, Newport R, Smith M. Phys. Chem. Chem. Phys. 1999;1:2527. 10. Brinker C, Keefer K, Schaefer D, Assink RKB, Ashley C. J. Non-Cryst. Solids 1984;63:45. 11. Klemperer W, Mainz V, Millar D. Mater. Res. Soc. Symp. 1986;73:15. 12. Franmery E, Mutin P, J. Sol–Gel Sci. Technol. 2002;24: 191. 13. Bose M, Basu A. Solid State Ion. 1986;18/19:902. 14. Hirschinger J, Mongrelet T, Marichal C, Granger P, Savariault J-M, D´eramond E, Galy J. J. Phys. Chems. 1993;97:10301. 15. Rozier P, Savariault J-M, Galy J, Marichal C, Hirschinger J, Granger P. Eur. J. Solid State Inorg. Chem. 1999;33:1. 16. Cocciantelli JM, Suh KS, S´en´egas J, Documerc JP, Pouchard M. J. Phys. Chems. Solids 1992;53:57k. 17. Cocciantelli JM, Suh KS, S´en´egas J, Documerc JP, Soubeyroux JL, Pouchard M. J. Phys. Chems. Solids 1992;53:51. 18. Holland G, Buttry D, Yarger J. Chem. Mater. 2002;14: 3875. 19. Hollland G, Yarger J, Buttry D, Huguenin F, Torresi R. J. Electrochem. Soc. 2003;150:A1718. 20. Park HK, Smyrl WH, Ressler J, Owens BB, Smyrl WH. J. Electrochem. Soc. 1994;141:L25. 21. Le DB, Passerini S, Guo J, Ressler J, Owens BB, Smyrl WH. J. Elecrochem. Soc. 1996;143:2099. 22. Yin S, Grondey H, Strobel P, Huang H, Nazar L. J. Am. Chem. Soc. 2003;125:326. 23. Taylor P, Baugher J, Kriz H. Chem. Rev. 1975;75:591. 24. Ganapathy G, Rajamohanan R, Ganguly P, Venkatraman T, Kumar A. J. Phys. Chems. A. 2000;104:2007.

1551

Sharon E. Ashbrook Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, UK Present address: School of Chemistry, University of St. Andrews, St. Andrews, Fife KY16 9ST, UK

Introduction Oxygen is the second most abundant element on Earth, constituting 30% by weight of the total Earth and 46% by weight of the Earth’s crust. Given also the abundance of silicon (15% by weight) and the strength of the Si–O bond, it is perhaps no surprise that the majority of rocks are composed of silicates. The study of these materials, therefore, has an important role in the understanding of both the physical and chemical properties of the Earth. The vertical stratification of the Earth, into a largely aluminosilicate crust, oxide mantle, metallic liquid outer core, and metallic inner core is well known. As shown schematically in Figure 1, the mantle itself may be further subdivided, with the principal component of the upper mantle (down to a depth of 410 km) being olivine, (Mg0.8 Fe0.1 )2 SiO4 , an iron-bearing form of forsterite (α-Mg2 SiO4 ). A series of phase transitions occurs with increasing depth and pressure, resulting in the formation of the β- and γpolymorphs of (Mg,Fe)2 SiO4 , wadsleyite, and ringwoodite, defining the transition zone [1,2]. At greater pressures still, in the lower mantle, disproportionation occurs forming (Mg,Fe)SiO3 (perovskite) and magnesiow¨ustite, (Mg,Fe)O. The presence of less abundant phases in the mantle, such as garnets, pyroxenes, and high-pressure silica, is also of crucial importance in determining the physical properties of the deep Earth. This simplified picture disguises a number of important questions within the Earth sciences, however. Although the compositions of the various mantle regions are somewhat constrained, we do not know the exact Si, Mg, Fe, Ca, or Al content of the various phases. We know about the interior of our planet only from secondary evidence, such as seismological measurements, meteorites, rocks brought to the crust from the mantle by geological processes, and the simulation of materials under extreme conditions [3]. In particular, there are many indications from these sources that the inner Earth contains many times the amount of hydrogen (colloquially termed water) than is present in the oceans, with estimates ranging from 200 to 600 ppm [4–7]. The presence of this water is of great interest owing to its influence on a range of physical Graham A. Webb (ed.), Modern Magnetic Resonance, 1551–1561.
C 2008 Springer.

and chemical properties such as mantle rheology and melting temperatures and pressures. It is now widely believed that nominally anhydrous minerals, such as olivine, wadsleyite, and ringwoodite can incorporate water within their structures [4–7]. The study of the structure and the dynamic behavior of high-pressure solid silicates and, in particular their hydration, is of vital importance.

Oxygen NMR The study of solid silicates by oxygen NMR would appear to be a natural choice given the presence of oxygen in the minerals present in the crust and mantle and the sensitivity of NMR to the local environment. However, the only NMR active isotope of oxygen, 17 O (I = 5/2), has both a low sensitivity (ω0 /2π = 54.2 MHz at 9.4 T) and an extremely low natural abundance (∼0.037%). The additional problem of difficult experimental syntheses for high-pressure silicates, which produce only small amounts of material, results in a considerable sensitivity problem. Although samples may be synthesized with isotopic enrichment, sensitivity remains a crucial factor. As 17 O has spin I > 1/2, spectra are broadened by the quadrupolar interaction, often over many megahertz [8]. Magic angle spinning (MAS) is able to remove the quadrupolar interaction to a first-order approximation but is unable to remove fully the second-order quadrupolar interaction [8] and resonances remain broadened. Much attention has been focused on the acquisition of highresolution NMR spectra for quadrupolar nuclei. One approach is to use composite sample rotation techniques, such as dynamic angle spinning (DAS) [9] and double rotation (DOR) [10], which achieve resolution enhancement though rotation around two different angles. The introduction of the multiple-quantum (MQ) MAS technique [11] in 1995 revolutionized high-resolution NMR of quadrupolar nuclei, and this method has become widespread owing to the ease of its implementation. In 2000, Gan introduced an alternative method for obtaining high-resolution spectra of quadrupolar nuclei, the satellite-transition (ST) MAS experiment [12]. Although slightly more difficult to implement than MQMAS, its high inherent

Part III

Solid-State 17O NMR Spectroscopy of High-Pressure Silicates

1552 Part III

Materials Science

Part III

component

depth/km 0 crust

aluminosilicates pyroxenes (MgSiO3)

upper mantle

forsterite (α-Mg2SiO4) pyroxenes (MgSiO3)

40

410

upper

wadsleyite (β-Mg2SiO4)

lower

ringwoodite (γ-Mg2SiO4)

530 transition zone 660

present, at least one oxygen species, denoted a “bridging oxygen,” is bonded to two silicons simultaneously. This is reflected in considerable changes in chemical shift and, in particular, in quadrupolar coupling [17]. Bridging oxygen species exhibit CQ values (∼4–6 MHz) that are significantly higher than those for non-bridging species (∼2–3.5 MHz). The largest quadrupolar couplings (∼6– 8 MHz), however, tend to be found for hydroxyl oxygens, providing an excellent opportunity to study hydration.

Sample Preparation lower mantle

perovskite (MgSiO3) magnesiowüstite (MgO) stishovite (SiO2)

2900 core

Fig. 1. Schematic diagram of the compositional variation of the Earth.

sensitivity has proved crucial in the study of high-pressure silicate phases, particularly in cases where sample volume is limited. Despite the problems outlined above 17 O NMR has enormous potential in the study of silicate minerals. NMR enables investigation of both the local structure and the dynamic behavior of a system. Both 29 Si and 1 H NMR have been employed in the study of mantle phases [13–16], but neither can offer the flexibility of 17 O NMR. Silicon NMR provides information only on the local Si environment while, on its own, 1 H NMR is unable to identify specific O–H pairings, is insensitive to dynamic effects and suffers from problems with a large “background signal.” In contrast, 17 O NMR (both single and double resonance techniques) provides information on the local O environment, can offer information on specific O–H pairings and often displays a sensitivity to dynamic behavior. Both the 17 O quadrupolar and chemical shift interactions are extremely sensitive to changes in the environment on a local scale [17], providing an excellent tool for studying silicates, a group of minerals with wide compositional diversity and structural modifications. Silicate structures tend to be dominated by the SiO4 tetrahedron. These units can be isolated from each other, and coordinated to other cation polyhedra present in the structure. The nature of the cation has a significant effect upon the Si–O bond length and, hence, the 17 O chemical shift. It has been shown for many systems that an increase in 17 O chemical shift reflects an increased Si–O bond length. Alternatively, SiO4 units can be connected to each other by corner sharing, forming Si2 O7 groups or even long chains. Further polymerization results in silicate sheets and finally three-dimensional silicate frameworks. When corner-sharing tetrahedra are

The low natural abundance of 17 O (0.037%) necessitates the use of isotopic enrichment in order to obtain good sensitivity. A wide range of enrichment schemes is possible, with both O2 and H2 O readily available as isotopically enriched starting material. The exact enrichment level employed is usually a compromise between cost and sensitivity. For the high-pressure silicates considered here, the preparation method of choice is usually synthesis from 17 O-enriched oxides prepared from H2 O (l), with common enrichment levels between 10 and 75%. The 17 O is fixed by reaction of H2 17 O with reagents such as SiCl4 , Mg3 N2 , Ca3 N2 , and AlN liberating a gas (HCl or N2 ) and leaving a solid oxide or hydroxide product. These may then be mixed in appropriate stoichiometric amounts to prepare the silicate required. The elements O, Mg, Al, Ca, and Si represent the major components of an Fe-free crust and mantle and allow all major phases to be prepared. Mantle phases may be prepared by simulating the pressure and temperature conditions present in the mantle using either a piston cylinder or a multi-anvil apparatus. Piston cylinder devices produce only relatively moderate pressures (up to 4 GPa, corresponding to a depth of only 120 km) but allow large volumes of sample (∼100 mg) to be produced. In comparison, multi-anvil devices allow access to a much greater pressure range (up to 22 GPa, or depths of >660 km) but the volume of material available is much reduced (often only 5–10 mg). Samples are quenched rapidly from the experimental high-pressure and hightemperature conditions as the high-pressure phases are usually kinetically stable at ambient conditions.

MQMAS NMR of Upper Mantle Silicates Forsterite (or α-Mg2 SiO4 ), is the Fe-free form of olivine, the principal component of the Earth’s mantle to a depth of ∼410 km. The crystal structure of forsterite (Pbnm) consists of isolated SiO4− 4 units with three distinct oxygen species (O1, O2, and two O3) [18]. Forsterite has become almost a “model” compound for 17 O NMR as: (i) isotopically enriched forsterite is relatively easily prepared at ambient pressures in reasonably large quantities [19]; (ii) olivine has a huge importance in the mantle; and

Solid-State 17 O NMR of High-Pressure Silicates

Part III

a

MQMAS NMR of Upper Mantle Silicates 1553

c

150

100

80

60

b

50 0 δ (ppm)

−50

1 kHz

δ1 (ppm)

20

40

60 40 δ2 (ppm)

20

0

Fig. 2. 17 O (54.2 MHz) MAS NMR of 35% isotopically enriched forsterite (α-Mg2 SiO4 ). (a) MAS NMR spectrum, (b) two-dimensional triple-quantum MAS NMR spectrum, recorded with a phase-modulated split-t1 shifted echo sequence, and (c) cross sections parallel to δ 2 , along the three second-order broadened ridge lineshapes and corresponding computer fits.

(iii) the non-bridging nature of the oxygen species results in only moderate quadrupolar couplings. As a result, a variety of NMR techniques have been employed in the study of forsterite. The extraction of information from the MAS spectrum alone, however, is hampered by the presence of second-order quadrupolar broadening [20], as shown in Figure 2a. This 17 O (54.2 MHz) MAS NMR spectrum of forsterite (enriched to 35% in 17 O) reveals a composite lineshape centered at ∼50 ppm, with features characteristic of second-order broadening, but resulting from the overlap of three different lineshapes. In order to obtain accurate chemical shift and quadrupolar information for the three oxygen species spectral resolution must be improved. Both DAS [21] and DOR [22] have been employed to obtain isotropic spectra of forsterite. However, a complete assignment of the 17 O high-resolution spectrum was first achieved using MQMAS [19]. Figure 2b shows the two-dimensional 17 O triple-quantum MAS NMR spectrum of forsterite (enriched to 35% in 17 O), recorded using a phase-modulated split-t1 shifted echo experiment [23]. The spectrum consists of a series of ridge lineshapes, one for each crystallographically distinct species, lying parallel to the δ 2 axis. The high-resolution or isotropic spectrum, obtained from a projection onto δ 1 , reveals three distinct resonances in the ratio 1:2:1. Although MQMAS efficiency varies as a function of the quadrupolar interaction [23], the small differences between the three distinct oxygen species enables the site populations to be determined directly from the relative intensities of the resonances. The

position of the center of gravity of each lineshape in the two-dimensional spectrum (δ 1 , δ 2 ) allows the extraction of the isotropic chemical shift, δ CS , and the secondorder quadrupolar shift, δQ [19]. From δQ , the quadrupolar product PQ = CQ (1 + η2 /3)1/2 can be obtained. The individual quadrupolar parameters, the magnitude (CQ ) and asymmetry (η) of the quadrupolar interaction, may be determined from cross sections taken along each ridge, parallel to δ2 . The lineshapes obtained may be fitted using a computer fitting routine to obtain values for CQ , η, and δCS , as shown in Figure 2c. The values of the chemical shift and quadrupolar parameters for the oxygen species in forsterite are given in Table 1 [19]. If the relative populations of the sites and the similarity in the coordination environment of O2 and O3 are considered a complete assignment of the resonances is possible, and is also shown in Table 1. MQMAS has also been employed to study a series of pyroxenes, an important group of rock-forming minerals which occur in nearly all igneous rocks and comprise up to 25% of the upper mantle [24]. Pyroxenes have the general formula M1M2Si2 O6 and consist of infinite chains of corner-sharing SiO4 tetrahedra. Each tetrahedron contains both bridging and nonbridging oxygen species, providing a more challenging target for 17 O NMR. Perhaps the most relevant pyroxene for mantle chemistry is enstatite, Mg2 Si2 O6 , of which there are three stable or metastable polymorphs under ambient conditions. Orthoenstatite (Pbca) and protoenstatite (Pbcn) are orthorhombic [25,26], whilst

1554 Part III

Materials Science

Part III

Table 1: 17 O isotropic chemical shifts (δCS ), quadrupolar products (PQ ), quadrupolar coupling constants (CQ ), asymmetries (η), relative populations, and tentative assignments of the oxygen species a variety of silicate minerals. Mineral Forsterite [19]

Orthoenstatite [24]

Clinoenstatite [24]

Protoenstatite [24]

Diopside [24]

Wadsleyite [29]

Ringwoodite [31] Chondrodite [37]

Clinohumite [37]

δCS (ppm)

PQ /MHz

C Q /MHz

η

Population

Assignment

48(1) 61(1) 64(1) 42(1) 46(1) 52(1) 54(1) 64(3) 73(3) 41(1) 45(1) 51(1) 54(1) 64(3) 75(3) 39(1) 52(1) 66(3) 63(1) 72(2) 84(2) 38(1) 65(1) 66(1) 76(1) 63(1) 52(1) 59(2) 60(2) 63(1) 25(3) 49(1) 52(1) 59(2) 60(2) 61(2) 64(1) 65(1) 25(3)

2.8(1) 2.4(1) 2.6(1) 2.8(1) 2.9(1) 3.0(1) 3.0(1) 4.3(1) 4.9(1) 2.8(1) 2.8(1) 3.0(2) 3.0(2) 4.3(3) 4.8(3) 2.7(1) 2.8(1) 4.3(2) 2.8(1) 4.3(2) 2.7(2) 1.3(1) 3.8(1) 4.4(1) 5.6(1) 4.9(1) 2.7(1) 2.3(2) 2.3(2) 2.6(1)

2.8(1) 2.5(1) 2.5(1)

0.3(1) 0.2(1) 0.4(1)

1 2 1

3.8(1) 4.4(1) 4.9(2) 4.9(2) 2.7(1) 2.3(1) 2.3(1) 2.5(2) 6.6(1) 2.7(1) 2.7(2) 2.4(1) 2.4(2) 2.3(2) 2.4(1) 2.5(2) 7.0(1)

0.3(1) 0.2(1) 0.9(1) 0.0(2) 0.2(1) 0.3(1) 0.2(1) 0.3(1) 0.1(2) 0.2(1) 0.2(2) 0.2(1) 0.2(1) 0.1(2) 0.2(1) 0.3(1) 0.2(2)

1 4 2 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1

O1 O3 O2 O21 O22 O11 O12 O31 O32 O21 O22 O11 O12 O31 O32 O2 O1 O3 O2 O3 O1 O1 O4 O3 O2 O1 O4 O2 O3 O1 O5 O1 O5 O8 O7 O3 + O4 O6 O2 O9

2.7(1) 2.7(1) 2.4(2) 2.3(2) 2.3(2) 2.5(1) 2.5(1)

clinoenstatite (P21/c) is monoclinic [27]. The related mineral diopside, MgCaSi2 O6 , is also monoclinic (C2/c) [28]. In both orthoenstatite and clinoenstatite there are two types of silicate chains, whilst in protoenstatite and diopside the two chains are related by symmetry reducing the number of distinct O species. The polymorphs may be prepared from 17 O-enriched oxides either at ambient pressures or using a piston cylinder apparatus [24].

The 17 O MAS NMR spectra of the pyroxene minerals (shown in Figure 3) contain broad and complex resonances centered between 30 and 50 ppm, with little indication of the number and nature of crystallographically distinct oxygen species present [24]. MQMAS yields high-resolution isotropic spectra, as displayed in Figure 3, which reveal the presence of many distinct oxygen species [24]. Orthoenstatite and clinoenstatite exhibit very similar spectra with six distinct species present.

Solid-State 17 O NMR of High-Pressure Silicates

Part III

MAS

MQMAS NMR of Upper Mantle Silicates 1555

MQMAS

a

b

* *

c

* *

d

150

100

50

0

−50

−100

80

δ (ppm)

60

40

20

0

−20

δ1 (ppm)

Fig. 3. 17 O (54.2 MHz) MAS and isotropic triple-quantum MAS NMR spectra of (a) orthoenstatite (MgSiO3 ), (b) clinoenstatite (MgSiO3 ), (c) protoenstatite (MgSiO3 ), and (d) diopside (CaMgSi2 O6 ). Nominal levels of isotopic enrichment are (a–c) 35% and (d) 29%.

In each case, four of the resonances possess small CQ values indicative of non-bridging oxygens, whilst two have the increased chemical shift and quadrupolar interactions characteristic of bridging species. In protoenstatite, owing to the increased symmetry only three resonances are present, one bridging and two non-bridging, although an impurity phase of clinoenstatite is also detected. Diopside also displays three distinct oxygens species, but two are

overlapped in the spectrum shown in Figure 3 (recorded at 9.4 T). Measurements at different field strengths allow these species to be separated. The values of the quadrupolar and chemical shift interactions for each species, along with their assignments are given in Table 1 [24]. The different chemical shifts observed in diopside are consistent with deshielding due to the increase in effective cation radius [17,24].

Materials Science

Part III

STMAS NMR of Dense Silicate Phases While α-Mg2 SiO4 in the form of olivine is the main component of the Earth’s upper mantle, the increase in pressure at greater depths results in the formation of the β- and γ-polymorphs, wadsleyite, and ringwoodite. These minerals are thought to be the major constituents of the transition zone (as shown in Figure 1), with the α−β transition responsible for the seismic discontinuity observed at a depth of 410 km and the transition from γ-Mg2 SiO4 to MgSiO3 (perovskite) responsible for the discontinuity around 660 km [1,2]. The β–γ transition may well contribute to the less pronounced discontinuity around 520 km. These dense phases are of particular interest because of their potential as the repository for the largest reservoir of water in the Earth. Before studies of hydration can be undertaken, however, the anhydrous counterparts must be fully characterized. Although high-resolution 17 O NMR spectra of forsterite were obtained easily using MQMAS [19], the sensitivity of this experiment, particularly at high MAS rates, is not sufficient for the study of wadsleyite and ringwoodite. These phases, found at much greater depths in the Earth, must be synthesized at much greater pressures (16 and 20 GPa) using a multi-anvil press, producing only 5–10 mg of material [29]. Furthermore, sample purity can also be a problem, with pressure or temperature gradients resulting in a mixture of phases, thereby reducing the sensitivity further. Recently, the introduction of the STMAS experiment [12], a method with inherently high sensitivity at both slow and fast MAS rates, has enabled the study of such small amounts of 17 O-enriched material on a reasonable timescale. Although of good sensitivity, STMAS does have stringent experimental requirements, such as a stable MAS rate (to within ±2 Hz) and an accurately adjusted spinning angle ( χ = 54.736◦ ± 0.002◦ ) [12,30]. STMAS has been successfully employed to obtain and assign high-resolution 17 O MAS NMR spectra of both wadsleyite and ringwoodite [29,31]. Figure 4a shows a 17 O MAS NMR spectrum of ∼10 mg of 35% 17 Oenriched wadsleyite [29]. The broad complex lineshape results from the overlap of resonances from distinct oxygen species, which are resolved in the STMAS spectrum shown in Figure 4b. In addition to a diagonal resulting from the autocorrelation of the central transition, four ridge lineshapes are observed with relative intensities approximately 1:4:2:1 in order of increasing δ 1 . As with MQMAS, the values of CQ , η, δCS and PQ may be extracted from either the position or shape of the ridge lineshapes, and are given in Table 1 [29]. Wadsleyite has a spinelloid structure (Imma) and contains Si2 O7 groups, which results in both bridging (O2) and non-bridging (O3 and O4) oxygens, in addition to an oxygen (O1) coordinated by five Mg [32]. The quadrupolar and chemical

a

150

100

50

0

−50

δ (ppm)

b 0

δ1 (ppm)

1556 Part III

20

40

60

80

40

0

δ2 (ppm)

−40

Fig. 4. 17 O (54.2 MHz) MAS NMR of ∼10 mg of 35% isotopically enriched wadsleyite (β-Mg2 SiO4 ). (a) MAS NMR spectrum, (b) two-dimensional STMAS NMR spectrum and corresponding isotropic projection, recorded with a phase-modulated split-t1 shifted echo sequence.

shift information extracted from an STMAS spectrum of a sample of ∼5 mg of 35% 17 O-enriched ringwoodite (with a substantial wadsleyite impurity) are also given in Table 1. Although ringwoodite has a spinel structure (Fd3m) with only a single non-bridging oxygen species [33], high-resolution NMR techniques were required to determine the exact nature of the impurity phases [31].

NMR of Hydrous Magnesium Silicates: Humite Minerals The suggestion that the mantle contains a large volume of water is supported by the potential of nominally anhydrous minerals, such as olivine and wadsleyite, to incorporate significant amounts of water into their structures [4–7]. The minerals of the hydroxyl-humite group, nMg2 SiO4 · Mg(OH)2 , have been proposed as possible models of the defect sites accommodating water in olivine [34]. These minerals can be described (rather simplistically) as comprising of n forsterite-like layers alternating with brucite layers, and so may be viewed as hydrated forms of forsterite. Both chondrodite (n = 2)

Solid-State 17 O NMR of High-Pressure Silicates

NMR of Hydrous Magnesium Silicates: Humite Minerals 1557

Part III

MQMAS

STMAS

a

b

c

50

40

30

20

δ1 (ppm)

50

40

30

20

δ1 (ppm)

Fig. 5. 17 O (54.2 MHz) isotropic triple-quantum MAS and STMAS NMR spectra of (a) forsterite (Mg2 SiO4 ), (b) chondrodite (2Mg2 SiO4 · Mg(OH)2 ), and (c) clinohumite (4Mg2 SiO4 · Mg(OH)2 ). Nominal levels of 17 O enrichment were 35% in all cases.

and clinohumite (n = 4) are monoclinic with space group P21/c, resulting in five (four non-bridging and one hydroxyl) and nine (eight non-bridging and one hydroxyl) distinct oxygen species, respectively [35,36]. Both minerals can be synthesized, isotopically enriched in 17 O, at pressures equivalent to the upper mantle using a piston cylinder apparatus [37]. The 17 O MAS NMR spectra of chondrodite and clinohumite are both extremely similar to that of forsterite, with a broad complex lineshape centered at ∼50 ppm [37]. Distinct oxygen species my be resolved, however, in MQMAS spectra, as shown in Figure 5, which displays isotropic 17 O triple-quantum MAS NMR spectra

of 35% isotopically enriched forsterite, chondrodite, and clinohumite [37]. The spectrum of chondrodite is broadly similar to that of forsterite, with a shift of some resonances and the splitting of the central resonance (δ1 = 36 ppm) into two as a result of the decrease in symmetry. This latter effect is clearer when a full two-dimensional spectrum is considered. The spectrum of clinohumite contains five peaks and can be clearly seen to resemble the summation of the spectra of forsterite and chondrodite, reflecting the (simplified) description of the crystal structure of clinohumite as layers of chondrodite and forsterite in equal proportion. The relative intensities of the peaks, 1:1:2:2:2, indicate that all eight non-bridging oxygens

1558 Part III

Materials Science

Part III

Fig. 6. (a) Conventional and (b) crosspolarized 17 O (54.2 MHz) spin-echo NMR spectra of static samples of (a) chondrodite and (b) clinohumite.

Spin echo

Cross-polarized spin echo

a

b

500

−500

0

δ (ppm)

are observed, but that some are too similar to be resolved at this field strength. A high-resolution 17 O triplequantum MAS NMR spectrum of clinohumite recorded at B0 = 16.4 T was able to resolve seven of the eight distinct species [37], with O3 and O4 still too similar to be distinguished. Although resolution of the non-bridging oxygen species in clinohumite and chondrodite was achieved with MQMAS, there is little evidence of the hydroxyl species in the brucite-like layer in these spectra. Hydroxyl oxygen species are predicted to possess much larger quadrupolar couplings [17] making their observation non-trivial at moderate B0 fields, slow MAS rates and low 1 H decoupling strengths. However, the 17 O spin-echo NMR spectra of static samples of chondrodite and clinohumite shown in Figure 6 [37], display a composite lineshape with both a narrow feature and a broader component typical of an oxygen species with a much larger quadrupolar interaction. The use of cross-polarization, a technique which edits the NMR spectrum on the basis of spatial proximity to 1 H, retains only this broad species, confirming its assignment as the hydroxyl oxygen. The quadrupolar and chemical shift parameters for this species in both chondrodite and clinohumite, given in Table 1, are very similar to those observed for brucite [37]. Owing to the similarity of the two techniques, the isotropic spectra produced by MQMAS and STMAS are expected to be very similar in nature, and for many crystalline materials this is indeed the case [38]. However, the isotropic 17 O STMAS NMR spectra of 35% isotopically enriched forsterite, chondrodite, and clinohumite,

500

0

−500

δ (ppm)

shown in Figure 5, do show significant differences from the MQMAS spectra [38]. For forsterite, the two spectra are extremely similar with the small increase in line width in the STMAS spectrum attributed to a very small misset in the magic angle. However, the STMAS spectrum of chondrodite contains a very broad resonance, in contrast to the sharp peaks observed in the MQMAS spectrum. For clinohumite, some of the resonances observed in the MQMAS spectrum remain sharp whilst others appear significantly broadened. Both chondrodite and clinohumite contain a brucite-like layer with two distinct H sites, each of which is half occupied. The large isotropic displacement parameters associated with both sites suggest significant motion of H relative to the other atoms, either vibrational motion within a cavity or even exchange between the sites [39]. This motion of H modulates the electric field gradient experienced by the oxygens and, hence, the 17 O quadrupolar interaction. For the central transition (and indeed the multiple-quantum transition employed in MQMAS) this modulation is small. However for the satellite transitions, where a first-order quadrupolar interaction is present, the change in frequency may be significant and a “motional broadening” is observed in STMAS spectra ˚ of [38]. In chondrodite, all oxygen species are within 3 A either H1 or H2, and will all most likely be involved in hydrogen bonding, producing a large effect in the spectrum. In clinohumite, the oxygens in the chondrodite-like layer are also to 1 H and display a significant motional broadening. However, those oxygens in the forsterite-like layer are sufficiently far from 1 H that they remain relatively unaffected by motional processes. It is possible to

Solid-State 17 O NMR of High-Pressure Silicates

b

Part III

a

Discussion and Conclusions 1559

80.0 70.0

O δcs (ppm)

Si-O-Si Si-O-3Mg

17

Si-O-2Mg Mg-O-4Mg

60.0 50.0 40.0 30.0 20.0

Mg-O-H

20.0

30.0

10.0 40.0

50.0

17O

c

60.0

70.0

80.0

δcs (ppm)

1.62

1.64

1.66

1.68

1.70

1.72

forsterite chondrodite clinohumite o-enstatite p-enstatite wadsleyite ringwoodite brucite

5.0

PQ / MHz

1.60

Si-O bond length / A

6.0

17O

0.0 1.58

4.0 3.0 2.0 1.0 0.0 1.58

1.60

1.62

1.64

1.66

1.68

1.70

1.72

Si-O bond length / A Fig. 7. (a) Variation of the 17 O isotropic chemical shift (δ CS ) with coordination environment, (b) plot of 17 O isotropic chemical shifts (δ CS ) against coordination environment, and (c) 17 O quadrupolar parameters (PQ ) against average Si–O bond length for a variety of silicate minerals.

observe changes in the line width of the resonances in STMAS spectra of both chondrodite and clinohumite by changing the temperature, confirming the motional origin of the broadening [38].

Discussion and Conclusions Both MQMAS and STMAS may be employed to obtain isotropic 17 O NMR spectra with resolution of distinct oxygen species, and allow extraction of accurate quadrupolar and chemical shift parameters. These values, coupled with a comparison of isotropic spectra between like compounds often allows a complete assignment of the spectrum with reference to the suggested crystal structure. It is then possible to investigate any correlations between

the NMR parameters and structural variables such as bond lengths or number and type of coordinating atoms. Figure 7a shows the ranges of the 17 O isotopic chemical shift (δCS ) with the local coordination environment in the magnesium silicate minerals considered here. Oxygen species with different environments are reasonably well separated, with bridging species found between 65 and 80 ppm and non-bridging species between 40 and 67 ppm. Those non-bridging species coordinated to 1 Si and 2 Mg are found at significantly lower δCS than those with 1 Si and 2 Mg. A further decrease in chemical shift is observed for species with no Si bonds, the O coordinated only by 5 Mg atoms in wadsleyite (∼38 pm) and Mg–O–H species (∼25 ppm). The 17 O isotopic chemical shift can be linked directly to changes in the Si–O bond length (the average bond length for bridging

1560 Part III

Materials Science

Part III

oxygens) as shown in Figure 7b. A general positive correlation is observed with an increase in the Si–O bond length resulting in an increase in δCS . A linear regres˚ with a sion analysis yields δCS /drSi−O = +325 ppm/A, correlation coefficient, R 2 , of 0.816. A slightly better fit (R 2 = 0.822) can be obtained with a second-order polynomial regression. This increase in chemical shift with Si–O bond length appears general for both bridging and non-bridging oxygens. Figure 7c shows the variation of the 17 O quadrupolar parameter, PQ , with Si–O bond length. Unlike the chemical shift, no real correlation is observed and the oxygen species tend to cluster into two distinct groups, non-bridging oxygens between 2 and 3.2 MHz and bridging species between 4 and 6 MHz, in general agreement with literature data for a range of different systems. However, it should be noted that the nonbridging oxygen species in the high-pressure polymorphs of Mg2 SiO4 , wadsleyite, and ringwoodite, show anomalously high PQ values (between 3.8 and 4.8 MHz), much more similar to those expected for bridging species. Work is ongoing both experimentally and through ab initio computer simulations in order to gain a deeper understanding of the origin of these effects. In conclusion, 17 O NMR offers great potential for the study of the silicate minerals that make up the Earth’s crust and mantle as it is very sensitive to the local structural environment. The low natural abundance of 17 O (0.037%) may be overcome through the use of isotopic enrichment, and materials may be synthesized at mantle conditions using either piston cylinder or multi-anvil apparatus. Although the presence of second-order quadrupolar broadening hinders the extraction of information directly from conventional MAS NMR spectra, both MQMAS and STMAS methods can be used to obtain high-resolution spectra where distinct oxygen species are resolved. STMAS is a more sensitive technique and although technically difficult to implement is a vital tool in cases where sensitivity is limited or where only small amounts of material may be available. While removing the anisotropic quadrupolar broadening, both techniques retain the isotropic quadrupolar and chemical shift information enabling, in many cases, the full assignment of the spectra. The correlation of these interactions with local structural parameters, such as bond lengths or coordination environments, will help to provide information in cases where crystal structures are less well known, for example, in the investigation of hydrated minerals. The use of 17 O/1 H double resonance techniques, such as crosspolarization, are also of significant use in the determination and characterization of hydrated oxygen species. The observation of motional broadening in STMAS spectra offers indications that the combination of both MQMAS and STMAS may be able to provide information on the dynamic behavior of hydrated mineral phases. A combination of 17 O NMR methods, therefore, really

does seem to offer an ideal opportunity to study mantle silicate minerals and their hydration processes.

Acknowledgments I would like to thank Dr. Stephen Wimperis (University of Exeter, UK) and Dr. Andrew J. Berry (ANU, Canberra, Australia) who contributed considerably to much of the work described here and to Stefan Steuernagel (Bruker, Karlsruhe, Germany) for access to high-field NMR instruments. I also acknowledge the Royal Society for the award of a Dorothy Hodgkin Fellowship.

References 1. Ringwood AE. Composition and Petrology of the Earth’s mantle. McGraw-Hill: New York, 1995. 2. Katsura T, Ito E. J. Geophys. Res. 1989;94:15663. 3. Navrotsky A. Physics and Chemistry of Earth Materials. Cambridge University Press: Cambridge, 1994. 4. Smyth JR. Am. Mineral. 1994;79:1021. 5. Bolfan-Casanova N, Keppler H, Rubie DC. Earth Planet. Sci. Lett. 2000;182:209. 6. Bell DR, Rossman GR. Science. 1992;255:1391. 7. Kohlstedt DL, Keppler H, Rubie DC. Contrib. Mineral. Petrol. 1996;123:345. 8. Ganapathy S, Schramm S, Oldfield E. J. Chem. Phys. 1982;77:4360. 9. Llor A, Virlet J. Chem. Phys. Lett. 1988;152:248. 10. Samoson A, Lippmaa E, Pines A. Mol. Phys. 1988;65:1013. 11. Frydman L, Harwood JS. J. Am. Chem. Soc. 1995;117:5367. 12. Gan Z. J. Am. Chem. Soc. 2000;122:3242. 13. Kohn SC. Am. Mineral. 1996;81:1523. 14. Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ. Am. Mineral. 2002;87:293. 15. Phillips BL, Burnley PC, Worminghaus K, Navrotsky A. Phys. Chem. Miner. 1997;24:179. 16. Stebbins JF, Kanzaki M. Science. 1991;251:294. 17. Kirkpatrick RJ. Spectroscopic Methods in Mineralogy and Geology. Reviews in Mineralogy, Volume 18. Mineralogical Society of America, Washington, 1988. 18. Hazen RM. Am. Mineral. 1976;61:1280. 19. Ashbrook SE, Berry AJ, Wimperis S. Am. Mineral. 1999;84:1191. 20. Schramm S, Oldfield E. J. Am. Chem. Soc. 1984;106:2502. 21. Mueller KT, Wu Y, Chmelka BF, Stebbins J, Pines A. J. Am. Chem. Soc. 1991;113:32. 22. Mueller KT, Baltisberger JH, Wooten EW, Pines A. J. Phys. Chem. 1992;96:7001. 23. Brown SP, Wimperis S. J. Magn. Reson. 1997;124:279. 24. Ashbrook SE, Berry AJ, Wimperis S. J. Phys. Chem. B. 2002;106:773. 25. Hawthorne FC, Ito J. Can. Mineral. 1977;15:321. 26. Murakami T, Takeuchi Y, Yamanaka T. Z. Kristallogr. 1982;160:299. 27. Ohashi Y, Finger LW. Carnegie Inst. Wash. Year Book. 1973;75:743.

Solid-State 17 O NMR of High-Pressure Silicates

34. M. Kitamura, Kondoh S, Morimoto N, Miller GH, Rossman GR, Putnis A. Nature. 1987;328:143. 35. Taylor WH, West J. Proc. R. Soc. Lond. A1928;117: 517. 36. Gibbs GV, Ribbe PH, Anderson CP. Am. Mineral. 1970;55: 1182. 37. Ashbrook SE, Berry AJ, Wimperis S. J. Am. Chem. Soc. 2001;26:6360. 38. Ashbrook SE, Antonijevic S, Berry AJ, Wimperis S. Chem. Phys. Lett. 2002;364:634. 39. Berry AJ, James M. Am. Mineral. 2001;86:181.

Part III

28. Cameron M, Sueno S, Prewitt CT, Papike JJ. Am. Mineral. 1973;58:594. 29. Ashbrook SE, Berry AJ, Hibberson WO, Steuernagel S, Wimperis S. J. Am. Chem. Soc. 2003;125:11824. 30. Ashbrook SE, Wimperis S. Prog. NMR Spectrosc. 2004;45:53. 31. Ashbrook SE, Berry AJ, Hibberson WO, Steuernagel S, Wimperis S. Am. Mineral. 2005;90:1861. 32. Horiuchi H, Sawamoto H. Am. Mineral. 1981;66:658. 33. Sasaki S, Prewitt CT, Sato Y, Ito E. J. Geophys. Res. 1982;87:7829.

References 1561

1563

Philip J. Grandinetti and Ted M. Clark Department of Chemistry, Ohio State University, 100 West, 18th Avenue, Columbus, OH 43210 USA

As a probe of local structure in network forming oxide glasses, 17 O nuclear magnetic resonance (NMR) spectroscopy has been an invaluable tool [1–10]. The analysis of oxygen in these glasses is vital because their structure predominately consists of well-defined cornersharing SiO4 tetrahedra connected in a continuous infinite three-dimensional network lacking long-range order. The interconnection of two tetrahedra involves a Si–O–Si bond angle, and two dihedral angles, and the variation in these angles is considered to be one of the main sources of disorder in a conventional melt quenched silicate glass. The best approach for measuring Si–O–Si bond angle distributions is to use methods that provide a detailed and direct measurement of the local environment around oxygen. Unfortunately, the information obtainable from X-ray absorption spectroscopies has been limited for light backscattering atoms like oxygen [11]. In contrast, solidstate 17 O NMR, specifically, the 17 O quadrupolar coupling and chemical shift parameters provide a simple and direct probe of electronic structure, and is well suited for measuring the local structure around bridging oxygen [12–21]. With the development of 2D NMR methods that separate and correlate anisotropic and isotropic lineshapes [22–24], it is becoming increasingly possible to obtain local structural information for glasses via 17 O NMR. In 1992, Farnan et al. [4] first demonstrated how the distribution of 17 O NMR parameters can be measured in a silicate glass using 2D dynamic-angle spinning (DAS) NMR without any assumptions about the shape of the NMR parameter distributions. Additionally, they demonstrated that the 17 O quadrupolar coupling parameter distribution can, in principle, be mapped into the Si–O–Si bond angle distribution in the glass. A crucial aspect in these structural determinations is an understanding of the relationship between the measurable NMR parameters and the variations in local structure that influence the NMR parameters. The development of such relationships has been an ongoing task for oxygen in oxide glasses, with significant gains being made by the combination of computational methods and experimental investigation of a variety of oxygen environments in crystalline silicates. For bridging oxygen, a number of structural features play a role in determining the 17 O quadrupolar coupling Graham A. Webb (ed.), Modern Magnetic Resonance, 1563–1568.
C 2008 Springer.

parameters, with the most significant features occurring in the first coordination sphere of the bridging oxygen. In the mid-1980’s Oldfield and coworkers [25–27] clearly established on the basis of their 17 O magic-angle spinning (MAS) NMR measurements that the electronegativity of the cations coordinated to a bridging oxygen atom is a primary factor influencing the magnitude of the Cq -parameter for the bridging oxygen. In a series of ab initio calculations on model clusters focusing on the silicate bridging oxygen linkage Tossell and coworkers [12–14] predicted simple trends for the 17 O quadrupolar coupling parameters with Si–O–Si angle. These trends were later experimentally confirmed by Grandinetti et al. [17] in 17 O DAS measurements on coesite. Vermillion et al. [19] and Clark et al. [20] further refined this understanding by examining the effect of additional coordination of the bridging oxygen by network modifier alkali cations. Their findings, obtained using ab initio methods on model clusters representing typical bridging oxygen environments in lithium, sodium, and potassium silicates, suggested that the previously established trend in 17 O Cq with Si–O–Si angle is systematically shifted to lower magnitudes with increasing number and field strength of coordinating alkali cations. They also found that the previously established trend [12–15] in 17 O quadrupolar coupling asymmetry parameter, ηq , with Si– O–Si angle is systematically shifted to higher values by the presence of one coordinating alkali cation, and only slightly shifted to higher values by the presence of two coordinating alkali cations. In two related papers, Clark and Grandinetti [18, 28] used ab initio methods to study a number of clusters with the coordinating cations varied from Group III to Group VI and from Periods 2 to 4, while simultaneously varying the network forming cation–oxygen distance. A general trend, shown in Figure 1A, was observed that the magnitude of Cq increases linearly with increasing network forming cation–oxygen bond distance, and also increases with cation group number, as shown Figure 1B. These findings suggest that group number and cation-bridging oxygen distance can serve as a better predictor of the bridging oxygen quadrupole coupling constant than electronegativity differences [26, 27]. The observation that the magnitude of Cq increases linearly with increasing network forming cation–oxygen

Part III

The Structure of Oxide Glasses: Insights from 17O NMR

1564 Part III

Materials Science

C

−16

Cq (MHz)

−14 −12 Ge

−10

Si

−8 −6 −4 1.2

1.4

1.3

1.5

1.6

1.7

1.8

d(X–O) (Å)

(B)

0

Period 3 −5

Cq (MHz)

Part III

(A) −18

−10

Period 4

−15

Peroid 2

−20 −25 2

3

4

5

6

Group Number Fig. 1. (A) Comparison of bridging oxygen 17 O Cq values for the clusters (OH)3 T–O–T(OH)3 , where T = carbon, silicon, and germanium, as a function of d(T–O). (B) Bridging oxygen Cq dependence on the group number of coordinating cations for each period. Cq values are calculated for identical cation–oxygen distances and a bridging oxygen angle of 180◦ .

bond distance was investigated in more detail by Clark and Grandinetti for bridging oxygen atoms in silicates [21]. In this work, ab initio calculations were performed for model clusters to examine the dependence of 17 O quadrupolar coupling parameters on both Si–O distance and Si–O–Si angle. This work demonstrated that the 17 O quadrupolar coupling constant is dependent on both of these structural features, while ηq is primarily dependent on the Si– O–Si angle. Also, it is noteworthy that the strong linear dependence of Cq on Si–O distance helped explain the rather different angular trends for Cq found in coesite [17] and ferrierite [29], since the Si–O–Si angle and Si–O

distance correlations for these two materials are quite different [21]. Relationships between 17 O quadrupolar coupling parameters and structure in silicates have also been examined using periodic density functional theory (DFT) calculations. These calculations, which are capable of describing quite accurately a system’s electronic structure, have been performed for crystalline sodium silicates, the siliceous zeolite ferrierite, and also for plausible sodium silicate glasses that were generated using molecular dynamics [30, 31]. These calculations generally support the previously established trends, i.e. Cq decreases in magnitude with decreasing Si–O–Si angle and Si–O distance, whereas ηq depends only on Si–O–Si angle. Additionally, these investigations found no dependence of the bridging oxygen Cq upon the number of coordinating alkali cations, a finding in contrast to that reported using modeling clusters [19, 20]. Overall, it has become clear that first coordination sphere structural features that appear to be most important in determining the 17 O quadrupolar coupling parameters of the bridging oxygen are the nature of the two coordinating network forming cations, the T–O–T linkage angle, the T–O bond distances, and the nature and number of coordinating network modifier cations. Contributions from beyond the first coordination sphere of the bridging oxygen appear to be secondary in importance. For example, Xue and Kanzaki [16] performed ab initio calculations employing clusters expanded out to four coordination spheres to model each of the silicate bridging oxygen linkages in coesite and obtained a slightly improved agreement with the experimental trends, with corrections on the order of a few percent in the 17 O Cq and ηq values. Equations proposed to summarize the important structural features for the determination of the 17 O quadrupolar coupling parameters are Cq (, n M ) = a

cos  1 + 2 cos  − 1

α + n M CqM

+ m d (dT O − dTo O )

β 1 cos  ηq () = b − + ηqM (n M ), 2 cos  − 1

(1) (2)

where  is the Si–O–Si bond angle, dT O is the average silicon–oxygen bond distance, and the terms n M CqM and

ηqM (n M ) account for contributions from modifier cations coordinated to the bridging oxygen. These expressions are not unique in their ability to describe these relationships, and other expressions have been used previously [4,14,15]. Clark and Grandinetti [21] performed a least-squares fit of Equations (1) and (2) to the experimental values for coesite [17], cristobalite [32], and α-quartz [33] as

The Structure of Oxide Glasses

Predicted Si–O–Si Angle (°)

180 170 160 150 140 130 120 120 130 140 150 160 170 180 Si–O –Si Angle (°)

Predicted Si–O Distance (Å)

(B)

Predicted Si–Si Distance (Å)

(C)

1.65 1.63 1.61 1.59 1.57 1.55 1.55 1.57 1.59 1.61 1.63 1.65 Si–O Distance (Å) 3.3

Ferrierite Cristobalite

3.2

Quartz Coesite

3.1 3.0 2.9 2.9

3.2 3.0 3.1 Si–Si Distance (Å)

3.3

Fig. 2. Comparison between 17 O quadrupolar coupling predicted (A) Si–O–Si angle, (B) Si–O distance, and (C) Si–Si distance with corresponding quantities reported from X-ray crystallography [40–42]. Solid diagonal lines represent perfect agreement. Best fit parameters for Equations (1) and (2) are a  = ˚ d ◦ = 1.654A, ˚ −6.53 MHz, α  = 1.80, m d = −12.86 MHz/A, TO b = 4.73, β = 1.12.

well as the ab initio predicted values for ferrierite [29] and obtained the parameter values given in the caption of Figure 2. The resulting parameterized equations describe the relationships between the 17 O quadrupolar coupling

parameters and the relevant structural features in the first coordination sphere for a bridging oxygen atom in the absence of modifier cations. The precision of these relationships can be seen in Figure 2, where the Si–O–Si angles and Si–O distances predicted by these expressions are shown versus the actual values. Close agreement between predicted and reported values are demonstrated, even though only two structural variables are considered, i.e. the average bond distance and the bond angle. Here, the Si–O–Si angles in Figure 2A are predicted from ηq alone and, with the Si–O–Si angle determined, Equation (1) is used to predict the Si–O distance from Cq , as shown in Figure 2B. After having determined the Si–O distance for a given angle, it is also possible to determine the Si–Si distance. It is noteworthy that not only the quadupolar coupling constant and asymmetry parameter can be used to measure the Si–O–Si angle and Si–O distance, but also the simultaneous measurement of Cq and ηq for each bridging oxygen site can be used to obtain the correlation between Si–O– Si angle and Si–O distance. This has significant impact for determining two-dimensional structural distributions around bridging oxygen in silicate glasses. This approach requires a good measurement of the quadrupolar asymmetry parameter, which has been historically been more difficult to measure than the quadrupolar coupling constant. Fortunately, this situation has been improved with the development of techniques such as Rotor Assisted Population Transfer (RAPT) [34–37] for enhancing the sensitivity of the MAS detected DAS [17,38,39]experiment, which allows overlapping anisotropic 17 O central transition lineshapes to be separated. Figure 3A is the recently reported 17 O RAPT/MASdetected DAS spectrum of silica glass [43]. From a least-square analysis of the 17 O 2D DAS spectrum a three-dimensional histogram, correlating the δcs (chemical shift), Cq , and ηq of 17 O, was obtained. A projection of this three dimensional histogram, showing the two-dimensional correlation between Cq and ηq for silica glass, along with corresponding one-dimensional projections is shown in Figure 3B. With the aid of Equations (1) and (2), the experimental Cq and ηq histograms were mapped into the two-dimensional histogram of the Si–O–Si angle versus the Si–O distance, shown in Figure 3C. This result is the first experimentally measured two-dimensional structural distributions in a glass, and illustrates nearly strong linear correlations between the Si–O–Si angle and Si–O distance in silica glass. Significant insights into the structure of silicate glasses are now made possible by an analysis using this methodology. For example, in the case of silica glass, it is remarkable that a strong positive correlation is observed between Si–O distance and Si–O–Si bond angle. This trend was unexpected since it is opposite of that generally found in crystalline SiO2 polymorphs, as shown in

Part III

(A)

The Structure of Oxide Glasses 1565

1566 Part III

Materials Science

Part III

Isotropic Dimension (ppm)

(A) Cq = 5.40 MHz ηq = 0.12

-80

Cq = 5.27 MHz ηq = 0.13

-60 -40

Cq = 5.13 MHz ηq = 0.14

-20

Cq = 5.01 MHz ηq = 0.16

0 20

Cq = 4.88 MHz ηq = 0.19

40 40

0

Cq = 4.73 MHz ηq = 0.21

-40 -80 -120 -160

MAS Dimension (ppm)

100 0 -100 -200 MAS Dimension (ppm)

(B) Quadrupolar Coupling Constant (MHz)

-7 -6 -5

180° 140° 150° 130°

120°

1.63Å 1.61Å 1.59Å 1.57Å 1.55Å 1.53Å

-4 -3 0.0

0.4

0.6

0.8

1.0

Alpha-Quartz Cristobalite

(C)

Trydimite (OP)

180

Ferrierite Trydimite (MP)

170

Coesite

Si–O–Si (°)

Fig. 3. (A) Experimental 2D 17 O RAPT/MAS-detected DAS spectrum of SiO2 at 9.4 T along with experimental 1D projections onto the MAS and isotropic dimensions. Shown on the right are selected experimental cross sections along with best fit simulations. (B) Two-dimensional histograms along with the corresponding one-dimensional projection showing the correlation between Cq and ηq . Grid lines shown were obtained from Equations (1) and (2) by varying Si–O distance with Si–O–Si angle held constant and the Si–O–Si angle with the Si–O distance held constant. (C) 2D histograms of Si–O–Si angle versus Si–O distance for silica glass derived from NMR parameter distributions, and for various crystalline polymorphs.

0.2

Quadrupolar Asymmetry Parameter

160 150 140 130 120 1.45

1.50

1.55

1.60

Si–O Distance (Å)

The Structure of Oxide Glasses

References 1. Jellison GE, Panek LW, Bray PJ, Rouse GB. Determinations of structure and bonding in vitreous B2 O3 by means of 10 B, 11 B, and 17 O, NMR. J. Chem. Phys. 1977;66:802. 2. Geissberger AE, Bray PJ. Determinations of structure and bonding in amorphous SiO2 using 17 O NMR. J. Non-Cryst. Solids. 1983;54:121. 3. Youngman RE, Haubrich ST, Zwanziger JW, Janicke MT, Chmelka BF. Short- and intermediate-range structural ordering in glassy boron oxide. Science. 1995;269:1416. 4. Farnan I, Grandinetti PJ, Baltisberger JH, Stebbins JF, Werner U, Eastman MA, Pines A. Quantification of the disorder in network-modified silicate glasses. Nature. 1992;358:31–5. 5. Wang S, Stebbins JF. On the structure of borosilicate glasses: A triple-quantum magic-angle spinning 17 O nuclear magnetic resonance study. J. Non Cryst. Solids. 1998;231:286–90. 6. Zhao P, Kroeker S, Stebbins JF. Non-bridging oxygen sites in barium borosilicate glasses: results from 11 B and 17 O NMR. J. Non-Cryst. Solids. 2000;276:122–31. 7. Florian P, Vermillion KE, Grandinetti PJ, Farnan I, Stebbins JF. Cation distribution in mixed alkali disilicate glasses. J. Am. Chem. Soc. 1996;118:3493–7. 8. Stebbins JF, Oglesby JV, Xu Z. Disorder among networkmodifier cations in silicate glasses: New constraints from triple-quantum 17 O NMR. Am. Mineral. 1997;82:1116–24. 9. Stebbins JF, Lee SK, Oglesby JV. Al–O–Al oxygen sites in crystalline aluminates and aluminosilicate glasses: High-resolution oxygen-17 NMR results. Am. Mineral. 1999;84:983–6. 10. Lee SK, Musgrave CB, Zhao P, Stebbins JF. Topological disorder and reactivity of borosilicate glasses: Quantum chemical calculations and 17 O and 11 B NMR study. J. Phys. Chem. B. 2001;105:12583–95. 11. Elliott SR. Non-diffraction spectroscopic probes of the structure of amorphous solids. J. Non-Cryst. Solids. 1990;123:149.

12. Tossell JA, Lazzeretti P. Ab Initio calculations of oxygen nuclear quadrupolar coupling constants and oxygen and silicon NMR shielding constants in molecules containing Si–O bonds. Chem. Phys. Lett. 1987;112:205. 13. Tossell JA, Lazzeretti P. Calculation of NMR parameters for bridging oxygens in H3 T–O–T’H3 linkages (T,T’ = Al, Si, P), for oxygen in SiH3 O− , SiH3 OH and SiH3 OMg+ and for bridging fluorine in H3 SiFSiH+ 3 . Phys. Chem. Minerals. 1988;15:564. 14. Lindsay CG,Tossell JA. Ab Initio calculations of 17 O and n T NMR parameters (n T = 31 P, 29 Si) in H TOTH dimers 3 3 and T3 O9 trimeric rings. Phys. Chem. Minerals. 1991;18: 191. 15. Sternberg U. The bond-angle dependence of the asymmetry parameter of the oxygen-17 electric field gradient tensor. Solid State NMR. 1993;2:181. 16. Xue X, Kanzaki M. An ab initio calculation of 17 O and 29 Si NMR parameters for SiO2 polymorphs. Solid State NMR. 2000;16:245–259. 17. Grandinetti PJ, Baltisberger JH, Werner U, Pines A, Farnan I, Stebbins JF. Solid-state 17 O magic-angle and dynamicangle spinning NMR study of coesite. J. Phys. Chem. 1995;99:12341–8. 18. Clark TM, Grandinetti PJ. Factors influencing the 17 O quadrupole coupling constant in bridging oxygen environments. Solid State NMR. 2000;16:55–62. 19. Vermillion KE, Florian P, Grandinetti PJ. Relationships between bridging oxygen 17 O quadrupolar coupling parameters and structure in alkali silicates. J. Chem. Phys. 1998;108(17):7274–7285. 20. Clark TM, Grandinetti PJ, Florian P, Stebbins JF. An 17 O NMR investigation of crystalline sodium metasilicate: Implications for the determinations of local structure in alkali silicates. J. Phys. Chem. B. 2001;105:12257–65. 21. Clark TM, Grandinetti PJ. Dependence of bridging oxygen O17 quadrupolar coupling parameters on Si–O distance and Si–O–Si angle. J. Phys. Condensed Matter. 2003;15:S2387– 95. 22. Bax A, Szeverenyi NM, Maciel GE. Correlation of isotropic shifts and chemical shift anisotropies by two-dimensional Fourier-transform magic-angle hopping NMR spectroscopy. J. Magn. Reson. 1983;52:147. 23. Terao T, Fujii T, Onodera T, Saika A. Switching-angle sample -spinning NMR spectroscopy for obtaining powder-pattern-resolved 2D spectra: Measurements of 13 C chemical-shift anisotropies in powdered 3,4dimethoxybenzaldehyde. Chem. Phys. Lett. 1984;107: 145. 24. Mueller KT, Sun BQ, Chingas GC, Zwanziger JW, Terao T, Pines A. Dynamic-angle spinning of quadrupolar nuclei. J. Magn. Reson. 1990;86:470. 25. Schramm S, Oldfield E. High-resolution oxygen-17 NMR of solids. J. Am. Chem. Soc. 1984;106:2502. 26. Timken HKC, Janes N, Turner GL, Lambert SL, Welsh LB, Oldfield E. Solid-state oxygen-17 nuclear magnetic resonance spectroscopic studies of zeolites and related systems. J. Am. Chem. Soc. 1986;108:7236. 27. Timken HKC, Schramm SE, Kirkpatrick RJ, Oldfield E. Solidstate oxygen-17 nuclear magnetic resonance spectroscopic studies of alkaline earth metasillicates. J. Phys. Chem. 1987;91:1054–8.

Part III

Figure 3C. Also, the resulting Si–O–Si bond angle distribution, which peaks at 147◦ with a standard deviation of 3.8◦ , is noteworthy for its narrowness in comparison with distributions determined by diffraction or Si-29 NMR measurements [44–46]. In conclusion, 17 O NMR is proving to be an extremely robust probe of structure in silicate glasses. As the relationship between the 17 O NMR parameters and structure has become increasingly well-understood, it is now possible to apply this knowledge for the interpretation of the two-dimensional NMR spectra of disordered materials. This has led to the experimental determination of correlated structural distributions for oxygen sites in glasses, a significant achievement that is not presently possible by other experimental techniques. With the continued application of this methodology, it will be possible to gain a more complete understanding of many glasses of scientific and technological importance.

References 1567

1568 Part III

Materials Science

Part III

28. Clark TM, Grandinetti PJ. Relationships between bridging oxygen 17 O quadrupolar coupling parameters and structure in germanates. J. Non-Cryst. Solids. 2000;265: 75–82. 29. Bull LM, Bussemer B, Anupold T, Reinhold A, Samoson A, Sauer J, Cheetham AK, Dupree R. A high-resolution 17 O and 29 Si NMR study of zeolite siliceous ferrierite and ab initio calculations of NMR parameters. J. Am. Chem. Soc. 2000;122:4948–58. 30. Charpentier T, Ispas S, Profeta M, Mauri F, Pickard CJ. First-principles calculation of 17 O, 29 Si, 23 Na NMR spectra of sodium silicate crystals and glasses. J. Phys. Chem. B. 2004;108:4147–61. 31. Profeta M, Mauri F, Pickard CJ. Accurate first principles prediction of 17 O NMR parameters in SiO2 : Assignment of the zeolite ferrierite spectrum. J. Am. Chem. Soc. 2003;125:541–8. 32. Spearing DR, Farnan I, Stebbins JF. Dynamics of the alpha–beta phase transitions in quartz and cristobalite as observed by in situ high temperature Si-29 NMR and O-17 NMR. Phys. Chem. Miner. 1992;19(5): 307–21. 33. Dupree R. Nuclear magnetic resonance as a structural probe of SiO2 . In: R. Devine, JP. Durand, E. Dooryhee (Eds), Structure and Imperfections in Amorphous and Crystalline SiO2 , John Wiley, pp. 107–20, 2000. 34. Kwak H-T, Prasad S, Clark TM, Grandinetti PJ . Selective suppression and excitation of solid-state NMR resonances based on quadrupole coupling constants. J. Magn. Reson. 2003;160:107–13. 35. Yao Z, Kwak H-T, Sakellariou D, Emsley L, Grandinetti PJ. Sensitivity enhancement of the central transition NMR signal of quadrupolar nuclei under magic-angle spinning. Chem. Phys. Lett. 2000;327:85–90.

36. Kwak H-T, Prasad S, Yao Z, Grandinetti PJ, Sachleben JR, Emlsey L. Enhanced sensitivity in RIACT/MQ-MAS NMR experiments using rotor assisted population transfer. J. Magn. Reson. 2001;150:71–80. 37. Prasad S, Kwak HT, Clark T, Grandinetti PJ. A simple technique for determining nuclear quadrupole coupling constants using RAPT solid-state NMR spectroscopy. J. Am. Chem. Soc. 2002;124(18):4964–5. 38. Chmelka BF, Mueller KT, Pines A, Stebbins J, Wu Y, Zwanziger JW . Oxygen-17 NMR in solids by dynamic-angle spinning and double rotation. Nature. 1989;339:42–3. 39. Mueller KT, Wooten EW, Pines A. Pure-absorption-phase dynamic-angle spinning. J. Magn. Reson. 1990;92:620. 40. Geisinger KL, Spackman MA, Gibbs GV. Exploration of structure, electron density distribution, and bonding in coesite with fourier and pseudo atom refinement methods using single-crystal X-ray diffraction data. J. Phys. Chem. 1987;91:3237. 41. Downs RT, Palmer DC. The pressure behavior of alphacristobalite. Am. Mineral. 1994;79(1–2):9–14. 42. Kihara K. An X-ray study of the temperature dependence of the quartz structure. Eur. J. Mineral. 1990;2(1):63–77. 43. Clark TM, Grandinetti PJ, Florian P, Stebbins JF. Correlated structural distributions in silica glass. Physical Rev. B. 2004;70. 44. Mozzi RL, Warren BE. The structure of vitreous silica. J. Appl. Cryst. 1969;2:164–72. 45. Poulsen HF, Neuefeind J, Neuman H-B, Schneider JR, Zeidler MD. Amorphous silica studied by high energy X-ray diffraction. J. Non Cryst. Solids. 1995;188:63–74. 46. Mauri F, Pasquarello A, Pfrommer BG, Yoon Y-G, Louie SG. Si–O–Si bond-angle distribution in vitreous silica from first-principles 29 Si NMR analysis. Phys. Rev. B. 2000; 62(8):4786–9.

1569

Jacco D. van Beek, Lilyane Beaulieu and Beat H. Meier Physical Chemistry, ETH Zurich, CH-8093 Zurich, Switzerland

Introduction The focus of the present article is the description of some NMR methods that can be used to structurally characterize partially disordered biopolymers in solid phase. Silk is not only an attractive and important material of this class, but also it turns out to be a convenient compound to be studied by NMR spectroscopy because it is quite easy to obtain the relatively large sample amounts needed (in the order of 10 mg) with amino-acid specific isotope labelling, in particular with 2 H, 13 C and 15 N. In the following, we will give a short introduction to silks, followed by a description of selected NMR techniques and the results of their application to silks, in particular to the silk of the Eri silkworm Samia cynthia ricini (S.c. ricini). Silks are fibrous proteins of considerable economic and scientific interest. They have found utility as textile fibre since thousands of years because of their strength, elasticity, luster, absorbency and affinity for dyes. The material is made by various animals for different biological purposes. Many silks exhibit interesting mechanical properties, e.g. a high toughness comparable to Kevlar combined with an extensibility of 30% or an extreme extensibility of more than 200% [1]. In our days, many new applications for these types of biopolymers are targeted [2–6]. The primary structure of silks is typically highly repetitive and rich in glycine and alanine [7–13]. Spider dragline silk and other silks like the one from the Eri silkworm, S.c. ricini, show a primary sequence of the silk protein with poly-alanine repeats of 6–12 monomers in between glycine-rich segments. In this article, we concentrate in particular on these types of silks and the spectra presented are all from S.c. ricini. The commercially most important silk from the domesticated silkworm Bombyx mori does not fall into this class and we refer the reader to a recent review of Zhao and Asakura [14]. Silks are partially disordered materials, and therefore, particularly interesting for solid-state NMR studies because there are few other methods for their structural characterisation. The linewidth in “high-resolution” magicangle spinning (MAS) solid-state experiments turns out to be in the order of 5–10 ppm, which is typically an order Graham A. Webb (ed.), Modern Magnetic Resonance, 1569–1579.
C 2008 Springer.

of magnitude more than obtained with microcrystalline proteins [15–18] and amyloid fibres [19,20] with the corresponding loss in the signal-to-noise ratio. Therefore, not all the benefits of MAS spectroscopy, as obtained for more highly ordered materials, can be realised on silk samples. In particular, it is often not possible to make distance measurements to determine the local structure because the signals from different sites are not spectrally resolved, except for special pairs introduced by selective labelling [21–23]. However, the measurement of backbone torsion angles is an attractive alternative. The knowledge of the (φ,ψ) torsion angle pair for each residue in principle fully determines the backbone structure. In practice, however, only secondary structure information can be obtained due to the cumulative errors and long-range distance constraints are necessary to address higher-than-secondary structural questions. In addition to the native fibres, we will also discuss studies of silk films. These films are formed when the liquid fibroin is directly casted from the posterior silk glands of the silkworm and can be used for immobilisation technology and especially as an immobilisation matrix for enzymes. Their functionality, including permeability, activity of entrapped enzymes and mechanical integrity depend on the orientation of the polymer chains and their conformation. Therefore, it is of interest to understand the relationship between the molecular structure and the diverse macroscopic particularities and utilities of this biopolymer.

The NMR Measurements of Torsion Angles Torsion angles can be measured through the orientational correlation of spin interactions which have a known orientation of the principal axis of the spatial tensor with respect to the molecular coordinate system. As an example, we use the 13 C chemical-shift tensor of the backbone carbonyl functionality of selected amino acids. The orientation of this tensor with respect to the molecular frame at the carbonyl site is known to a precision of a few degrees and is indicated schematically in Figure 1. Therefore, if we are able to selectively transfer the magnetisation between a

Part III

Studies of the Local Structure of Silk Using Solid-State NMR

1570 Part III

Materials Science

Part III Fig. 1. Orientation of the carbonyl chemical-shielding tensor (yellow ellipses) with respect to the molecular coordinate system in a polypeptide and its relation to the backbone torsion angles. The following empirical rules apply [55–59]: The principal axis associated with the most shielded CSA principal value σ33 is perpendicular to the peptide plane. The principal axis associated with the intermediate component of the CSA tensor σ22 is almost colinear with the C=O vector. More precisely, for proteins, experimental determination of the angle between the bound direction and the principal axis σ22 falls into a range 0–12◦ . (See also Plate 122 on page XVII in the Color Plate Section.)

selected pair of carbon spins, or to excite double-quantum coherence involving the two selected spins, the twodimensional CSA–CSA correlation spectrum would fully characterize the relative orientation of the two molecular segments described by the angle pair (φ, ψ), as schematically illustrated in Figure 2. The “selected” amino acids mentioned above are usually singled out by isotopic labelling schemes [24–27]. Obviously, a complex material like silk is not composed of isolated two-spin systems and the correlation spectrum will critically depend on the number of spins involved as well as on the experimental methods applied. Coherent 13 C–13 C polarization transfer experiments (e.g. rf-driven spin diffusion or double-quantum excitation/reconversion) typically favour the strongest coupling at the expense of weak coupling, not only through the inverse sixth power dependence of the spectral intensity from the intermolecular distance but also, in addition, through the effect of “dipolar truncation”

[28]. They can in fact be set up, such that predominantly next–neighbour interactions are observed, which is particularly useful for double-quantum selection in the double-quantum–single-quantum correlation experiment (DOQSY) [29–32]. Nevertheless, even in this favourable situation, the evaluation of the structural parameters (φ,ψ) from the spectra is not trivial because of disorder in the material: there is no single angle-pair (φ,ψ) which could explain the spectra but there is a distribution of such angles which must be characterised. Notwithstanding these complications, a “model free” determination of the parameter distributions is often possible [30,31,33,34]. Alternatively, information about (φ,ψ) may also be obtained from the chemical-shift distribution of the 13 C resonances in a MAS spectrum using empirical rules [27,35–38] or from initial-rate studies in spin-diffusion spectra (vide infra) [39]. Experimental DOQSY spectra of alanine-labelled silk from S.c. ricini fibre and cast-film samples are presented in Figure 3 along with the simulated spectra and the distribution in (φ,ψ), which leads to the simulated spectrum [30]. The resulting distribution in (φ,ψ) allows for a relatively detailed characterisation of the alanine sites. Similar work for the glycine sites will is described elsewhere [31] and yielded results comparable to the measurements on spider dragline silk [40]. Therefore, S.c. ricini seems, in fibrous form, to have a similar structure to that observed for spider dragline silk depicted in Figure 4, while it adapts rather a different form in the cast film with α-helical structures for the alanine-rich parts [30] and with substantially disordered glycine-rich parts. [31]

Geometrical Information on the Molecular-to-Nanometer Scale Proton-driven “incoherent” 13 C–13 C polarisation transfer is another option to probe the relative orientation of molecular segments. Proton-driven spin-diffusion (PDSD) on static samples is a well established technique for determining local structure (for a review see [41,42]). Because the length of the spin-diffusion period of the experiment is only limited by T1 relaxation, it can be up to 10 s and the method allows to probe structures with sizes greatly exceeding a single bond length. The polarisation transfer follows a kinetic master equation in this experiment [43] and the dependence on the internuclear separation is between the inverse third and sixth powers of the distance, depending on the details of the experiment [42]. In the limit of quasi-equilibrium (QE), when the spectrum does not depend on the mixing time anymore, the spectrum becomes particularly easy to interpret as it reflects only the relative orientation of the CSA tensors of the spins that are correlated, and no longer the distances. Figure 5 shows a

Local Structure of Spider Silk Using Solid-State NMR

Geometrical Information on the Molecular-to-Nanometer Scale 1571

B0 Ωj Ωi ω2 ω1

powder spectrum

Fig. 2. Principles of two-dimensional CSA–CSA correlation spectroscopy using, e.g. PDSD. The single-crystal spectrum (B) depends on the relative orientation of the two CSA tensors symbolised in (A) and on the crystal orientation. For a static sample the powder spectrum (C) depends only on the relative tensor orientation (and, of course, on the principal values). The illustration uses the correlation of two axially symmetric tensors that yields only a single Euler angle β. For general tensors, all three Euler angles can be determined. (See also Plate 123 on page XVIII in the Color Plate Section.)

β

ω2

comparison of the DOQSY and PDSD experiments for the proposed labelling strategy, as a function of the backbone torsion angles (φ,ψ). In PDSD, spins not involved in magnetisation exchange processes on the time scale of the mixing time give rise to a (often dominant) diagonal signal in the 2D spectrum. This can overshadow the off-diagonal signal, which contains the structural information. In biological materials like silk, where the degree of isotopic labelling may be considerably less than 100% (e.g. our S.c. ricini examples are at around 15% for the enriched amino acid), this feature leads to a complication of the spectral interpretation and, if one is interested in determining structural information on a molecular scale (like torsion angles) it is beneficial to apply the DOQSY approach mentioned above. This method actively selects out the signals from close neighbours and suppresses the signals from spins not involved in a double-quantum coherence. Conversely, the PDSD experiment, is, for longer mixing times, sensitive to longer distances exceeding the ones probed by DOQSY by one or two orders of magnitude. For fully disordered materials, the PDSD QE structure would be a pure “amorphous” spectrum where each cross section along t1 as well as t2 has the same shape (but not the same integral) [42]. In materials with sizeable

domains (consisting of several monomers in each spatial dimension), the QE spectra can characterise both the local structure within a domain and the domain size [41]. If well-isolated domains exist, they can lead to a QE where the spectra do not change appreciably with the mixing time anymore. However, if the mixing time would be increased drastically, the “amorphous” spectra would again result. A disadvantage to the PDSD experiment on static samples is the slow exchange of magnetisation coupled to a strong dependence on the frequency difference of the corresponding diagonal signals, which may lead to spectra difficult to interpret. The rate constant for the magnetisation exchange can be significantly enhanced, and made rather independent of the frequency-difference, if the sample is rotated slowly (10–100 Hz) during the mixing period of the experiment (slow magic-angle spinning (SMAS; [44]). The SMAS-PDSD experiment can also be performed with short mixing times in the initial rate regime. In this approach the buildup of each point in the spectrum, as a function of the mixing time, is followed. The local structure can then be obtained by fitting a structural model to the data [39]. Since the isotopic labelling degree in

Part III

single crystal spectrum ω1

1572 Part III

Materials Science

Part III

Fiber

PPM

200

Film

A

200

250

250

300

300

350

350

400

400

450

450 300

PPM

200

B

200

100

300

C

200

250

250

300

300

350

350

400

400

450

450 300

200

100

D 100 80 60 %

100

300

PPM

200

100

40

PPM

180

180

E 120

60

60

0

0

-60

-60

-120

-120

-180 -180

20

F

120

ψ (degrees)

200

-120

-60

φ (degrees)

0

-180 -180

0

-120

-60

0

φ (degrees)

Fig. 3. Experimental 2D DOQSY spectra of (A–B) two different silk samples from the silkworm S.c. ricini with (C–D) corresponding to the best fits of these spectra and (E–F) to the distribution functions obtained from the fit. The samples were labelled with 1-13 C alanine and prepared either as a fibre or a cast film. The samples were measured a 13 C resonance frequency of 100 MHz. The grid resolution in (φ,ψ) was 15◦ . (See also Plate 124 on page XIX in the Color Plate Section.)

silk samples is often relatively low and non site-specific, i.e. all glycines or all alanines are labelled to a certain statistical degree, the intra- and inter-chain distances expected are comparable in length. This is in contrast to the DOQSY approach, where one, to good approximation, selectively filters out the signals from spin-pairs with the minimal possible distance. This distance is usually in fact the one between intramolecular neighbours. While a full 3D structural model is required to explain the data for the SMAS-PDSD experiments, the DOQSY Spectra can

Fig. 4. Schematic model for the structure of spider dragline silk. Reproduced from Ref. [40], with permission. (See also Plate 125 on page XIX in the Color Plate Section.)

approximatively be described by two-spin systems. In our experience, the SMAS-PDSD spectra are often not sensitive enough towards the local structure in such samples, partly by the low signal-to-noise (SN) in the vital offdiagonal part of the spectrum in the initial rate regime, and partly by the large number of degrees of freedom of such a 3D model. While it is often possible to exclude certain packing models and torsion angles a unique solution is usually not available. These statistical arguments do not hold, for siteselectively labelled peptides and good results have been obtained using PDSD on labelled model peptides at offmagic-angle spinning (OMAS) conditions [22,23,45–47].

Local Structure of Spider Silk Using Solid-State NMR

180 150 120 90

ψ (degrees)

60 30 0

-30 -60 -90 -120 -150 -180 -180 -150 -120 -90 -60 -30 0 30 60 φ (degrees)

90 120 150 180

-180 -150 -120 -90 -60 -30 0 30 60 φ (degrees)

90 120 150 180

B) 180 150 120 90

ψ (degrees)

60 30 0 -30 -60 -90 -120 -150 -180

Fig. 5. Torsion-angle dependences of the (A) DOQSY and (B) PDSD experiments (at QE) when applied to 13 C=O labelled proteins, assuming an intra-chain contact between neighbouring amino acids. Typical β-sheet and right- and left-handed α-helical regions are marked by orange, dark blue and bright blue boxes, respectively. (See also Plate 126 on page XX in the Color Plate Section.)

Similar to SMAS, the OMAS experiment has a homogenous SD rate constant across the spectrum, but the spectrum is distributed over a smaller spectral area, which leads to a higher signal-to-noise ratio. Figure 6 shows an

example of such spectra obtained in specifically labelled model peptides. In such systems QE may be quite easily reached and the analysis is comparatively simple (typically, a 2-spin simulation with only the torsion angles as variables). The PDSD technique was first applied to static spider dragline silk samples by K¨ummerlen et al. [26] and, by correlating 13 C=O labelled spins, an average value for (φ,ψ) was determined for alanine–alanine and glycine– glycine peptide segments. It was concluded that the segments were part of a β-sheet structure for the alanines and were close to a 31 -helical arrangement for glycine. Whilst the main conclusions are identical to the ones obtained later using DOQSY [40], the PDSD results for both static [26] and slowly rotating [48] samples were less specific and would not support a similarly detailed analysis in terms of torsion-angle distributions. SMAS-PDSD spectra from S.c. ricini silk are shown in Figures 7a and 8a for various mixing times for casted film and fibre samples, respectively. In the film (Figure 7a), significant cross peaks appeared at clearly shorter mixing times than for the fibre. After 20 s, the spectrum indicates the presence of substantial disorder in this sample on the nanometer length scale. For the fibre (Figure 8a), the buildup of cross-peak intensity is slow, indicating that either the polarisation transfer between the alanines is very slow, because they are well separated in space, or that the alanines are highly ordered with all the 1-13 C CSA tensors close to parallel or anti-parallel to each other. The first explanation is improbable because the primary structure of S.c. ricini silk consists of alanine rich segments [47] and it is also incompatible with the DOQSY results discussed above. The second explanation is consistent with the DOQSY findings [30] and X-ray studies [49], which show the occurance of micro-crystalline domains with β-sheet structure. In contrast to the PDSD results of K¨ummerlen et al. [26], the SMAS-PSDSD results of Figure 8 are able to detect some weak cross-peak intensity due to the enhanced efficiency of spin diffusion under SMAS. A quantitative interpretation of the SMAS-PDSD spectra is rather involved but it is interesting to compare the measured spectra with the predicted spectra of simple model structures using numerical simulations. For the film sample, the spectrum at 250 ms shows the characteristic of a single α helix (b) or of an ordered bundle of helices (c). The spectra at longer mixing times indicate, that the cross-peaks cannot be explained by a single helix. The description by a bundle of biaxially ordered helices is better but still not fully convincing and further disorder is certainly present. Note that in the interpretation of the experimental and simulated data, the diagonal is usually not considered because it does not give information and is complicated by the presence of signals from nonexchanging spins.

Part III

A)

Geometrical Information on the Molecular-to-Nanometer Scale 1573

1574 Part III

Materials Science

Part III

Gly12

(ppm) 150

Ala19

(ppm)

150

Obsd.

Obsd.

Gly25

(ppm)

Obsd.

160 160

160

170 170

170 180

180

180 190

190

190 190

180 170 160 150 (ppm)

190

B.

A. (ppm)

170

180

160

190

160

160

170

160

150 (ppm)

160

150 (ppm)

(ppm) 150

Cald.

150

170

C.

(ppm)

Cald.

180

(ppm)

Cald.

160 170

170 180

180

180 190

190 190

180 170 160 150 (ppm)

190

190

180

D.

170

160 (ppm)

E.

190

180

170

F.

Fig. 6. 2D spin-diffusion NMR spectra of selectively doubly labelled model peptides, (A) GGAGGGYGGD[113 C]G [1-13 C]G (A) GGAGDGYGAG, (B) GGAGGGYGGDGG(A) [1-13 C]A [1-13 C]A 11 12 12 5 18 19 (A)5 GGAGDGYGAG, and (C) GGAGGGYGGDGG (A)11 [1-13 C]A24 [1-13 C]G25 GAGDGYGAG, after TFA treatment, and their corresponding simulated spectra. Reproduced from Ref. [60], with permission.

For the fibrous sample, the simulations of a single β strand, or an idealised parallel β-sheet, indicate that very little cross-peak intensity is expected even if standard (φ,ψ) values for β-sheets are used instead of the more extended form postulated by K¨ummerlen [26]. Again, the SMAS data are in good accordance with the DOQSY results. While the simulated spectra for a parallel or an anti-parallel arrangement of the β-sheets differ, it was not possible to determine the intermolecular arrangement from the spectra. The buildup rate of the cross-peaks in the SMAS spectra of both the film and fibre sample have been estimated from the initial rate regime and are reported in Figure 9, together with estimates from the simple models described above. Considering the approximations, the spectra represent a nice reconfirmation of the results

extracted from the DOQSY spectra and they allow for the qualitative assessment of the disorder on a supramolecular scale.

Conclusions To determine the local structure in partially disordered solid proteins, it is often attractive to perform NMR experiments without sample rotation. This reduces the signal-to-noise ratio in the experimental spectra, but on the other hand, there is often a much stronger dependence of the spectra on the parameters of interest and the increased spectral area typically yields a more strongly overdetermined problem in the data

Local Structure of Spider Silk Using Solid-State NMR

Conclusions 1575

Part III

Fig. 7. (A) 2D 13 C PDSD spectra of S.c. ricini 1-13 C alanine labelled silk film recorded under SMAS using mixing periods of 12.5 ms, 250 ms, 5 s and 20 s. (B) Corresponding simulations for a (Ala)100 segment organised in an α-helical structure. (C) Corresponding simulations for (Ala)12 segments organised in an ordered array of α-helical structures. For a model peptide of poly(alanine)n (n = 12 or n = 100) arranged in an α-helix, the set of Euler angles used for the simulations was (−27.4◦ , 96.6◦ , −1.1◦ ). As a first model (in B), one poly(alanine)100 helix has been considered. As a second model (C) 30 short α-helices (n = 12) have been taken into account: the helices were parallel and translationally symmetric to each other and the inter-helical distance between them has been set to ˚ (See also Plate 127 on page XXI in the Color Plate Section.) 7 A.

analysis. In particular, for the case of less than 100% isotopic enrichment of the amino acid to be investigated, DOQSY experiments are best suited to give the backbone torsion angles connecting neighbours in primary sequence. The knowledge of these angles determines the secondary structure. Information about the further

environment can be obtained by spin-diffusion experiment and by DECODER [50,51] type experiments on oriented silk samples, not discussed here [27,40,52–54]. The SD long-range information is, however, not specific enough to be able to construct an atomic model of the structure.

1576 Part III

Materials Science

Part III Fig. 8. (A) 2D 13 C PDSD spectra of S.c. ricini 1-13 C alanine labelled silk fibre recorded under SMAS using mixing periods of 12.5 ms, 500 ms, 5 s and 20 s. (B) Corresponding simulations for a (Ala)100 segment organised as a β-strand structure. (C) Simulations corresponding to 30 strands of 12 residues each. The three Euler angles between neighbouring carbonyl tesnors used in the model were (173.2◦ , 120.5◦ , −12.4◦ ). Two structures have been constructed. The first one is a long β-strand containing 100 alanine residues (B). As a second model, 30 β-strands containing 12 alanine residues have been taken into account (C). The relative orientation of all five layers is the same and the strands are anti-parallel. Such a model leads to a very ordered β-sheet. The inter-strand ˚ respectively. (See also Plate 128 on page XXII in the Color Plate and inter-layer distances have been set to 6.10 and to 5.33 A, Section.)

The silk fibres of the Eri silkworm S.c. ricini were found to be structurally quite similar to spider dragline silk from Nephila edulis: the alanine residues are arranged in β-sheets, while the glycine residues form predominantly 31 helical structures. If cast films are produced, the alanine-rich domains adapt an α-helical form instead.

Acknowledgments We acknowledge support by the Swiss National Science Foundation, the ETH Zurich, the Fonds pour la Formation de Chercheurs et l’Aide a` la Recherche, Qu´ebec, Canada (FCAR) and the European Science Foundation through the Network: Silk, Properties and Production. We

Local Structure of Spider Silk Using Solid-State NMR

References 1577

Fig. 9. Experimental average SD rate constants under SMAS between carbonyl-carbons in S.c. ricini 1-13 C alanine-labelled silk (A) film and (B) fibre as a function of the chemical shifts (upper part). The lower parts show corresponding simulations for (A) an (Ala)100 segment organised in a α helix structure and (B) a set of parallel (Ala)12 segments in a β sheet stucture. The offdiagonal region was fitted in the initial regime from four spectra with mixing times of 12.5, 75, 150 and 250 ms for the film and three spectra with mixing times of 12.5, 250 and 500 ms for the fibre. The fit determined the linear slope a in the equation y = ax + bx 2 used to approximate the initial regime of the buildup curve, where y are the data points, x the mixing times and a and b the coefficients. (See also Plate 129 on page XXIII in the Color Plate Section.)

thank Dr Pierre Robyr, Dr Edme Hardy, Kousuke Ohogo and Dr Phil. Williamson, Dr Aswin Verhoeven and Professor Michele Auger for constructive discussions. S.c. ricini silk samples were obtained from Professor Tetsuo Asakura.

1. Hinman MB, Jones JA, Lewis RV. Synthetic spider silk: a modular fiber. Trends Biotechnol. 2000;18:374–9. 2. Yao J, Asakura T. Synthesis and structural characterization of silk-like materials incorporated with an elastic motif. J. Biochem. (Tokyo) 2003;133:147–54. 3. Asakura T, Tanaka C, Yang M, Yao J, Kurokawa M. Production and characterization of a silk-like hybrid protein, based on the polyalanine region of Samia cynthia ricini silk fibroin and a cell adhesive region derived from fibronectin. Biomaterials. 2004;25:617–24. 4. Jin HJ, Park J, Valluzzi R, Cebe P, Kaplan DL. Biomaterial films of Bombyx mori silk fibroin with poly(ethylene oxide). Biomacromolecules. 2004;5:711–7. 5. Kim UJ, Park JY, Li CM, Jin HJ, Valluzzi R, Kaplan DL. Structure and properties of silk hydrogels. Biomacromolecules. 2004;5:786–92. 6. Nazarov R, Jin HJ, Kaplan DL. Porous 3-D scaffolds from regenerated silk fibroin. Biomacromolecules. 2004;5:718– 26. 7. Mita K, Ichimura S, James TC. Highly repetitive structure and its organization of the silk fibroin gene. J. Mol. Evol. 1994;38:583–92. 8. Zhou CZ, Confalonieri F, Medina N, Zivanovic Y, Esnault C, Yang T, Jacquet M, Janin J, Duguet M, Perasso R, Li ZG. Fine organization of Bombyx mori fibroin heavy chain gene. Nucleic Acids Res. 2000;28:2413–9. 9. Yukuhiro K, Kanda T, Tamura T. Preferential codon usage and two types of repetitive motifs in the fibroin gene of the Chinese oak silkworm, Antheraea pernyi. Insect Mol. Biol. 1997;6:89–95. 10. Xu M, Lewis RV. Structure of a protein superfiber: spider dragline silk. Proc. Natl. Acad. Sci. USA. 1990;87: 7120. 11. Hinman MB, Lewis RV. Isolation of a clone encoding a 2nd dragline silk fibroin—nephila-clavipes dragline silk is a 2protein fiber. J. Biol. Chem. 1992;267:19320–4. 12. Guerette PA, Ginzinger DG, Weber BHF, Gosline JM. Silk properties determined by gland specific expression of a spider fibroin gene family. Science. 1996;272:112–5. 13. Hayashi CY, Lewis RV. Evidence from flagelliform silk cDNA for the structural basis of elasticity and modular nature of spider silks. J. Mol. Biol. 1998;275:773–84. 14. Zhao C, Asakura T. Structure of silk studied with NMR. Prog. Nucl. Magn. Reson. Spectrosc. 2001;39:301–52. 15. B¨ockmann A, Lange A, Galinier A, Luca S, Giraud N, Juy M, Heise H, Montserret R, Penin F, Baldus M. Solid state NMR sequential resonance assignments and conformational analysis of the 2 × 104 Da dimeric form of the Bacillus subtilis protein Crh. J. Biomol. NMR. 2003;27: 323–39. 16. Pauli J, Baldus M, van Rossum B, de Groot H, Oschkinat H. Backbone and side-chain C-13 and N-15 signal assignments of the alpha-spectrin SH3 domain by magic-angle spinning solid-state NMR at 17.6 tesla. Chembiochem. 2001;2:272–81. 17. Castellani F, van Rossum B, Diehl A, Schubert M, Rehbein K, Oschkinat H. Structure of a protein determined by solidstate magic-angle-spinning NMR spectroscopy. Nature. 2002;420:98–102.

Part III

References

1578 Part III

Materials Science

Part III

18. Igumenova TI, McDermott AE, Zilm KW, Martin RW, Paulson EK, Wand AJ. Assignments of carbon NMR resonances for microcrystalline ubiquitin. J. Am. Chem. Soc. 2004;126:6720–7. 19. Tycko R. Applications of solid state NMR to the structural characterization of amyloid fibrils: methods and results. Prog. Nucl. Magn. Reson. Spectrosc. 2003;42:53– 68. 20. Jaroniec CP, MacPhee CE, Bajaj VS, McMahon MT, Dobson CM, Griffin RG. High-resolution molecular structure of a peptide in an amyloid fibril determined by magic-anglespinning NMR spectroscopy. PNAS. 2004;101:711–6. 21. Michal CA, Jelinski LW. Rotational-echo double-resonance on complex biopolymers: a study of Nephila clavipes dragline silk. J. Biomol. NMR. 1998;12:231–41. 22. Asakura T, Ashida J, Yamane T, Kameda T, Nakazawa Y, Ohgo K, Komatsu K. A repeated beta-turn structure in poly(Ala-Gly) as a model for silk I of Bombyx mori silk fibroin studied with two-dimensional spin-diffusion NMR under off magic-angle-spinning and rotational echo double resonance. J. Mol. Biol. 2001;306:291–305. 23. Ashida J, Ohgo K, Komatsu K, Kubota A, Asakura T. Determination of the torsion angles of alanine and glycine residues of model compounds of spider silk (AGG)(10) using solid-state NMR methods. J. Biomol. NMR. 2003;25:91–103. 24. Hijirida DH, Do KG, Michal C, Wong S, Zax D, Jelinski LW. C-13 NMR of Nephila clavipes major ampullate silk gland. Biophys. J. 1996;71:3442–7. 25. Simmons AH, Michal CA, Jelinski LW. Molecular orientation and two-component nature of the crystalline fraction of spider dragline silk. Science. 1996;271:84–87. 26. K¨ummerlen J, van Beek J, Vollrath F, Meier BH. Local structure in spider dragline silk investigated by twodimensional spin-diffusion nuclear magnetic resonance. Macromolecules. 1996;29:2920–8. 27. Asakura T , Ito T, Okudaira M, Kameda T. Structure of alanine and glycine residues of Samia cynthia ricini silk fibers studied with 15N and 13C NMR. Macromolecules. 1999;32:4940–6. 28. Reif B, Griffin RG. 1H detected 1H,15N correlation spectroscopy in rotating solids. J. Magn. Reson. 2003;160:78– 83. 29. Schmidtrohr K. A double-quantum solid-state NMR technique for determining torsion angles in polymers. Macromolecules. 1996;29:3975–81. 30. van Beek JD, Beaulieu L, Schafer H, Demura M, Asakura T, Meier BH. Solid-state NMR determination of the secondary structure of Samia cynthia ricini silk. Nature 2000;405:1077–9. 31. van Beek JD, Meier BH. A DOQSY approach for the elucidation of torsion angle distributions in biopolymers: application to silk. J. Magn. Reson. 2006;178:106–20. 32. Utz M, Robyr P, Suter UW. Solid-state NMR investigation of the structural consequences of plastic deformation in polycarbonate. 2. Local orientational order. Macromolecules. 2000;33:6808–14. 33. Utz M. Measurement of structural distribution functions in disordered systems: A general approach for sensitivity estimation. J. Chem. Phys. 1998;109:6110–24.

34. van Beek JD, Meier BH, Schafer H. Inverse methods in two-dimensional NMR spectral analysis. J. Magn. Reson. 2003;162:141–57. 35. Ando S, Yamanobe T, Ando I, Shoji A, Ozaki T, Tabeta R, Saito H. Conformational characterization of glycine residues incorporated into some homopolypeptides by solid-state 13C NMR spectroscopy. J. Am. Chem. Soc. 1985;107:7648–52. 36. Saito H, Iwanga Y, Tabeta R, Narita M, Asakura T. Chem. Lett. 1983;4:427. 37. Saito H, Tabeta R, Shoji a, Ozaki T, Ando I, Miyata T. A high-resolution 13C-NMR study of collagenlike polypeptides and collagen fibrils in solid-state studied by the cross-polarization magic angle-spinning method— manifestation of conformation-dependent C-13 chemicalshifts and application to conformational characterization. Biopolymers. 1984;23:2279–97. 38. Ishida M, Asakura T, Yokoi M, Saito H. Solventand mechanical-treatment-induced conformational transition of silk fibroins studies by high-resolution solidstate carbon-13 NMR spectroscopy. Macromolecules. 1990;23:88–94. 39. Robyr P, Tomaselli M, Grob-Pisano C, Meier BH, Ernst RR, Suter UW. Characterization of local order in atactic polystyrene using 2D NMR and atomistic simulations. Macromolecules. 1995;28:5320–4. 40. van Beek JD, Hess S, Vollrath F, Meier BH. The molecular structure of spider dragline silk: Folding and orientation of the protein backbone. Proc. Natl. Acad. Sci. USA. 2002;99:10266–71. 41. Schmidt-Rohr K, Spiess HW. Multidimensional Solid-State NMR and Polymers, Academic Press: London, 1994. 42. Ernst M, Meier BH. In: I. Ando, T. Asakura (Eds). Solid State NMR of Polymers. Elsevier, 1998. 43. Abragam A. The Principles of Nuclear Magnetism, Clarendon Press: Oxford, 1961. 44. Gan ZH, Ernst RR. Shift-independent nuclear spin diffusion by slow magic-angle sample spinning for the exploration of solids. Chem. Phys. Lett. 1996;253:13–9. 45. Asakura T, Yao JM, Yamane T, Umemura K, Ultrich AS. Heterogeneous structure of silk fibers from Bombyx mori resolved by C-13 solid-state NMR spectroscopy. J. Am. Chem. Soc. 2002;124:8794–5. 46. Nakazawa Y, Bamba M, Nishio S, Asakura T. Tightly winding structure of sequential model peptide for repeated helical region in Samia cynthia ricini silk fibroin studied with solidstate NMR. Protein Sci. 2003;12:666–71. 47. Nakazawa Y, Asakura T. Structure determination of a peptide model of the repeated helical domain in Samia cynthia ricini silk fibroin before spinning by a combination of advanced solid-state NMR methods. J. Am. Chem. Soc. 2003;125:7230–7. 48. van Beek JD. Methods for Studying Heterogeneous Solid Proteins and the Application to Silk, Diss. ETH No. 14637, ETH-Zurich, Zurich, 2002. 49. Warwicker JO. J. Mol. Biol. 1960;2:350–62. 50. Schmidt-Rohr K, Hehn M, Schaefer D, Spiess HW. Twodimensional nuclear magnetic resonance with sample flip for characterizing orientation distributions, and its analogy to X-ray scattering. J. Chem. Phys. 1992;97:2247.

Local Structure of Spider Silk Using Solid-State NMR

56.

57. 58.

59. 60.

by high-resolution solid state 13C NMR. Magn. Reson. Chem. 1986;24:835–52. Hartzell CJ, Pratum TK, Drobny G. Mutual orientation of three magnetic tensors in a polycrystalline dipeptide by dipole-modulated nitrogen-15 chemical shift spectroscopy. J. Chem. Phys. 1987;87:4324–31. Separovic F, Smith R, Yannoni CS, Cornell BA. Molecular sequence effect on the carbon-13 carbonyl chemical shift shielding tensor. J. Am. Chem. Soc. 1990;112:8324–8. Teng Q, Iqbal M, Cross TA. Determination of the carbon13 chemical shift and nitrogen-14 electric field gradient tensor orientations with respect to the molecular frame in a polypeptide. J. Am. Chem. Soc. 1992;114:5312–21. Mehring M. Principles of High Resolution NMR in Solids, 2nd ed., Springer: Berlin, 1983. Yao JM, Nakazawa Y, Asakura T. Structures of Bombyx mori and Samia cynthia ricini silk fibroins studied with solidstate NMR. Biomacromolecules. 2004;5:680–8.

Part III

51. Chmelka BF, Schmidtrohr K, Spiess HW. Molecularorientation distributions in poly(ethylene-terephthalate) thin-films and fibers from multidimensional decoder NMRspectroscopy. Macromolecules. 1993;26:2282–96. 52. Demura M, Minami M, Asakura T, Cross TA. Structure of Bombyx mori silk fibroin based on solid-state NMR orientational constraints and fiber diffraction unit cell parameters. J. Am. Chem. Soc. 1998;120:1300–08. 53. Eles PT, Michal CA. Strain dependent local phase transitions observed during controlled supercontraction reveal mechanisms in spider silk. Macromolecules. 2004;37: 1342–5. 54. Eles PT, Michal CA. A DECODER NMR study of backbone orientation in Nephila clavipes dragline silk under varying strain and draw rate. Biomacromolecules. 2004;5: 661–5. 55. Saito H. Conformation-dependent 13C chemical shifts: A new means of conformational characterization as obtained

References 1579

1581

Siegfried Stapf1 , A. Khrapitchev1 , C. Heine1 and Bernhard Bl¨umich2 1 RWTH

Aachen, Lehrstuhl f¨ur Makromolekulare Chemie, Sammelbau Chemie, Worringer Weg 1, 52074 Aachen, Germany 2 RWTH Aachen, Institut f¨ ur Technische Chemie und Makromolekulare Chemie, (ITMC), Worringer Weg 1, D-52056 Aachen, Germany

Introduction Granular materials are a class of systems of widespread interest, touching fields as diverse as food and pharmaceutical production processes, storage and packing of agricultural products, waste disposal and recycling, or powder technology; but also natural phenomena of equal economic importance such as avalanches, sand transport by wind or water, and sedimentation. Motion of a granular substance frequently evokes the impression of chaotic behavior, but granular particles experience the same physical forces like any other moving object, in particular gravity, friction and interparticle forces. It is the correlated motions of a large number of particles which require their treatment as a coupled system, and therefore, a statistical description of the motion processes. However, such a statistical description of motion is not restricted to granular materials. In fact, fluid dynamics uses the same type of formalism, describing the transport of finite fluid volume elements, while the trajectory of an individual molecule is ill-defined, with only time- and ensemble-averaged properties being available. In many senses, a granular medium can be considered as a closeup shot of a moving fluid, with a much reduced number of particles. The disadvantage of this could be that the particle number under some experimental conditions is not sufficient any longer for a proper averaging in the statistical sense; but at the same time, the observation of small numbers of particles, or even single ones, becomes feasible, at least in principle. The resemblance of granular material dynamics to certain characteristics of phase behavior in “normal” matter (solids, liquids and gases) has quite naturally triggered much theoretical work attempting to find proper means to model and describe granular motion, but the ubiquity of granular media themselves and their huge economic importance have provided further impetus. Questions arising are either of everyday nature (the famous case of the shaken cornflakes box, an exercise in particle segregation), or comparatively far-fetched (much has been learnt by analyzing and modeling the shape of desert and underwater dunes that could be applied to understand the Graham A. Webb (ed.), Modern Magnetic Resonance, 1581–1587.
C 2008 Springer.

current and ancient atmospheric conditions on Mars and several moons in the solar system). Many phenomena of interest in granular media are connected to transport properties. While transport through granular media, such as unconsolidated soils, falls into this category, it is similar to the well-covered field of flow in structured media, which has been investigated in great detail by NMR methods for a long time. In this article, we want to point out the difficulties and possibilities in NMR investigations of transport of granular media, i.e., the particle motion itself. Despite the large importance of such works, they remain scarce in the literature, one reason being the limited applicability of conventional NMR methods.

NMR of Transport in Granular Media—an Overview The essential difficulty in tackling granular motion by NMR is the fact that the particles are very often solids in the NMR meaning of the word, i.e., their spectral features contain very broad lines, and sometimes sensitive nuclei like 1 H are even mostly absent, such as in dry sand grains. Provided that the sample indeed contains protons in sufficient quantity, pure imaging is still possible employing methods which had been specifically developed for solids (single-point acquisitions or multi-pulse line narrowing techniques). The encoding of velocity with pulsed field gradients (PFGs), however, consumes a minimum time of a few hundred µs in the best case, which is still one order of magnitude longer than the transverse relaxation time of solids. Only permanent-gradient techniques bear the potential to measure motion in some, but not all, solid granular particles directly, but no such effort has yet been made. The way out of this dilemma is either to obtain indirect information about the granular motion by measuring the transport of the surrounding fluid (which can be a liquid but frequently is a gas) or the particles have to be labeled with some narrow-line compound; this could be either a liquid imbibed into porous particles and held by capillary forces (particles made from a catalyst carrier material were used in [1]), or a liquid core covered by

Part III

Velocity Imaging of Granular Materials

1582 Part III

Materials Science

Part III

a solid shell. Nature provides the latter particles in the shape of oil-filled seeds, such as mustard, sesame [2, 3] or poppy [1], which possess relatively even shapes and sizes. Liquid-filled hollow spheres are also manufactured for a wide range of pharmaceutical applications [4, 5]. Of the many conceivable geometries, research has been concentrated on only a few. Mueth et al. have looked into the velocity distribution of spherical particles inside a Couette cell [3, 6]. Shaken or vibrated beds were investigated by Kuperman et al. [7–9] and more recently by Yang et al. [10, 11]. Both segregation processes and localized velocities were measured for rotating horizontal cylinders half-filled with granular particles [2,4,5,12,13]. Finally, gas-fluidized beds were investigated because of their proximity to a common reactor type used in chemical engineering [1]. For a complete survey, see [14]. This contribution will focus on the latter two geometries as examples.

dynamics in the granular bed can then be characterized by the following approaches: (a) 1D imaging measures either the vertical extension of the bed or the signal intensity at any given layer, both being related to the particle density in the array; (b) the average particle mobility can be characterized by a number of parameters of increasing complexity, starting from dispersion coefficients via distribution functions of displacements within a given time (propagators) to higher-order information such as velocity changes, eventually leading to estimates of the collision time and the time-dependence of displacements; (c) velocity imaging combines both approaches and assigns average velocities or dispersion coefficients to the spatial position inside the bed. In Figure 1, measurements of the average bed density normalized to the case in the absence of air flow are

Gas-Fluidized Bed A gas-fluidized bed consists of a granular system filled into a vessel which is subjected to a continuous gas stream. The intense exchange between the gas and the surface of the solid pellets favor gas phase reactions, hence its large industrial potential. The different stages of motion can be classified similar to phases in condensed matter [15, 16]. In the absence of flow, the medium behaves like a solid— if the whole array is tilted relative to the vertical axis, it will keep its overall shape until the angle of repose is reached. Above a certain threshold of gas flow rate, the bed becomes “fluidized”, i.e. it will follow a tilt by liquid-like deformation so that the surface remains horizontal. In this stage, the particles have gained a slightly increased free volume, so that the total density of the bed is reduced. At higher flow rate, bubbles are formed, which drive those particles immediately above them upward, until dispersive components make them slide down around the bubble. The situation very much resembles a boiling liquid, the density is reduced further. Increasing the gas flow rate usually has the effect of increasing the average bubble size, which eventually reaches the diameter of the confining cell. Under these conditions, a pulsatile motion known as “slug phase” is observed. Depending on the cohesive forces and the inertia of the particles, they will eventually be driven out of the reactor cell. It is easy to imagine that boundary conditions, such as wall effects, even distribution of the air stream, but also particle size and shape, “stickiness” and size dispersion affect the transition between the phases. Because optical methods can access only the vicinity of the container wall, where the unwanted wall effects are strongest, the advantage of NMR which can sample arbitrary volumes of the bed is obvious. The state and

Fig. 1. Densities of gas-fluidized beds for different particle sizes, normalized by the density in the absence of flow. v indicates the average air velocity. (a) poppy seeds, (b) catalyst particles of different diameter ranges.

Velocity Imaging of Granular Materials

 D( ) = 0





1−

τ

v(0)v(τ )dτ

Part III

(a)

Deff [10−6 m2s−1]

4 3 2 1 112-140 µm 140-200 µm 200-250 µm

0 0

3

12

9

6

15

18

v [cm/s] (b)

7 6

Deff [10−7 m2s−1]

presented. As for all other experiments on fluidized beds, the measurements were taken out on a Bruker DSX 200 console equipped with a vertical 4.7 T magnet at MARC, RWTH Aachen, Germany. A glass tube with an inner diameter of 21 mm was filled with either poppy seeds (diameter 500–800 µm but not spherical), or ground and sieved commercial catalyst pellets (Al2 O3 ) saturated with acetone having size ranges given in the figure. The stepwise density changes are readily identified for the poppy seeds, but blurred for the smaller catalyst pellets. According to the categories given, e.g. in [16], the full spectrum of “phases” is not always expected; small particles with their relatively larger cohesive forces can be difficult to fluidize, while wall effects can become dominant for larger particles. Keeping these uncertainties in mind, all presented results must be regarded as somewhat imperfect states of granular flow which, however, are suitable for demonstrating the typical features of the dynamics. It should be noted that, in order to obtain a value for the relative density, which is unbiased by relaxation effects, a single point acquisition technique had to be used which allowed signal acquisition during 10 µs following a low flip-angle pulse; the 1D profile was then reconstructed from 128 such encoding steps after sufficient averaging. A simple frequency-encoding sequence turned out to be unsuitable to provide a quantitative measure for the spin density; the relatively high particle velocities and short transverse relaxation times generate a combination which makes the use of a time-consuming velocity-compensated pulse sequence difficult, particularly for the catalyst pellets. From an experimental point of view, the adsorbate in these pellets was evaporating on a timescale of many hours. This is one example for experimental difficulties which require a careful consideration of the use of standard methods for moving samples when being applied to granular media. The simplest motion-related quantity, which can be measured by NMR, is the dispersion coefficient. It is related to the velocity autocorrelation function (VACF) by

Gas-Fluidized Bed 1583

5 4 3 2 1 0 0

50

100

150

200

250

300

∆ [ms] Fig. 2. (a) Dispersion coefficients for catalyst particles as a function of air velocity and particle size. (b) Dispersion coefficients for poppy seeds as a function of encoding time.

vector q = (2π )−1 γ δG. Analyzing the initial slope of the ˜ ) = S(q, )/S(0, ), normalized signal intensity, S(q, renders the dispersion coefficient [17]:

(1)

if the VACF depends on the time differences τ only [5]. This condition can be violated, e.g. in the bubbling phase, when the system is macroscopically heterogeneous and each measurement probes a different particle distribution. It is assumed, however, that sufficient averaging does provide a reasonable approximation to the dispersion coefficient. D( ) was determined using a standard diffusion encoding sequence, including a pair of pulsed field gradients of duration δ and strength G, defining the wave

D( ) ≈ −(4π 2 )−1 lim

q→0

˜ )| ∂ ln | S(q, . ∂q 2

(2)

Figure 2a compares the increase of D( ) with air velocity for catalyst pellets with different size ranges. As expected, smaller and lighter particles are more affected by the driving air flow. By comparison with the density measurements in Figure 1, the sudden increase of the dispersion coefficient at an air velocity of 12 cm/s for the largest particles is indicative of the transition to the bubbling phase—when higher velocities begin to appear due

Materials Science

and is obtained via inverse Fourier transformation. Already the visual appearance of the bubbling phase in fluidized beds hints to a strongly asymmetric velocity distribution: a small number of fast particles moving upward is compensated by the remaining particles slowly moving downward. This impression is confirmed and quantified by the shape of the propagator averaged over the whole sample volume inside the resonator. Figure 3 demonstrates the influence of an increasing air flow rate on the propagator shapes for a bed of poppy seeds at a relatively short encoding time, = 10 ms. As expected, the distribution of vertical (Z ) velocities is asymmetric, with a long tail of positive (upward) velocities

(a) 0,5

vz 0,4

probability

Part III

to the fast-moving bubbles—, this value being shifted to lower air velocities for the smaller particles. For the results of Figure 2a, encoding took place during a fixed interval = 1.4 ms, which is much longer than the average time between collisions, but short compared to the time required for an individual particle to sample a region with a statistical representation of all possible void structures and velocity distributions. Such a representative elementary volume had been identified as the characteristic structure beyond which dispersion processes become time-independent. The initial concept was formulated for fluid flow in fixed beds [18], but a similar volume can be defined for the fluidized bed, if wall effects can be neglected. An example of the time-dependence of the dispersion coefficient is sketched in Figure 2b, which was obtained for poppy seeds, using a multi-PFG technique compensating for coherent motion contributions. It is this long-time limit, approached around ≈ 100 ms for the case shown, which is of relevance for understanding processes that affect the efficiency of the fluidized-bed reactor as a whole. It is preceded by a regime which possesses some ballistic characteristics, where particles move in groups, for instance driven by an individual gas bubble; this regime shows an almost linear increase of D( ) with . The truly ballistic regime would be observed on time scales much shorter than the time between collisions. Ongoing research attempts to access this time scale by reducing the encoding intervals to well below 1 ms and choosing the granular system appropriately. The next, more complex function describing motion in the system is the so-called propagator [19]. It still represents an averaging over a large region or the whole bed, but provides the distribution of displacements R during a time , and thus the velocities are averaged over this period. The propagator P(R, ) appears in the definition of the signal intensity as a function of encoding by wave vector q:  ˜ ) = P(R, ) exp[i2πq · R( )] d R, (3) S(q,

p (hPa)

∆ = 10 ms

35.0 45.0 55.0 65.0

0,3 0,2 0,1 0,0 −30 −20 −10

0

10

20

30

40

50

60

70

particle velocity [mm/s]

(b) 0,5

vx 0,4

probability

1584 Part III

p (hPa)

∆ = 10 ms

35.0 45.0 55.0 65.0

0,3 0,2 0,1 0,0 −30

−20

−10

0

10

20

30

particle velocity [mm/s] Fig. 3. Propagators for poppy seeds at different air pressure drops (a) parallel (Z ) and (b) perpendicular (X ) to the air flow direction.

gaining in weight for higher flow rates. For some conditions, the most probable velocity is actually negative, while mass balance requires the average to remain zero. The transverse velocities (X ) have a symmetric distribution about zero because of the absence of a net transverse motion. Finally, the combination of imaging sequences with a velocity-encoding module allows the local description of motion in terms of pixel-averaged velocities, or indeed propagators. One example, again for poppy seeds in the bubbling phase, is shown in Figure 4. Two-dimensional imaging (phase- and frequency-encoding) was combined with a preceding pair of velocity-encoding gradients at a separation of = 4.7 ms. Data were taken in a horizontal slice of 5 mm thickness near the center region of the bed of 27 mm height. For the chosen air flow

-50 mm/s x-position

Fig. 4. Map of the vertical velocity v z for a bed of poppy seeds of 44 mm diameter in the bubbling phase obtained in a horizontal slice of 5 mm thickness. The white line denotes zero vertical velocity. The vertical stripes are artifacts of the imaging reconstruction process. (See also Plate 130 on page XXIII in the Color Plate Section.)

rate, motion is characterized by a fountain-like behavior, where a relatively small fraction of particles in the central region is transported upwards with large velocities, while the particles further out and toward the cylinder wall move downward with smaller velocity values. The white line in Figure 4 separates the region with positive and negative velocities. The result was obtained from the difference of two differently encoded images to remove systematic phase errors; each image took 90 min. These results first demonstrate the feasibility of obtaining velocity images in a fluidized particle bed, but for the relatively small number of particles used in this particular case, the strong fluctuations in the velocity field, and therefore, the need for a large number of data accumulations becomes evident.

Rotating Drum The first NMR measurements on granular flow in a partially filled rotating drum have been reported in 1993 [2]. Since then, continuous efforts have been made to either visualize the complex patterns arising from segregation effects between two types of particles distinguished by size and/or mass [12, 20, 21], or the actual velocity pattern by means of tagging sequences [13] or velocity encoding techniques [4, 5]. In particular, [5] uses the example of granular flow for a detailed demonstration of the

Rotating Drum 1585

experimental techniques and interpretations of the role of coherent and incoherent motion, providing maps of velocities, VACF correlation times and dispersion coefficients with spatial resolution. As a demonstration of the essential features of granular flow in a rotating cylinder, Figure 5 presents velocity components in the two orthogonal directions normal to the rotation axis, being combined to a magnitude image and a vector plot. A PMMA cylinder of 88 mm inner diameter and 560 mm length has been filled to half its volume with mustard seeds (diameter 1.5 ± 0.2 mm) and was rotated at /2π = 0.453 per s. Experiments were performed on a horizontal 310 mm bore Oxford 1.9 T magnet at New Mexico Resonance, Albuquerque, NM. A double-echo pulse sequence consisting of one sliceselective π/2 pulse and two refocussing π hard pulses was used in order to compensate for background gradients [5]. The results were taken from [22]. The experimentally obtained velocity fields can be understood with the help of the theoretical interpretation given in Figure 6 [22] for an idealized situation. For small rotation angles, the granular bed will follow the motion like a solid object until the particles at the highest (top right in Figure 6) position will fall downward, generating an avalanche of other particles. Above a certain threshold rotation rate , this series of individual avalanches will turn into a periodic process where the velocity field of all particles can reasonably be well described by average quantities which depend only on position. In Figure 6a, this situation is explained: all particles within the top layer of thickness r0 (subset “1”, r < r0 ) are sliding parallel to the surface, i.e. in direction d, with a velocity dependence v d (r ) = v max (1 − r/r0 )2 − r

(4)

while the particles in the bottom (subset “2”, r > r0 ) rotate like a solid body with v d (r ) = r . An increase of  will lead to deviations from this ideal case, in particular the surface becomes curved [23]. The experimental results of Figure 5 can be interpreted by these different influences. The surface is clearly warped, but remains stationary during the course of the experiment. Figure 5a and b shows the expected smooth behavior of v x and v y in the bottom part of the granular bed. Figure 5c features the almost immobile region near the rotation axis and the strong dependence of velocities near the surface. Figure 5d finally visualizes the vectorial components of motion in the plane perpendicular to the rotation axis; the velocity vectors perfectly parallel to the surface can be easily identified, as the significant wall can slip, particularly at the bottom of the cylinder, where the column height of the granular material, normal to the cylinder wall, is largest.

Part III

y-position

Velocity Imaging of Granular Materials

1586 Part III

Materials Science

Part III Fig. 5. Velocity maps of a bed of mustard seeds inside a cylinder of 88 mm inner diameter rotating at /2π = 0.453 per s; the images were taken from an axial slice of 20 mm width. (a) component v x , (b) component v y , (c) magnitude (v x2 + v 2y )1/2 , (d) vector representation of the sum of (a) and (b). Numbers indicate velocities in m/s. (See also Plate 131 on page XXIV in the Color Plate Section.)

Summary a)

b)

Ω

r0

d

1

2R

Ω

1 2

2

y r x

Fig. 6. Velocities for a granular bed inside a cylinder rotating with . “1” denotes the region where granular flow occurs, “2” represents solid body rotation. (a) The surface orientation, which defines the vectors d and r relative to the gravity axis along −y, is determined by the angle of repose. (b) For larger , the surface plane is deformed into a sinusoidal shape.

Granular systems in many ways still are a challenge for the theoretician, particulary if their dynamic behavior shall be understood. To develop and verify suitable methods, however, getting experimental evidence is inevitable. While this is feasible for averaged quantities, such as flow resistance or dispersion parameters, a more detailed knowledge of particle motion might become necessary. Most techniques are optical and are thus limited by the opacity of the granular medium and the presence of a large number of interfaces, restricting accessibility to the surface region of the granular array. NMR does not have this disadvantage and is, at least in principle, able to provide any kind of information about the dynamics of granular systems. It does, however, require systems containing a liquid-like marker component which has a relaxation time that is sufficiently long to allow all necessary manipulations of the spin system for making the encoding schemes applicable.

Velocity Imaging of Granular Materials

Acknowledgments We are grateful to R. Savelsberg for preparing and performing the first fluidized bed experiments, to E. Fukushima for helpful comments to this manuscript, and to M. K¨uppers for continuing support in the current part of this project. Part of this work was done during a productive stay of one of the authors (C.H.) at New Mexico Resonance, Albuquerque, and we would like to thank the whole team of NMR for their assistance and helpful contributions. Financial support by the DFG (Bl 231/25-2) is gratefully acknowledged.

References 1. Savelsberg R, Demco DE, Bl¨umich B, Stapf S. Phys. Rev. E 2002;65:020301. 2. Nakagawa M, Altobelli SA, Caprihan A, Fukushima E, Jeong EK, Exp. Fluids. 1993;16:54. 3. Mueth DM, Debregeas GF, Karczmar GS, Eng PJ, Nagel SR, Jaeger HM, Nature. 2000;406:385. 4. Seymour JD, Caprihan A, Altobelli SA, Fukushima E, Phys. Rev. Lett. 2000;84:266. 5. Caprihan A, Seymour JD, Magn. J. Reson. 2000;144:96.

6. Mueth DM, Phys. Rev. E. 2003;67:011304. 7. Ehrichs EE, Jaeger HM, Karczmar GS, Knight JB, Kuperman VY, Nagel SR, Science. 1995;276:1632. 8. Kuperman V Yu, Phys. Rev. Lett. 1996;77:1168. 9. Knight JB, Ehrichs EE, Kuperman V Yu, Flint JK, Jaeger HM, Nagel SR, Phys. Rev. E. 1996;54:5726. 10. Yang XY, Huan C, Candela D, Mair RW, Walsworth RL, Phys Rev. Lett. 2002;88:044301. 11. Huan C, Yang X, Candela D, Mair RW, Walsworth RL, Phys. Rev. E. 2004;69:041302. 12. Hill KM, Caprihan A, Kakalios J, Phys. Rev. Lett. 1997;78:50. 13. Yamane K, Nakagawa M, Altobelli SA, Tanaka T, Tsuji Y, Phys. Fluids. 1998;10:1419. 14. Fukushima E, Adv. Complex Syst. 2001;4:503. 15. Davidson DF, Clift R, Harison D (Eds). Fluidization. Academic Press: London, 1985. 16. Geldart D (Ed). Gas Fluidization Technology. WileyInterscience: New York, 1986. 17. Callaghan PT, Stepiˇsnik J, Adv. Magn. Opt. Reson. 1996;19:325. 18. Bear J, Dynamics of Fluids in Porous Media. American Elsevier: New York, 1972. 19. K¨arger J, Heink W, Magn. J. Reson. 1983;51:1. 20. Nakagawa M, Chem. Eng. Sci. 1994;49:2540. 21. Hill KM, Caprihan A, Kakalios J, Phys. Rev. E. 1997;56:4386. 22. Heine C, NMR von rotatorischer und translatorischer Dynamik, PhD thesis, RWTH Aachen, Germany, 2001. 23. Nakagawa M, Altobelli SA, Caprihan A, Fukushima E. In: RB Behringer, JT Jenkins (Eds). Power & Grains’ 97. AA Balkema: Rotterdam, 1997, p. 447.

Part III

Such systems will frequently be of model character and are not identical to the real industrial application, but they are entirely sufficient to provide an extensive database with which theoretical descriptions can be developed.

References 1587

Part III

Part III, Section 2: Applications in Food Science

1591

Foreword to Applications in Food Science

The range of chemicals, structures and materials represented by foodstuffs is enormous. At one extreme sugar is the purest mass produced chemical available. Beverages such as coffee, tea or wine are structurally homogeneous but may contain thousands of chemicals covering a vast range of concentration. Solid foods represent a higher level of complexity combining high levels of heterogeneity and chemical complexity. All these levels represent a challenge to magnetic resonance, and it is not surprising that in this section all available magnetic resonance methods are represented. The challenge that foods represent to magnetic resonance have been amply exemplified in the biennial series of international conferences “Applications of Magnetic Resonance to Food Science”, the proceedings of which are a summary of state of the art of the subject [1–5]. However these books are summaries of work in progress they do not represent a review of the knowledge of particular areas. Within this section the authors have been selected as authorities in their fields who are in a position to write an overview of their subjects. Given the scale of the field some selectivity and sharp focus has been required. However each time each author has produced an article which is complete in itself. Overlap, therefore, is inevitable, and

a necessity, when each article is a comprehensive survey of its subject matter. Whilst the entries have been divided by the magnetic resonance technique in the main, I have included section on the applications of a variety of techniques to specific systems. This is needed if the reader is to appreciate the range of problems that confront the spectroscopist and the way in which an approach using a range of techniques can tackle them. Peter Belton

References 1. Belton PS, GA Webb and MJ McCarthy, ”Annual Reports on NMR Spectroscopy”, 1995;31. 2. Belton PS, I Delgadillo, AM Gil and GA Webb, ”Magnetic Resonance in Food Science”, RSC, Cambridge, 1995. 3. Belton PS, GA Webb, and BP Hills “Advances in Magnetic Resonance in Food Science” RSC, Cambridge, 1999. 4. Webb GA, PS Belton, AM Gil and I Delgadillo “Magnetic Resonance in Food Science A View to the Future” RSC, Cambridge, 2001. 5. Belton PS, AM Gil, GA Webb and D Rutledge “Magnetic Resonance in Food Science, Latest Developments” RSC, Cambridge, 2003.

1593

Glossary

agar: A polysaccharide obtained form sea weeds containing galactose units

high molecular weight subunits: Protein subunits of gluten mainly responsible for the elasticity of dough

anthocyanins: Water soluble pigments occurring in many plants

isoflavones: A sub group of the flavonoid group of compounds

Brix: A system of calculating sugar concentrations based on specific gravity

kaempferol: One of the flavonoid group of compounds

carrageenans: Sulphated polysaccharides derived from seaweeds casein: General name for a group of proteins occurring in milk catechins: generic name for a group of polyphenolic (QV) compounds couette: A viscometer consisting on concentric cylinders cutin: A waxy compound occurring in plants especially on leaves cytoplasm: region within a cell in which the organelles are embedded durum: Of wheat; a hard type of wheat used in making pasta epicatechin: a member of the catechin group of compounds

lignins: A highly condensed polymer of phenolic compounds occurring in plants lintnerization: Partial acid hydrolysis of starch Maillard reaction: A non enzymic reaction between amino acids and sugars leading to brown colourations and complex chemical products myofibril: The contractile unit of muscle fibre myricetin: One of the flavonoid group of compounds nutrigenomics: The study of how dietary chemicals interact with the genotype and gene expression oleaginous: Of plants, oil bearing organoleptic: The sensory properties of a food pectin: A polysaccharide occurring in plants, widely used in food as thickener phloretin: A flavonoid present in apples

extrusion: A thermomechanical process in which a screw system forces a product through a restricted opening

polyphenols: Compounds consisting of a number of phenolic groups, most of which are thought to have antioxidant activity in vivo

flavonoid: Polyphenolic (QV)compounds occurring in plants, thought to have role in protection against degenerative disease

quercetin: A compound belonging to the flavonoid group

galactomannans: A polysaccharide based on the sugar units galactose and mannose gellan gum: A polysaccharide produced by bacterial fermentation

renneting: Process of clotting of milk by enzymic action used in cheese making resveratrol: a phenolic compound found in grape skins and wines syneresis: separation of a liquid layer from gel

gliadins: A subset of gluten proteins soluble in ethanol water mixtures

theaflavins: A complex molecule occurring in black tea formed by the oxidation of flavonoids

gluten: A mixture of proteins which constitute the major storage proteins of wheat. The determining factor in the viscoelastic behaviour of dough

thearubigins: A group of chemicals similar to theaflavins also occurring in black tea

hemicelluloses: Complex carbohydrates occurring in plant cell walls

theophylline: A purine derivative which can act as muscle relaxant

1594 Glossary

Food Science

trigonelline: A derivative of nicotinic acid present in green coffee

transport and storage, in plants it contributes to waste disposal and turgor (QV)

turgor: pressure within cells that contributes to the mechanical rigidity

vanillin: The active principle in the flavour vanilla

vacuole: A sac within the cell that has functions of

water-holding capacity: The amount of water that may be held in a system without drip loss

Part III

High Resolution Solution State Methods

1597

Jurkica Kidriˇc1 and Iztok Joˇze Koˇsir1,2 1 National

Institute of Chemistry, SI-1000 Ljubljana, Slovenia; and 2 Slovenian Institute for Hop Research and Brewing, SI-3310 Zalec, ˇ Slovenia

Introduction Characterization of the chemical composition of beverages is important for establishing their quality and authenticity and in connection with their organoleptic properties. It enables screening for the possible addition of toxic materials and the determination of compounds important for human health. A fine example of the latter are polyphenols found in wine, fruit juices, and tea; they have a wide range of biological and pharmacological effects, including anticarcinogenic, anti-inflammatory, and antioxidant activity [1,2]. Beverages are complex mixtures of different classes of compounds present at a broad range of concentrations. In the past, various analytical techniques have been used in the analysis of beverages, most frequently HPLC, GC, capillary electrophoresis, and conductometry. Because of the compositional complexity of beverages and the wide concentration range of components classical chemical analysis requires time consuming sample preparation. In the case of compounds at low concentration, separation, derivatizaion, and preconcentration are common steps in the procedure. To date, NMR has assumed an outstanding position in the field of chemical analysis of food products. Though NMR is less sensitive than the mentioned methods, sample preparation for NMR is simpler and less time consuming. NMR is nondestructive, selective, and capable of simultaneous detection of a great number of low molecular mass components in complex mixtures. The great advantage of NMR also is the possibility of detecting different nuclei in different spatial and electronic environments. As a consequence, it is the preferable technique for molecular structure determination and the study of molecular interactions in solution. With development of high-field (500–900 MHz) NMR spectrometers with superconducting magnets and the possibility of recording two- and multi-dimensional spectra, NMR has become a powerful method for analyzing at the molecular level complex mixtures like beverages. The use of 2D homoand hetero-nuclear experiments, pulse sequences for the suppression of strong signals, and special probes, such as the so-called nano probe for µl quantities of samples, has Graham A. Webb (ed.), Modern Magnetic Resonance, 1597–1603.
C 2008 Springer.

enabled the characterization of minor compounds like anthocyanins [3]. The cold or cryogenic probe also is very promising. By cooling the key probe components to cryogenic temperature significant gains in sensitivity (up to 4 times in comparison to the room temperature probe) can be achieved and for a given amount of sample the experiment time is reduced by a factor of 16. The problems with signal overlapping and the limit detection may be overcome by the use of NMR hyphenation with techniques such as HPLC and MS spectrometry (LC-NMR, fully automated capillary scale LC-NMR, and LC-NMR/MS) [4–6]. Another way to overcome the problem of overlapping in NMR spectra of complex liquid mixtures is the application of diffusion-ordered spectroscopy (DOSY) [7]. NMR sensitivity can be further improved by combining hyphenated techniques with cryoprobe and by the incorporation of an online postcolumn solid-phase extraction (SPE) system for online preconcentration of analytes prior to transfer to NMR tube [4].

Alcoholic Beverages Alcoholic beverages comprise a large group of beverages containing varying amounts of alcohol (ethanol). Large scale of production includes beer, wine and China rice wine, and distilled spirits such as brandy, whisky, rum, gin, cognac, vodka, and China distilled spirit. The components of alcoholic beverages are very complex and over 1300 compounds have been identified in various beverages. Using NMR for studying the chemical composition of beverages attention was paid mostly to wine and more recently to beer.

Wine Water is the dominant component in wine. Ethanol, glycerol and other higher alcohols, sugars, organic acids, and various ions are present at high concentrations, while other aliphatic and aromatic alcohols, amino acids, phenolic compounds, and esters are present at much lower concentrations.

Part III

Characterization of the Chemical Composition of Beverages by NMR Spectroscopy

1598 Part III

Food Science

Part III

Already in 1976, it was demonstrated that NMR could be successfully used as a quick test for the determination of ethanol in wine, wine like beverages, and liquors [8]. Low-field (hence low cost) NMR spectroscopy has found application for the direct and rapid analysis of ethanol in wine and other alcoholic beverages [9]. A procedure to determine adulterations with diethylene glycol or methanol in natural wine has been described [10,11]. 13 C NMR was introduced in wine analysis by Rapp et al. [11]. It has been shown that 13 C NMR can successfully be used for the detection of sugars, alcohols, organic acids, and amino acids [12,13], and that amino acids can be used as fingerprints for monitoring of European wines [14]. Spraul and Hofman have demonstrated the usefulness of LC-NMR for detecting sugars in wine concentrate, and for distinguishing α from β glucose [15]. 1 H and 13 C NMR spectra of wines of different variety, vintage, and geographical area differ in the intensity of particular signals and in the appearance of some signals [16]. This offers the possibility to follow the variability in their chemical composition on the ground of 1 H and 13 C NMR spectra. Use of 1D and 2D homo- and hetero-NMR experiments enables a reliable assignment of 1 H and 13 C resonances of some organic acids and alcohols normally present in wines in higher amounts. Using WET suppression for simultaneous suppression of strong signals 1 H and 13 C assignment of 17 amino acids and of γ-aminobutyric acid commonly present in wine has been possible [17] (Figure 1). Brescia et al. have characterized the geographical origin of Italian red wines from Apulia using analytical and

A

NMR data (1 H shifts) in combination with multivariate statistical methods. Heavy metals and amino acids seem to be the species most responsible for the discrimination of the analytical and the NMR data sets, respectively [18]. The NMR technique of high-resolution DOSY has been applied to the complex mixture of compounds that constitute Port wine in order to achieve improved interpretation of the composition changes between wines of different ages [19]. Wine polyphenols have attracted much attention in connection with the so-called “French Paradox” (i.e. despite of the fact that the French consume a high fat diet they have low rates of coronary heart disease). This has been attributed to the phenolic compounds found in red wine frequently consumed in France. SPE-1 H NMR in combination with chemometric analysis has been used for monitoring changes in phenolic composition of Cabernet Sauvignon produced in the Southern region of Brazil over 1986, 1987, and 1995 vintages in order to corroborate the standardization process of this beverage [20]. The combination of NMR with LC and MS methods enabled the characterization of phenolics in the phenolic extract from red wine of the variety Bordˆo [6]. Anthocyanins, the most important group of plant pigments, are responsible for the color of red wines. They are also considered to be safe color additives in the food industry. Considerable research efforts, concerning their stability, includes structural determinations, the studies of multiequilibria in solution, and the characterization of exchange processes among the different species by NMR

B

Fig. 1. Part of GHSQC (A) and WETGHSQC (B) spectra of wine in the region between 0.40 and 5.00 ppm in the 1 H dimension and between 0.00 and 100.00 ppm in the 13 C dimension. In the WETGHSQC experiment WET suppression was applied on water frequency at 4.80 ppm, on ethanol at 3.64 ppm, on glycerol at 3.62 ppm, and on unassigned signals at 3.55 and 3.53.

NMR of Beverages

Beer Beer is a fermented beverage made from malted grains (usually barley), hops, yeast, and water. Fruits, herbs, and spices may also be used to give beer a particular character. The different combination of ingredients, production processes, and storage conditions give rise to an enormous variety of beers [34]. There has been great interest in studying the chemical composition of beer as this information is valuable for the assessment of beer quality and the development of new products. Besides water and ethanol, the major components of beer are carbohydrates that comprise fermentable sugars (e.g. glucose, maltose, and maltotriose), glucose oligosaccharides (dextrins), and arabinoxylanes. Important components in beer also are iso-α- and β-acids responsible for the bitterness of beer and a range of polyphenols. The latter have an impact on foam stability, taste, and clearness of beer and, being antioxidants, also on human health. The direct characterization of beer by NMR has been carried out by Duarte et al. [35]. With the aid of 1D and 2D NMR experiments about 30 compounds were identified. Many different metabolites such as adenosine, cystidine, histidine, phenylalanine, polyphenols, tryptophan, tyrosine, and uridine were also identified using 1D and 2D NMR experiments [6]. Since beer is a very complex mixture of a lot of different compounds occurring in a wide concentration range the signal overlapping and problems with the limit of detection may arise. These problems may be overcome by the use of LC-NMR/MS and DOSY technique. LCNMR/MS has been used for the identification of dextrins with degree of polymerization up to nine monomers [36]. The preliminary application of DOSY technique to beer has been presented by Gil et al. [37]. It was found that this technique was useful mainly for aliphatic and aromatic compounds in beer. HPLC-NMR has been used for the characterization of hop and beer bitter acids [38,39]. NMR data show clear distinction between cis and trans forms of the iso-α-acids, and unequivocally assign the structure of their acyl side chains. Differentiation between different types of beer and the classification of different types of beer regarding different raw materials used in their production has become important since the final product should be clearly declared if it is not produced only from the malt originating from barley. For this purpose, NMR spectra can be used as fingerprints. Using the combination of 1 H NMR spectra and PCA method most ales, lagers, and alcohol-free samples could be distinguished based on their aromatic composition [35,40]. Also, it has been shown that the low-field NMR region is highly sensitive toward different types of beer fermentation [40].

Part III

methods [21]. The identification of vitisin A, an anthocyanin occurring in some wines has been reported [22]. The anthocyanins in a particular LC fraction of red wines from the Coastal wine-growing region in Slovenia have been identified by NMR [3]. The structures of malvidin-derived wine pigments were determined by 1 H NMR and MS [23]. 1 H and 13 C analyses of anthocyaninderived pigments isolated from red wine has been reported [24,25]. Special attention has been paid to resveratrol found in high content in some European and USA red wines, and particularly in Brazilian red wines. Resveratrol is known for its biochemical and pharmacological effects, including anticarcinogenic, antioxidant, and antiinflammatory. trans-Resveratrol in commercial Brazilian red wines has been determined by 1 H NMR spectroscopy [26]. The assignment of 1 H and 13 C NMR spectra of resveratrol derivatives has been published [27]. The NMR studies of the structures of norisoprenoid glucosidic precursors of wine flavor have been reported [21]. The suitability of proton NMR for the quantitative determination of polydimethylsiloxanes (PDMS) in foodstuffs, particularly in wine, has been demonstrated [28]. PDMS are used in various stages of food processing as structure-shaping food additives and existing regulations of FDA and EU require the amount of added PDMS to be controlled. The nature of undesirable insoluble deposits adhering to the inner glass surface of bottles of red wine has been investigated by solid-state 13 C NMR [29]. The deposit is composed of a phenolic polymer of anthocyanins, procyanidins, and proteins. Weekley et al. have extended high-resolution NMR spectroscopy to the analysis of full bottle wine samples [30]. In this work examples of full bottle 13 C NMR spectra are provided and the application of the method to the measurement of acetic acid content, the detection of complex sugars, phenols, and trace elements in wine is discussed. It has been shown that using 5 mm diameter probehead with a 400 MHz spectrometer it was possible to measure fluorine-containing compounds directly in wine with a detection limit of 1.0 mg/l[31]. 11 B NMR has been used for determining boric acid esters in wine [32]. Boron was found in at least three different molecular species that occur in variable proportions. Troup and Hunter have discussed the achievements and uses of electron paramagnetic resonance in wine industry [33]. They presented the first observation of free radicals in red and white wines and found that the radical concentration, and therefore the antioxidant action, of white wines was increased by skin and oak wood exposure.

Alcoholic Beverages 1599

1600 Part III

Food Science

Part III

SNIF-NMR method has been used in the studies of different stable isotope ratios to investigate their values in connection with raw materials from which the sugars used in the fermentation process have originated [41].

Nonalcoholic Beverages The main difference in the chemical composition between alcoholic and nonalcoholic beverages is the absence of ethanol in nonalcoholic beverages. This simplifies the NMR characterization of composition since it is only the water signal that dominates in the spectra and should be suppressed by one of the suppression techniques.

Tea Tea is made from the tender leaves of two varieties of the plant Camellia sinensis: assamica and sinensis. Tea was discovered in China where it has been consumed for its medicinal properties since 3000 BC. Tea was introduced to Europe in 17th century and, later, to America and Africa. Tea has become the most popular beverage after water throughout the world and has attracted more and more attention because of reported health benefits owing to its content of antioxidants and, particularly, for its potential for preventing cancer and cardiovascular diseases. Many kinds of tea are produced that can be classified principally into three types: green (unfermented), oolong (semifermented), and black (fully fermented). The important components of tea are the methylxanthine alkaloids, including caffeine and theophylline, polyphenols-flavonoids anthocyanins, catechins, quercetin, kaempferol and myricetin, and volatile and aromatic components. Aminoacids, carbohydrates, lipids, vitamins, minerals, and trace elements were also found. In the manufacture of black tea, catechins can be oxidized to form orange-colored theaflavins and brownish thearubigins. The overall chemical content of tea is influenced by genetic strain, climate, altitude of growth, type of soil, use of fertilizers, and even the processing of tea leaves. The complete and unequivocal 1 H and 13 C assignments for a range of green tea polyphenols have been reported [42]. The same authors also assigned 1 H and 13 C resonances of theaflavin, theaflavin monogallate, and theaflavin digallate [43]. Using 2D NMR the structures of procyanidin and proanthocyanidin oligomers have been elucidated [44,45]. NMR has been used to clarify the antioxidative molecular mechanism of catechins [46]. It has been shown that NMR can be used to analyze simultaneously catechins, amino, organic, phenolic, and fatty acids, and sugars from a single green tea extract [47]. In this study, the data of just under 200 green teas from China, Japan, Vietnam, India, Indonesia, and Bangladesh were

collected. It has been pointed out that the geographical origin may be indicated by the chemical composition of green teas provided that a representative number of samples for each population is available. Solution 1 H and 13 C spectra of commercial samples of green and black teas in acetone-d6 and DMSO-d6 and the solid-state NMR spectra have been recorded and 13 C chemical shifts of (+)catechin in solution and in the solid state assigned [48]. Differences observed in chemical shifts are mainly due to the formation of stronger, intermolecularly hydrogen bonded species in the solid, with respect to those formed in solution. Catechins have received considerable attention due to their diverse pharmacological activities, the commonly discussed mechanism of which is their antioxidative activity. ESR study on the structure-antioxidant activity relationship of tea catechins and their epimers has been performed [49]. Significant differences were found between the scavenging effects of the catechins and their epimers if their concentrations were low. The scavenging abilities of catechins were stronger than those of their corresponding epimers. Tea infusions and dry tea leaves contain substantial amounts of aluminum, a powerful neurotoxicant that also has a potential for skeletal and hematopoietic toxicity. Hence, the consumption of large amounts of tea may significantly contribute to direct intake of aluminum. It has been shown that 27 Al NMR can be used nondestructively to determine aluminum at the ppm level in 0.5 cm3 of complex matrices such as tea infusions and leaf digests [50]. 27 Al NMR has been used for characterization of aluminum complexes in tea extracts [51,52]. These studies provide evidence that Al forms complexes with catechins and organic acids.

Coffee The beverage commonly called coffee is made from the roasted and ground beans of the coffee “tree”—a tropical evergreen shrub. There are three species of coffee “tree,” arabica, originating from Ethiopia, liberica from West Africa, and robusta from the Congo. There are many different varieties of coffee beans due to the wide range of climatic differences in these areas. Coffee is composed of more than 700 compounds. Coffee’s chemical composition is determined by a complex interaction of agricultural factors, roasting, blending, and brewing. In the order of abundance the water-soluble constituents ranging from 8% to 0.1% are phenolic polymers, polysaccharides, chlorogenic acids, minerals, water, caffeine, trigonelline, organic acids, sugars, lipids, and aroma compounds. An extensive study using high-resolution 1 H NMR has been performed on expresso cups of two arabica and one robusta samples, and one water extract from ground green

NMR of Beverages

Coffee solutions exhibit the strong free radical scavenging ability.

Fruit Juices Fruit juices contain organic acids, sugars, amino acids, phenolics-particularly anthocyanins, vitamins, βcarotene, mineral salts, flavors originating from essential oils, and may contain some alcohol. The concentrations of organic acids, sugars, and amino acids are of comparable order of magnitude to those in wine. Composition and concentration of phenolics vary with the sort of the juice. Rapp et al. have demonstrated the usefulness of 13 C NMR for the identification and quantification of amino acids in fruit juices and juice concentrates [12]. Belton et al. have assigned the 1 H NMR signals of different fruit juices to characterize several classes of compounds [57,58]. Significant spectral differences between cultivars were found. The results point to the usefulness of NMR in speciation and authentication studies and in the analysis of biochemical changes occurring in fruits and their juices. Vogels et al. have demonstrated the usefulness of NMR (peaks in the 1 H spectra) combined with PCA for detecting the adulteration in orange juices [59]. Colquhoun has studied the chemical composition of plum, raspberry, and strawberry purees/juices by 1 H NMR and chemometric methods [60]. For pure juices spectral differences are so great that visual identification is possible. To distinguish pure from adulterated juices (addition of small, fixed amount of plum, or apple juice to raspberry juices), multivariate analysis of NMR spectra is necessary. Several quercetin and phloretin glycosides in an apple peel extract have been identified by HPLC-NMR/MS [61]. NMR provided the information on the identity of the compounds, the α and β conformations and the position of the glycosides on quercetin and phloretin. The paper by Gil et al. describes the application of NMR spectroscopy and LC-NMR/MS to the direct analysis of grape juice [6]. It has been shown that LC-NMR/MS can overcome the problems of signal overlap and the low peak intensity and aids significantly in the identification of aromatic compounds. DOSY NMR has been used for analyzing the composition of apple and grape juices [37]. Particularly useful is DOSY analysis in cases were NMR alone leads to ambiguous assignments. A series of compounds having uncoupled methyl signals but differing in size have been identified; the sugar region only allows interpretation in the anomeric region where signal overlap is reduced; in the aromatic region the larger compounds are identified by slower diffusivities. Liquid- and solid-state NMR spectroscopy have been used to follow the compositional changes in mango juice and mango pulp during ripening [62]. It has been shown that NMR can aid in controlling postharvest quality and in

Part III

coffees and the chemical changes due to the roasting process have been followed [53]. 1 H spectra of coffee in aqueous solution show a complicated pattern of signals. From the 1 H spectrum in H2 O/DMSO/HCL (HCL was added for better resolution and DMSO for dissolution of less hydrophobic molecules) it was possible to identify lactic, citric, free quinic, acetic, and formic acids whose concentration could be directly measured. The caffeine molecule was clearly identified as well as chlorogenic acids. The signals due to trigonelline and the signals of the methylene groups of the fatty acids were also observed. 1 H spectra of chloroform extracts are dominated by the signals from caffeine and triglycerides. Many other signals can be related to hydrophobic aromatic substances which were not detectable in the aqueous phase. Many aldehyde carbonyl signals from aldehydes were found. The signal found at 1.84 ppm seems to be a possible marker of the robusta variety. It was also possible to observe the resonances due to sterols. The effect of roasting has been studied by observing the relative integrals of several selected resonances. The most pronounced is an increase of the concentration of free quinic acid during the roasting and a decrease of the concentrations of citric, chlorogenic, and formic acids. A decrease of the amount of trigonelline and an increase of the amount of its degradation product N -methyl pyridine is notable. Significant differences in the relative intensities of signals originating in aldehydes from the three coffee samples have been observed. The Columbian sample, with a typical sweet and fruity taste, shows a higher content of aldehydes with respect to the Kenya sample. The 1 H NMR spectra of a complex mixture such as coffee are difficult to interpret in details. Using multivariate statistical analysis it is possible to simplify the spectral data into a series of principal components that cover almost all the spectral variability. This allows for rapid comparison between many spectra, and the identification of differences within or between sample sets from the NMR data. The inherent ability of NMR to provide information about the chemical composition of a sample will allow the identification of source of variation. In this way the presence of inherent differences between coffees produced by different manufacturers and those produced by the same manufacturer has been demonstrated by identifying 5-(hydroxymethyl)2-furaldehyde as a marker compound using the structural characteristics determined by NMR [54]. Using NMR and LC-MS it has been shown that antibacterial components in the coffee extract are protocatechuic acid (3,4-dihydroxy benzoic acid), chlorogenic acid, and caffeic acid [55]. Electron paramagnetic resonance has been used for the characterization of free radicals in soluble coffee [56].

Nonalcoholic Beverages 1601

1602 Part III

Food Science

Part III

developing appropriate technologies to enhance the marketable lifetime of the mango, one of the most important commercial fruit crops. NMR with chemometrics has successfully been used in the discrimination between the orange juice and the pulp wash which is paler, more bitter, and regarded as a lower quality product [63]. It has been shown that dimethylproline is a marker compound for discrimination between two types of juice and can be used for orange juice adulteration.

References 1. Broustet JP. Heart. 1999;81:459. 2. Fr´emont L. Life Sci. 2000;66:663. 3. Koˇsir IJ, Lapornik B, Andrenˇsek S, Golc Wondra A, Vrhovˇsek U, Kidriˇc J. Anal. Chim. Acta. 2004;513:277. 4. Corcoran O, Stone Wilkinson P, Godejohann M, Braumann U, Hofmann M, Spraul M. Am. Lab. 2002;34:18. 5. Smith JP, Hinson-Smith V. Anal. Chem. 2001;73:155A. 6. Gil AM, Duarte IF, Godejohann M, Braumann U, Maraschin M, Spraul M. Anal. Chim. Acta. 2003;488:35. 7. Barjat H, Morris GA, Smart S, Swanson AG, Williams SCR. J. Magn. Reson. B. 1995;108:170. 8. Anders U, Titgemeier F, Hailer G. Z. Lebensm. Unters. Forsch. 1976;162:21. 9. Guillou M, Tellier C. Anal. Chem. 1988;60:2182. 10. Rapp A, Spraul M, Humpfer E. Z. Lebensm. Unters. Forsch. 1986;182:419. 11. Rapp A, Markowetz A, Spraul M, Humpfer E. Dtsch. Lebensm. Rundsch. 1987;83:375. 12. Rapp A, Markowetz A, Niebergall H. Z. Lebensm. Unters. Forsch. 1991;192:1. 13. Rapp A, Spraul M, Humpfer E. In Vino Analytica Scientia, Bordeaux;359:1997. 14. Holland MV, Bernreuther A, Reniero F. In: PS Belton, I Delgadillo, AM Gil, G Webb (Eds). Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 1995, p 136. 15. Spraul M, Hofman M. In: PS Belton, I Delgadillo, AM Gil, G Webb (Eds). Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 1995, p 77. 16. Koˇsir I, Kocjanˇciˇc M, Kidriˇc J. Analusis. 1998;26:97. 17. Koˇsir IJ, Kidriˇc J. J. Agric. Food Chem. 2001;49:50. 18. Brescia MA, Caldarola V, De Giglio A, Benedetti D, Fanizzi FP, Sacco A. Anal. Chim. Acta. 2002;458:177. 19. Nilsson M, Duarte IF, Almeida C, Delgadillo I, Goodfellow BJ, Gil AM, Morris GA. J. Agric. Food Chem. 2004; 52:3736. 20. Maraschin RP, Ianssen C, Arsego JL, Capel LS, Dias PF, Cimadon AMA, Zanus C, Caro MSB, Maraschin M. In: PS Belton, AM Gil, GA Webb, D Rutledge (Eds). Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 2003, p 255. 21. Ramos A, Santos H. In: GA Webb (Ed). Annual Reports on NMR Spectroscopy, Vol. 35. Academic Press: London, 1999, p 179. 22. Baker J, Timberlake CF. J. Agric. Food Chem. 1997; 45:35.

23. Fulcrand H, Cameira dos Santos P, Sarmi-Manchado P, Cheynier V, Favre-Bonvin J. J. Chem. Soc. Perkin Trans. 1996;1:735. 24. Mateus N, Silva AMS, Santos-Buelga C, Rivas-Gonzalo JC, De Freitas V. J. Agric. Food Chem. 2002;50:2110. 25. Mateus N, Carvalho E, Carvalho ARF, Melo A, Gonz´alezParam´as AM, Santos-Buelga C, Silva AMS, De Freitas V. J. Agric. Food Chem. 2003;51:277. 26. Maraschin M, Passos R, Duarte da Silva JMO, Dias PF, Araujo PS, Oltramari J, Fontana JD, Caro MSB. In: G Webb, PS Belton, AM Gil, I Delgadillo (Eds). Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 2001, p 136. 27. Koh D, Park KH, Jug J, Yang H, Hun Mok K, Lim Y. Magn. Reson. Chem. 2001;39:768. 28. Mojsiewicz-Pienkowska K, Jamr´ogiewicz Z, Lukasiak J. Food ´ Addit. Contam. 2003;20:438. 29. Waters EJ, Peng Z, Pocock KF, Jones GP, Clarke P, Williams PJ. J. Agric. Food Chem. 1994;42:1761. 30. Weekley AJ, Bruins P, Sisto M, Augustine MP. J. Magn. Reson. 2003;161:91. 31. Mortimer AD, Dawson BA. J. Agric. Food Chem. 1995;39:1781. 32. Lutz O, Humpfer E, Spraul M. Naturwissenschaften. 1991;78:67. 33. Troup GJ, Hunter CR. Ann. N.Y. Acad. Sci. 2002;957:345. 34. Baxter ED, Hughes PS (Eds). An Overview of the Malting and Brewing Processes. In beer Quality, Safety and Nutritional Aspects. RSC: Cambridge, 2001, pp 1–13. 35. Duarte IF, Barros A, Belton PS, Righelato R, Spraul M, Humpfer E, Gil AM. J. Agric. Food. Chem. 2002;50:2475. 36. Duarte IF, Godejohann M, Braumann U, Spraul M, Gil AM. J. Agric. Food. Chem. 2003;51:4847. 37. Gil AM, Duarte I, Cabrita E, Goodfellow BJ, Spraul M, Kerssebaum R. Anal. Chim. Acta. 2004;506:215. 38. Nord LI, Sorensen SB, Duus JO. Magn. Reson. Chem. 2003;41:660. 39. Pusecker K, Albert K, Bayer E. J. Chromatogr. A. 1999;836:245. 40. Duarte IF, Barros A, Almeida C, Spraul M, Gil AM. J. Agric. Food. Chem. 2004;52:1031. 41. Rossmann A. Food Rev. Int. 2001;17:347. 42. Davis AL, Cai Y, Davies AP, Lewis JR. Magn. Reson. Chem. 1996;34:887. 43. Davis AL, Cai Y, Davies AP. Magn. Reson. Chem. 1995;33:549. 44. Balas L, Vercauteren J. Magn. Reson. Chem. 1994;32:386. 45. Balas L, Vercauteren J, Laguerre M. Magn. Reson. Chem. 1995;33:85. 46. Sawai Y, Sakata K. J. Agric. Food Chem. 1998;46:111. 47. Le Gall G, Colquhoun IJ, Defernez M. J. Agric. Food Chem. 2004;52:692. 48. Martinez-Richa A, Joseph-Nathan P. Solid State Nucl. Magn. Reson. 2003;23:119. 49. Guo Q, Zhao B, Shen S, Hou J, Hu J, Xin W. Biochim. Biophys. Acta. 1999;1427:13. 50. Koch KR. Analyst. 1990;115:823. 51. Nagata T, Hayatsu M, Kosuge N. Phytochemistry. 1992;31:1215. 52. Mhatre SN, Iyer RK, Moorthy PN. Magn. Reson. Chem. 1993;31:169.

NMR of Beverages

59. 60. 61. 62. 63.

Colquhoun IJ, Dennis MJ, Spraul M. Magn. Reson. Chem. 1997;35:S52. Vogels JTWE, Terwel L, Tas AC, van den Berg F, Dukel F, van der Greef J. J. Agric. Food Chem. 1996;44:175. Colquhoun IJ. Spectrosc. Eur. 1998;10:1. Lommen A, Godejohann M, Venema DP, Hollman PCH, Spraul M. Anal. Chem. 2000;72:1793. Gil AM, Duarte IF, Delgadillo I, Colquhoun IJ, Casuscelli F, Humpfer E, Spraul M. J. Agric. Food Chem. 2000;48:1524. Le Gall G, Puaud M, Colquhoun IJ. J. Agric. Food Chem. 2001;49:580.

Part III

53. Bosco M, Toffanin R, de Palo D, Zatti L, Segre A. J. Sci. Food Agric. 1999;79:869. 54. Charlton AJ, Farrington WHH, Brereton P. J. Agric. Food Chem. 2002;50:3098. 55. Dogasaki C, Shindo T, Furuhata K, Fukuyama M. Yakugaku Zasshi. 2002;122:487. 56. Pascual EC, Goodman BA, Yeretzian C. J. Agric, Food Chem. 2002;50:6114. 57. Belton PS, Delgadillo I, Holmes E, Nicholls A, Nicholson JK, Spraul M. J. Agric. Food Chem. 1996;44:1483. 58. Belton PS, Delgadillo I, Gil AM, Roma P, Casuscelli F,

References 1603

1605

Fred van de Velde1,2 , and Harry S. Rollema2 1 Wageningen

Centre for Food Sciences, PO Box 557, 6700 AN Wageningen, The Netherlands; and 2 NIZO Food Research, Kernhemseweg 2, 6710 BA Ede, The Netherlands

Carrageenan Structure Carrageenan is the generic name for a family of linear, sulphated galactans, obtained by extraction from certain species of marine red algae (Rhodophyta) [1]. They are composed of alternating 3-linked β-d-galactopyranose (G-units) and 4-linked α-d-galactopyranose (D-units) or 4-linked 3,6-anhydro-α-d-galactopyranose (DA-units), forming the disaccharide repeating unit of carrageenans (Figure 1). The sulphated galactans are classified according to the presence of the 3,6-anhydro-bridge on the 4linked galactose residue and the position and number of sulphate groups. The most common types of carrageenan are traditionally identified by a Greek prefix indicating the major component of the sample. To describe more complex carrageenan structures a uniform letter code nomenclature has been developed by Knutsen et al. [2]. The three commercially most important carrageenans are called ι (iota)-, κ (kappa)-, and λ (lambda)-carrageenan, the corresponding letter codes are G4S-DA2S, G4S-DA, and G2S-D2S,6S. Two other types, called µ (mu)- and ν (nu)-carrageenan, which are the biological precursors of respectively κ- and ι-carrageenan, are often encountered in commercial carrageenan.

Experimental Setup Sample Preparation In general, carrageenan samples are sonicated prior to recording the spectra and the NMR experiments are carried out at elevated temperature to reduce the viscosity of the solution. Samples for 13 C NMR are prepared at relatively high concentrations (50–100 mg/ml in D2 O) as compared to 1 H NMR samples (5–10 mg/ml in D2 O). The solution is buffered by the addition of 20 mM Na2 HPO4 to avoid hydrolysis at the elevated temperature used during the acquisition. Sonicated carrageenan for 13 C NMR spectroscopy is preferably prepared according to the following procedure. A carrageenan solution (5 mg/ml in 20 mM Na2 HPO4 ) is sonicated for at least three times 30 min in melting ice (Heat Systems XL 2020 sonicator, 12 mm tip, power Graham A. Webb (ed.), Modern Magnetic Resonance, 1605–1610.
C 2008 Springer.

475 W, frequency 20 kHz). After centrifugation at elevated temperature, the sonicated solution is dialyzed against phosphate buffer (20 mM Na2 HPO4 ) and water and lyophilized. 13 C NMR-samples are prepared by dissolving the sonicated carrageenans (70 mg/ml) in D2 O containing 20 mM Na2 HPO4 and 10 mM internal standard (DSS). Sample preparation for the 1 H NMR experiments typically involves dissolving the carrageenan sample (5 mg/ml) at 80 ◦ C in D2 O containing 1 mM DSS and 20 mM Na2 HPO4 , followed by sonication for at least 1 h in a sonicator bath (Branson 2510).

Internal Standards The use of an appropriate internal standard is of importance to obtain valid chemical shift data. The IUPAC commission for molecular structure and spectroscopy recently recommended the use of 2,2-dimethyl-2-silapentane3,3,4,4,5,5-d6 -5-sulfonate sodium salt (DSS) as the primary reference for both 1 H and 13 C NMR spectroscopy in polar solvents. For most purposes the difference between DSS and TMS when dissolved in the same solvent are negligible and, therefore, the data from DSS and TMS scales may be validly compared without correction [3]. Recently, the application of DSS as internal standard for NMR spectroscopy of carrageenans is studied and reported by Van de Velde et al. [4]. Moreover, this publication summarizes the chemical shift data of common internal standards, such as DMSO, MeOH, TSP, and acetone, relative to DSS and as a function of temperature and pD. Chemical shifts (δ) given throughout this chapter are relative to internal DSS standard (δ = 0.000 ppm for both 1 H and 13 C according to the IUPAC recommendations [3]).

Methodology 1 H and 13 C NMR spectra are generally taken at 65 ◦ C on a 500 or 600 MHz spectrometer. For 1 H NMR typically 64 scans are taken using 90◦ pulses with an interpulse delay of 5 s (T1 values for the resonances of the anomeric protons of κ- and ι-carrageenan at 600 MHz are approximately 1.5 s [5]). 1 H decoupled 13 C NMR spectra are generally recorded using 90◦ pulses and interpulse delay

Part III

High Resolution NMR of Carrageenans

1606 Part III

Food Science

Part III

HO

OH

OH

HO O

O

O

OH

O O

OH

O

O

O O

OH

HO

OH

OH

G

D

G

DA

Carrageenan 3-linked 4-linked

Carrageenan 3-linked 44-linked

µ (mu) ν (nu) λ (lambda) ξ (ksi))

κ (kappa) ι (iota) θ (tetha (tetha) β (beta)

G4S G4S G2S G2S

D6S D2S,6S D2S,6S D2S

G4S G4S G2S G

DA DA2S DA2S DA

Fig. 1. Schematic representation of the repeating units of carrageenan. The letter codes refer to the nomenclature of Knutsen et al. [2].

of 1.5 s and acquiring 30,000 scans (T1 values for the anomeric carbons of κ- and ι-carrageenan at 400 MHz are approximately 0.5 s [5]). Quantitative applications of NMR techniques require a number of special precautionary measures. It is of paramount importance to ensure that the inter-pulse delay amounts to at least 4–5 times the longest T1 value of the resonances used for the analysis [5,6]. For 1 H NMR of samples containing mono- or disaccharides relatively long inter-pulse delays have to be applied. On the other hand 13 C NMR spectra of carrageenans can be recorded using relatively short inter-pulse delays. An additional complicating factor for 13 C NMR is a possible variation of the NOE effects for the different carbons. However in the case of carrageenans all carbon atoms of the repeating unit have one proton attached. Based on this, one would expect little variance in the NOE effect for the different carbons. Experiments with a mixture of κ- and ι-carrageenan using continuous composite pulse decoupling and gated decoupling produced apart from signal to noise ratio (S/N), the same result for the intensity ratio of the anomeric carbon resonances [5].

Analysis of the Major Carrageenan Types 1

H-NMR Spectroscopy

Proton NMR spectroscopy is the preferred method to quantify the composition of carrageenan samples. 1 H NMR has the advantage of a relatively high sensitivity. Therefore, spectra of samples with low carrageenan

concentration can be recorded in a couple of minutes. Generally, 5 mg sample is sufficient to record a 1 H NMR spectrum. Moreover, minor components can be detected at concentration down to 1%, due to the high S/N ratio. The quantification of different carrageenan types in a sample by 1 H NMR spectroscopy is based on the resonances of the α-anomeric protons (D- and DA-units) in the region from 5.0 to 5.7 ppm (Table 1 and Figure 2A) [4,6]. The signals for the β-anomeric protons (G-units) are less suitable for either identification or quantification purposes. This is the main drawback of proton NMR spectroscopy for carrageenan identification. Moreover, this limits this method to rather homogeneous samples, Table 1: Chemical shifts (ppm) of the α-anomeric protons of carrageenans referred to DSS as internal standard at 0 ppma Carrageenan β (beta) ι (iota) κ (kappa) λ (lambda) ν (nu) µ (mu)

Monosaccharideb

Chemical shift (ppm)

DA DA2S DA D2S,6S D2S,6S D6S

5.07 5.29 5.09 5.55 5.50 5.24

Source: van de Velde et al. [4]. a Carrageenan (30 mg/ml), DSS (10 mM) and Na HPO (20 mM) 2 4 in D2 O recorded at 65 ◦ C. b Codes refer to the nomenclature developed by Knutsen et al. [2].

High Resolution NMR of Carrageenans

B

5.8 5.6 5.4 5.2 5.0 4.8 4.6 4.4 4.2 4.0 3.8 3.6 3.4 (ppm)

Fig. 2. 1 H NMR spectra (A) and κ-carrageenan and λ-carrageenan.

13 C

108 104 100 96

92 88 84 80 (ppm)

76

72 68

64

60

NMR spectra (B) of the major carrageenan types from bottom to top: ι-carrageenan;

as resonances of some common additives and contaminants can interfere with the major carrageenan types (vide infra). 13

Part III

A

Analysis of Minor Components 1607

C-NMR spectroscopy

A typical 13 C NMR experiment takes approximately 12– 18 h to reach a reasonable S/N. However, due to the high chemical shift dispersion in a 13 C NMR spectrum each carbon atom of the repeating unit of a carrageenan variant gives rise to one single signal and the various carrageenan variants show unique and characteristic patterns (Table 2 and Figure 2B). Therefore, 13 C NMR spectroscopy is needed for an unambiguous identification of carrageenan samples. The 13 C NMR spectra of gel forming κ- and ι-carrageenan have been interpreted, using studies on model synthetic monosaccharide derivatives [7] or oligomeric polysaccharide fragments [8,9]. A wellresolved spectrum of λ-carrageenan was obtained much

later [10,11] due to some technical difficulties, which are usually explained by the high viscosity of λ-carrageenan solutions. It should be noted that the biological precursors (µ- and ν-carrageenan) and many other possible carrageenan diads given in Table 2 are usually found only as components of hybrid polymeric molecules.

Analysis of Minor Components In addition to the common carrageenan repeating units, the NMR spectra of carrageenan samples may show resonances originating from minor constituents and/or contaminants. These signals can result from substituents of the carrageenan backbone, such as pyruvic acid acetals and O-methyl groups. Contaminants originate either from the red algal biomass, which were not removed in the carrageenan isolation procedure [1,12] or from compounds (inorganic salts, sucrose, galactomannans) added by manufacturers to improve and/or control some

1608 Part III

Food Science

Part III

Table 2:

13

C NMR chemical shifts for the most common carrageenan structural unitsa Chemical shift (ppm) relative to DSS as internal standard

Carrageenan

Unitb

β (beta)

G DA G4S DA2S G4S DA G2S D2S,6S G4S D6S G4S D2S,6S G2S DA2S G2S D2S

ι (iota) κ (kappa) λ (lambda) µ (mu) ν (nu) θ (tetha) ξ (ksi)

C-1

C-2

C-3

C-4

C-5

C-6

104.8 96.8 104.4 94.3 104.7 97.3 105.6 93.9 107.0 100.3 107.0 100.5 102.6 97.8 105.4 94.9

71.7 72.4 71.5 77.2 71.7 72.1 79.6 77.0 72.7 70.7 72.4 78.6 79.8 77.1

82.6 81.6 79.1 80.0 81.0 81.4 78.0 71.8 80.5 72.8 82.4 70.4 79.4 79.6

68.6 80.3 74.3 80.6 76.3 80.5 66.4 82.6 76.3 81.4 73.4 82.1 67.0 81.8 66.9

77.6 79.3 77.0 79.3 77.0 79.1 76.5 70.9 77.1 70.5 77.2 70.4 77.1 79.2

63.5 71.7 63.5 72.0 63.5 71.7 63.5 70.3 63.5 69.9 63.6 70.0 63.4 72.4

Source: van de Velde et al. [4]. a Carrageenan (30 mg/ml), DSS (10 mM) and Na HPO (20 mM) in D O recorded at 65 ◦ C. 2 4 2 b Codes refer to the nomenclature developed by Knutsen et al. [2].

functional properties of carrageenan samples (solubility, viscosity, gel strength, etc.). These additives are generally encountered in standardized commercial carrageenan samples. Low molecular weight contaminations and additives can be removed by dialysis, prior to recording the spectra. The identification of these components is described below and their chemical shifts are summarized in Table 3.

the methyl, acetal, and carboxyl carbon atoms. The pyruvic acid acetals are also detected in the 1 H NMR spectra by the methyl proton resonances with a chemical shift of 1.44 ppm relative to DSS [4]. In addition to the methyl

Table 3: NMR chemical shifts for minor components and additives observed in carrageenan samples Chemical shift (ppm) relative to DSS as internal standard

Carrageenan Substituents NMR spectra of carrageenans may reveal the presence of O-methyl substitution of the carrageenan backbone. For example, as 3-linked 6-O-methyl-d-galactose residues in κ-carrageenan from Kappaphycus alvarezii [13] and in methylated ι-carrageenan from Clavivlonium ovatum [14]. These residues give specific signals for OMe at 61.0 ppm in 13 C NMR and 3.41 ppm in 1 H NMR relative to DSS [4,14]. Pyruvic acid is a common component of many complex carrageenans. It forms a cyclic acetal at positions 4 and 6 of 3-linked galactose residues. This substituent can be identified by characteristic signals of its carbons together with specific substitution effects on the corresponding carbon atoms of 3-linked d-galactose [15]. The 13 C NMR chemical shifts of these characteristic signals are 27.6, 103.6, and 178.0 ppm for, respectively,

Component/substituent

Observed unit

6-O-methyl-d-galactose Methyl group C-6 C-5 Pyruvic acid acetal Methyl group Acetal Carboxyl Floridean starch C-1 C-1 C-5 H-1 iso-Propanol Methyl group

13 C

NMR

61.0 73.9 79.4 27.6 103.6 178.0 102.6 74.3 74.0 26.4

1H

NMR

3.41

1.44

5.35 1.17

High Resolution NMR of Carrageenans

Contaminants Floridean starch, a branched (1-4,1-6)-α-d-glucan structurally related to plant amylopectins, is a reserve polysaccharide of red seaweeds. It is soluble in water and can accompany carrageenans in the extraction and precipitation steps. Apart from the detection of glucose in an acid hydrolysate, the presence of floridean starch can be confirmed by the well-known set of signals of 4linked α-d-glucopyranose residues in the 13 C-NMR spectrum (Table 3) [16]. In the 1 H NMR spectra, floridean starch is detected by the signal of the anomeric proton of the α(1 → 4)-linked d-glucopyranosyl at 5.37 ppm [16]. The resonance of the anomeric proton of the α(1 → 6)-linked d-glucopyranosyl appears at 4.98 ppm. Floridean starch can be removed by starch degrading enzymes or by any procedure of separation of acid and neutral polysaccharides. iso-Propanol is generally applied in industrial processing to precipitate carrageenans from the extraction liquid [1]. In the 1 H NMR spectra isopropanol gives a characteristic triplet at 1.17 ppm and in the 13 C NMR spectra the methyl carbon is detected at 26.4 ppm. Galactans of the agar group (differ from the carrageenan as the 4-linked units have the l-configuration) may be present in carrageenans, when mixed algal populations containing both agarophytes and carrageenophytes are used for carrageenan manufacture. The 13 C NMRspectral data of these so-called d/l-hybrids [6], may require complementary chemical analysis to unambiguously identify the agar-like structures. Some red algae contain water-soluble sulfated xylomannans and neutral xylans, but these polysaccharides are usually absent from the common carrageenophytes and only accidentally may be found in carrageenans. 13 C NMR spectra of both types of polysaccharides were published and may be used for their identification [17–19].

Additives Commercial carrageenan preparations are often blended with additives such as sucrose and glucose to adjust their viscosity. In 13 C NMR spectroscopy, the anomeric signal of α-d-glucopyranose residue of sucrose may overlap with that of DA2S residue of ι-carrageenan at about 94 ppm. Nevertheless, the presence of sucrose can be unambiguously detected by its other signals in the 13 C-NMR spectrum of the mixture [6]. These additives do not disturb the NMR analysis and can in

general be detected apart from the different carrageenan forms. Only the resonance of the anomeric proton of glucose overlaps with the resonance of µ-carrageenan in 1 H NMR. If necessary, glucose and sucrose together with inorganic salts can be removed by dialysis or ethanol precipitation. The gelling properties of carrageenans can be considerably improved by addition of several galactomannans or mannans of some higher plants [20,21]. NMR studies were used to elucidate the possible modes of interaction between polysaccharide components in these blends [22]. Since the 13 C NMR spectra of galactomannans are well-known [23], the corresponding spectra of mixed preparations recorded at elevated temperatures (above the melting points of gels) can be used to detect the presence of galactomannan additives and to calculate their content relative to carrageenan. In 1 H NMR spectra the anomeric proton of the α-d-galactopyranose is observed at 5.0 ppm and that of the β-d-mannopyranose at 4.8 ppm [23].

References 1. Van de Velde F, De Ruiter AG. In: A Steinb¨uchel, S DeBaets, EJ VanDamme (Ed). Biopolymers Vol. 6. Polysaccharides II: Polysaccharides from Eukaryotes, Wiley-VCH: Weinheim, 2002, p 245. 2. Knutsen SH, Myslabodski DE, Larsen B, Usov AI. Bot. Mar. 1994;37:163. 3. Harris RK, Becker ED, CabralDeMenezes SM, Goodfellow R, Granger P. Pure Appl. Chem. 2001;73:1795. 4. Van de Velde F, Pereira L, Rollema HS. Carbohydr. Res. 2004;339:2309. 5. Van de Velde F, Peppelman HA, Rollema HS, Tromp RH. Carbohydr. Res. 2001;331:271. 6. Van de Velde F, Knutsen SH, Usov AI, Rollema HS, Cerezo AS. Trends Food Sci. Technol. 2002;13:73. 7. Usov AI. Bot. Mar. 1984;27:189. 8. Rochas C, Rinaudo M, Vincendon M. Int. J. Biol. Macromol. 1983;5:111. 9. Greer CW, Rochas C, Yaphe W. Bot. Mar. 1985;28:9. 10. Falshaw R, Furneaux R. Carbohydr. Res. 1994;252:171. 11. Stortz CA, Bacon CE, Cherniak R, Cerezo AS. Carbohydr. Res. 1994;261:317. 12. Rudolph B. In: RE Martin, EP Carter, LM Davis, GJ Flich (Ed). Marine and Freshwater Products Handbook. Technomic Publishing Company, Inc.: Lancaster, U.S.A., 2000, p 515. 13. Bellion C, Brigand G, Prome J-C, Bociek DWS. Carbohdr. Res. 1983;119:31. 14. Chiovitti A, Bacic A, Craik DJ, Kraft GT, Liao M-L. Carbohydr. Res. 2004;339:1459. 15. Chiovitti A, Bacic A, Craik DJ, Munro SLA, Kraft GT, Liao M-L. Carbohydr. Res. 1997;299:229. 16. Knutsen SH, Grasdalen H. Bot. Mar. 1987;30:497. 17. Kovac P, Hitsch J, Shashkov AS, Usov AI, Yarotsky SV. Carbohydr. Res. 1980;85:177.

Part III

signal, specific substition effects on the chemical shifts of the other protons are observed in the anomeric proton region of the 1 H-NMR spectra [15].

References 1609

1610 Part III

Food Science

Part III

18. Usov AI, Dobkina IM. Bioorg. Khim. 1991;17:1051. 19. Kolender AA, Pujol CA, Damonte EB, Matulewicz MC, Cerezo AS. Carbohydr. Res. 1997;304:53. 20. Therkelsen GH. In: RL Whistler, JN BeMiller (Ed). Industrial Gums: Polysaccharides and Their Derivatives. Academic Press Inc.: San Diego, 1993, p 145.

21. Imeson AP. In: GO Phillips, PA Williams (Ed). Handbook of Hydrocolloids. Woodhead Publishing Ltd.: Cambridge, 2000, p 87. 22. Rochas C, Taravel F-R, Turquois T. Int. J. Biol. Macromol. 1990;12:353. 23. Grasdalen H, Painter TJ. Carbohydr. Res. 1980;81:59.

1611

C. Moreau and E. Guichard Unit´e Mixte de Recherche sur les Arˆomes (UMRA, INRA-ENESAD), INRA, 17 rue Sully, BP 86510, 21065 Dijon, France

Interactions between flavor compounds and non-volatile food components (proteins, polyphenols, carbohydrates, polyolosides, etc.) could significantly affect their impact on flavor perception [1,2]. Indeed, the presence of protein or polysaccharides can decrease aroma intensity by forming complexes depending on the nature, the concentration, and the physicochemical properties of aroma compound [3]. Headspace or chromatographic methods have been extensively used to investigate the retention of aroma compounds in food matrices [4,5]. However, these techniques only give indirect information on aroma binding and no information at a molecular scale. Spectroscopic techniques have also been used to study interactions between small molecules and macromolecules since they can give information on the binding site localization and on the nature of the interactions [6–9]. Of all of the spectroscopic techniques, NMR spectroscopy is probably the most effective in investigating ligand–macromolecule interactions as protein–protein, drug–protein, DNA–protein, lipid–protein or small molecules–macromolecules [10– 14]. Compared to other spectroscopic methods, NMR is a non-invasive technique and ligand–macromolecule complexes can generally be investigated in their natural environment (pH, temperature, buffer conditions) without spin labeling or fluorescent probe. A wide range of NMR methods can be used to characterize macromolecule–ligand complexes and gives access to structural and dynamic information on both the ligand and macromolecule [13,15–17]. The choice of which of these NMR methods to use, depends on various factors such as macromolecule and ligand size and nature, equilibrium binding constant, and the complex dissociation rate (K D ). They depend on the facts that, after complex formation, mean perturbations are observed on NMR parameters such as chemical shifts, resonance broadening, relaxation constants, diffusion coefficients, or the nuclear Overhauser effect (NOE). The equilibrium constant, K , for the ML complex, where M is the macromolecule and L the ligand is defined as: kon

M + L −→ ←− ML; koff

K =

kon [ML] = koff [M] [L]

Graham A. Webb (ed.), Modern Magnetic Resonance, 1611–1615.
C 2008 Springer.

The ligand binds to the macromolecule at a rate of kon , the association rate, and leaves its binding site at a rate of koff , the dissociation rate which depends on the ligand concentration, [L], and the interaction strength between the ligand and the binding site, K D . The dissociation constant of the ML complex is defined as K D = koff /kon = [M][L]/[ML]. For such a ML complex, the appearance of the NMR spectrum depends principally on the exchange kinetic determined by the koff rate, but also on the dissociation constant, K D . Depending on the NMR method used, the ligand and/or the macromolecule are observed and information on the dissociation constant and complex stoichiometry, binding site localization, and dynamic parameters (diffusion coefficient, relaxation times) can be obtained from the complex. The most popular method of NMR used to investigate interactions of a specific ligand–macromolecule complex is based on the chemical shift changes (and/or on resonance broadening) of the macromolecule using 1D and/or 2D NMR spectroscopy. The spectrum of the macromolecule alone is recorded and is then compared to the one in the presence of the ligand. Chemical shift changes ( δ) of the macromolecule are induced by the presence of the ligand when it binds to the macromolecule. The observed spectral changes will depend on the exchange rate (τ ) at which the complex is formed [18]. Three exchange regimes can be defined on the NMR timescale. 1. The slow exchange if 1/τ | δ| in which two resonance sets are observed corresponding to the free and complex molecule resonances. 2. The rapid exchange if 1/τ | δ| in which one set of resonances is observed but progressively shifted with the increase concentration of the ligand. Observed chemical shifts (δ obs ) of implicated groups in the complex are a weighted average of the free and bound molecule chemical shifts. 3. The intermediate exchange is observed if 1/τ ≈ | δ| in which a broadening of resonances is observed. For most of the studied systems, the rapid exchange is observed.

Part III

Flavor–Food Compound Interactions by NMR Spectroscopy

1612 Part III

Food Science

Part III

Other information can be obtained using this NMR method. Firstly, the chemical shift changes of specific chemical groups on the macromolecule provide information on the binding site localization. For example, 1 H–1 H TOCSY NMR spectra give information on shortand long-range coupling through chemical bonds and particularly for a protein, on signals of the NH–CHα region which is a fingerprint of the amino acid sequence. 1D and 2D NOE NMR experiments give information on proton coupling through space, thus showing which protons are ˚ The binding sites and the spatial spatially closed (500 MHz) [3] enabling comprehensive compositional information to be gathered for grape juice [2,4], apple juice [2,5], mango juice [6], orange juice [2,7], and tomato juice [8,9]. Figure 1 shows high-resolution 500 MHz 1 H NMR spectra of apple juice and grape juice, whereas Figure 2 shows an expansion of the 1 H-13 C heteronuclear single-quantum coherence (HSQC) spectrum of a Red Setter tomato juice. Graham A. Webb (ed.), Modern Magnetic Resonance, 1617–1621.
C 2008 Springer.

Due to the continuous need to detect low abundance compounds, often in crowded spectral, improvements in the qualitative analysis of juices by NMR have become increasingly necessary. Diffusion Ordered SpectroscopY (DOSY) has been explored in tomato juice (Figure 3) [9], apple juice and grape juice [10] as a potential tool to assign compounds with basis on their diffusivity values, in addition to NMR parameters. This method was found to be of value when identifying compounds in less overlapped spectral regions, the diffusivity information loosing resolution in the most crowded regions such as the carbohydrate region in juices. Hyphenated NMR involves the tandem use of NMR with liquid chromatography (LC) and mass spectrometry (MS), in the forms of LC-NMR or LC-NMR/MS. These methods achieve an extreme simplification of the full 1D NMR spectra as a result of chromatographic separation of compounds at the first stage of the experiment. Figure 4 shows an on-flow record obtained for grape juice, under conditions suitable for the analysis of aromatic compounds [4]. The lines shown correspond to spectra of single compounds or simple mixtures. The spread of NMR signals in the aromatic region as a function of retention time clearly shows the extension of spectral simplification achieved. In the case of grape juice, LC-NMR/MS enabled several cinnamic acids and their derivatives to be identified for the first time by direct analysis of the juice. The use of improved methods for qualitative analysis of fruit juices may, naturally, be explored with numerous aims in mind, including the study of biochemical changes with ripening [5,6] or with storage [5], detection of adulteration or contamination. However, the practical use of NMR in particular towards the latter applications necessarily involves vast numbers of samples and this brought about the tandem use of NMR with chemometrics. The potential of NMR for the authentication of orange juices was explored using 400 MHz 1D 1 H NMR together with Principal Component Analysis (PCA) and discriminant analysis [11]. In this way, authentic and non-authentic samples could be distinguished with basis on their NMR spectra, taking into account parallel measurements by alternative analytical methods [high-pressure liquid chromatography (HPLC) and atomic absorption spectroscopy]. As a result,

Part III

High-Resolution Nuclear Magnetic Resonance Spectroscopy of Fruit Juices

1618 Part III

Food Science

Part III

malic

a)

Val, Leu, Ile

quinic Ala Fru 9

8

7

Suc

2.5

ppm

2.0

1.5

ppm

Glc

9

8

7

6

5

4

3

b)

2

ppm

Succinic Malic Acetic Fru

Thr

GAB

Glc

9

8

9

7

8

7

Arg

2.5

ppm

6

5

Ala

4

3

2

2.0

1.5

Val, Leu, Ile

ppm

ppm

Fig. 1. High-resolution 1 H NMR spectra of (a) apple juice and (b) grape juice samples. Vertical expansions are shown for aliphatic and aromatic regions and some assignments are indicated. Reprinted with permission from Ref. [10].
C 2004 Elsevier.

additions of sucrose, beet medium invert sugar or sodium benzoate could be detected, as well as other deviations from authenticity requirements. The problem of orange juice adulteration with pulp wash is, at present, a pressing one in the fruit juice industry and this was addressed using PCA and discriminant analysis of 1 H NMR spectra of more than 300 orange and pulp wash juices [7]. It was shown that dimethylproline plays a key role in discriminating between the two sample types, with higher

levels in pulp wash. In addition, at least 21 more peaks were identified in the spectra as potential markers of this type of adulteration. An additional study aimed at distinguishing apple juices of different varieties (Spartan, Bramley, and Russet) using PCA and discriminant analysis of their 500 and 600 MHz 1 H NMR spectra [12]. Samples were successfully separated in terms of variety, through the analysis of different subregions of the spectra. Different levels of malic acid and of sucrose were found

High-Resolution NMR of Fruit Juices

1

50

3

4

2

5 55 7 8

60 10

7

9

65

13 70

12

14

16

15 19

17 18

75

6 11

80 28

22 23

20

21 24

25

Fig. 2. Expansion of carbohydrates region of 1 H-13 C HSQC spectrum of a Red Setter tomato juice. Key: α-Glc, α-Dglucose; β-Glc, β-D-glucose; α-FF, α-Dfructofuranose; β-FF, β-D-fructofuranose; β-FP, β-D-fructopyranose; 1, CH3 of methanol; 2, N(CH3 )3 choline; 3, α-CH asparagine; 4, α-CH aspartate; 5, α-CH glutamine and glutamate; 6, α-CH threonine; 7, C6 β-glc; 8, C6 α-Glc; 9, C6 β-FF; 10, C6 β-FP; 11, C1 β-FF; 12, C1 β-FP; 13, C3 β-FP; 14, C5 β-FP; 15, C4 α- and β-Glc; 16, α-CH malate; 17, C4, β-FP; 18, C5 α-Glc; 19, C2 α-Glc; 20, C3 α-Glc; 21, C4 β-FF; 22, C2 β-Glc; 23, C3 and C5 β-Glc; 24, C3 β-FF; 25, C4 α-FF; 26, C5 β-FF; 27, 5 α-FF; 28, C3 α-FF; 29, C1 α-Glc; 30, C1 β-Glc. Reprinted with permission from Ref. [9].
C 2003 Wiley.

26

27

85 90

29 30

95 100 5.0

4.5

4.0

3.5

ppm

0.4

D ×1010 (m2 s−1)

0.6

0.8

1

1.2

1.4 4.5

4

3

3.5 ppm

2.5

2

Fig. 3. DOSY spectrum of Red Setter tomato juice. Note that the error in the determination of diffusion coefficients was more substantial in the case of weaker resonances, which show spots stretched out along the f 1 direction in the DOSY map. Reprinted with permission from Ref. [9].
C 2003 Wiley.

Part III

ppm

High-Resolution NMR of Fruit Juices 1619

1620 Part III

Food Science

Part III

RT (min)

Fig. 4. On-flow NMR chromatogram obtained for a grape juice sample (flow rate 1.00 ml/min; injection volume 500 µl). Reproduced with permission from Ref. [4].
C 2003 Elsevier.

30 25

26.2 min 21.9 min

20 18.1 min 15

15.8 min 12.8 min

10 5

8

7

to be important discriminating variables, as well as some variations in minor constituents. The potential of NMR to quantify and identify a large number of compounds makes it a leading technique in the emerging area of metabolomic studies. Within this area, a matter of present and increasing interest is the safety testing of genetically modified organisms (GMO) and their food derivatives. The NMR detection of possible unintended effects following a genetic modification has been significantly explored for tomato [8,13–15], a major transgenic crop. Such studies have involved mostly the analysis of tomato extracts and the methods used comprise combination of 1 H NMR and statistics [8,13,14] as well as offline combination of LC and NMR[13,14]. The methodology proposed in ref. [13] and [14] is shown to enable the comparison of compositional alterations in a novel tomato crop with respect to related non-transgenic reference lines. The authors point out, however, that such differences must be statistically demonstrated to be significant compared to those arising from natural genetic and/or physiological variations [14]. The compositional effects of upregulating flavonoid biosynthesis in tomato (Lycopersicon esculentun), in order to enhance its antioxidant capacity, was specifically investigated [8]. 1 H NMR of tomato extracts, in tandem with multivariate and univariate methods, was shown to distinguish between modified and nonmodified tomatoes, not only with basis on the expected flavonoid increases but also on the levels of, at least, 15 additional compounds. Again, the authors point out that, although meaningful, such changes may fall into the same order of magnitude as natural variations observed in a field-grown crop. Further work along similar lines describes the application of 1 H NMR to the detection and quantification of metabolites related to the genetic origins

6

5

4

3

2

1

8.8 min 4.6 min

0 ppm

of strawberry fruit quality and to tomato transformants overexpressing hexokinase [15]. In this study, absolute quantification of metabolites was performed by signal integration, with a synthesised electronic reference signal and validated by enzymatic or HPLC measurements. Absolute quantitation by NMR usually involves signal integration, which is often not a straightforward subject due to extensive signal overlap and to possible intercomponent interactions that may affect visible signal areas. In spite of these difficulties, useful methods have been developed recently for the absolute quantitation of chlorogenic acid and of (-)-epicatechin in cider apple juices by NMR [16,17]. In both cases, 1,3,5-benzenetricarboxylic acid was added to the juice sample as an internal standard, allowing successful quantification of the selected metabolites by signal integration.

References 1. Eads TM, Bryant RG. J. Agric. Food Chem. 1986;34(5):834. 2. Belton PS, Delgadillo I, Holmes E, Nicholls A, Nicholson JK, Spraul M. J. Agric. Food Chem. 1996;44(6):1483. 3. Belton PS, Delgadillo I, Gil AM. Semin. Food Anal. 1998;3(3):223. 4. Gil AM, Duarte IF, Godejohann M, Braumann U, Maraschin M, Spraul M. Anal. Chim. Acta 2003;488(1):35. 5. Belton PS, Delgadillo I, Gil AM, Roma P, Casuscelli F, Colquhoun IJ, Dennis MJ, Spraul M. Magn. Reson. Chem. 1997;35:S52. 6. Gil AM AM, Duarte IF, Delgadillo I, Colquhoun IJ, Casuscelli F, Humpfer E, Spraul M. J. Agric. Food Chem. 2000;48(5):1524. 7. Le Gall G, Puaud M, Colquhoun IJ, J. Agric. Food Chem. 2001;49(2):580.

High-Resolution NMR of Fruit Juices

13. Lommen A, Weseman JM, Smith GO, Noteborn HPJM. Biodegradation 1998;9(6):513. 14. Noteborn HPJM, Lommen A, van der Jagt RC, Weseman JM. J. Biotechnol. 2000;77(1):103. 15. Moing A, Maucourt M, Renaud C, Gaudillere M, Brouquisse R, Lebouteiller B, Gousset-Dupont A, Vidal J, Granot D, Denoyes-Rothan B, Lerceteau-Kohler E, Rolin D. Funct. Plant. Biol. 2004;31(9):889. 16. Berregi I, Santos JI, del Campo G, Miranda JI, Aizpurua JM. Anal.Chim. Acta. 2003;486(2):269. 17. Berregi I, Santos JI, del Campo G, Miranda JI, Talanta 2003;61(2):139.

Part III

8. Le Gall G, Colquhoun IJ, Davis AL, Collins GJ, Verhoeyen ME. J. Agric. Food Chem. 2003;51(9):2447. 9. Sobolev AP, Segre A, Lamanna R. Magn. Reson. Chem. 2003;41(4):237. 10. Gil AM, Duarte I, Cabrita E, Goodfellow BJ, Spraul M, Kerssebaum R. Anal. Chim. Acta 2004;506(2):215. 11. Vogels JTWE, Terwel L, Tas AC, van den Berg F, Dukel F, van der Greef J. J. Agric. Food Chem. 1996;44(1): 175. 12. Belton PS, Colquhoun IJ, Kemsley EK, Delgadillo I, Roma P, Dennis MJ, Sharman M, Holmes E, Nicholson JK, Spraul M. Food Chem. 1998;61(1–2):207.

References 1621

1623

Huiru Tang1,2 and Yulan Wang3 1 School

of Chemical Science, Shaanxi University of Science and Technology, Xianyang, Shaanxi, P. R. China; 2 State Key Laboratory of Magnetic Resonance and Molecular and Atomic Physics, Wuhan Institute of Physics and Mathematics, the Chinese Academy of Sciences, Wuhan 430071, P.R. China; and 3 Biological Chemistry Section, Biomedical Sciences Division, Imperial College London, South Kensington, London SW7 2AZ, UK

Introduction Analysis of metabolites has always been an essential component in life science research since metabolites are the downstream and even endpoint products of a given biological process, thus carrying rich information about the process [1–8]. Whilst genes and proteins set the stage for what is likely to happen [1–8], metabolites indicate what has actually happened in the levels of proteins, transcriptions, and genes. Therefore, metabolite analyses will undoubtedly be vital to gain insights into life processes and to understand in greater depth the complexity of the whole organisms and their interactions with the environment. In the “post-genome” era, it is the total metabolite complement [3–8] (metabolome), not some parts of metabolites, that becomes the focus of attention. The changes of the metabolome, i.e., metabonomics [7], provide unique, dynamic, integrated, and holistic insights into the changes of the given biological process in metabolic level. These changes can be traced back, in turn, to the activities in the levels of proteins, transcriptions, and genes. The metabolomic studies of organisms (metabonomics or metabolomics) thus have increasingly become the inseparable part of the integrated systems biology strategy. The NMR-based metabonomics approach has become an established method and proven powerful for studying animal systems [7] by scrutinizing the changes of the metabolites in biofluids, such as blood plasma [8–12], urine [4– 9,13] and cerebrospinal fluid [14–19], and biopsy tissues [9,20–26]. Since biofluids contain information about biochemistry status of human body as a whole, analysis of the metabolite compositions in biofluids provides information on the responses of human systems to the effects of the internal physiological changes and external stimuli, such as diets, drugs, and diseases. Furthermore, metabolome of tissue samples [20–24] provides information about what is happening ex vivo or in situ and sometimes can provide direct evidences in vivo.

Graham A. Webb (ed.), Modern Magnetic Resonance, 1623–1630.
C 2008 Springer.

Amongst biofluids, blood plasma consists of macromolecules, such as lipoproteins and immunoglobulin, and a range of low molecular weight metabolites such as amino acids, hydroxycarboxylic acids, and sugars [10– 12]. In order to maintain consistency of the internal environments of human body, the content of blood plasma has to be kept reasonably stable until such homeostasis is perturbed by some “abnormal activities.” Therefore, changes of the blood plasma metabolome offer a sensitive handle to monitor the changes in human physiology. On the other hand, some waste products have to be disposed for the same purpose of homeostasis and such wastes are excreted and disposed in the form of human urine. Urine contains small amount of proteins, amino acids, hydroxycarboxylic acids, considerable amount of salts, and metabolites of exogenous compounds such as those from diets and drugs [8,27]. Therefore, human urine contains information reflecting the superposition of many biochemical processes taking place in the human body. Urine analysis also provides embedded information about the functioning state of some organs such as kidneys [28], liver [29,30], and heart [31] as well. High-resolution 1 H NMR spectroscopy can be used to detect the signals of a wide range of low molecular weight metabolites simultaneously in the complex samples such as biofluids. These signals carry information about the identity, concentration, and interactions of metabolites in the samples. If required, the endogenous and exogenous metabolites detected in such samples can be related to the pathophysiological effects such as the state of health [31–33], dietary, and drug interventions. Apart from analysis for human and animal samples, recently, high-resolution NMR spectroscopy have been employed to analyze food of plant origins [34–41]. Both liquid samples and tissues have been examined and metabolite resonances were assigned. Some plant extracts [42– 46] were also examined for various purposes. Such concept can be easily extended to studies of the plant metabolites or metabolome.

Part III

High-Resolution NMR Spectroscopy in Human Metabolism and Metabonomics

1624 Part III

Food Science

Part III

Most of the metabolites are organic compounds and consist of the elements such as carbon, hydrogen, nitrogen, oxygen, and sometimes phosphorus. It is therefore conceivable that 1 H, 13 C, and 31 P are natural reporting probes for NMR studies of metabolism. Sometimes, 2 H, 13 C, and 15 N labeling are also used to provide probes with little or no background signals. 19 F NMR has also been used in drug metabolism studies for the similar reasons. However, high-resolution 1 H NMR methods are by far the most widely employed methods due to the high sensitivities and many reviews have been published already on 1 H NMR studies of biological samples [2–8]. Therefore, this chapter is not aimed to provide a comprehensive review of this literature. Instead, it is aimed to discuss problems related in such studies and possible solutions to them. In almost all the biofluids and tissue samples, water is the nature’s chosen solvent and accounts for more than 80% (w/w) of the whole samples. The concentration of water protons are therefore in the level of 100 M in contrast to that of the metabolites in 10−6 –10−3 M or even lower. For nuclei other than 1 H, NMR detection is not affected by the amount of water. For 1 H NMR and related correlation spectroscopy, however, such drastic difference in concentration becomes a huge dynamic range problem and water signals have to be suppressed even though in certain cases these signals also carries important information.

Water Suppression For biological fluid samples, the ideal water suppression method should have (1) high effectiveness to suppress the water signals that are 105 –109 times more intense than metabolite signals, (2) high selectivity (or sharp suppression profile) to suppress the water signal without suppressing other signals close to water, and (3) high efficiency to suppress water signals within a short duration to minimize magnetization transfer with the exchangeable or interactive protons [12]. Besides, the method should be easy to be implemented in both 1D and multidimensional experiments. For the time being, the water suppression methods in use can be classified into three categories, spectral editing methods based on spin relaxation or translational diffusion properties between water and solutes, presaturation methods based on actively irradiating the water signals, and tailor-excitation methods by using the pulsed field gradient (PFG). The relaxation- and diffusion-based methods take advantage of the difference between water and solutes in T1 , T2 , or self-diffusion rates. The T1 and diffusionbased methods are only useful for solutes having T1 and self-diffusion coefficients drastically different from these of water. For example, in aqueous solution of macromolecules, one can observe signals of macromolecules by the choice of an appropriate relaxation recovery time in

the inversion-recovery experiment [47,48]. The gradientHMQC/HSQC experiments [49] can also be used to saturate the water signals by repeating the (13 C or 15 N) experiment rapidly. In this case, the water signals can be saturated due to its long T1 but can be done so only partially. The DRYCLEAN method [50] was based on the difference in diffusion coefficients between water and macromolecules by using a diffusion filter prior to detection. Whilst these methods are of some values for aqueous solution of macromolecules such as proteins, they are not useful for the biofluids (i.e. the solution of small metabolites) since the T1 and diffusion rates of solutes and water are on the similar scale. A T2 editing method has also been reported [51,52] and can be applied to some biofluids. This method, however, requires artificially shortening the T2 of water by adding contaminants [51,52] to the sample, is thus invasive. The irradiation methods are implemented by applying a secondary irradiation on resonance to the water peak. Amongst them, the simplest one is often referred to as “presaturation method,” which combines a single pulse with a weak continuous wave (CW) irradiation on resonance to water signals [53] during recycle delays and necessary durations. The strength of such irradiation has to be controlled to a reasonable level to avoid heatinginduced stability problems. Since the chemical shift of water is dependent on the sample temperature, the CW water suppression method requires good spectrometer stability, excellent magnetic field homogeneity and reasonable temperature controlling facilities. Generally, modern spectrometers can meet such demands. The presaturation method can be readily implemented in most pulse sequences and does not affect quantification of spectra with exception for the resonances close to water and interacting with water (via. NOE for instance). The other variant, called 1D NOESY [54] or NOSEYPRESAT, is equivalent to the first increment of the NOESY experiment [54] (90◦ -t1 -90◦ -tmix -90◦ -AQ-RD), where t1 is set to about 3 µs, the mixing time, tmix , to 0.1– 0.15 s combined with irradiating the water resonance during both the tmix and the recycle delay (RD). This method works well with systems where the exchange between water and metabolites is either slow or absent, resulting in much better suppression of the residual “hump” from the water signals [11,55]. However, unlike single-pulse version, this method is restricted to the flip angle of 90◦ and can only be used for quantitative work with an extremely long RD employed. Nevertheless, the FLIPSY (FLIP-angle adjustable one-dimensional NOESY) experiment [55] (θ-t1 -180-t2 -180-AQ-RD) can overcome the quantification problem by using a small flip angle (θ ∼ 10–30◦ ) and short t1 and t2 delays (∼2 µs). It has to be noted that the water suppression efficiency is inferior to the 1D NOESY method. It is obvious that the above methods have only one suppressed region (or “null point”) in

NMR Spectroscopy in Human Metabolism and Metabonomics

Assignments of the Metabolite Resonances The water suppressed 1 H NMR spectra of biological samples, such as urine, blood plasma, and tissues, are information-rich but data complex owing to (1) large number of resonances and (2) superposition of the resonances from a multitude of different chemical entities in multiphase compartments. To make meaningful use of such spectra, the first step is often to assign as many resonances

as possible. Normally, the assignments of metabolite signals rely on the positions (chemical shifts), multiplicity patterns (spin–spin coupling), and the intensity of the signals. However, extra care has to be taken when assigning the 1D 1 H NMR spectra since the NMR parameters can be affected by many factors such as pH, concentration, temperature, ionic strength, and interactions. To overcome the problems of resonance overlapping, 2D NMR experiments are normally employed to spread the resonances into greater space and to establish the atomic connectivity. Amongst them, 1 H–1 H 2D COSY [64], TOCSY [65], 1 H–13 C HSQC [66], HMQC [67], and HMBC [68] NMR experiments are useful for routine metabolite identification. Due to complex interactions the metabolites experiencing in the biological samples, “spiking” with the authentic compounds offers a simple but extremely useful way for unambiguous assignment. In practice, however, the complexity of the spectra of the heterogeneous samples such as blood plasma [2– 12,69] and tissues [20–23,70,71] are amplified by the presence of a large number of overlapping resonances resulting from a wide range of molecules, such as proteins, lipids, lipoproteins, and metabolites [2–12,20–23,69–73], which often have different NMR line shapes resulting from the biomolecules with a wide distribution of molecular weights and mobility. Consequently, the predominating peaks arising from the relatively high concentration molecules often obscure some substances at a low concentration. However, these low-concentration molecules can sometimes be extremely important for classifying toxicological or disease processes. For example, in blood plasma, the intense broad lipoprotein resonances often overwhelm the signals from small metabolites; in tissues, the dominant lipid resonances often obscure signals for the low-concentration metabolites. To overcome these problems, the hyphenated HPLC-NMR-MS can be employed to separate each component before spectroscopic detection. One can also use the non-invasive spectral editing methods to separate NMR signals without separating samples.

Spectral Editing in Biological NMR Spectroscopy A number of NMR spectral editing methods have been developed based on the variation in spin relaxation [11,69], diffusion [74], and spin coupling properties. For molecules experiencing weak interactions, the diffusion rate measured as the apparent diffusion coefficients is inversely correlated with the molecular size. Therefore, larger molecules diffuse slowly whereas small ones diffuse fast (unless they are bound to macromolecules). Thus, the complex NMR spectra of biofluids can be edited by taking advantage of the differences in diffusivity (or size) of molecules. Diffusion-edited NMR spectroscopy can

Part III

the whole spectrum. These irradiation methods can be implemented on any spectrometers. The other type of water suppression methods have been developed using PFGs in the form of so-called tailored excitation [56,57] or excitation sculpting [58] and therefore requires gradient facilities. Amongst them, WATERGATE (water suppression by gradient-tailored excitation) schemes are widely employed in the form of either three pairs [56] (W3) or five pairs [57] (W5) of symmetric pulses. The latter showed much improved (or narrower) suppression profiles [57]. However, such methods often employ lengthy pulse sequences and cause some phase distortions to multiplets with a large J couplings even with perfectly optimized spectrometers [58]. Furthermore, more than one “null points” are produced [57] and signals close to such points are attenuated. Another selective excitation sequence, WET [59–62], has also been reported, in particular for LC-NMR studies, with capability of suppressing several signals [61]. Whilst these sequences are reasonably efficient, elaborate optimization is often required posing a great hurdle for the non-specialists. On the other hand, a method called double PFG spin-echo (DPFGSE) have been reported [58] to suppress water resonance efficiently without causing any phase distortions and can be implemented in any sequences readily without elaborative optimization [58]. These gradient dependent methods require gradient facilities and often attenuate, to some extent, the broad signals such as those of lipids in plasma and tissues due to T2 relaxation or J modulations in the spin-echo-based methods. For all water suppression methods, the metabolite signals close to the “null points” (e.g. water resonance) will also be affected, if not suppressed completely. If these signals are essential for the purpose of studies, they can be recovered with magnetization transfer through scalar or dipolar coupling as in so-called RECUR-NMR methods [63]. RECUR-NMR methods also reduce, to some extent, the phase distortion caused by WATERGATE. From the above discussion, it becomes clear that the methods meet all criteria for an ideal method remains to be developed and in reality, the choice of water suppression method ought to be as simple as possible if the method is sufficient for the purpose of studies.

Spectral Editing in Biological NMR Spectroscopy 1625

1626 Part III

Food Science

Part III

be implemented by inserting a “diffusion filter” to attenuate or even eliminate resonances from the fast-diffusing molecules [74–77] prior to the detection. Some 1D and 2D diffusion-edited NMR methods have been employed to aid the assignments of resonances arising from the lipoproteins [74–76] in human blood plasma. Spin relaxation properties can also be exploited in spectral editing. For example, “relaxation-filters” can be inserted, before detection, to attenuate or eliminate the resonances selected according to their relaxation times. Such editing methods have been widely used in high-resolution NMR of biopolymers in the solid state [78–83] for many years and have proven useful. Three major spin relaxation phenomena can be encountered in high-resolution NMR of biological samples, spin–lattice relaxation characterized by T1 and T1ρ , and spin–spin relaxation described by T2 . Whilst the three relaxation times are identical for small molecules tumbling rapidly and isotropically, they can be different in multiphase, multicompartment systems, such as biofluids [12] and tissues [23,84]. To use T1 editing, an inversion-recovery sequence (180◦ x -τ 1 -90◦ x ) is used to prepare spins in such a way that NMR signal intensities are dependent on the length of τ 1 relative to T1 . The 90◦ pulse can serve as a read-pulse in 1D experiments or the excitation pulse for the 2D experiments. One can purposefully make a choice of τ 1 to have a positive peak (τ 1 > T1 *ln 2), negative peak (τ 1 < T1 *ln 2), or no peak at all (τ 1 = T1 *ln 2). The T1 editing has also been reported in 1D NMR as a method of solvent suppression [47,48]. Similarly, the magnetization can be prepared using a spin-locking sequence (90◦ x -τ SPy ) such that resonances with different T1ρ can be separated by choosing an appropriate spin-lock time, τ SP , and spin-lock power [12]. In the case of T2 editing, the magnetization can be prepared using a spin-echo sequence, typically a CPMG sequence [85], [90◦ x -(τ E -180◦ y -τ E )n ]. By choosing appropriate τ E and n, the signals can be separated according to their T2 (i.e. inverse of the line width). Generally, the broad NMR signals from macromolecules or bound small molecules are attenuated or even eliminated, leaving the sharper resonances from the mobile small molecules unaffected. The resonances having intermediate T2 will be attenuated to an intermediate extent. However, the T1 -, T1ρ -, and T2 -edited 1D NMR spectra of biological samples can still be too complex [12], since NMR spectral peaks from the low concentration metabolites can still be severely overshadowed. Consequently, it is necessary to extend the relaxation-editing approach to multidimensional NMR spectra to establish atomic connectivity and to achieve resonance assignments unambiguously. Examples for T1 -, T1ρ -, and T2 -edited 2D NMR methods have already been reported in studying blood plasma and biological tissues [12,86–90]. NMR spectra can also be edited according to the multiplicity using J -resolved spectroscopy or the nature of carbons

(CH3 , CH2 , CH, or C) by using DEPT [91], INEPT [92] for 13 C NMR and MAXY [93–98] for 1 H NMR. These spectral editing techniques can be used in conjunction with most of 2D NMR sequences.

Other Useful Techniques For most heterogeneous biological samples such as cells and tissues, 1 H NMR lines are broadened by residual dipolar interactions, chemical shift anisotropy and susceptibility [99,100]. With the magic angle spinning, these linebroadening effects can be eliminated and high-resolution magic-angle spinning NMR (HRMAS) [20,21,101] spectra can be obtained with similar quality to that of the solution state NMR. In addition, gradient is also available at the magic angle, allowing the gradient-enhanced NMR spectroscopy to be carried out effectively as well. However, automation in terms of shimming and sample exchanges is still in their infancy even though some automatic sample exchangers are commercially available. Therefore, throughput of the HRMAS NMR remains low (two to three samples per hour). In contrast, high throughput facilities for solution state NMR are well developed and can be in the forms of either flow –injection [102] or automatic sample exchanger. Both methods can screen up to 400 samples per day with options of sample recovery. For flow NMR, the samples are handled with a liquid handler and a sample transfer system together with a flow NMR probe. The whole process can be automatically executed including sample preparation, delivering samples to the detection cells, matching and tuning of probe, pulse length calibration, shimming, data acquisition, and processing. The commercial flow probes include normal 5 mm probe which can handle 300–500 µl samples and micro-flow probes such as micro-capillary probes [103,104] that can handle samples with a volume as small as 3 µl. Automatic sample exchangers can also be used to handle samples in NMR tubes and, in this case, an exchanger, a lot of spinners, and NMR tubes are required, even though there is no need for a liquid handler, sample transfer systems, and Flow NMR probes. NMR techniques are intrinsically insensitive and any attempts to improve the detection sensitivity have to consider ways to reduce noises (or increase signal-to-noise ratio). There are two types of “noises” in NMR spectroscopy, electronic noises, and chemical noises [8]. The chemical noises arise from the signal overlapping in particular in complex mixtures. To reduce the chemical noises, NMR has to be done with higher magnetic field to increase the chemical shift dispersion or with reduced complexity, ideally, in the form of a single compound by using separation techniques. Such separation can be done either separately or by hyphenating with NMR. The possible hyphenation between separation techniques and NMR

NMR Spectroscopy in Human Metabolism and Metabonomics

NMR-Based Metabonomics Techniques The advantages of NMR spectroscopy include (1) detecting all the NMR visible molecules simultaneously, non-invasively, and non-destructively, (2) providing rich molecular information such as structure, dynamics, interactions, pH, and concentrations, and (3) making it possible to do in vivo, in vitro,ex vivo, and in situ studies. The molecular information so obtained includes that from both the endogenous and exogenous metabolites related to the human metabolisms. However, thousands of resonance signals so generated, e.g. from 1 H NMR of a urine sample, present a challenge to data interpretation. Such challenge is amplified even further by the requirements of multiple parallel sampling. For example, multiple samples are re-

quired for both controls and non-control samples in order to obtain statistically meaningful results. In such cases, visualizations of the spectra by conventional techniques (e.g. human naked eyes) are no longer adequate or possible at all for the data analysis. The large density of the NMR spectral data, nevertheless, can be regarded as fingerprints of the biological systems. Then multivariate data analysis can be readily applied on the NMR data to reduce the complexity and to visualize patterns inherent in the data. Such combination provided the basic methodology for the NMR-based metabonomics/metabolomics studies. In fact, the combination of NMR with the multivariate statistical data analysis, such as principal components analysisand partial least square-based methods [30,116,117], has been extensively applied to study the human inborn errors of metabolism [118,119], the biochemical consequences of genetic strain differences in mice [120,121], and the responses of the biological systems to drugs [23,122] or other xenobiotics [123]. NMR detection strategies for studying the dietary effects on human metabolism are similar to that for drug effects and some results have already been reported using NMR-based metabonomics. For example, an elevation in urinal excretion of TMAO [124] is reported for subjects who consumed fish meals. Zuppi et al. [125] reported that the level of lipids and carbohydrates in diet profoundly affects the levels of urinary metabolites such as alanine, lactate, glycine, and hippuric acid. In another case, after 6 weeks of Atkins’ diet containing high lipid, protein but no carbohydrate, a male subject showed elevation of citrate and ketobodies (e.g. β-hydroxybutyrate, acetoacetate) and depletion of alanine in his blood plasma compared to the same subject with normal diet. Furthermore, the level of the low-density lipoproteins (or bad cholesterols) was also elevated after Atkins’s diet (Y.L. Wang, unpublished results). The NMR-based metabonomics methods have also been used to monitor the human metabolomic changes following dietary intervention with soy isoflavones [126]. Substantial intersubject variations have been detected in the plasma metabolomes of five subjects [126]. Although the extent of metabolic responses to soy intervention, amongst volunteers, was subject dependent, the nature of such responses were consistent across subjects by showing changes in plasma lipoproteins, amino acids, and carbohydrates [126]. Similar observations were also obtained from a study of the effects of chamomile tea ingestion on human metabolism using metabonomics approach [127]. Clear differentiation between the samples obtained before and after chamomile ingestion was observed based on subtle changes in the spectral signature relating to increased excretion of hippurate, glycine, together with depletion in the levels of urinary creatinine. It was concluded that perturbation of the aromatic species might be indicative of altered intestinal microbial activity. Samples obtained in the 2-week period after daily chamomile ingestion formed

Part III

detection now includes HPLC-NMR [105,106], HPLCNMR-MS [107–109], HPLC-SPE-NMR-MS [110], CENMR or CEC-NMR [111,112], and SFC-NMR [113– 115]. In these hyphenations, the purified fractions (peaks) from HPLC, capillary electrophoresis (CE), capillary electrochromatography (CEC), and supercritical fluid chromatography (SFC) can be measured with NMR or/and MS to obtain structural information. In LC-NMRMS, the whole chromatogram can be detected in on-flow, direct stop-flow, or timed slice modes. One can also store interesting chromatographic peaks into separate loops and then carry out NMR/MS studies by transferring the loop content. Similarly, one also can store these peaks on a solid-phase extraction (SPE) cartridge and then elude the cartridges individually to determine the molecular characteristics of the peaks [110]. The SPE method offers higher detection sensitivity than loops due to much more compact peak elution and a great opportunity to trap the same peak several times to increase concentration of solutes, increasing detection sensitivity drastically. However, appropriate hardware is required and concentration of the solvent impurities will also be increased with multiple trapping. Hyphenated SFC and NMR and CE-NMR are less common but have enormous potentials for future applications. On the other hand, in order to reduce the electronic noises, one has to reduce the temperature at which the detection electronic devices work. This involves cooling the detection electronics down to about 20K to increase the sensitivity for about fourfolds and such facilities are now available from almost all the major manufacturers. The latest cryogenic probe can also be used as a flow probe (for flow-injection and LC-NMR work) as well as a conventional tube probe. With a cryogenic probe on a moderate (say, 600 MHz) spectrometer, the detection limit can go down to the nanogram level. However, for the cryogenic probes, the radiation damping remains a massive problem especially when dealing with biological samples.

NMR-Based Metabonomics Techniques 1627

1628 Part III

Food Science

Part III

an isolated cluster in the discriminant analysis map inferring that the metabolic effects of chamomile ingestion were not reversible during the 2-week post-dosing period. Such studies highlighted the potential for NMR-based metabonomics technology in the assessment of nutritional interventions, and demonstrated the sensitivity of the approach for characterizing subtle metabolomic changes in the presence of a high degree of variation from genetic and environmental sources.

Future Perspectives It is clear that NMR-based metabonomics will be one of the most important applications of NMR spectroscopy in studies at the interface of chemistry and biology. Increasingly, our understandings to biological systems are required to treat such systems as an integrated entity or to use systems biology approaches. The information in the metabolism level is therefore essential to understand the functional aspects. In nutritional sciences, the NMRbased metabonomics, integrated together with nutrigenomics, nutritranscriptomics, and proteomics, will undoubtedly provides more complete understandings to the specificity of human responses to nutrients and assists the identification of surrogate tissues and biomarkers that can be used to predict a response. In such a revolutionary era, however, the shortcomings of NMR spectroscopy will have to be further addressed. From the hardware aspects, it is predictable that the NMR frequency will move beyond 1 GHz (1000 MHz) to further increase the chemical shift dispersion and to reduce the overlapping (or chemical noises). In addition, detection sensitivity will also be increased by developing the higher field spectrometers perhaps coupled with further reduction of the electronic noises of the detection circuits. For both aspects, the development of high temperature (e.g. liquid nitrogen or even room temperature) superconductors will be extremely helpful since not only the development of magnet production technology will benefit from such materials but also, when such materials are used for the detection circuits, the electronic noises will be drastically reduced. From the spectroscopy point of view, substantial efforts will be needed to overcome the radiation damping effects as the sensitivity increases. The other important area includes developing ultra-rapid methods to record the multidimensional NMR. Some recent reports of ultrarapid multidimensional NMR spectroscopy are leading toward such progress already [128–135].

References 1. Fan WMT. Prog. NMR Spectr. 1996;28:161. 2. Nicholson JK, Wilson ID. Prog. NMR Spectr. 1989;21:449.

3. Lindon JC, Holmes E, Nicholson JK. Prog. NMR Spectr. 2001;39:1. 4. Nicholson JK, Connelly J, Lindon JC, Holmes E. Nat. Rev. Drug Discov. 2002;1:153. 5. Lindon JC, Nicholson JK, Holmes E, Everett JR. Concepts Magn. Reson. 2000;12:289. 6. Lindon JC, Holmes E, Nicholson JK. Curr. Opin. Mol. Ther. 2004;6:265. 7. Nicholson JK, Lindon JC, Holmes E. Xenobiotica. 1999;29:1181. 8. Lindon JC, Nicholson JK, Everett JR. Ann. Rep. NMR Spectr. 1999;38:1. 9. Nicholson JK. J. Anat. 1981;132:467. 10. Foxall PJD, Spraul M, Farrant RD, Lindon JC, Neild GH, Nicholson JK. J. Pharm. Biomed. Anal. 1993;11:267. 11. Nicholson JK, Foxall PJD, Spraul M, Farrant RD, Lindon JC. Anal. Chem. 1995;67:793. 12. Tang HR, Wang YL, Nicholson JK, Lindon JC. Anal. Biochem. 2004;325:260. 13. Keun HC, Ebbels TMD, Antti H, Bollard ME, Beckonert O, Schlotterbeck G, Senn H, Niederhauser U, Holmes E, Lindon JC, Nicholson JK. Chem. Res. Toxicol. 2002;15:1380. 14. Bloom M, Sternin E. Biochem. 1987;26:2101. 15. Sweatman BC, Farrant RD, Holmes E, Ghauri FY, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1993;11:651. 16. Koschorek F, Offermann W, Stelten J, Braunsdorf WE, Steller U, Gremmel H, Leibfritz D. Neurosurg. Rev. 1993;16:307. 17. Lynch J, Peeling J, Auty A, Sutherland GR. Can. J. Neurol. Sci. 1993;20:194. 18. Wevers RA, Engelke U, Wendel U, DeJong JGN, Gabreels FJM, Heerschap A. Clin. Chem. 1995;41:744. 19. Nicoli F, VionDury J, ConfortGouny S, Maillet S, Gastaut JL, Cozzone PJ. Compt. Rend. Acad. Sci. III-Sci. Life Sci. 1996;319:623. 20. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall PJD, Nicholson JK, Lindon JC. Anal. Commun. 1997;34:107. 21. Moka D, Vorreuther R, Schicha H, Spraul M, Humpfer E, Lipinski M, Foxall PJD, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1998;17:125. 22. Garrod S, Humpfer E, Spraul M, Connor SC, Polley S, Connelly J, Lindon JC, Nicholson JK, Holmes E. Magn. Reson. Med. 1999;41:1108. 23. Wang YL, Bollard ME, Keun H, Antti H, Beckonert O, Ebbels TM, Lindon JC, Holmes E, Tang HR, Nicholson JK. Anal. Biochem. 2003;323:26. 24. Bollard ME, Garrod S, Holmes E, Lincoln JC, Humpfer E, Spraul M, Nicholson JK. Magn. Reson. Med. 2000;44:201. 25. Cheng LL, Chang IW, Smith BL, Gonzalez RG. J. Magn. Reson. 1998;135:194. 26. Cheng LL, Chang IW, Louis DN, Gonzalez RG. Cancer Res. 1998;58:1825. 27. Holmes E, Foxall PJD, Spraul M, Farrant RD, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1997;15:1647. 28. Foxall PJD, Nicholson JK. Exp. Nephrol. 1998;6:409. 29. Sequeira S, So PW, Everett JR, Elcombe CR, Kelvin AS, Nicholson JK. J. Pharm. Biomed. Anal. 1990;8:945. 30. Holmes E, Nicholls AW, Lindon JC, Ramos S, Spraul M, Neidig P, Connor SC, Connelly J, Damment SJP, Haselden J, Nicholson JK. NMR Biomed. 1998;11:235.

NMR Spectroscopy in Human Metabolism and Metabonomics

61. Smallcombe SH, Patt SL, Keifer PA. J. Magn. Reson. A. 1995;117:295. 62. Zhang SM, Yang XB, Gorenstein DG. J. Magn. Reson. 2000;143:382. 63. Liu ML, Tang HR, Nicholson JK, Lindon JC. J. Magn. Reson. 2001;153:133. 64. Aue WP, Bartholdi E, Ernst RR. J. Chem. Phys. 1976;64:2229. 65. Bax A, Davis DG. J. Magn. Reson. 1985;65:355. 66. Bodenhausen G, Ruben DJ. Chem. Phys. Lett. 1980;69:185. 67. Bax A, Griffey RH, Hawkins BL. J. Magn. Reson. 1983;55:301. 68. Bax A, Summers MF. J. Am. Chem. Soc. 1986;108:2093. 69. Nicholson JK, Buckingham MJ, Sadler PJ. Biochem. J. 1983;211:605. 70. Griffin JL, Walker LA, Garrod S, Holmes E, Shore RF, Nicholson JK. Comp. Biochem. Physiol. B-Biochem. Mol. Biol. 2000;127:357. 71. Bollard ME, Holmes E, Lindon JC, Mitchell SC, Branstetter D, Zhang W, Nicholson JK. Anal. Biochem. 2001;295:194. 72. Foxall PJD, Parkinson JA, Sadler IH, Lindon JC, Nicholson JK. J. Pharm. Biomed. Anal. 1993;11:21. 73. Wevers RA, Engelke U, Heerschap A. Clin. Chem. 1994;40:1245. 74. Liu ML, Nicholson JK, London JC. Anal. Chem. 1996;68:3370. 75. Liu ML, Nicholson JK, Parkinson JA, Lindon JC. Anal. Chem. 2002;40:S83. 76. Liu ML, Tang HR, Nicholson JK, Lindon JC. Magn. Reson. Chem. 2002;40:S83–8. 77. Wu DH, Chen AD, Johnson CS. J. Magn. Reson. A. 1995;115:260. 78. Tang HR, Wang YL, Belton PS. Solid State NMR. 2000;15:239. 79. Tang HR, B. P. Hills in Webb GA, Belton PS, Gil AM, Delgadillo I (Eds). Magnetic Resonance in Food Science— A View to the Future. Royal Society of Chemistry: Cambridge, 2001, p 155. 80. Schmidtrohr K, Clauss J, Spiess HW. Macromolecules. 1992;25:3273. 81. Zumbulyadis N. J. Magn. Reson. 1983;53:486. 82. Tang HR, Hills BP. Biomacromolecules. 2003;4:1269. 83. Newman RH, Davies LM, Harris PJ. Plant Physiol. 1996;111:475. 84. Rooney OM, Troke J, Nicholson JK, Griffin JL. Magn. Reson. Med. 2003;50:925. 85. Meiboom S, Gill D. Rev. Sci. Instrum. 1958;29:688. 86. Williams PG, Saunders JK, Dyne M, Mountford CE, Holmes KT. Magn. Reson. Med. 1988;7:463. 87. Saunders JK, Stevens JD. Magn. Reson. Chem. 1986;24:1023. 88. Rabenstein DL, Fan S, Nakashima TT. J. Magn. Reson. 1985;64:541. 89. Trehout D, Desille M, Doan BT, Mahler S, Fremond B, Malledant Y, Campion JP, Desbois J, Beloeil JC, Certaines J, Clement B. NMR Biomed. 2002;15:393. 90. Rabenstein DL, Srivatsa GS, Lee RWK. J. Magn. Reson. 1987;71:175. 91. Doddrell DM, Pegg DT, Bendall MR. J. Magn. Reson. 1982;48:323.

Part III

31. Brindle JT, Antti H, Holmes E, Tranter G, Nicholson JK, Bethell HWL, Clarke S, Schofield PM, McKilligin E, Mosedale DE, Grainger DJ. Nat. Med. 2002;8:1439. 32. Nicholson JK, Sadler PJ, Bales JR, Juul SM, MacLeod AF, Sonksen PH. Lancet. 1984;2:751. 33. Foxall PJD, Mellotte GJ, Bending MR, Lindon JC, Nicholson JK. Kidney Int. 1993;43:234. 34. Belton PS, Delgadillo I, Gil AM, Roma P, Casuscelli F, Colquhoun IJ, Dennis MJ, Spraul M. Magn. Reson. Chem. 1997;35:S52–60. 35. Gil AM, Duarte IF, Delgadillo I, Colquhoun IJ, Casuscelli F, Humpfer E, Spraul M. J. Agric. Food Chem. 2000;48: 1524. 36. Le Gall G, Colquhoun IJ, Davis AL, Collins GJ, Verhoeyen ME. J. Agric. Food Chem. 2003;51:2447. 37. Sobolev AP, Segre A, Lamanna R. Magn. Reson. Chem. 2003;41:237. 38. Ni QX, Eads TM. J. Agric. Food Chem. 1993;41:1035. 39. Defernez M, Colquhoun IJ. Phytochemistry. 2003;62:1009. 40. Belton PS, Colquhoun IJ, Kemsley EK, Delgadillo I, Roma P, Dennis MJ, Sharman M, Holmes E, Nicholson JK, Spraul M. Food Chem. 1998;61:207. 41. Belton PS, Delgadillo I, Holmes E, Nicholls A, Nicholson JK, Spraul M. J. Agric. Food Chem. 1996;44:1483. 42. Wang YL, Tang HR, Nicholson JK, Hylands PJ, Sampson J, Whitcombe I, Stewart CG, Caiger S, Oru I, Holmes E. Plant. Med. 2004;70:250. 43. Fan TWM, Lane AN, Pedler J, Crowley D, Higashi RM. Anal. Biochem. 1997;251:57. 44. Fan TWM, Lane AN, Shenker M, Bartley JP, Crowley D, Higashi RM. Phytochemistry. 2001;57:209. 45. Bailey NJC, Sampson J, Hylands PJ, Nicholson JK, Holmes E. Plant. Med. 2002;68:734. 46. Bailey NJC, Oven M, Holmes E, Zenk MH, Nicholson JK. Spectr. Int. J. 2004;18:279. 47. Benz FW, Feeney J, Roberts GCK. J. Magn. Reson. 1972;8:114. 48. Price WS, Hayamizu F, Arata Y. J. Magn. Reson. 1997;126:256. 49. Ruiz-Cabello J, Vuister GW, Moonen CTW, van Gelderen P, Cohen JS, and Van Zijl PCM. J. Magn. Reson. 1992;100:282. 50. Van Zijl PCM, Moonen CTW. J. Magn. Reson. 1990;87:18. 51. Connor S, Nicholson JK, Everett JR. Anal. Chem. 1987;59:2885. 52. Connor S, Everett J, Nicholson JK. Magn. Reson. Med. 1987;4:461. 53. Hoult DI. J. Magn. Reson. 1976;21:337. 54. Kumar A, Ernst RR, Wuthrich K. Biochem. Biophys. Res. Commun. 1980;95:1. 55. Neuhaus D, Ismail IM, Chung CW. J. Magn. Reson. A. 1996;118:256. 56. Piotto M, Saudek V, Sklenar V. J. Biomol. NMR. 1992;2:661. 57. Liu ML, Mao XA, Ye CH, Huang H, Nicholson JK, Lindon JC. J. Magn. Reson. 1998;132:125. 58. Hwang TL, Shaka AJ. J. Magn. Reson. A. 1995;112:275. 59. Crofton DJ, Pethrick RA. Polymer. 1981;22:1048. 60. Ogg RJ, Kingsley PB, Taylor JS. J. Magn. Reson. B. 1994;104:1.

References 1629

1630 Part III

Food Science

Part III

92. Morris GA, Freeman R. J. Am. Chem. Soc. 1979;101:760. 93. Liu M, Farrant RD, Nicholson JK, Lindon JC. J. Magn. Reson. A. 1995;112:208. 94. Liu M, Farrant RD, Nicholson JK, Lindon JC. J. Magn. Reson. B. 1995;106:270. 95. Liu M, Nicholson JK, Lindon JC, Sanderson PN, Tranter GE. Magn. Reson. Chem. 1996;34:865. 96. Liu ML, Mao XA, Ye CH, Nicholson JK, Lindon JC. J. Magn. Reson. 1997;129:67. 97. Liu ML, Ye CH, Farrant RD, Nicholson JK, Lindon JC. Mol. Phys. 2001;99:1701. 98. Liu ML, Lindon JC. Curr. Org. Chem. 2001;5:351. 99. Mehring M. Principles of High Resolution NMR in Solids. Springer-Verlag: Berlin, 1983. 100. Fyfe CA. Solid State NMR for Chemists. C.F.C. Press: Guelph, ON, 1983. 101. Humpfer E, Spraul M, Nicholls AW, Nicholson JK, Lindon JC. Magn. Reson. Med. 1997;38:334. 102. Spraul M, Hofmann M, Ackermann R, Nicholls AW, Damment JP, Haselden JN, Shockcor JP, Nicholson JK, Lindon JC. Anal. Commun. 1997;34:339. 103. Webb AG. Prog. NMR Spectr. 1997;31:1. 104. Ciobanu L, Webb AG, Pennington CH. Prog. NMR Spectr. 2003;42:69. 105. Spraul M, Hofmann M, Dvortsak P, Nicholson JK, Wilson ID. J. Pharm. Biomed. Anal. 1992;10:601. 106. Spraul M, Hofmann M, Wilson ID, Lenz E, Nicholson JK, Lindon JC. J. Pharm. Biomed. Anal. 1993;11:1009. 107. Shockcor JP, Unger SE, Wilson ID, Foxall PJD, Nicholson JK, Lindon JC. Anal. Chem. 1996;68:4431. 108. Lenz EM, Greatbanks D, Wilson ID, Spraul M, Hofmann M, Troke J, Lindon JC, Nicholson JK. Anal. Chem. 1996;68:2832. 109. Lindon JC, Nicholson JK, Wilson ID. Prog. NMR Spectr. 1996;29:1. 110. Corcoran O, Spraul M. Drug Discov. Today. 2003;8:624. 111. Schewitz J, Gfrorer P, Pusecker K, Tseng LH, Albert K, Bayer E, Wilson ID, Bailey NJ, Scarfe GB, Nicholson JK, Lindon JC. Analyst. 1998;123:2835. 112. Pusecker K, Schewitz J, Gfrorer P, Tseng LH, Albert K, Bayer E, Wilson ID, Bailey NJ, Scarfe GB, Nicholson JK, Lindon JC. Anal. Commun. 1998;35:213. 113. Albert K, Braumann U, Tseng LH, Nicholson G, Bayer E, Spraul M, Hofmann M, Dowle C, Chippendale M. Anal. Chem. 1994;66:3042.

114. Albert K, Braumann U, Streck R, Spraul M, Ecker R. Fresen. J. Anal. Chem. 1995;352:521. 115. Albert K. J. Chromatogr. A. 1995;703:123. 116. Holmes E, Nicholson JK, Nicholls AW, Lindon JC, Connor SC, Polley S, Connelly J. Chemometr. Intel. Lab. Syst. 1998;44:245. 117. Tate AR, Foxall PJD, Holmes E, Moka D, Spraul M, Nicholson JK, Lindon JC. NMR Biomed. 2000;13:64. 118. Spraul M, Neidig P, Klauck U, Kessler P, Holmes E, Nicholson JK, Sweatman BC, Salman SR, Farrant RD, Rahr E, Beddell CR, Lindon JC. J. Pharm. Biomed. Anal. 1994;12:1215. 119. Holmes E, Foxall PJD, Nicholson JK, Neild GH, Brown SM, Beddell CR, Sweatman BC, Rahr E, Lindon JC, Spraul M, Neidig P. Anal. Biochem. 1994;220:284. 120. Gavaghan CL, Holmes E, Lenz E, Wilson ID, Nicholson JK. FEBS Lett. 2000;484:169. 121. Gavaghan CL, Wilson ID, Nicholson JK. FEBS Lett. 2002;530:191. 122. Holmes E, Nicholls AW, Lindon JC, Connor SC, Connelly JC, Haselden JN, Damment SJP, Spraul M, Neidig P, Nicholson JK. Chem. Res. Toxicol. 2000;13:471. 123. Azmi J, Griffin JL, Antti H, Shore RF, Johansson E, Nicholson JK, Holmes E. Analyst. 2002;127:271. 124. Alwaiz M, Mitchell SC, Ayesh R, Idle JR, Smith RL. Br. J. Clin. Pharmacol. 1987;23:614. 125. Zuppi C, Messana I, Forni F, Ferrari F, Rossi C, Giardina B. Clin. Chim. Acta. 1998;278:75. 126. Solanky KS, Bailey NJC, Beckwith-Hall BM, Davis A, Bingham S, Holmes E, Nicholson JK, Cassidy A. Anal. Biochem. 2003;323:197. 127. Wang YL, Tang HR, Holmes E, Nicholson JK. J. Agric. Food Chem. 2005;53:191. 128. Freeman R, Kupce E. J. Biomol. NMR. 2003;27:101. 129. Frydman L, Lupulescu A, Scherf T. J. Am. Chem. Soc. 2003;125:9204. 130. Shrot Y, Frydman L. J. Magn. Reson. 2004;167:42. 131. Mishkovsky M, Frydman L. Chemphyschem. 2004;5: 779. 132. Kupce E, Nishida T, Freeman R. Prog. NMR Spectr. 2003;42:95. 133. Kupce E, Freeman R. J. Magn. Reson. 2003;162:300. 134. Kim S, Szyperski T. J. Am. Chem. Soc. 2003;125:1385. 135. Atreya HS, Szyperski T. Proc. Natl. Acad. Sci. USA. 2004;101:9642.

1631

Josefina Belloque ´ Instituto de Fermentaciones Industriales, Consejo Superior de Investigaciones Cientficas (C.S.I.C.), Juan de la Cierva 3, 28006 Madrid Spain

General Remarks Milk contains basically water, carbohydrates, fat, proteins, and salts, in the form of a solution, a colloidal dispersion and an emulsion (Figure 1). There are excellent books that cover most aspects of milk and milk components [1,2]. High-resolution NMR has been applied to milk, and such studies include quali- and quantitative chemical analysis, the identification of isolated compounds, the in vivo monitoring of dairy bacterial metabolism, relaxometry studies, and structural studies of milk components, particularly proteins [3,4]. Initially, milk should be an ideal food product for NMR analysis because it is liquid. However, there are several considerations that have to be taken into account: 1. The presence of compounds in uneven proportions causes significant overlapping or masking of resonances. To solve this problem some components can be removed: fat can be removed by centrifugation in cold conditions, caseins can be precipitated by acidification to pH 4.6, both fat and caseins can be simultaneously removed by ultracentrifugation, whey proteins can be fractionated by ultrafiltration, and lactose can be partially removed by ethanol precipitation. Regular laboratory methods are found in specialized scientific publications, and official standard methods can be obtained from the web page of the International Dairy Federation (http://www.fil-idf.org). 2. The differences in size and mobility of milk components (Figure 1) leads to a wide range of relaxation times and mixed resonance line widths. Fat globules and casein micelles are large aggregates of variable sizes, with very low mobility compared with small soluble components, free in solvent, such as lactose. If better signal and resolution of the aggregated components is needed, and structural information is not required, it is convenient to disaggregate these large structures. For instance, since casein micelles are held together by calcium phosphate, and orthophosphate is partially in the form of insoluble calcium salts, chelation or removal of calcium, with EDTA or CHELEX resin, disaggregates both of them.

Graham A. Webb (ed.), Modern Magnetic Resonance, 1631–1635.
C 2008 Springer.

3. The natural composition changes significantly among species but, within the same species, there are also significant differences related to factors such as the breed, geographical location, stage of lactation, and even individuals. Bovine raw milk of a great mix of cows is available from dairy farms. Processed milk, purchased in supermarkets, comes from mixed milk, and has usually been subjected to a standardization process. However, differences can still be found among milks from different countries, farms, or milking seasons. 4. The different processing steps that milk undergoes along the processing chain leads to changes in composition and physico-chemical properties, for instance: r Storage: proteolysis, lipolysis, and other enzymatic reactions change the composition. r Heating: protein unfolding and aggregation, changes in phosphate distribution, and formation of products derived from the Maillard reaction. These alterations occur to different degrees as a function of the type and intensity of process (e.g. pasteurization, sterilization, UHT, drying). r Homogenization: Disruption of large structures, micelles and fat globules, into smaller ones, due to the pressure treatment.

NMR Spectra of Milk Water, lactose, and wide lipid signals dominate the 1 Hand 13 C-spectra of whole milk, and little attention has been paid to these types of spectra. Recently, Hu and cols. [5] have applied 1D and field gradient HSQC and HMBC experiments to whole milk and milk extracts, and have assigned the observable resonances (Figure 2). On the other hand, 31 P-spectra of milk are very clean (Figure 3) and the resonances assignments are available [3]. 31 P-NMR has the capability to distinguish among different forms of phosphate present naturally in milk (orthophosphate, casein SerP, phosphorylated carbohydrates, phosphoglycerydes, creatine-P, nucleotides), as well as diphosphates, added as stabilizers. Many of the phosphorylated compounds visible in the spectrum are

Part III

High-Resolution NMR of Milk and Milk Proteins

1632 Part III

Food Science

Part III

Fig. 1. Basic composition and structure of bovine milk.

found in the serum, while most SerP, a significant proportion of phosphate, and diphosphate additives, are present in the micellar phase. 31 P-NMR spectra has allowed to analyse the distribution of phosphates between the micellar and the soluble phases, the evaluation of differences among species and during lactation, and has revealed new enzymatic reactions induced by pseudomonas. Quantitative analysis by 31 P-NMR is a good alternative to the classical spectrophotometric method, which is tedious and

a)

poorly informative. However, quantitation can be tricky, since the relaxation of the different compounds, even when disaggregated, can differ largely. For instance, in a 400 MHz spectrometer, 31 P spectra of a homogeneous sample, obtained from milk after fat and calcium removal, shows T1 values of ∼2 and ∼12 s for casein SerP and orthophosphate, respectively. The use of the relative area to an internal standard can provide quantitative results, provided that full relaxation between transients is achieved,

1H

b)

Lecithin

6

Creatine

5

ppm

Unsaturated acyl chains Citrate

20 40

2

80 100

Gl-1,3

0 ppm

Gl-1,3 -CH2Citrate

13C Acyl chains

1

CH3

60

D-lactose

H2 O

3

Gl-2

3.0 2.9 2.8 2.7 2.6 2.5 2.4 2.3

CH2=CH2

4

N-acetyl carbohydrate Unsaturated acyl chains

Lactose Lactose

Lecithin

120 Unsaturated acyl chains 6

Fig. 2. (a) glycerol.

5 1 H-NMR

4

3

2

spectrum of whole milk. (b)

1

ppm

1 H-13 C

HSQC spectrum of ethanol-extracted-milk. Reproduced from [5]. Gl:

NMR of Milk

SerP GPC

Gal-1P P-creatine

GPEA

5

4

3

2

1

0

−1

−2

which may require a long time. If a shorter pulse (e.g. 45◦ ) is used, relaxation is faster, thus shortening the experiment time. If the samples are homogeneous and relaxation times do not vary among samples, a 90◦ pulse and a short delay time (i.e. partial relaxation) can be used, obtaining reproducible spectra with good S/N (Figure 3). 31 P-NMR has also widely been applied to the study of the structure of caseins and casein micelles (see below). In addition to proteins, the milk fat has become one of the most interesting fractions to study in milk. Qualiand quantitative NMR analysis of lipid components is being applied to develop methods for authentication of dairy products. The lipid analysis can discriminate among species [6], but also among different types of feeding and/or geographic origin [7], since the feeding of an animal modifies the milk fat composition. 13 C- is used to analyse the distribution of mono-, di- and unsaturated fatty acids, as well as the positional isomers of fatty acids in triacylglycerols. These compounds are fractionated from milk and dissolved in an organic solvent, usually CDCl3 . 31 P-NMR is used for the analysis of phospholipids, previously extracted from the fat globule membrane, and prepared in solvents such as dimethylformamidetriethylamine-guanidinum hydrochloride.

NMR Studies of Milk Proteins The structure and structural changes of milk proteins have been a major research subject in the dairy field. First of all, the structure of milk is closely related to the structure of the casein micelles, which still remains a matter of controversy, although most accepted models agree that κ-casein is located on the surface of the micelle. Second, many milk proteins are allergens, and this activity is closely related to their structure and conformational stability. Third, the physiological and biological activities of peptides derived from milk proteins, that depend on the peptide structure, are a current matter of study. Fourth, the structural changes induced to milk proteins are related to their functional

−3

−4

ppm

properties (e.g. gelling, foaming, emulsifying, thickening capabilities), that are used to improve dairy and non-dairy products. The 3D structure of milk proteins and peptides have been elucidated by NMR, X-ray diffraction and theoretical modelling, and have provided the basis for the study of induced structural changes. Whey proteins that are soluble and with well-defined structures, have been subjected to considerable research by NMR compared to caseins.

Caseins They are difficult to study since their native form is the micelle. There is evidence to support that caseins are rheomorphic, i.e. they do not have a defined structure, and some authors have even further suggested that they are in an open/unfolded state [2,8]. Caseins are multiphosphorylated proteins that contribute to the solubilization and transport of calcium and phosphate, due to their ability to interact with calcium phosphate, through SerP residues, clustered in limited regions of their polypeptide chain. Phosphopeptides carrying these SerP residues are studied because they also bind calcium phosphate and are easier to study. NMR and structure models have been reported for phosphopeptides, such as β-CN f(1–25), αs2 CN f(2– 20), casein macropeptide (κ-CN f(106–169) and, the most recent one, αs1 -CN f(59–79) [9].

Casein Micelles Calcium phosphate, particularly the micellar calcium phosphate (MCP), plays an essential role in the stabilization of micelles, by cross-linking SerP residues of caseins, but at least part of the calcium phosphate is in equilibrium with serum (SCP). When casein micelles are subjected to 31 P-NMR, the observable resonances belong to casein SerP signals and the phosphate present in the micelle. Using liquid-state NMR, casein SerP appears as

Part III

Fig. 3. 31 P-NMR spectrum of non-fat milk, pre-treated with CHELEX, at pH 9.5. SerP: casein phosphoserine, PO43- : orthophosphate, Gal-1P: galactose-1-phosphate, NAGA-1P: Nacetylglucosamine-1-phosphate, GPEA: glycerophosphoethanolamine, GPC: glycerophosphocholine, P-creatine: phosphocreatine.

PO43−

NAGA-1P

NMR Studies of Milk Proteins 1633

1634 Part III

Food Science

Part III

a broad peak providing little information, unless the micelles are disaggregated by calcium chelation. The use of solid-state magic angle spinning (MAS)-NMR provides higher resolution. 31 P single-pulse- and cross-polarization MAS NMR have been applied to casein micelles and milk powder. Using model compounds, and with the aid of simulation software, this technique has shown that phosphate is in the form of hydroxyapatite (HAP), has allowed the identification of mobile and immobile forms of HAP, and has made possible to identify Ca2+ binding SerP sites from different caseins [10].

Whey Proteins Opposite to caseins, these are globular proteins that are soluble in the milk serum. The major proteins are βlactoglobulin (BLG) and α-lactalbumin (ALA), although other proteins, such as BSA, lactoferrin, immunoglobulins and lysozime, are present in minor quantities. Even though many studies have been done with whey bovine proteins, the only structure that has been elucidated by NMR is that of BLG. This protein is easy to purify, it has a MW of 18 KDa, and gives reasonably sharp NMR signals, especially at pH 2–3 since, under these conditions, it is a monomer, and very stable. The structure has been solved by both X-ray crystallography and NMR (Protein DataBase (PDB) ID 1CJ5), and complete 1 H- and 13 C and 15 N assignments are available. The structure of bovine ALA was obtained from X-ray diffraction (PDB ID 1F6S), although 1 H, 13 C and 15 N NMR assignments are available from the BioMagRes data Bank (BMRB ID 4811 and 4332). Similarly, the structure of bovine lactoferrin comes from X-ray diffraction (PDB ID 1BLF). In some cases, such as BSA, lysozyme and immunoglobulin, crystal structures are available for other species different than bovine (e.g. human serum albumin (PDB ID 1BM0), or hen egg lysozyme (PDB ID 193L). Crystal structures can be very useful for partial NMR assignments of bovine whey proteins.

Peptides Peptides present in milk are the result of proteolysis of caseins or whey proteins by enzymes, either indigenous in milk or from microorganisms. While proteolysis is, in general, detrimental to milk, it enhances the quality of some dairy products, particularly fermented milks and cheese. In the last years there has been an increased interest in the identification of milk peptides with antimicrobial, immunomodulant, ACE-inhibiting and other biological activities that can be used to improve the quality of fermented dairy products. NMR can contribute to the evaluation of the structure-function relationship. For instance,

lactoferricin is a peptide derived from lactoferrin that has higher antimicrobial properties than its precursor. While this peptide has a helix shape in the native protein (PDB ID 1BLF), NMR studies have shown that, in solution, it has a hairpin conformation (PDB ID 1LFC), which is more effective for microbial membrane permeabilization.

Milk Protein Structural Changes Because of the nature of caseins, structural changes studies have been limited to relaxation and diffusion measurements. However, whey proteins and, above all, BLG, has been subjected to detailed structural studies, analysing the influence of denaturing agents, heating, high pressure, lactosylation, site-directed mutagenesis, and ligand binding on its structure and/or stability [11]. One of the most interesting areas has been to elucidate the unfolding pathways of the protein under physical treatments. The knowledge of the unfolded intermediates helps to understand the nature of the interactions of the protein with other proteins, water and co-solutes. In the last years, there has been increasing evidence to support that protein folding is a hierarchical process that starts with the building of robust regions and allow the correct folding of the rest of the protein [12]. In the same manner, the unfolding of the protein is a stepwise process. In BLG, it can be observed that particularly robust regions, that seem to be the initiation-folding site, are the most resistant to unfolding [11]. 1 H-NMR coupled with deuterium exchange give very useful and detailed information about the unfolding process, by perturbing the protein, and observing the disappearance of backbone NH resonances. If the structure of the protein is known and the 1 H resonances are assigned, the unfolding process of precise regions of the protein can be followed by using 2D spectra (Figure 4). The effect of heat, cold and high pressures on BLG has been studied this way [13–16]. Some NMR probes allow performing experiments in situ, which provides the advantage of observing the real unfolding mechanism, avoiding the refolding of the protein and the re-equilibration of the system. However, if this is not available, an off-line scheme can be used, capturing different intermediate unfolded states by their degree of deuterium substitution (Figure 4). The recent progress on NMR techniques and the interest of researchers in the folding/unfolding processes have led to the development of methods that allow 2D spectra acquisition in real time. Checking whether the structure is refolded after the physical treatment or, on the contrary, whether it shows changes from the original form, can be done by comparing the spectra taken before and after the treatment. If relaxation or diffusion parameters are measured before and after processing, information about aggregation can also be obtained.

NMR of Milk

References 1635

Part III

a) ppm 37°C 4.0

55°C

b)

75°C

c)

4.5 5.0 5.5

ω1

6.0 6.5 65°C

T(°C) Struct.

4.0

55-60 D-E c-term

60-65 C-D

4.5

α-helix 65-70 A-B A-I

5.0 N-term

5.5 α

6.0

E-F 75-80 A-H B-C

6.5

F-G

10

9

8

7

10 ω2

9

8

7 ppm

Fig. 4. Effect of heat on BLG. (a) Fingerprint region (NH-Hα) of TOCSY spectra of BLG heated at different temperatures in a D2 O solution. Grey contours are from a non-heated non-deuterium-exchanged BLG reference sample. Reproduced from [13]. (b) 3D structure of BLG. (c) Organization of β-strands (A-I) and the α-helix of BLG, and structural interactions that are lost at different temperatures [13].

References 1. Walstra P, Geurts TJ, Noomen A, Jellema A, Van Boekel MAJS, Dairy Technology. Principles of milk properties and processes. Marcel Dekker Inc. 1999. 2. Fox PF, McSweeney PLH (Eds). Advanced Dairy Chemistry, vol. 1: Proteins, Elsevier Applied Science, 2003. 3. Wahlgren NM, Drakenberg T. In: Webb GA, Belton PS, McCarthy MJ (Eds). Annual reports on NMR spectroscopy, vol. 31, Academic Press, London, 1995, p. 275 4. Belloque J, Ramos M. Tr. Food Sci. Tech. 1999;10:313. 5. Hu F, Furihata K, Ito Ishida M, Kaminogawa S, Tanokura M. Agric. J. Food Chem. 2004;52:4969. 6. Andreotti G, Lamanna R, Trivellone E, Motta A. J. Am. Oil Chem. Soc. 2002;79:123. 7. Renou JP, Deponge C, Gachon P, Bonnefoy JC, Coulon JP, Garel JP, Verit´e R, Ritz P. Food Chem. 2004;85:63.

8. Smyth E, Clegg RA, Holt C. Int. J. Dairy Tech. 2004;57:121– 126. 9. Huq NL, Cross KJ, Reynolds EC. J. Dairy Res. 2004;71:28. 10. Bak M, Rasmussen LK, Petersen TE, Nielsen NC. J. Dairy Sci. 2001;84:1310. 11. Sawyer L, Barlow PN, Bol, MJ, Creamer LK, Denton H, Edwards PJB, Holt C, Jameson GB, Kontopidis G, Norris GE, Uhrinova S, Wu S-Y. Int. Dairy J. 2002;12:299. 12. Dinner R, Sali A, Smith LJ, Dobson CM, Karplus M. Tr. In Biol. Sci. 2000;25:331. 13. Edwards PJB, Jameson GB, Palmano KP, Creamer LK. Int. Dairy J. 2002;12:331. 14. Katou H, Hoshino M, Kamikubo H, Batt CA, Goto Y. J. Mol. Biol. 2001;310:471. 15. Kuwata K, Li H, Yamada H, Batt CA, Goto Y, Akasaka K. J. Mol. Biol. 2001;305:1073. 16. Belloque J, Smith G. J. Agric. Food Chem. 1998;46:1805.

1637

Maria Antonietta Brescia and Antonio Sacco Dipartimento di Chimica, Universit`a di Bari, via Orabona 4, 70126 Bari, Italy

Introduction High-resolution 13 C NMR spectroscopy has been widely used in lipid and vegetable oil analysis [1,2]. This is due to the electronic configuration of 13 C nucleus, that makes the chemical shift spread over a wide range of frequencies (200 ppm), permitting a minor overlapping of the signals compared to proton spectra, and, consequently, the acquisition of structural information. The most important advantage of 13 C NMR spectroscopy is the possibility to determine the positional distribution of the acyl chains on the glycerol moiety. This determination cannot be done with 1 H NMR spectra due to the strong overlap of the peaks. However, a limitation connected with the use of 13 C spectroscopy for the analysis of oils is due to low natural abundance of 13 C and small gyromagnetic ratio that results in its peculiar low sensitivity. As a consequence, longer analytical time is required for acquisition and only major components, like triacylglycerides can be analyzed and quantified. Quantitative information from 13 C NMR spectra can be acquired if resonances are not affected by the effect of nuclear Overhauser enhancement (NOE) that occurs during proton decoupling of 13 C, through the mechanism of dipole–dipole relaxation. NOE can be suppressed by using the inverse-gated proton-decoupled sequence, which gates the decoupler only for the acquisition time. Analysis of the different regions of the 13 C NMR spectrum and/or the entire spectrum can provide useful insight on oil composition, enabling quality and authenticity controls.

Quantitative Determination of the Oils Major Components Fatty Acids The 13 C NMR spectrum of a virgin olive oil, dissolved in CDCl3 , is shown in Figure 1. Four signal regions can be recognized due to different types of carbons of triacylglycerides. The assignments were performed by comparing the signals of the 13 C spectrum with the chemical Graham A. Webb (ed.), Modern Magnetic Resonance, 1637–1643.
C 2008 Springer.

shift of standard glycerides. This procedure is generally followed [3–5] even though it has been recently demonstrated that a slight chemical shift of the signals in the 13 C NMR spectra of oils is a function of their concentration, especially in the carbonyl region [6]. Table 1 illustrates some case studies based on analytical applications of 13 C NMR spectroscopy. A quantitative determination of MUFA, linoleic, and linolenic acids of various vegetable oils from the resonances present in the methyl and methylene regions (14–35 ppm) of the 13 C NMR spectrum [7]was proposed. It was found that the optimization of the recovery delay was necessary in order to get data in agreement with those obtained by the classical gas chromatography method, in particular for the linoleic acid determination. It resulted that a total experimental time of 30 min per sample was required. The areas of signals resonating in this region were also used for the determination of iodine value of palm oil [8]. An alternative method for the quantification of the major fatty acids (saturated, oleic, and linoleic) based on the analysis of the signals due to carbonyl and ethylenic carbons, resonating in the 195–105 ppm region was proposed for Greek olive oils [9]. This quantification procedure required the presence of an internal reference (pyrazine). The overlap of peaks in the carbonyl region was reduced applying a Lorentz Gaussian enhanced filter function to the FID and a peak deconvolution procedure to allow a precise integration. This methodology was extended investigating mixtures of virgin olive oils and seed oils. In this case, the quantitative analysis was performed only on signals resonating in the olefinic region, due to the minor overlapping presented [10]. A linear increase of linoleic acids contents was observed with increasing addition of seed oils to olive oils. This evidence allowed to construct reference curves, proposed as a diagnostic tool for the estimation of olive oil adulteration. The adulteration of olive oil with soybean oil having high linoleic acid content was also examined by 13 C NMR distortionless enhancement by polarization transfer (DEPT). This technique employs a polarization transfer able to enhance signal intensities [11], and shorten the experimental times required when compared to 13 C NMR spectra. The signals of unsaturated carbons of

Part III

High-Resolution 13C Nuclear Magnetic Resonance in the Study of Oils

1638 Part III

Food Science

Part III

Alkyl carbons

Vinyl carbons

Solvent CHCl3

Carbonyl carbons Glycerol carbons

190

180

170

160

150

140

130

120

110

100

90

80

70

60

50

40

30

20

10 ppm

Fig. 1.

13 C

NMR 500 MHz spectral pattern of an olive oil.

oleyl, linoleyl, and linolenyl chains, observed in the 132– 127 ppm region, were found to give two resonances for each carbon signal, depending on the chain position on the glycerol backbone. Integrals of these resonances were used to calculate the calibration curves based on a linear relationship between resonance intensities and soybean concentration in olive oil. The methylene, olefinic, and carbonyl regions in the 13 C NMR spectra were considered to obtain quantitative

information about the fatty acid distribution of fish oils [12]. The addition of paramagnetic compounds to the analyzed samples was provided in order to decrease the experimental time for performing quantitative measurements. Qualitative and quantitative determination of fatty acids in vegetable seeds is necessary to assess their value. Magic angle spinning 13 C NMR technique has shown to be very attractive in this case because it is a non-destructive

Table 1: Quantitative 13 C NMR determinations on oils Samples

Parameters

Analyzed spectral region

Reference

Canola seeds Fish oils Fish oils Mixtures of olive and seed oils Olive oils Olive oils Palm oils Palm oils Vegetable oils Vegetable oils Vegetable seeds

Fatty acids Fatty acids Fatty acids and positional distributions Unsaturated fatty acids Free fatty acids and partial aylglycerols Diglycerides Iodine value Free fatty acids and partial acylglycerols Unsaturated fatty acids Fatty acids and positional distributions Unsaturated fatty acids

29–20 ppm Entire spectrum 178–172 and 75–50 ppm 130–127 ppm 176–173 and 75–60 ppm 75–60 ppm 30–24 ppm 174–172, 75–60, and 30–20 ppm 36–14 ppm 173–127 ppm 130–127 ppm

12 10 25 9, 32 16 17 7 18 6 8, 19–20, 22–24 11

13 C NMR of Oils

Glycerol Esters Glycerol can form mono-, di-, and triacylglycerols, resonating in the spectral region from 60 to 72 ppm (Figure 2). The symmetrical esters (2-monoacyl, 1,3-diacyl, and triacyl) give only two signals at 1:2 intensity ratio, while the asymmetrical esters give three separate signals of equal intensity. Assignments of standard glycerides were obtained using DEPT experiments [18]. The total amount of diacylglycerols and the ratio between 1,2 and 1,3-diacylglycerols can be evaluated in terms of molar ratios (%) through the analysis of the peaks intensities: R1 = 100 × DG/(DG + TG) R2 = 1,2-DG/1,3-DG R3 = MG × 100/(MG + DG + TG) In the above-reported relationships, DG represents the sum of the intensities of the resonances due to carbons

of 1,2- and 1,3-diacylglycerides; TG represents the sum of the intensities of carbons of triacylglycerides; MG indicates the intensities of signals due to αmonoglycerides (69.9 and 63.1) and β-monoglycerides (61.02 and 72.2) (signals not present in the spectrum shown in Figure 2). The spectrum must be acquired using appropriate relaxation delays and pulse widths in order to carry out the quantification procedure, since full recovery of magnetization between subsequent pulses is necessary. This determinations can be useful to detect the freshness of the oils and to reveal the presence of partially neutralized oils to olive oils that are given out as “virgin”. Diacylglycerols (mainly 1,2) can come either from incomplete triacylglycerols biosynthesis or from limited lipase action on triacylglycerols and their content, indicated by R1, should not exceed 2–3% in virgin olive oil. The glycerol content increases with olive oil acidity. Large amounts of diacylglycerols (mainly 1,3-diacylglycerols) are found in neutralized olive oils, produced from a starting material with a high level of free fatty acids and in oils obtained from overripe olives or after several months of oil storage due to lipolysis and 1,2- and 1,3-isomerization. In both the cases, R2 is always lower than 1. A quantitative 13 C NMR study carried out on olive oil showed that significant differences in diacylglycerols content occurred among cultivars [19], probably due to its relation with maturation stage. In fact, it was observed that lower total diacylglycerol content is characteristic of late maturing cultivars, whereas early maturing fruits produce higher levels of total diacylglycerides. The determination of mono- and diacylglycerol content is also important, since they are commonly used as emulsifiers in the food industry. For this reason, the quantitative analysis of these compounds was performed on palm oil [20]. Since glycerol carbons are near the T1 minimum for dipole–dipole relaxation mechanism, spectra were acquired at low magnetic fields (300 MHz) and at an elevated temperature (50 ˚C) to obtain narrow signals.

Fatty Acids Positional Distribution In the low-field region of the spectrum, carbonyl carbons of the triacylglycerides resonate. Oil samples have two sets of carbonyl signals: the first, centered at ∼173.2 ppm, includes the fatty acids esterified at 1,3-glycerol positions (α), while the second, centered at ∼172.8 ppm, is due to fatty acids bonded to the 2-glycerol position (β). In olive oil, four signals were observed in the first group due to saturated (S), eicosenoic (E), oleic (O), polyunsaturated fatty acids (L), while in the second group there are two signals, respectively due to oleic and polyunsaturated acids (Figure 3). The analysis of this region enables the

Part III

method, being applied directly on the seed, and because it permits to obtain a spectral resolution that approached that of liquid oils. However, it is worth to consider that for the analysis of seeds having a low oil content (like corn and rape seeds) or a complex unsaturated fatty acid composition, DEPT technique was necessary to increase the sensitivity [13]. From the analysis of the methylene region of canola seed oil spectrum, it was possible to detect trans olefins in oils [14] because allylic methylenes adjacent to cis double bonds are shifted by 5 ppm to frequency lower than that of trans isomer. This determination is advantageous respect to the classical chromatographic method, because it detects all trans isomers simultaneously, while in a chromatogram different isomers give rise to different small signals that are not accurately measurable. The detection of trans fatty acids can be extremely helpful to reveal the addition of refined oils to virgin olive oils. In fact, oil refining processes (bleaching and/or deodorizing) are carried out at high temperatures that favor the isomerization of oleic acid (cis) to elaidinic (trans). This acid is absent in virgin olive oils, while its content in olive and olive-pomace oils varies from 0.1 to 0.6%, therefore its presence in olive oil can be considered an index of adulteration [15]. The detection of the amount of trans and cis fatty acids is also important for the quality control of hydrogenated vegetable oils and to determine the effect of the experimental conditions of the hydrogenation process in double bonds isomerization and migration. Gunstone [16] and Miyake and Yokomizo [17] obtained important results on the position and configuration of double bonds by investigating the olefinic region of the 13 C spectrum.

Quantitative Determination of the Oils Major Components 1639

1640 Part III

Food Science

Part III Fig. 2. The glycerol carbon region 60–72 ppm of the 13 C spectrum of a standard mixture of glycerides at different esterification degree. The resonances of glycerol carbons are labeled as –CH2 OAc and –CH2 OH in correspondence of an acylated and a non-acylated alcoholic group of the glycerol molecule (from Ref. [2]).

detection of addition of synthetic esterified oils, considered non-edible by the European Union, to virgin olive oils. Esterified oils show a random distribution of saturated fatty acids, leading the 15% of saturated acids in the 2-glycerol position. The biosynthesis of triacylglycerols in the vegetable oils leads to the preferential esterification of unsaturated fatty acids in the 2-glycerol position; therefore, the amount of oleic and linoleic acids in this position represents the 98–99% of the total fatty acid content. As a consequence, if the saturated fatty acids exceed 1.5% in this position, the analyzed oil contains a certain amount of esterified oils. The EU official analytical procedure for the analysis of esterified oils mixed with olive oils is based on

an enzymatic method requiring several steps, while NMR making use of 13 C resonances yields immediate results, operating directly on the oil samples. In fact, in the carbonyl region of the 13 C NMR spectrum of virgin olive oil only one resonance is observed for the 2-glycerol position, corresponding to the unsaturated fatty acids, while in the same spectral region an esterified oil shows a minor peak due to saturated acids. This signal was quantified by means of a curve resolution program, enabling the detection of a 10% addition of esterified oils to olive oil [21]. The areas of the signals resonating in the carbonyl region of the spectrum were used to perform a regiospecific

13 C NMR of Oils

Minor Oil Components 1641

Part III

1,3-positions O

2-positions O

S

L

L

E

173.50

173.40

173.30

173.20

173.10

173.00

172.90

172.80

172.70 ppm

Fig. 3. 500 MHz 13 C NMR spectrum of the carbonyl region of an olive oil.

analysis of triacylglycerides of Italian monovarietal olive oils, calculating the distribution of saturated, oleyl, and linoleyl chains between 1,3 and 2 positions of glycerol [22]. Using the same approach, the fatty acid composition and distribution in triacylglycerols of other vegetable oils was determined [23–25]. The examination of acyl positional distribution in canola and other non-edible vegetable oils is more complex than for olive oil, due to the presence of eicosenoic and erucic acids, which signals result overlapping to those of linoleic acid in the carbonyl region. For this reason, the analysis of the olefinic region of the spectra was also necessary to obtain a qualitative and quantitative evaluation of these acids [26]. The study of the carbonyl region was also performed on the 13 C NMR spectra of fish oils revealing that glycerol β-position is esterified preferentially by docosahexaenoic acid [12]. In the 13 C NMR spectrum of an oil with high content of free fatty acids, the signals of free COOH groups appear at 176–178 ppm. The determination of free acidity can be easily performed by comparing the intensity of this signal to the sum of free and esterified carbonyl groups. This method was applied to quantify free fatty acids in olive oils [18] and in fish lipids [27]. Quantification of

free fatty acids in palm oils was performed from the C3 peak of free fatty acids resonating in the region of aliphatic carbons [20].

Minor Oil Components Among minor olive oil components, phenolic compounds play an important role in the sensory characteristics and in the oxidation stability of virgin olive oil (see Section 2.4 of “1 H NMR of oils”). In particular, it has been demonstrated that the esters of tyrosol (Ty) and hydroxytyrosol (OHTy) with elenolic acid contribute to the sensory characteristics of the oil, being responsible of the pungent taste, while OHTy and its derivatives are correlated with oil resistance to oxidation [14]. 13 C NMR spectroscopy of the polar fraction extracted from virgin olive oils permits to calculate the ratio of free and esterified Ty + OHTy comparing the intensities of C1 carbon of free Ty + OHTy (64.5 ppm) with the C1 signal of the corresponding esterified compounds (66.8 ppm). The molar ratio between total (free and esterified) OHTy and Ty can be calculated by comparing the intensities of aromatic signals at 121.3 and 116.1 ppm and provides a useful parameter for oil stability.

1642 Part III

Food Science

Part III

13

C NMR Spectroscopy As a Discriminating for the Varietal, Geographical, and Botanical Origin of Vegetable Oils

Since olive oil is a product of high nutritional and commercial value, its adulteration with different quality and origin oils represents a big temptation. A rapid screening method is therefore requested to detect oil adulteration. Recently, methods based on 13 C NMR spectroscopy have been considered advantageous in this field because of compositional and structural information contained in the NMR signals. The proposed methods provide models which are obtained applying statistical methods of multivariate analysis to spectroscopic measurement of representative sets of samples. Principal component analysis, principal component regression, and partial least squares were applied to 35 signals areas from the 13 C NMR spectra of six olive oil cultivars sampled in different Italian regions with variable selection methods of Fisher ratio and the ratio of inner variance to outer variance [28]. The classification of the oils by variety was found to be more successful than classification by region of origin. In order to improve the geographical origin discrimination, the potential of DEPT sequence was also investigated, since it permits to obtain better S/N ratio of 13 C NMR spectra. This approach was applied to olive oils coming from 13 Italian PDO areas [29] located in different regions, and to olive oils coming from three close PDO areas in Apulia region (Southern Italy) [30]. The acquisition and processing conditions were optimized in order to perform accurate and reproducible measurements. Multivariate statistical analysis was applied to the intensities of selected peaks present in the 13 C NMR spectra and permitted to distinguish the analyzed oils according to their origin. In both studies the cultivar factor seemed very important, since classification worked well especially when oils of different PDO where obtained from olives of different cultivars. Cultivars differentiation of virgin olive oils coming from different PDO areas in Apulia region, was also explored with classical chromatographic analyses (fatty acid, sterol, and triacylglyceride composition) and with 13 C NMR spectroscopy, that was used for comparative purposes to official methods. Therefore, classical and NMR parameters were separately evaluated but the same statistical approach was used for both data sets. As far as NMR data are concerned, normalized heights of signals present in the 13 C NMR spectrum, selected with analysis of variance, were used as variables for statistical elaboration [31].The models obtained from the application of discriminant analysis to the data sets of chromatographic and 13 C NMR results were found to be in a good agreement, being statistically similar. This study confirmed that 13 C NMR spectra contain the information about acyl

composition and positional distribution of glycerol moiety; since for the analytical data, fatty acids and triacylglycerols were the most responsible for distinguishing oils’ cultivars, and all the selected peaks for 13 C NMR were relevant. This technique could therefore be effectively used for classifying cultivars. A combined 13 C NMR and GC approach was also successfully used to determine the cultivar effect on fatty acid composition of monovarietal Sicilian olive oils [32]. Some authors studied the potential of 13 C NMR analysis to distinguish oils from different cultivars, geographical and botanical origins, as well as mixtures of olive oils from different geographical origins and of hazelnut oil in olive oil [33]. The intensities of peaks most effective in the discrimination of the different types of oil were selected as variables by stepwise discriminant analysis (SDA). The difference between virgin olive oils and vegetable oils having a high content of linoleic and oleic acid was obtained. In addition, the statistical method classified, with 97% of classification ability, hazelnut oil samples and mixtures of hazelnut oil in olive oil. 13 C NMR of Olive Oil Minor Compounds to Determine Oil Authenticity 13

C NMR spectroscopy cannot be directly used for the analysis of minor compounds due to its low sensitivity. Nevertheless, the information acquired by 13 C NMR in the analysis of triacylglyceridic composition of oils have prompted some authors [34] to explore the unsaponifiable matter of virgin and refined olive and pomace oils, after previous separation from triacylglycerides by treatment with a potassium hydroxide solution. The low-field region of the spectra contained signals that were assigned to squalene that was the major component of the unsaponifiable matter of virgin olive oil and was present, in lower quantity, also in refined olive oils and in refined olive-pomace oils. Other signals were assigned to sterols, β-sitosterol, triterpenic, and long-chain monounsaturated alcohols. The heights of these peaks were evaluated by means of SDA permitting to achieve a discrimination between the different types of investigated oils. However, the use of unsaponifiable matter requires a time-consuming preparation and does not hold information about triacylglycerols and other saponifiable components, which can also be related to oil quality. To overcome this limitation, a chromatographic method was used to isolate an oil fraction containing 15% of triacylglycerols and 85% of polar compounds [35]. 13 C NMR spectra of this fraction were obtained from 109 vegetable oils of different grade and various botanical and geographical origin. Moreover, mixtures of virgin olive oils from different geographical regions were analyzed. Spectra were similar

13 C NMR of Oils

References 1. Gunstone FD, Shukla VKS. Annu. Rep. NMR Spectrosc. 1995;31:219. 2. Vlahov G. Prog. Magn. Reson. Spectrosc. 1999;35:341. 3. Ng S. Lipids 1985;20:778. 4. Shaw AD, Di Camillo A, Vlahov G, Jones A, Bianchi G, Rowland J, Kell DB. Anal. Chim. Acta 1997;348:357. 5. Mannina L, Luchinat C, Emanuele MC, Segre AL. Chem. Phys. Lipids 1999;103:47. 6. Mannina L, Luchinat C, Patumi M, Emanuele MC, Rossi E, Segre AL. Magn. Reson. Chem. 2000;38:886. 7. Miyake Y, Yokomizo K, Matsuzaki N. J. Am. Oil Chem. Soc. 1998;75:1091. 8. Ng S, Gee PT. Eur. J. Lipid Sci. Technol. 2001;103:223. 9. Mavromoustakos T, Zervou M, Theodoropoulou E, Panagiotopulos D, Bonas G, Day M, Helmis A. Magn. Reson. Chem. 1997;35:S3. 10. Mavromoustakos T, Zervou M, Bonas G, Kolocouris A, Petrakis P. J. Am. Oil Chem. Soc. 2000;77:405.

11. Vlahov G. Magn. Reson. Chem. 1998;35:S8. 12. Aursand M, Rainuzzo JR, Grasladen H. J. Am. Oil Chem. Soc. 1993;70:971. 13. Wollenberg K. J. Am. Oil Chem. Soc. 1991;68:391. 14. Hutton WC, Garbow JR, Hayes TR. Lipids. 1999;34:1339. 15. Sacchi R, Addeo F, Paolillo L. Magn. Reson. Chem. 1997;33:133. 16. Gunstone FD. J. Am. Oil Chem. Soc. 1993;70:965. 17. Miyake Y, Yokomizo K. J. Am. Oil Chem. Soc. 1998;75:801. 18. Sacchi R, Addeo F, Giudicianni I, Paolillo L. Riv. Ital. Sost. Grasse. 1990;LXVII:245. 19. Vlahov G. Fett/Lipid. 1996;98:203. 20. Ng S. J. Am. Oil Chem. Soc. 2000;77:749. 21. Sacchi R, Addeo F, Giudicianni I, Paolillo L. Ital. J. Food Sci. 1992;2:117. 22. Vlahov G. J. Am. Oil Chem. Soc. 1996;73:1201. 23. Howarth OW, Vlahov G. Chem. Phys. Lipids. 1996;81:81. 24. Vlahov G. Phytochemistry. 1996;42:621. 25. Vlahov G, Chepkwony PK, Ndalut PK. J. Agric. Food Chem. 2002;50:970. 26. Wollenberg KF. J. Am. Oil Chem. Soc. 1990;67:487. 27. Sacchi R, Medina I, Aubbourg SP, Giudicianni I, Paolillo L, Addeo F. J. Agric. Food Chem. 1993;41:1247. 28. Shaw AD, di Camillo A, Vlahov G, Jones A, Bianchi G, Rowland J, Kell DB. Anal. Chim. Acta. 1997;348:357. 29. Vlahov G, Schiavone C, Simone N. Magn. Res. Chem. 2001;39:689. 30. Vlahov G, Del Re P, Simone N. J. Agric. Food Chem. 2003;51:5612. 31. Brescia MA, Alviti G, Liuzzi V, Sacco A. J. Am. Oil Chem. Soc. 2003;80:945. 32. Mannina L, Dugo G, Salvo F, Cicero L, Ansanelli G, Calcagni C, Segre AL. J. Agric. Food Chem. 2003;51:120. 33. Zamora R, Alba V, Hidalgo FJ. J. Am. Oil Chem. Soc. 2001;78:89. 34. Zamora R, Navarro JL, Hidalgo FJ. J. Am. Oil Chem. Soc. 1994;71:361. 35. Zamora R, G´omez G, Hidalgo FJ. J. Am. Oil Chem. Soc. 2002;79:267. 36. Hidalgo FJ, G´omez G, Navarro JL, Zamora R. J. Agric. Food Chem. 2002;50:5825.

Part III

to those obtained for olive oils but contained the double number of signals, since minor compounds like diacylglycerols, monoacylglycerols, sterols, and fatty alcohols were visible. The information contained in the spectra was employed to classify olive oils by using SDA and variable selection. 13 C NMR spectra of oil fractions obtained with the same method for 66 vegetable oils were also analyzed to evaluate the potential use of these fractions in predicting oil stabilities. A 0.99 correlation index was found by stepwise linear regression analysis between oil stability, determined by Rancimat method, and the heights of several signals present in the 13 C NMR spectra [36]. The statistical method selected the signals more correlated to oil stability. These signals were associated to fatty acid composition and to the presence of minor compounds with antioxidant characteristics.

References 1643

1645

Maria Antonietta Brescia and Antonio Sacco Dipartimento di Chimica, Universit`a di Bari, via Orabona 4, 70126 Bari, Italy

Introduction High-resolution NMR plays an important role in oil analysis. This is firstly due to the absence of sample pretreatment, since oils are simply added to a deuterated solvent before analysis. 1 H NMR studies of oils have increased because of the great amount of information that high field NMR instruments can provide with a single experiment and in a very short time. This entry is mainly focused on the qualitative and quantitative information that can be obtained from the oils 1 H NMR spectra regarding major and minor components. Some of the most interesting implications in the control of quality and authenticity will be also discussed.

Triglycerides A typical 1 H NMR spectrum of an olive oil is reported in Figure 1. It shows 10 signals of significant intensity due to protons of triglycerides—the main oil components. These signals are also present in the spectra of other vegetable oils, but their shapes and intensities are different due to different fatty acid composition. These signals do not allow the complete discrimination of single fatty acid components, unlike the traditional gas chromatographic method, considering their overlapping in the spectral area of interest. Anyway, they can be used for various useful applications. For example, the oil unsaturation degree can be calculated by determining olefinic protons, resonating at 5.29 ppm (signal A in the spectrum), relative to the number of glyceride methylene protons (signal B in the spectrum) [1]. This parameter is an important indicator of the development of rancidity and was employed to monitor the lipid oxidation process in vegetable oils [2]. The traditional determination of the unsaturation degree of an oil consists in the quantitative reaction of iodine to fatty acids double bonds, giving as a result the iodine value (IV). It can be obtained directly from the NMR spectrum by means of the following equation: IV = 253.8[(A − B/4)/2 + (E/4 + C/2)]/2/100 · average molecular weight, Graham A. Webb (ed.), Modern Magnetic Resonance, 1645–1650.
C 2008 Springer.

where 253.8 is the molecular weight of iodine, and the average molecular weight of the considered oil is obtained as follows: average molecular weight = 15.034(G + H )/3 + 14.026(C + D + E + F + I )/2 + 173.100B/4 + 26.016(A − B/4)/2, where A–I are the integrals of the signals present in the spectrum in Figure 1. The IV was determined for several vegetable oils and coincided with that obtained by the conventional method [3]. The measurement of the ratio of olefinic to aliphatic protons was proposed to follow the oxidative deterioration of fish oils [4], since the common peroxide value is not easily measurable due to the high content of unstable polyunsaturated acids. A good correlation between the determined NMR index and peroxide value was also verified. From the integrals of the signals A–I , the quantification of classes of fatty acids (saturated, monounsaturated, linoleic, and linolenic acids) [5] can also be carried out (Table 1). These data are well correlated with those obtained with the official chromatographic method, although they are less detailed, but these determinations are sufficient for preliminary quality and authenticity evaluations. Recently, other procedures based on 1 H NMR spectra have been proposed for the quantification of fatty acids in prepared triacylglycerols mixtures [6] and in mixtures of methyl esters and triacylglycerols [7]. The methyl group of linolenic acid (signal G in Figure 1) can be used to quantify this acid but, in corn and olive oil, this signal is quite small, therefore accurate quantification through the formula given in Table 1 is precluded. A more suitable quantification method is based on the comparison of signal G and the nearby 13 C satellites of signal H , whose amount is exactly 0.57% of the intensity of signal H . This parameter is important because it permits to detect the adulteration of olive oil with seed oils as soybean and rapeseed oils, since they have a different linolenic acid content. Signal G can also be used to obtain a global measure of the ω-3 polyunsaturated fatty acids

Part III

High-Resolution 1H Nuclear Magnetic Resonance in the Study of Oils

1646 Part III

Food Science

Part III

F methylene protons

H methyl protons

D A

methylene protons bonded to C2

vinyl protons

E

I

allylic protons

B C

glycerol -CH2

Glycerol -CH

G

methylene protons bonded to C3

diallylic protons

5.4

5.2

5.0

4.8

4.6

4.4 4.2

4.0

3.8 3.6

3.4

3.2

3.0

2.8

2.6

2.4

2.2

2.0

1.8

1.6

1.4

methyl protons of linolenic acid

1.2

1.0

0.8 ppm

Fig. 1. 1 H NMR 500 MHz spectrum of an olive oil.

in fish oils [8]. In order to reduce errors in the quantitative determination, careful attention must be paid to avoid oxidative deterioration of sample oil, because oxidized oils may give signals interfering with measurements. Oils from 24 samples of raw, cooked, and canned albacore tuna were quantified and compared. Since vegetable oils contain various proportions of saturated, oleic, and polyunsaturated acyl groups, the resulting signals have different chemical shifts and shapes depending on these proportions. A careful observation of the shape and number of peaks of each signal present in the spectra allows to find significant differences between vegetable oils of different composition [9,10]. For example, sunflower oil can be easily distinguished from olive oil due to its high linoleic acid content that is particularly evident in the intensity of the signal resonating at

Table 1: Calculations of fatty acid composition of oils by signal intensities in the 1 H NMR spectrum Fatty acids (%)

Formula

Linolenic acid Linoleic acid Unsaturated acids MUFA

100[G/(G + H )] 100(3C−4G)/2(G + H ) 100(E/2D) 100{E/2D − [G(G + H )] − [(3C − 4G)/ 2(G + H )]} 100[H /(G + H ) − E/2D]

Saturated acids

2.76 ppm. Protons of linoleic acid give visible signals at 0.89 ppm, and at 2.06 ppm, due, respectively, to the terminal methyl group and to allylic protons of linoleic acid. A careful observation of the spectra can be also useful to distinguish oils of similar composition, such as olive and hazelnut oils. In fact, the presence of a very small proportion of linolenic acid in olive oil (∼1%) permits to distinguish it from hazelnut oil, whose amount of linolenic acid is much lower (∼0.1%). Moreover, the difference in saturated acids, higher in olive than in hazelnut oil, is evident also observing the shape of the signal due to methylene protons, that is formed by three signals, resonating at slightly different chemical shifts, due to saturated, oleic, and linoleic acids. These slight differences in unsaturated and polyunsaturated fatty acids content between olive and hazelnut oil become less remarkable when a real adulteration problem is concerned, and mixtures of hazelnut and olive oil have to be detected. In this case, pure and adulterated samples cannot be distinguished by a visual inspection of the spectra. To solve this problem, a multivariate statistical method was developed and evaluated first on a big number of pure and adulterated samples [11]. The integrals of peaks resonating at 5.29, 2.76, and 1.6 ppm were chosen as variables for statistical analysis, since they reflect, respectively, the amount of unsaturated, polyunsaturated fatty acids, and the total content of fatty acids. Moreover, the ratio of linolenic respect to linoleic acid and the ratio of linolenic to all other fatty acids were determined calculating ratios of suitable signal heights. Discriminant analysis was applied to these

1 H NMR of Oils

Minor Compounds 1647

Part III

Root 1 vs. Root 2 12 10 Hazelnut Oil 8 6

Root 2

4 35 %

2

29 % 55 %

0

23 %

21 %

−2 −4

27 %

17 %10 %

55 %

Sunflower Oil

Olive Oil

−6 −8 −35

−30

−25

−20

−15

−10

−5

0

5

10

15

Root 1 Fig. 2. Plot of discriminant functions determined by 1 H NMR parameters. Circles: olive oils; squares: hazelnut oils; rhombus: sunflower oils; crosses: mixtures (m/m%) of olive oil and sunflower oil; triangles: mixtures of olive oil and hazelnut oil. The ellipses display the 95% confidence range for each group of oils. The big cross indicates one olive oil which was repeated four times (from Ref. [11]).

variables permitting the detection of mixtures at a minimum level of 25% hazelnut oil (Figure 2). Some analytical methods for the detection of mixtures of hazelnut oil in olive oil are based on the observation of the differences in δ and γ tocopherol and in β/δ tocopherol ratio [12], that are normally higher in hazelnut oil, and on LC/GC analysis of the esterified sterolic fraction [13]. These methods reveal the addictions of 5–10% of hazelnut oil, but need long times of sample preparation. 1 H NMR spectroscopy can be interesting as a screening method, because it allows much shorter times for sample preparation and analysis. A statistical approach was also applied to the analysis of 13 types of vegetable oils and mixtures of virgin olive oil with seed oils (corn, sunflower, and soybean oil) in an attempt to establish the lowest possible detection level of adulteration. In this case, it was necessary to combine several oil compositional parameters obtained by two NMR approaches. 1,2-diglycerides, 1,3-dyglicerides, and sterols were determined by analysis of 31 P NMR spectra after derivatization of diglycerides with 2-chloro-4,4,5,5tetramethyl dioxaphospholane. On the same phosphitylated samples, 1 H NMR spectra were carried out and

fatty acids were quantified [14]. With this method, it was possible to detect olive oil adulteration at 5%. The analysis of the 1 H NMR spectrum of fish oils permits the quantitative determination of docosahexaenoic acid (DHA), whose positive influence on human health has been emphasized. Its content depends on the fish species, on age, on storage and processing, due to the oxidative deterioration of the highly unsaturated lipids. Due to their proximity to a double bond, methylene C2 protons of DHA resonate downfield to C2 signals of other fatty acids. It was thus possible to determine the DHA content in mg/g in fish oils by comparing the signal areas of the C2 protons of DHA with those of other fatty acids, in the presence of an internal standard solution of ethylene glycol dimethyl ether and 1,4-dioxane [15].

Minor Compounds Phenolic Compounds Phenolic compounds play an important role in the nutritional characteristics and oxidation stability of edible

1648 Part III

Food Science

Part III

vegetable oils. The majority of seed oils contains only traces of phenolic compounds, while virgin olive oil is particularly rich in these antioxidants. Moreover, these compounds contribute in a characteristic way to oil sensory qualities, since they are responsible for its bitter and pungent flavor. The bitter taste is particularly related to the presence of oleuropein glucoside and its aglycon. The content of phenolic compounds in olive oils is heavily affected by the variety, origin, and maturity degree of olives [16]. Phenyl acids and phenyl alcohols (like tyrosol and hydroxytyrosol) have been found prevalently in olives [17], but are also present in olive oil. The prevalent phenolic constituents of olive oil are secoiridoid derivatives, such as the dialdehydic form of the elenolic acid linked to hydroxytyrosol or to tyrosol (ligstroside) and the elenolic acid linked to hydroxytyrosol (oleuropein aglycon). Their structure has been identified using NMR [18], after extraction in absolute methanol and separation by HPLC. Other compounds detected in olive oil phenolic fraction, whose structure was elucidated with NMR, are demethyl-oleuropein, glucosides of hydroxytyrosol [19] and also two soluble lignans—pinoresinol and its derivative, 1-acetoxypinoresinol [20].

reaches a certain level, and when the thermal process is intensified, second oxidation occurs with the formation of aldehydic compounds that are responsible of the undesirable rancid flavor. These aldehydes can be easily detected in the low-field region of the 1 H NMR spectrum [25] and can be used to follow the oxidation process of an oil and to evaluate its oxidative stability. The composition of volatiles plays an important role in defining the sensory quality of virgin olive oils. Positive (green-fruity odor, etc.) and negative attributes (bitter tasting, etc.) have been correlated with the presence of aldehydes, alcohols, polyphenols, terpenes, and acetic acid. The very low concentration of these compounds in virgin olive oil makes possible their direct determination only by high field (500–600 MHz) 1 H NMR [26]. In order to observe these polar minor compounds, they should be soluble in the used solvent. For this reason, a trace of DMSO-d6 is added to chloroform-d in the NMR tube. Moreover, a high number of scans (4000) is a necessary acquisition parameter to obtain these minor compounds with appreciable intensity (Figure 3).

Sterols Aldehydes and Volatile Compounds The aldehydic compounds are important in the evaluation of oil quality for two reasons: first of all they permit to follow the oils oxidative processes; secondly, due to their volatility, they play an important role in the sensory attributes of olive oils. The first 1 H NMR studies regarding oxidation processes of edible oils have been conducted on pure triacylglycerols, like trilinolein, trilinolenin, and triolein. When these model systems were brought to 40 ◦ C in presence of pure oxygen, they formed principally monohydroperoxides, together with bis-hydroperoxides and tris-hydroperoxides [21,22] that produce a characteristic signal at 8–8.9 ppm, due to the hydroperoxide group. Other authors [23,24] have directly monitored by 1 H NMR spectroscopy the deterioration process of edible oils submitted to thermal stressing, in order to study the nature and levels of potentially toxic products, which were produced during heating at frying temperature. They detected the presence of signals due to hydroperoxides and saturated, mono-unsaturated, di-unsaturated, and hydroxy mono-unsaturated aldehydes. While saturated and monounsaturated aldehydes were present in various vegetable oils, the last two were also detected in oils containing a high quantity of polyunsaturated fatty acids (as linseed oil). From these studies, it was concluded that the oxidation process occurs in two stages. In the first stage of thermal oxidation, the fatty acids take oxygen-forming hydroperoxides. When the hydroperoxide concentration

The signal due to methyl CH3 in position 18 of the sterols resonates between 0.6 and 0.7 ppm. The signal due to methyl 18 signal of β-sitosterol, resonating at 0.62 ppm, has been clearly identified in virgin and refined olive oils. Its intensity can be compared with the resonance of 13 C satellites of the methyl group of fatty acids, permitting the quantification of β-sitosterol [5], that can be useful to distinguish seed and pomace oils from olive oils, since the former have shown a higher amount of β-sitosterol and other sterolic compounds. Signals due to stigmasterol [26] and to cyclo-arthenol [27] have been also detected in olive oils.

Diglycerides Diglyceride content and the ratio of 1,2-diglycerides and 1,3-dyglicerides represent an important index of olive oil quality and freshness. The quantitative evaluation of these compounds can be performed with medium field 1 H NMR (200–400 MHz) after oil derivatization with trichloroacetyl isocyanate that reacts with hydroxyl groups of diglycerides leading to the quantitative formation of trichloroacetyl carbamates [28]. After derivatization the CH2 and CH glycerol resonances of 1,2-diacylglycerides are well resolved from the triacylglyceride multiplets, therefore an integration of the relative signal intensities is possible, permitting to evaluate the amounts of 1,2- and 1,3-diglycerides. In addition, the diglyceride proton is converted by derivatization in

1 H NMR of Oils

a

3.9 3.8 3.7 3.6 3.5

ppm

Y b

4.3

4.2

4.1

4.0

3.9

P

P

3.8

3.7 ppm

P

7.6

7.4

7.2

7.0

6.8

6.6

c 6.4

ppm

W

W

d 7.7 7.6 7.5 7.4 7.3 7.2 7.1 7.0 6.9 6.8 6.7

ppm

F e 8.05

8.00

7.95

7.90

ppm

K

H

f

9.7

9.6

9.5

9.4

9.3

ppm

1 H NMR spectrum of olive oil. (a) X are the characteristic

Fig. 3. resonances of a bitter oil; (b) Y of a vinegary tasting oil; (c) and (d) P and W are soluble phenols; (e) F is formaldehyde present in pungent oils; (f) K and H are aldehydes in green-fruity olive oils (from Ref. [26]).

the NH proton, that resonates at 8.4 ppm and it can be integrated to give a quantitative measure of the free hydroxyl groups. The results obtained by this method agree with those obtained with 13 C NMR approach (see Section “Minor oil components” in “13 C NMR of oils”). Anyway, the same information can be achieved with 600 MHz 1 H NMR, since resonances of glyceryls protons of 1,2diglycerides and 1,3-dyglicerides can be detected without the need of derivatization [5]. The same signals were also easily monitorable in the 1 H NMR spectra of encapsulated marine oil supplements [29].

Use of 1 H NMR Spectroscopy to Characterize Olive Oil Geographical Origin Extra virgin olive oil is very valuable among edible oils due to its high nutritional value. Its characteristics are determined by a series of factors: nature of the soil, climate, variety of the plant, cultivation, and extraction techniques. In order to protect producers of high quality olive oils and ensure consumer awareness of product quality, European Community legislation has introduced the Protected Designation of Origin (PDO) mark that allows the labeling of virgin olive oils with the names of the areas where they are produced. This certification increases the oil value. Several studies, based on the combination of 1 H NMR data and multivariate statistical analysis, have been performed for the systematic characterization of olive oils composition. One of the first works was carried out on 55 olive oils coming from four Italian regions [30]. The samples were dissolved in a mixture of 0.7 ml of chloroform-d and 20 µl of DMSO-d6 and the spectra were run on a 600 MHz spectrometer. The heights of 30 resonances present in the spectra were measured and normalized to the height of the signal due to the methylene resonance at 1.26 ppm, to give an index proportional to the molar ratio between each compound and the total amount of fatty acids. Afterwards, a variable selection method permitted to identify peaks which showed significant variations for oils of different geographical origins, that are due to β-sitosterol, saturated and unsaturated aldehydes, and volatile compounds. A cluster analysis was applied to the selected resonances permitting to discriminate the oils in relation to their geographical origin. This procedure demonstrated its reliability when applied on an enlarged data set formed by 216 oils collected in 3 years from different Italian regions, using the same set of selected resonances [31] (Table 2). The question of geographical identification becomes more interesting when it regards small production areas, where high quality products are often obtained. Moreover, often there are several PDO oils in one region, produced in areas not far from each other, as it happens in some Italian regions, like Apulia and Tuscany, where four and three PDO were, respectively, designated for olive oil. For this

Part III

X

X 4.4 4.3 4.2 4.1 4.0

Use of 1 H NMR Spectroscopy to Characterize Olive Oil Geographical Origin 1649

1650 Part III

Food Science

Part III

Table 2: Chemical shift assignment of the selected resonances used for geographical origin discrimination of olive oils (according to Ref. [31]) δ (ppm)

Group

Compound

9.70 9.61 9.54 9.45 8.00 4.88 4.65 4.63 4.53 1.62 0.62

–CHO –CHO –CHO –CHO –CH Nd Nd Nd Nd –CH3 –CH3

Hexanal Unsaturated aldehyde Unsaturated aldehyde Trans-2 hexenal Formaldehyde Cycloarthenol Terpene Terpene Terpene Squalene β-sitosterol

reason, the same previously described statistical approach was used to discriminate 42 extra virgin olive oils from three different zones in Tuscany [32]. Several hierarchical clustering procedures and discriminant analysis were applied on the signal intensities confirming the potential of 1 H NMR in the differentiation of PDO oils. It was also interesting to notice that when oils differed in geographical origin and in cultivar, a better separation of the oils resulted, indicating that both cultivar and geographical origin contributed to the differentiation of oils. A different approach consisted in the application of 1 H NMR to the phenolic extracts of olive oil samples coming from three close and restricted geographic areas. The peak intensities were determined and signals due to aldehydic, vinyl, and aromatic protons of the phenolic compounds were selected for the statistical analysis according to their capability of discriminating the origin of the samples. Discriminant analysis was applied to the selected intensities of the NMR signals [33] and permitted to separate the oils according to their geographical origin obtaining a predictive capacity of 88%.

References 1. Sacchi R, Addeo F, Giudicianni I, Paolillo L. Riv. Ital. Sostanze Grasse 1989;66:171. 2. Cengarle L, Carta A, Pranzetti P. Riv. Ital. Sostanze Grasse 1992;69:463. 3. Miyake Y, Yokomizo K, Matsuzaki N. J. Am. Oil Chem. Soc. 1998;75:15. 4. Saito H. Agric. Biol. Chem. 1987;51:3433.

5. Sacchi R, Patumi M, Fontanazza G, Barone P, Fiordiponti P, Mannina L, Rossi E, Segre AL. J. Am. Oil Chem. Soc. 1996;73:747. 6. Guill´en MD, Ruiz A. Eur. J. Lipid Sci. Technol. 2003;105: 688. 7. Knothe G, Kenar JA. Eur. J. Lipid Sci. Technol. 2004;106: 88. 8. Igarashi T, Aursand M, Hirata Y, Gribbestad IS, Wada S, Nonaka M. J. Am. Oil Chem. Soc. 2000;77:737. 9. Guill´en MD, Ruiz A. J. Sci. Food Agric. 2003;83:338. 10. Guill´en MD, Ruiz A, Cabo N, Chirinos R, Pascual G. J. Am. Oil Chem. Soc. 2003;80:755. 11. Fauhl C, Reniero F, Guillou C. Magn. Res. Chem. 2000;38: 436. 12. Morchio G, Pellegrino A, Mariani C, Bellan G. Riv. Ital. Sostanze Grasse 1999;76:115. 13. Mariani C, Bellan G, Morchio G, Pellegrino A. Riv. Ital. Sostanze Grasse 1999;76:297. 14. Vigli G, Philippidis A, Spyros A, Dais P. J. Agric. Food Chem. 2003;51:5715. 15. Sacchi R, Medina I, Aubourg SP, Addeo F, Paolillo L. J. Am. Oil Chem. Soc. 1993;70:235. 16. Amiot MT, Fleuriet A, Macheix JT. J. Agric. Food Chem. 1986;34:823. 17. Gariboldi P, Jommi G, Verotta L. Phytochemistry 1986;24: 865. 18. Montedoro GF, Servili M, Baldioli M, Selvaggini R, Miniati E, Macchioni A. J. Agric. Food Chem. 1993;41:2228. 19. Bianco A, Mazzei RA, Melchioni C, Romeo G, Scarpati ML, Soriero A, Uccella N. Food Chem. 1998;63:461. 20. Brenes M, Hidalgo FJ, Garcia A, Rios JJ, Garcia P, Zamora R, Garrido A. J. Am. Oil Chem. Soc. 2000;77:715. 21. Neff WE, Frankel EN, Miyashita K. Lipids 1990;25:33. 22. Frankel EN, Neff WE, Miyashita K. Lipids 1990;25:40. 23. Moya Moreno MCM, Mendoza Olivares D, Am`ezquita L´opez FJ, Peris Mart´ınez V, Bosch Reig F. J. Mol. Struct. 1999;482–483:557. 24. Schiller J, S¨uß R, Petkovi´c M, Arnold K. Eur. Food Res. Technol. 2002;215:282. 25. Sacchi R. In: GA Webb, PS Belton, AM Gil, I Delgadillo (Eds). Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 2001, p 213. 26. Mannina L, Segre A. Grasas y Aceites 2002;53:22. 27. Segre AL, Mannina L. Recent Res. Dev. Oil Chem. 1997;1:297. 28. Sacchi R, Paolillo L, Giudicianni I, Addeo F. Ital. J. Food Sci. 1991;4:253. 29. Siddiqui N, Sim J, Silwood CJL, Toms H, Iles RA, Grootveld M. J. Lipid Res. 2003;44:2406. 30. Sacchi R, Mannina L, Fiordiponti P, Barone P, Paolillo L, Patumi M, Segre AL. J. Agric. Food Chem. 1998;46:3947. 31. Mannina L, Patumi M, Proietti N, Bassi D, Segre A. J. Agric. Food Chem. 2001;49:2687. 32. Mannina L, Patumi M, Proietti N, Segre AL. Ital. J. Food Sci. 2001;1:53. 33. Sacco A, Brescia MA, Liuzzi V, Reniero F, Guillou C, Ghelli S, Van Der Meer P. J. Am. Oil Chem. Soc. 2000;77:619.

1651

G´erard J. Martin1 , Serge Akoka2 , and Maryvonne L. Martin1 1 Eurofins 2 University

Introduction

A and R are related by the simple relations:

Applications of high resolution deuterium-NMR in organic and bioorganic chemistry have regularly expanded in the last decades, accompanying the development of high field spectrometers which are required to overcome, at least in part, the severe drawbacks resulting from the nuclear properties of deuterium. In the technological conditions of the seventies both the small gyromagnetic ratio (4.107 × 107 rad/s/T), responsible for poor intrinsic sensitivity (9.6 × 10−3 with respect to 1 H), and the very low natural abundance of deuterium (1.5 × 10−4 ) were serious impediments to the study of natural abundance deuterium spectra. Fortunately, the predominance of quadrupolar relaxation avoids perturbations in signal intensities due to nuclear Overhauser effects. Since 1980, more than 300 articles have been published in the field of high resolution 2 H-NMR and, in this number, about 200 are concerned with quantitative determinations of isotope ratios, for which we have proposed the terminology SNIF-NMR (site-specific natural isotope fractionation studied by nuclear magnetic resonance). Both the first paper dealing with the NMR determination of non-random distributions of deuterium and the corresponding patent concerning applications appeared in 1981 [1,2].

Isotopic Abundances and Isotopic Ratios In a closed system at equilibrium (chemical, plant, animal) the distribution of the different isotopes of a given atom may be described by overall and specific parameters which must satisfy mass and isotopic balance. For a given atom, such as carbon or hydrogen, having both a heavy and a light stable isotope, the isotopic abundance,A, is the ratio of the number of heavy isotopes, H , to the number of heavy and light, L, isotopes: A = H/(H + L)

(1)

The isotope ratio, R, is the ratio of the numbers of heavy and light isotopes: R = H/L Graham A. Webb (ed.), Modern Magnetic Resonance, 1651–1658.
C 2008 Springer.

Scientific, 44323 Nantes, France; and of Nantes, LAIEM, 44322 Nantes, France

(2)

A = R/(R + 1)

(3a)

R = A/(1 − A)

(3b)

In principle, a given molecular species is composed of a large number of isotopomers containing one or several isotopes. Thus, for a given chemical species Cx H y Oz , the theoretical number of isotopomers, N , is: N = (XH3 + 1)a (XH2 + 1)b (XH1 + 1)c (2x )(3z )

(4)

where a, b, and c are the numbers of XH3 , XH2 , and XH1 groups contributing respectively to 4, 3, and 2 isotopomeric possibilities. For example, vanillin, C8 H8 O3 , has 884,736 isotopomers. In the case of hydrogen, owing to the low natural abundance of the deuterium isotope, the presence of bideuterated (and a fortiori multi-deuterated) molecules may be neglected at the usual precision of the NMR determinations and only mono-deuterated isotopomers are generally considered. In these conditions, the A and R values are equal. However, in the case of carbon for instance the A and R parameters should not be confused and bi-labeled molecules may no longer be ignored. It may also be convenient to express the isotopic content in terms of the relative parameter, δ, commonly used in isotope ratio mass spectrometry (IRMS). The isotopic deviation, expressed in ‰, refers the isotopic content of the considered sample or molecule, i, to that of an international reference, ref: δi = 1000(Ri − Rref )/Rref

(5)

The reference is a compound carefully calibrated by interlaboratory campaigns of measurements conducted by international organizations: IAEA (Vienna), IRMM (Geel), NIST (Washington, DC). The certified values of the isotopic ratio and isotopic abundance of the ViennaStandard Mean Ocean Water (V-SMOW) [3] are 155.76 and 155.74 ppm, respectively. A secondary reference, the Standard Light Antarctic Precipitation (SLAP), exhibiting an isotope ratio of 89.0 ppm, is also available from IAEA in Vienna.

Part III

SNIF-NMR—Part 1: Principles

1652 Part III

Food Science

Part III

The site-specific isotope ratio, Ri , of a given molecular site, i, has been defined in the same way [4]: (D/H )i = Di /Hi

(6)

where Di is the number of deuterium atoms at site i. Since the presence of bi-deuterated molecules can usually be neglected Di represents the number of isotopomers mono-deuterated at site i. Hi , the total number of hydrogen atoms of type i, is equal to Pi NH where Pi is the number of equivalent positions at site i and NH the number of fully protonated molecules. The 2 H-NMR investigation of SNIF is presently the only way to access directly and simultaneously to hydrogen isotope ratios, (D/H )i , associated with different positions, i, in a given molecule. It is also convenient to define a relative parameter, Ri/j : Ri/j = F j Si /S j

(7)

where S is the intensity of signal i or j. Ri/j represents the number of heavy isotopes at site i in a situation where site j is arbitrarily given its stoichiometric number of atoms, F j . In principle, S is the area of the NMR signal. However, Ri/j may be semi-empirically defined on the basis of signal heights which are usually measured with better precision than signal areas. Obviously, when computed from signal heights, the Ri/j parameters measured in different samples of a given molecular species may be safely compared only on condition that constant transverse relaxation times are maintained by strictly identical preparation of the samples. In the case of ethanol, for instance, the line widths of the methyl and methylene deuterium signals being nearly equal, the ratio of signal heights is close to the ratio of signal area. When the methylene site, i ≡ II, is compared to the methyl site, j ≡ I, characterized by the stoichiometric number of hydrogens FI = 3, the value of RII/I usually differs strongly from the statistical value 2 which would correspond to a random distribution of the deuterium atoms. Values of RII/I higher than 2.7 are determined for instance in ethanol samples resulting from the fermentation of beet sugar.

Isotopic Fractionation As discussed in Part 2, isotopic fractionation may occur at many steps of the elaboration of a product as a result of thermodynamic or kinetic isotope effects. The site-specific isotopic ratios of a molecule are therefore governed by the isotope contents in the precursor sites of the starting materials and by the subsequent individual fractionating events.

The Mass and Isotopic Balance In a physical, chemical, or biochemical transformation the parameters of the starting and resulting compounds are connected by mass and isotopic balance relationships. Considering a complete transformation of a given compound, S, into two products P and Q, with mole fractions x and (1 - x), the mass and isotope balance requires that the isotopic abundances are related by: AS = xAP + (1 − x)AQ

(8)

Denoting α P and α Q the fractionation factors of P and Q, the disproportionation of isotopes from S to P and Q is written: AS = αP AP = αQ AQ

(9)

More generally the overall isotopic abundances, AS and AQ of initial, S, and final, Q, states composed, respectively of i = 1 to n and j = 1 to m components are: A S = A 1 x 1 + A 2 x 2 + . . . + Ai x i + . . . + A n x n

(10a)

AQ = A1 x1 + A2 x2 + . . . + A j x j + . . . + Am xm (10b) Obviously, in a closed system where the S components have been completely transformed into Q: AQ = AS

Site-Specific Hydrogen Fractionation If the deuterium isotope was randomly distributed within a given molecule, S, containing P hydrogen atoms, the molar fraction of isotopomer i would be equal to the statistical value Fi = Pi /P where Pi denotes the number of equivalent hydrogen positions at site i. In fact, we have shown that most molecules exhibit strong deviations with respect to a statistical distribution of deuterium [1] and that the true molar fractions of the mono-deuterated isotopomers, denoted f i , can be directly determined by quantitative 2 H-NMR. For a molecule with n different sites, the site-specific isotope ratio, Ri , is related to the overall isotope ratio, R, by: Ri = f i R/Fi R=

in Fi

Ri

(11a) (11b)

If we consider again the complete transformation of a set of starting materials, S, containing n sites, into products, Q, the deuterium content, R Qj , in every site, j, of an end product, Q, can be computed with a good approximation, from the deuterium contents, RiS , in the different

SNIF-NMR Principles

R Qj =

n i

a ji RiS

(12a)

or, in matrix notation: RQ = [A]RS

(12b)

where RQ and RS are the column vectors of the deuterium contents in the end and starting molecules respectively and [A] is the matrix of the transfer coefficients, a ji, which connect selectively every site j of an end product to every site i of the reactants. These redistribution coefficients are powerful sources of mechanistic information (cf. Part 2). In particular, the a ji value enables to determine whether a given hydrogen j of a reaction product is connected or not to hydrogen at position i of a starting compound. Intermolecular hydrogen exchanges occurring in the course of the reaction pathway for instance are thus detected [5].

Isotopic Fractionation in the Preparation of the Sample Thermodynamic and kinetic isotope effects associated with physical transformations (cf. Part 2) have important repercussions on the reliability of the experimental isotopic parameters. Usually, a compound extracted from a chemical or natural medium has been subjected to various technological or chemical treatments, each of them being likely to be responsible for isotope effects. Consequently, the SNIF values measured on the extracted product may be considered as representative of the true isotopic parameters of the molecule only on condition that the compound has been completely recovered and that no hydrogen exchange with the medium has occurred [6]. In comparative studies of a given molecular species incompletely extractable from different materials it is therefore necessary either to elaborate a non-fractionating analytical chain or to define standardized analytical procedures in which isotopic fractionation is strictly controlled. In the last case, fractionation effects may be considered as reproducible and the experimental isotope ratios may be safely discussed on a relative basis.

Quantitative Deuterium-NMR Experimental Parameters Firstly it should be emphasized that, in a given magnetic field, the chemical shift discrimination expressed in frequency units is 6.5 times lower in 2 H-NMR as compared to proton NMR. With a spectrometer characterized

by a proton nominal frequency of 400 MHz, for example, deuterium-NMR has therefore a resolution potential somewhat analogous to that of 1 H-NMR at 60 MHz. Consequently, in order to optimize peak separation it is advantageous to resort to high field equipments (11.4 T spectrometers are currently used). Neglecting small intrinsic isotope shifts, the deuterium chemical shifts expressed in ppm are identical to the corresponding proton chemical shifts. This property is very useful for assigning deuterium signals. However, differences in solvent shifts affecting the 1 H and 2 H spectra must be avoided or taken into account. In a first step, a complete analysis of the proton spectrum has therefore to be achieved by exploiting, if necessary, all resources of available one- and two-dimensional NMR sequences. As previously noted, the relaxation of the deuterium nuclei (spin 1) is dominated by the relatively efficient quadrupolar mechanism. Consequently, nuclear Overhauser effects, which are mainly due to dipolar relaxation components, may usually be neglected. This property has important repercussions on the technical conditions of spectrum acquisition and on the methods of quantitative analysis. In particular, it is not necessary to resort to time-consuming gated decoupling sequences aimed to eliminate Overhauser enhancements. For simple organic molecules reorienting relatively rapidly, the quadrupolar longitudinal, T1 , and transversal, T2 , relaxation times are equal and relatively short. Short relaxation times are favorable from the sensitivity point of view since they authorize fast signal accumulation. However line widths increase when T2 decreases. Fortunately, in the case of deuterium, the quadrupolar coupling constant is relatively small and the relaxation times of small molecules are frequently higher than 0.1 s. These values are convenient for efficient accumulation of signals with moderate line widths. For example, in the case of ethanol, accumulation of 90˚ pulses with an acquisition time At equal to 6.8 s fits conveniently the condition At >6 T1 required to recover at least 99.8% of the magnetization of the methyl isotopomer characterized by a relaxation time T1 equal to 1 s. To recover the same proportion of the methylene isotopomer (T1 = 1.25 s) an acquisition time At = 7 s would be required. Moreover, the signal is sampled finely enough to ensure correct curve fitting. In order to minimize random fluctuations due to short-term instabilities of the spectrometer, each spectrum recorded with a given number of scans, NS , is repeated NE times, leading to a duration NS × NE × At of the experiment. For ethanol or vanillin studied at 11.75 T (500 MHz 1 H spectrometer) in a 10 mm OD cell with an acquisition time of 6.8 s, 2.4 and 5.8 h, respectively, are needed. At 7.05 T (300 MHz 1 H spectrometer) and in the same conditions the experimental times are, respectively, equal to 9.7 and 23.2 h. In order to weaken these drastic conditions the acquisition time for ethanol can be reduced

Part III

sites, i, of the starting molecules [5]:

Quantitative Deuterium-NMR 1653

1654 Part III

Food Science

Part III

from 6.8 to 3.4 s but at the price of a small decrease of the internal repeatability. As in most linear analytical techniques, the standard deviation of a series of measurements decreases with the number of repetitions, and obviously the precision increases with the signal-to-noise ratio. From these properties, it is clear that the method is subjected to severe limitations when the compound to be investigated is available only in small quantity (20), and in the hypothesis of Gaussian noise, this algorithm was theoretically proved to ensure optimum precision compared to all other exact and differentiable algorithms. The data are modeled as a sum of Gaussian noise, unphased Lorenztians, and zero-, first-, and second-order baseline distortions. The method, which is well suited for quantitative determination of spectra containing a limited number of signals ( 0 is r 2  ∼ DSf t.

of the spin-echo amplitude with Ig and without applied gradient field I0 [2,3] is given by (2)

Ig /I0 = exp{−γ 2 δ 2 g 2 [ − (1/3)δ]DS }.

(5)

For restricted diffusion it holds [20] r 2  ∼ DSf t 2/(2+θ) .

(3)

The θ effective dimensionality of the random walk vanishes for Brownian motion [3,14,21]. For sufficiently long times, the mean square distance r 2  becomes linear as a function of time r 2  ∼ DSr t.

(4)

The self-diffusion coefficient DSr for restricted diffusion is also called effective diffusion coefficient Deff for long times. The time dependence of the self-diffusion coefficient contains information about the structure of porous media [18,21–24]: (a) characteristic dimensions of the gel structure (critical pore radius rc ), (b) free diffusion for short times, dependent on the pore –structure, and the pore fluid, DSf , and (c), if different from case (b), restricted diffusion behavior for long time DSr . DSr is a measure for the macroscopic permeability of the pore structure, DSf characterizes the viscosity of the serum [8], generally smaller than that of water due to solutes. Ig is the signal (e.g. PGSE or PGTSE), with applied gradient magnetic field g, I0 is the corresponding signal without gradient. The ratio

Figure 1 shows the ratio of the spin-echo amplitudes with and without gradient, Ig /I0 , (a) for a fluid with free diffusion, (b) an emulsion with monodisperse drops (the signal of the matrix has been suppressed), and (c) a gel-network with free diffusion for short times and restricted diffusion for long times. Figure 1 illustrates the method of how to derive the self-diffusion coefficients at free DSf (∼b: slope for short ) and restricted DSr (∼slope d: for long

) diffusion and of the critical pore radius rc of the matrix (∼[DSf t0 ]0.5 ) with the evolution of Ig /I0 . The intersection point of both straight lines, which are asymptotes for the short- and long-time diffusion behavior, defines the critical time t0 . t0 depends on the characteristic length scales of the pore system and DSf . The critical pore radius rc (spherical cavities) is given by [4,25] rc = [5DSf t0 ]0.5 .

(6)

Besides DSf , DSr , and rc , further parameters characterizing the structure the pore system can be derived from DS as a function of : the tortuosity T = DSf /DSr [21] or [(1 − φ)DSf /DSr ]0.5 (φ: mean porosity) [26], the surfaceto-pore-volume ratio S/V , derived from this, the pore size distribution [23], characteristic lengths rch [27], and

Diffusion and Relaxation in Gels

Relaxation TD-NMR [40,41] are applied in food technology in order to (a) characterize the food composition, the chemical structure/physical state (solid/liquid, crystalline/amorphous), or molecular dynamics/structural changes caused by processing [42–44] and (b) derive correlations between NMR parameters and macroscopic quality factors applicable as tools for process/quality control (on-line process NMR spectroscopy). Molecular/microscopic (molecular mass [45]) and macroscopic properties (η [45,46], permeability [47]) can be obtained (averaged, partially resolved).

Ad (a): Structural Characterisation The interpretation of relaxation time data for an unambiguous characterization of different physical states of water or other components present in heterogeneous systems, derived from spectroscopic data [48–50], has not been clarified to its full extent due to chemical [51,52], diffusion [53,54], or spin quantum exchange [55] processes. The relaxation in aqueous systems has been studied by 1 H, 2 H [56,57], 13 C [58], and 17 O [59,60] NMR. 1 H NMR is applied in the majority of the experiments because of the wide spreading of 1 H spectrometers and the possibility to analyze without preparation. 2 H and 17 O NMR, however, provide many fundamental data on water relaxation mechanisms in heterogeneous media [5,61–63]. T1 and T2 allow one to distinguish between free (no interaction with solid particles or dissolved molecules) and immobilized (crystallization, hydration, or other chemically/physically bound) water. Assuming models for exchange processes (chemical, diffusion, spin quantum), the relaxation times and fractions of the corresponding water types can principally be derived. Usually, up to four different groups of spins (two correlated with water, two with the dissolved macromolecules) can be distinguished in protein and polysaccharide solutions [5,64]. Hydroxyl or amino protons can exchange with water protons, other protons are not able to exchange. Combining 1 H, 23 Na, 87 Rb, and 133 Cs NMR and adding the corresponding cations three water phases (free bulk, strongly interacting with κ-carrageenan, and weakly bound) were detected in κ-carrageenan gels [65]. For one-phase relaxation the observable T2 relaxation times T2obs (analogous T1 ) can be used as characteristic T2 . Reviews of empirical or theoretically established models of the correlation between the concentration of dissolved molecules and the observed or the average T2 of solutions are given by [8,83]. Thus, conclusions concerning the influence of macromolecules and their concentrations on the existence of multiple water phases T2,i and their fractions d2,i can be drawn [5,43,66]. Structural Changes Including Phase Transitions and Denaturation Many foods and food components are thermally treated (pasteurization, sterilization, rectification, distillation, drying, freezing, freeze drying, and freeze concentrating) [67]. The variation of the temperature generally causes changing NMR parameters (relaxation times, spectra, or DS ). For simple fluids like water T1 , T2 and DS increase with the temperature. High-resolution NMR is an independent method besides X-ray analysis for the characterization of protein structures (denaturation, pH) [68–70]. Investigations concerning the heat treatment of proteins by means of LR-NMR (T1 , T2 , DS ) have been performed

Part III

the permeability [28]. Many disperse systems are not only heterogeneous at microscopic, but also at meso-/ macroscopic lengths. In some cases, the heterogeneities are caused by phase separation (creaming up) including syneresis (dairy products, mustard), mechanical loading, swelling, the use of fillers/fibers (yoghurts) or processspecific heterogeneities like different morphology of core and shell zones (chocolate, dairy products). Hence, it is desirable to infer structural data (T , S/V ), which are spatially resolved by combining MRI and dynamic imaging methods (diffusion/flow, ENMR), especially, if the characteristic lengths are below the spatial resolution of common MRI techniques [29]. PFG NMR allows one to quantify the effect of the composition, the pH value, the concentration of ions and sugars, the denaturation fraction of proteins and the impact of the processing/process conditions on the structure in gels [30,31]. In dairy products, DSf , DSr , and rc do not seem to correlate with one another. It is plausible to assume that rc , like the average drop radius of water in w/o emulsions, is relevant for the microbiological stability of the gel. The diffusion coefficient of the serum DSf and, even to a higher extent, the permeability DSr presumably correlate with the texture, the sensorial (e.g. sandy), the filtration/washing and syneresis behavior of gels. Furthermore, it is possible to observe structural changes due to sol–gel transitions (hydrocolloids, denaturation), their temporal evolution, dependence on denaturation temperature/duration/fraction or ageing-processes during storage (syneresis) [32]. Combining PFG and T2 two fractions of water (bound to collagen or retained in channels traversing collagen fibers) were observed in collagen gels [9]. DS of aqueous proteins systems increases with temperature and water content [33]. Analogous experiments have been performed in starch/dextran [32,34,35] or casein systems [36], cheese [37], curdlan [38], and gellan gum [39]. The correlation between DSf , DSr , and rc characterizes different sol/gel states. A review of diffusion experiments in disperse media (emulsions, porous foods) is given in Ref. [8].

Relaxation 1715

1716 Part III

Food Science

Part III

[71–76] after and not during the thermal treatment (at temperatures below the denaturation temperature in order to prevent further changes) varying the process temperature, the pH value and the concentrations of protein, ions and sugars. Reversible (increased T2 : higher mobility at higher temperature) and irreversible transformations (decreased T2 : aggregation) can thereby not be differentiated. This procedure provides only reaction kinetic information using a series of tests with different samples. This is a great disadvantage for natural substances due to their compositional variety. In aqueous protein systems generally four phases can be distinguished. There are both proteins whose T2,i of the phases with higher mobility are increased due to the thermal denaturation and proteins with T2,i reduced after the thermal denaturation, whereby only two phases have been detected [72]. The dependence between T1 or T2 and the temperature (fixed treatment duration) and the T2 /T1 ratio indicate structural changes (sol–gel transitions, freezing damage [73]). Online NMR-measurements offer possibility to study denaturation the during the thermal treatment (average time T2m , relaxation times of the single phases T2,i , and fractions di ). The advantage is that several measurements can be performed with the same sample (without interrupting the thermal treatment and/or continuing with another sample), thus achieving higher accuracies of measurements. The reaction orders and rate constant of the participating reactions can be derived. If the relaxation data vary after thermal/pressure treatments, reversible and irreversible structural changes can be distinguished. WHC Including Sol–Gel Transitions and Syneresis In food technology, the bonding type and mobility of water molecules are relevant with regard to the yield, sensory evaluation, stability, texture, flow behavior, and processing. There are many measuring methods available to characterize the WHC of a mixture [77–80]; many techniques originally developed in practice, provide quick, and easy indications concerning the studied material. Most of the available methods usually provide differing results, which are not absolute values and are, therefore, generally not comparable or transferable to technological processes. These difficulties are due to (i) different measuring principles and experimental conditions and (ii) the frequent destruction/modification of the structure due to the measuring procedure (centrifugation). A precise characterization of the WHC for food constituents like proteins and hydrocolloids is required for a reliable design of products and processes. For different materials, the relationship between the WHC and the other macroscopic (flow/deformation/sensoryl/syneresis behavior) and microscopic properties (fine stranded/particulate network) is not identical, thus complicating a practicable definition of the WHC. Rennet acid gels, for example, have a coarser

network with thicker protein strings than pure acid gels [81]. Rennet gels are firmer, with a lower WHC. As mentioned before, the definition of water types depends on the applied measuring technique. The fraction of bound water is defined in rheology as bound physically/chemically at solid surfaces or retained in the interior of solid particles. With regard to NMR water is bound, if it interacts with solid material or if the water molecules are in sufficiently small cavities. For free water it is not of any importance, if fluid molecules are inside or outside of solid particles. Another difficulty is that disperse media usually have different microstructures during rheological (flowing) and NMR experiments (at rest), apart from Rheo-NMR [82] including flow experiments. In order to differentiate free water, which can be removed through capillaries (non-stirred) or out of the pore system (stirred), from water retained in micelles/agglomerates, the special setup was developed [8,83–85]. The sample is filled into an NMR glass closed at the bottom with a PTFE-thread. T2 experiments are performed at the original sample and afterward at the sample being washed with D2 O. Between sample and thread is a filter membrane. The mobile free water in the sample is substituted by D2 O non-detectable by 1 H NMR. Exchange processes between water and OH/NH/SH groups are neglected. Thus, with the help of the washout test, immobilized, retained free and mobile free water can be distinguished. The fraction of water/serum that contributes to phase-separation phenomena like syneresis possibly correlates with the phase distribution and the amount of the water, which can be washed out. If the water content is known, it is possible to define the WHC (physically/chemically bound plus water retained in pores) with the fraction pr of washable water WHCNMR =

m Wnw m mW = Wnw = pr F, mS mW mS

(7)

where m Wnw is the mass of non-washable water, m W the total mass of water, m S the dry mass, pr the m Wnw /m W , averaged over all phases, and F the water content of a gel (w/w). It is remarkable that not only water of the phase with the highest mobility, but also of phases with reduced mobility (T2 ) is washable. This means that the molecular mobility characterized by T2 and the washability (Equation 7) are different, but complementary characteristic gel features. The washout test was used to study the influence of the recipe and processing on the evolution of the WHC and pr of yoghurts (fermented in glasses or cut out) [31]. Hydrocolloids Hydrocolloids are used (e.g. jellied and dairy products, cocoa, ice-cream, marmalade, sausages, convenience food) as thickeners, gelling agents, or stabilizers. It is to incorporate water into foods (higher yield and lower fat

Diffusion and Relaxation in Gels

Ad (b): Correlations The motivation to derive correlations between NMR and material properties is that these correlations are ideal to determine industrially relevant quantities (material process parameters) by partly methodologically/apparatively more comfortable, non-destructive,

often less time-consuming and on-line capable NMR methods (no trace analysis). This is a necessary prerequisite for an effective process/quality control [109–112]. Continuous- and stop-flow process NMR spectroscopy are realized [113]. The continuous-flow requires usually premagnetization for a complete polarization of the sample. The sensor is integrated into a pipe, belts, or chutes. Applications are the monitoring of the fermentation (yoghurts) [83], mashing [46], and cooking (rice species, shelling/polishing process: G¨otz et al., 2004). Besides DS determined by PFG NMR, relaxation times are another independent approach to characterize the gel structure and to study the dynamics of de-/ hydration of fluids. The relaxation times of fluids in gels (partial/completely saturated) are generally reduced compared to the free and bulk fluid due to interactions between liquid molecules and the liquid–solid interface. In many cases, a simple two-site model is used to interpret the observed enhanced relaxation rates in gels [48,53,114], partly a homogeneous water layer is assumed at the solid surface [24,115]. A practicable approach based on T1 is the assumption of rapid exchange of liquid molecules between (i) the zone at the solid surface (interactions between the liquid and the solid material), and (ii) the molecules in the bulk. The change of T1 (porous medium) compared to T1f in the bulk phase is used to derive the fraction f of the fluid interacting with the solid. In order to characterize the density of reactive surface sites in a fluid-saturated porous system, the influence of ion concentrations from different salts on T1 was measured [28]. Fluid molecules near solid walls are exposed to relaxation mechanisms different from those in the bulk fluid. Various models have been developed to derive pore size distributions from relaxation measurements. Due to the complex interactions, formalisms generally valid for all porous media have not been developed yet [116]. In spite of these ambiguities, relaxation measurements enable one to characterize individual pore systems. Based on a twosite model, the mean relaxation time T1 is related to the pore radius rp = 2V /S. Thus, it is possible to derive the pore size distribution from the T1 frequency distribution [118–120]. If the gel structures do not significantly alter in the temperature regime above and below the freezing phase transition of the fluid, measurements are performed at temperatures above and below the freezing phase transition. While the bulk fluid is frozen, a non-frozen fluid layer exists at the solid surface due to the different and often incommensurable crystal structures of the fluid adsorbate and the surface structure. The fluid molecules in the adsorbed layer are, even in the frozen state of the bulk fluid, surprisingly mobile (translationally along the surface). In order to obtain information concerning the size distribution of crystallites in frozen gels the size distribution of pores in freeze-dried starchgels was determined by T2

Part III

content). Many functional properties of proteins in foods are attributed to interactions with water and other food constituents [79]. Many hydrocolloids exhibit hysteresis of their TD-NMR or rheological parameters for cyclic temperature variations [31,86,87] and a strong pH dependence [88]. Polymer networks are usually thermoreversible due to the H-bridges breaking up at sufficiently high temperatures. The influence of temperature and gelatine concentration on the structure of gelatine solutions was studied with T1 [89] and T2 [90]. The water relaxation in biopolymer systems was modeled assuming proton exchange cross-relaxation [91,92]. The mean T2 relaxation times in gelatine systems increases dramatically at the temperature-induced sol/gel transition. Chemical/diffusive exchange in water/gelatine systems was studied by variation of the CPMG pulse spacing [93]. 1 H and 23 Na NMR (T1 , T2 ) enable one to differentiate up to four water phases in aqueous gelatine gels [94,95]. Properties of starch relevant for food technology are the swelling (temperature, moisture: 1 H: T1 , T2 ; 2 H: T1 , T2 , spectra [96,97]), plastisication/gelatinization (1 H T2 , 2 H spectra [98]), the degree of polymerization, glass transitions (T1 , T2 , spectra [99]), retrogradation (T1 , T2 [100]), conformations of dissolved starch molecules like multiple/doublehelices or eggbox [86], fraction of crystallinity (CP-MAS [101], 13 C [66]), the ratio of amylose/amylopectin [102], water content (T1 , T2 [89,99,103]), 2 H spectra [104], and storage behavior (T2 , MRI [105]). After a heat treatment the carrageenan molecules reveal a disordered coil form. When reducing the temperature, the conformation of κand ι-carrageenan changes to ordered helical structure (e.g. double-helices) and aggregates of helices coupled to gelation [87]. Various parameters [concentration, temperature, type (T1 , T2 [86,89]), molecular mass (1 H spectra [105]), composition of carrageenan molecules (κ-, ι-units: 1 H, 13 C spectra [107]), ionic type/content [87]/composition [108]] have a strong impact on macroscopic (texture, flow/sensorial behavior, sol–gel transition) and microscopic properties (T1 , T2 , spectra). Gelling (hysteresis) and only viscosity-increasing carrageenan types (λ) are differentiable by the temperature dependence of the T2,i and di [8,31]. This provides a deeper insight into the microstructure than T2m alone [86]. T2m for the solutions correlates with their voluminosity (i.e. immobilized water). The voluminosity in gels is not determinable with rheology yet.

Relaxation 1717

1718 Part III

Food Science

Part III

relaxometry using acetone (no matrix collapse) instead of water [121,122]. Besides the possibility to calculate the gel permeability (T, S/V ) [23,28,123] with empirical models (permeability porosity φ) [124], permeabilities by correlating T1 , T2 , T1ρ , DS , and φ are derived [125]. Further correlations between macroscopic properties (water activity, microbial stability/shelf-life, solutionrate of spray-dried powders, caking behavior of powders, drug release in proteineous matrices, and sensory texture) or capillary/pressure curves and NMR data are given in Ref. [8].

References 1. Le Bihan D. In: DM Grant, RK Harris (Eds). Encyclopedia of NMR, Vol. 3. Wiley: Chichester, 1996, p 1645. 2. Stejskal EO, Tanner JE. J. Chem. Phys. 1965;41:288. 3. Callaghan PT. Principles of NMR Microscopy. Oxford University Press: Oxford, 1991. 4. K¨arger J, Fleischer G. TRAC. 1994;12:145. 5. Colquhoun IJ, Goodfellow BJ. In: RH Wilson (Ed). Spectroscopic Techniques for Food Analysis. John Wiley & Sons: New York, 1994, p 87. 6. Hills BP. Magnetic Resonance Imaging in Food Science. John Wiley & Sons: New York, 1998. 7. Bartusseck I. NMR-Untersuchungen von Diffusionsvorg¨angen in Mehrkomponentensystemen. RWTH, Diss., Aachen, 2002. 8. G¨otz J. Applications of NMR to Food and Model Systems in Process Engineering. Habilitation, http://tumb1.biblio. tu-muenchen.de/publ/diss/ww/2004/goetz. html, TUM, 2004. 9. Knauss R, Fleischer G, Gr¨under W, K¨arger J, Werner A. Magn. Reson. Med. 1996;36:241. 10. Ohtsuka A, Watanabe T. Carbohydr. Polym. 1996;30:135. 11. Watanabe H, Fukuoka M. In: TL Barsby, AM Donald, PJ Frazier (Eds). Starch—Advances in Structure and Function. RSC: Cambridge, 2001, p 53. 12. Holz M. Nachr. Chem. Tech. Lab. 1986;34:858. 13. Velan SS, Chandrakumar N. J. Magn. Reson. 1996;123:122. 14. Appel M, Fleischer G, K¨arger J, Fujara F, Siegel S. Europhys. Lett. 1996;34:483. 15. Callaghan PT. J. Magn. Reson. Ser. A. 1995;113:53. 16. Packer KJ. In: DM Grant, RK Harris (Eds). Encyclopedia of NMR, Vol. 3. Wiley Press: Chichester, 1996, p 1615. 17. Johnson CS. In: DM Grant, RK Harris (Eds). Encyclopedia of NMR, Vol. 3. Wiley Press: Chichester, 1996, p 1626. 18. Mitra PP. Phys. A. 1997;241:122. 19. Mandelbrot BB. The Fractal Geometry of Nature. Freemann, San Francisco, 1982. 20. Banavar JR, Lipsicas M, Willemsen JF. Phys. Rev. B. 1985;32:6066. 21. K¨arger J, Heitjens P, Haberlandt R (Eds). Diffusion in Condensed Matter. Vieweg: Wiesbaden, 1998. 22. Hallmann M, Unger KK, Appel M, Fleischer G, K¨arger J. J. Phys. Chem. 1996;100:7729. 23. Latour LL, Kleinberg RL, Mitra PP, Sotak CH. J. Magn. Reson. Ser. A. 1995;112:83.

24. Watson AT, Chang CT. Prog. Nucl. Magn. Reson. Spectrosc. 1997;31:343. 25. Appel M, Fleischer G, K¨arger J, Dieng AC, Riess G. Macromolecules. 1995;28:2345. 26. Epstein N. Chem. Eng. Sci. 1989;44:777. 27. K¨arger J, Pfeifer H, Heink W. Adv. Magn. Res. 1988;12:1. 28. Frosch GP, Tillich JE, Haselmeier R, Holz M, Althaus E. Geothermics. 2000;29:671. 29. Tillich J. NMR-Methoden zum Studium von Struktur und Transport in por¨osen Medien. Uni., Diss., Karlsruhe, 2002. 30. G¨otz J, Zick K, Hinrichs R, Weisser H. Z. Lebensm. Unters. FA. 2004;218:323. 31. Hinrichs R, G¨otz J, Weisser H. Proc 2nd Int Workshop Eurofood’s Water. Reims, 2002, Specif. Food Chem., 2003. 32. Ohtsuka A, Watanabe T, Suzuki T. Carbohydr. Polym. 1994;25:95. 33. Kimmich R, Klammler F, Skirda VD, Serebrennikova IA, Maklakov AI, Fatkulin N, Appl. Magn. Reson. 1993;4:425. 34. Farhat IA, Loisel E, Saez P, Derbyshire W, Blanshard JM. Int. J. Food Sci. Tech. 1997;32:377. 35. Kwak S, Viet MT, Lafleur M. J. Magn. Reson. 2003;162: 198. 36. Mariette F, Topgaard D, J¨onsson B, S¨oderman O. J. Agric. Food Chem. 2002;50:4295. 37. M´etais A, Mariette F. J. Magn. Reson. 2003;165:265. 38. Kwak S, Lafleur M. Colloid Surface A. 2003;221:231. 39. Watanabe T, Ohtsuka A, Murase N, Barth P, Gersonde K. Magn. Reson. Med. 1996;35:697. 40. Belton PS. In: RH Wilson (Ed). Spectroscopic Techniques for Food Analysis. John Wiley & Sons: New York, 1994. 41. Alberti A, Belton PS, Gil AM. Ann. Rep. NMR Spectrom. 2002;47:111. 42. Belton PS. Prog. Biophys. Mol. Biol. 1994;61:61. 43. Cornillon P. Semin. Food Anal. 1998;3:235. 44. Pedersen HT, Ablett S, Martin DR, Mallett MJ, Engelsen SB. J. Magn. Reson. 2003;165:49. 45. G¨otz J, Altmann S, Weisser H. Proc. 1st Eur. Silicon Days, M¨unchen, 2001. 46. G¨otz J, Schneider J, F¨orst P, Weisser H. J. Am. Soc. Brewing Chem. 2003;61:37. 47. Watson AT, Chang CT. Prog. Nucl. Magn. Reson. Spectrosc. 1997;31:343. 48. Zimmerman JR, Brittin WE. J. Phys. Chem. 1957;61:1328. 49. Hills BP, Takacs S, Belton P. Food Chem. 1990;37:95. 50. Belton PS. Int. J. Biol. Macromol. 1997;21:81. 51. Carver JP, Richards RE. J. Magn. Reson. 1972;6:89. 52. Berliner LJ, Reuben J. Biological Magnetic Resonance. Plenum Press: New York, 1980. 53. Brownstein KR, Tarr CE. Phys. Rev. A. 1979;19:2446. 54. Belton PS, Hills BP. Mol. Phys. 1987;61:999. 55. Edzes H, Samulski E. J. Magn. Reson. 1987;31:202. 56. Hallenga K, Koenig SH. Biochemistry. 1976;15:4255. 57. T. Kumosinski F, Pessen H, Farrell HM. In: H Levine, L Slade (Eds). Water Relationships in Foods, Advances in Experimental Medicine and Biology, Vol. 302. Plenum Press: New York, 1991, p 541. 58. Cassin G, de Costa C, van Duynhoven J, Agterof WG. Langmuir. 1998;14:5757. 59. Halle B, Denisov VP, Venu K. In: LJ Krishna Berliner (Ed). Biological Magnetic Resonance, Vol. 17. Kluwer/Plenum Publishers: New York, 1999.

Diffusion and Relaxation in Gels

88. Kerr WL, Wicker L. Carbohydr. Polym. 2000;42:133. 89. Labuza TP, Busk GC. J. Food Sci. 1979;44:1379. 90. Lambelet P, Berrocal R, Desarzens C, Froehlicher I, Ducret F. J. Food Sci. 1988;53:943. 91. Hills BP. Mol. Phys. 1992;76:489/509. 92. Djabourov M. G`elification Thermor`eversible du Syst`eme Eau-G`elatine. Paris: Uni., PhD, 1986. 93. Traor´e A, Foucat L, Renou J-P. Eur. Biophys. J. 2000;29:159. 94. Vackier M-C, Hills BP, Rutledge DN, J. Magn. Reson. 1999;138:36. 95. Vackier M-C, Rutledge DN. Food Chem. 1996;57:287. 96. Lechert H, Maiwald W, K¨othe R, Basler W-D. J. Food Process. Preserv. 1980;3:275. 97. Lechert H. Proc. 52nd Conference FEI, Hamburg, 1994. 98. Tang H-R, Brun A, Hills BP. Carbohydr. Polym. 2001; 46:7. 99. Kou Y, Dickinson LC, Chinachoti P. J. Agric. Food Chem. 2000;48:5489. 100. Farhat IA, Blanshard JM, Mitchell JR. Biopolymers. 2000;52:411. 101. Blanshard JM, Jarozkiewicz EM, Gidley MJ. In: JW Finley, SJ Schmidt, A Seriani (Eds). NMR Applications in Biopolymers. Plenum: New York, 1990, p 155. 102. Dunn LB, Krueger WJ. In: H H¨ocker (Ed). Macromolecular Symposia. Wiley-VVH: Weinheim, 1999, p 179. 103. Leung HK, Steinberg MP, Wie LS, Nelson AI. J. Food Sci. 1976;41:297. 104. Yakubu PI, Baianu IC, Orr PH. In: DS Reid (Ed). The Properties of Water in Foods ISOPOW 6. Blackie Academicand International: London, 1998, p 160. 105. Ruan RR, Almaer S, Huang VT, Perkins P, Chen P, Fulcher RG. Cereal Chem. 1996;73:328. 106. van de Velde F, Peppelman HA, Rollema HS, Tromp RH. Carbohydr. Res. 2001;331:271. 107. Borgstrom J, Egermayer M, Sparrman T, Quist PO, Picullell L. Langmuir. 1998;14:4935. 108. McCarthy MJ, McCarthy KL. Magn. Reson. Imaging. 1996;14:799. 109. Hills BP. Trends Food Sci. Tech. 1995;6:111. 110. Rutledge DN. Food Control. 2001;12:437. 111. Cho SI, Chung CH. TASAE. 2001;44:377. 112. Nordon A, McGill CA, Littlejohn D. Analyst. 2001;126: 260. 113. Barrie PJ. Ann. RNMR S. 2000;41:265. 114. Tellier C, Mariette F, Guillement JP, Marchall P. J. Agric. Food Chem. 1993;41:2259. 115. Smith DM, Hua D-W, Earl WL. MRS Bull. 1994;19:44. 116. Chui MM, Philipps RJ, McCarthy MJ. J. Colloid Interface Sci. 1995;174:336. 117. Stapf S, Kimmich R. J. Chem. Phys. 1995;103:2247. 118. Zavada T, Kimmich R. Phys. Rev. E. 1999;59:5848. 119. Hills BP, Snaar JE. Mol. Phys. 1992;76:979. 120. Hills BP, Wright KM, Snaar JE. J. Magn. Reson. Anal. 1996;2:305. 121. H¨urlimann MD, Latour LL, Sotak CH. Magn. Reson. Imaging. 1994;12:325. 122. Gupte AR. Widerstandsgesetz der Porenstr¨omung. Uni., Diss., Karlsruhe, 1970. 123. Friedmann TE, Windhab E. Separ. Sci. Technol. 1998;33:2221.

Part III

60. Mora-Gutierrez A, Kumosinski TF, Farrell HM. J. Agric. Food Chem. 1997;45:4545. 61. Ruan RR, Chen PL. Water in Foods and Biological Systems. A Nuclear Magnetic Approach. Technomic Publishing: Lancaster, 1998. 62. Belloque J, Ramos M. Trends Food Sci. Tech. 1999;10:313. 63. N¨olting B. Protein Folding Kinetics, Springer, Berlin, 1999. 64. Nusser W. NMR-Untersuchungen zur molekularen Dynamik in Protein-Wasser-Systemen. Uni., Diss., Ulm, 1990. 65. Hikichi K. Polym. Gels Netw. 1993;1:19. 66. Gidley MJ, McArthur AJ, Darke AH, Ablett S. In: E Dickinson (Ed). New Physico-Chemical Techniques for the Characterisation of Complex Food Systems. Blackie Academic Professional: London, 1995, p 296. 67. Harz H-P. Untersuchungen zum Gefrierverhalten fl¨ussiger Lebensmittel. Uni., Diss., Karlsruhe, 1987. 68. Jonas J, Jonas A. Annu. Rev. Biophys. Biomol. Struct. 1994;23:287. 69. Cavanagh J, Fairbrother WJ, Palmer AG, Skelton NJ. Protein NMR Spectroscopy, Principles and Practice. Academic Press: New York, 1996. 70. Hancock TJ, Hsu JT. Biotech. Prog. 1996;12:494. 71. Lambelet P, Berrocal R, Renevey F. J. Dairy Res. 1992;59:517. 72. Lambelet P, Ducret F, Leuba JL, Geoffroy M. J. Agric. Food Chem. 1991;39:287. 73. Lambelet P, Renevey F, Kaabi C, Raemy A. J. Food Sci. 1995;43:1462. 74. Hills BP, Takacs S, Belton P. Mol. Phys. 1989;67:919. 75. Barfod N. Proc. Conference, Rebild, Denmark, 1991. 76. LeDean A, Mariette F, Lucas T, Marin M. Lebensm. Wiss. Technol. 2001;34:299. 77. Kinsella JE. Crit. Rev. Food Sci. 1984;21:197. 78. Kinsella JE, Fox PF. In: Int. Dairy Federation (Ed). IDF Bulletin 209, 1987, p 12. 79. Kneifel W, Paquin P, Abert T, Richard J-P. J. Dairy Sci. 1991;74:2027. 80. Kulicke W-M, Eidam D, Kath F, Kix M, Kull AH. Starch/St¨arke. 1996;48:105. 81. Schkoda P. Serumbindung und Rheologie fermentierter Milchprodukte. TU, Diss., M¨unchen, 1998. 82. G¨otz J. Rheo-NMR with applications on food. In: G Webb (Ed). Handbook of Modern Magnetic Resonance. Kluwer Academic Publishers, London, 2005. 83. Hinrichs R. NMR-Messungen zum Charakterisieren der Wasserbindung und der Struktur in Systemen aus Milchproteinen und Hydrokolloiden.: TU, Diss., M¨unchen, 2004. 84. Hinrichs R, G¨otz J, Noll M, Wolfschoon A, Eibel H, Weisser H. Characterisation of Different Treated Whey Protein Concentrates by Means of Low Resolution Nuclear Magnetic Resonance (LR NMR). To appear in International Dairy Journal, 2006. 85. Hinrichs R, G¨otz J, Noll M, Wolfschoon A, Eibel H, Weisser H. Characterisation of the Water-Holding Capacity of Fresh Cheese Samples by Means of Low Resolution Nuclear Magnetic Resonance. Food Res. Int. 2004;37:667–676. 86. Suggett A. In: F Franks (Ed). Water, Vol. 4. Plenum Press: New York, 1975, p 550. 87. Lewis GP, Derbyshire W, Ablett S, Lillford PJ, Norton IT. Carbohydr. Res. 1987;160:397.

References 1719

1721

B.P. Hills Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 7UA,UK

Introduction The importance of the worldwide production of fruit and vegetables both for human health and as an economic force should not be underestimated. In the United States alone, the fruit and vegetable industry generates over $6 billion in income annually. Much research effort is aimed at maximizing production by optimizing the factors affecting growth in the field such as irrigation, nutrients, and harvesting times as well as finding optimum storage conditions, especially for fruit kept in long-term storage for export and/or off-season marketing. It also involves finding the best ways to sort and grade fruit coming out of storage which means developing non-invasive sensors of fruit and vegetable quality. Because it is non-invasive, NMR has been widely used to research each of these stages. High-resolution NMR spectroscopy and imaging have been used to study physiological changes during plant growth, development, and storage and these aspects have been reviewed [1]. MRI has also been used to monitor water transport during preservation processes such as drying and freezing and examples of this work can be found in Refs. [2–4]. However, the main focus of this chapter will be the correlation between proton NMR relaxometry and diffusometry and fruit and vegetable quality. As we shall see, these NMR techniques are especially sensitive to the (sub)-cellular compartmentation of water and this is affected by physiological changes during ripening and storage and also holds promise for the development of non-invasive NMR sensors during the sorting and grading stage. We therefore begin by reviewing the relationship between NMR water proton relaxometry and diffusometry and cell tissue structure.

NMR Relaxation and Water Compartmentation In the absence of paramagnetic ions, there are three main phenomena that affect water proton relaxation in plant tissue, namely, (a) Fast proton exchange on a (sub-)millisecond timescale between water and exchangeable protons Graham A. Webb (ed.), Modern Magnetic Resonance, 1721–1727.
C 2008 Springer.

on cell metabolites, especially sugars and biopolymers, including the cell wall biopolymers and starch granules. (b) Dephasing of transverse magnetization by the diffusion of water molecules through internal field gradients created by magnetic susceptibility discontinuities in the tissue, especially across the interface of inter-cellular air gaps. This effect is especially important at high field strengths when gradients are largest. (c) Diffusion of water between the various subcellular (vacuolar, cytoplasmic) and extra-cellular water compartments which averages the magnetization to an extent that depends on cell morphology and membrane permeability. The fact that plant cells are compartmentalized means that, in general, the transverse and longitudinal relaxation are multiple exponential. Moreover, if transverse relaxation occurs by the proton exchange and internal field gradient mechanisms then, in general, the observed CPMG transverse relaxation behavior depends on the field strength and the CPMG pulse spacing. However, relating the NMR data to factors such as organelle size and membrane permeability is far from straightforward though a start has been made by combining the proton exchange model [5] with a simplified spherical cell model of a plant cell [6]. This has proved useful in fitting the changing relaxation data during the drying and freezing of apple [7], carrot [8], and potato [9] parenchyma tissue. In principle, the same theoretical approach could be used to model the changes in relaxation behavior observed as plant tissue ripens, though this has yet to be attempted. The potential of multi-dimensional relaxation experiments for exploring the relationship between water compartmentation and fruit/vegetable quality has only just begun to be explored. Figure 1 shows the T1 –T2 crosscorrelation spectrum for parenchyma tissue of fresh red delicious apple. The dependence of the peak positions and intensities on physiological conditions such as mealiness are being researched in the authors laboratory and hold considerable promise [10]. Given the likelihood that online NMR sensors will be based on low rather than high field strengths it is

Part III

NMR Relaxation and Diffusion Studies of Horticultural Products

1722 Part III

Food Science

101

2D Laplace transform of REDD-FT20-T1T2-23-200-01 (Log intensity)

100

T2 (secs)

Part III

Fig. 1. A typical T1 –T2 correlation spectrum of parenchyma tissue of red delicious apple acquired at room temperature at 23 MHz with a CPMG pulse spacing of 200 µs. The diagonal line denotes T1 = T2 . The vertical and horizontal dotdashed lines denote the limits of reliability based on the acquisition parameters.

10−1

10−2

10−3

10−4 10−4

surprising that so little field cycling relaxometry has been reported for plant tissue. The dependence of the longitudinal relaxation time on field strength gives a direct handle on the spectral densities describing molecular mobility in the tissue, data that could provide a useful test of the numerical cell model. It would be especially useful to see how the spectral densities change during tissue breakdown for conditions such as mealiness and chilling injury. Although correlations between relaxation and quality naturally focus on simple T2 and T1 measurements, it should not be forgotten that these are not the only types of relaxation that can be studied. Relaxation in the rotating frame, characterized by T1ρ , is another potentially valuable handle on tissue quality and this, in general, depends on the strength of the spin-locking field and the spin-locking angle. Possible correlations between cross-polarization rates and quality factors are an additional aspect that appears to have been overlooked, yet image contrast based on cross relaxation is an important diagnostic tool in clinical imaging and therefore, potentially, in plant imaging. In this context, cross relaxation refers to the transfer of longitudinal proton magnetization between water and more rigid biopolymers in the plant tissue matrix. These rates can be

10−3

10−2

10−1 T1 (secs)

100

101

measured by sequences such as those of Goldman and Shen [11].

NMR Diffusometry and Water Compartmentation In contrast to water proton relaxation, there have been very few attempts to correlate water diffusivity in cellular tissue as measured with PGSE sequences to plant quality factors. An exception is the early work of Keener et al. [12] who found a useful correlation between the apparent water diffusion coefficient and apple maturity as measured by the refractometer Brix value. The paucity of work on water diffusivity is surprising because the restriction of water diffusion by cell membrane and cell wall barriers is a useful and sensitive probe of changes in cell morphology. The potential of PGSE studies of cellular tissue was illustrated by Hills and Snaar in their studies of fresh parenchyma apple tissue [6]. In a PGSE, Stejskal–Tanner experiment the echo amplitude, S(q, , τ ) is a function of both the echo time, 2τ , the diffusion time, , defined as the time between gradient pulses and the wave vector, q, defined as the area of the gradient pulse, γ Gδ/2π. For

NMR in Horticulture

Fruit and Vegetable Quality 1723

i

ai (experimental)

ai (theory)

pi (experimental)

pi (theory)

1 2 3

0.292 0.032 0.007

0.302 0.044 0.017

0.78 0.13 0.09

0.80 0.10 0.10

variation of at fixed τ we can write S(q, , τ ) = i pi (q, τ ) exp(− /ai (q, τ )),

(1)

where the ai (q, τ ) are generalized diffusion times and pi are the associated fractional populations. Both these coefficients depend on the wave vector, q. In principle, these additional coefficients can be analyzed and related to quality factors in an analogous manner to the relaxation times discussed in the previous section, though no one has yet attempted to do this. Table 1 shows a comparison of the experimental coefficients for fresh apple tissue with those calculated with the numerical cell model. The agreement is quite reasonable and supports the general theoretical framework. It would be interesting to apply this approach to mealy apple and to other types of fruit and vegetable. Quite clearly there remains a great deal of basic NMR and modeling work to be done before we can claim to understand the relaxation and diffusion behavior of horticultural products. The next section surveys the literature up to 2004 with no claim to completeness.

Fruit and Vegetable Quality The literature to the end of 1999 relating to internal defects and storage disorders has been summarized by Clark et al. [1] and Koizumi et al. [13]. The more recent literature to 2003 has been reviewed by Hills and Clark [14]. The types of **product studied and the quality disorders investigated are summarized in Table 2. Space does not permit each product in Table 2 to be discussed in detail, so we illustrate the NMR-quality correlations for the most intensively studied examples, namely apple and potato.

Apple Maturity Up to three peaks have been reported in the distribution of water proton transverse relaxation times in parenchyma apple tissue measured with the CPMG sequence with a short 90–180◦ pulse spacing at 2.3 T (100 MHz) [7]. These peaks have been assigned to vacuolar water and water associated with cytoplasm and cell wall. The effect of freezing and drying on this relaxation time spectrum has

been modeled with a numerical cell model with moderate success. Keener et al. [12] have investigated the relationship between low-field (0.13 T; 5.4 MHz) apparent water proton diffusion coefficients (Dw ), the dominant CPMG T2 measured with a pulse spacing of 2 ms, and degree of ripeness (soluble solids content) as measured by refractometry. The following trends were reported: (a) In healthy “Golden Delicious” and “Granny Smith” apples, the T2 decreased with increasing soluble solid content (Brix) in healthy apples. (b) The water diffusivity decreased with increasing soluble solid content in all samples, both bruised and healthy. This is no doubt a result of the increasing viscosity of more concentrated sugar solutions. (c) Internal defects caused by bruising, watercore, and internal browning resulted in a decrease in T2 in all varieties at this low field strength. It remains to be seen whether T1 also shows useful correlations with soluble solids and internal defects. Cho et al. [15] reported that the amplitude of the sugar proton peak following water suppression with a T1 null sequence correlated well with the sugar content in apples (as well as banana and muskmelon). This exploited the observation that the T1 of exchangeable water protons was at least twice as long as that of the non-exchanging sugar protons, so a simple inversion recovery sequence with a delay time chosen to null the water signal, leaves a substantial positive sugar proton signal.

Apple Bruises A combination of spin-echo and gradient reversed echo imaging in a high field magnet at 1.4 T, showed that the redistribution of intra- and extra-cellular water arising from cell and membrane damage in the bruised region actually results in a decrease in the mean transverse relaxation rate, T2 , as measured by the CPMG sequence [16]. This established that the enhanced image intensity observed for bruised tissue with high-field MRI could not be a result of intrinsic T2 changes. Instead, the enhanced brightness results from a reduction in the signal attenuation arising from water diffusing through internal magnetic field

Part III

Table 1: Comparison of theoretical and experimental pi and ai in Equation (1) for apple parenchyma tissue

1724 Part III

Food Science

Part III

Table 2: Summary of references to NMR studies of quality factors in the major types of fruit and vegetables Heat and chill injury Product Fruit Apple Avocado Banana Kiwifruit Mandarin Mango Melon Nectarine Orange Peach Pear Pineapple Tangerine Berries Blueberry Grape Strawberry Small, stone fruit (drupes) Cherry Olive Plum/prune Vegetables Potato Squash Tomato

Maturity

Bruises/voids

Tissue breakdown

Heat

Chill

Infections

12,15,31 40–42 43 44 47 49–52 55,56 41 41 – 49 56 63

12,16,18,31,32 – – – – – 55–57 – – – 18 – 63

10,12,17,19–21,31,33–39 – – – – – 58 59 18 59 61,62 18 63

– – – – – 53,54 – – – – – – –

– – – 45,46 – – – – – – – – –

– – – – 48 – – 60 – – – – –

64 65–67 68,69

– – 68,69

– – –

– – –

64 – –

70 – 73

– – –

– – –

– – –

– – –

– – 68,69 Pit detection 71 72 –

74

75

75 79–81

76–78 56

18,51,82–84

gradients created by the magnetic susceptibility difference between air in extra-cellular spaces and cellular fluid. This is an important observation because the effect of internal gradients on the effective relaxation rate is expected to vary as the square of the applied field as well as on the pulse spacing. Reducing the field strength to say 0.14 T (a proton frequency of ca. 6 MHz) would be expected to reduce the effect of the internal gradients on the relaxation rate by at least two orders of magnitude so that bruises in a low-field imaging system could actually appear with lower rather than higher intensity. This prediction remains to be investigated but highlights the importance of repeating the NMR-internal quality correlations over a range of frequencies, especially the low-field range.

Internal Browning Disorders in Apples Internal browning disorder manifests itself as light and dark brown patches throughout the cortex and core. The

disorder is induced by high CO2 concentrations, especially during storage in modified atmospheres, but can also appear in unpicked fruit on the tree. As its name implies, the brown patches are not visible from the outside but give contrast in MRI. The light brown regions have a lower signal intensity than normal tissue because of a reduced proton density and a shorter T2 . Dark brown regions have higher intensity than normal tissue because of a longer T2 [17].

Watercore in Apples In normal apple tissue, the inter-cellular spaces are ordinarily filled with air. However, in watercore-affected tissue, the air is replaced by a sorbitol solution giving the cut tissue a translucent watery appearance. Watercore is also associated with elevated water content and decreased reducing sugars. The disorder mostly occurs late in the growing season in over-mature unpicked fruit and cannot

NMR in Horticulture

Mealiness in apples Mealiness in apples results from a breakdown in adhesion between cells so that chewing the apple tissue results in cell separation rather than cell rupture. Eating a mealy apple is therefore associated with the unpleasant sensory perception of a lack of juiciness, loss of crispness, and hardness. The first significant correlations between NMR measurements and mealiness were reported by Barreiro et al. [19,20], who found that the histogram of T2 values (i.e. the number of pixels having a particular T2 ) for mealy apples was skewed to shorter relaxation times with a significant tail located in the maximum T2 range. In contrast, non-mealy apples had normal, non-skewed T2 histograms. More recently, increasing mealiness has been shown to cause a systematic reduction in the non-spatially resolved CPMG-T2 values undertaken at the slightly lower field of 2.32 T [21].

Potato Most NMR studies of potato have focused on understanding the effects of processes such as boiling, frying, freezing, and drying [9,22,23]. Exceptions are the NMR microimaging of healthy and diseased potato tubers reported by Goodman and co-workers [24–26] and relaxation changes in healthy tubers following induced wounding. Most recently, a number of papers have appeared where NMR relaxometry on raw potato has been used to predict the sensory texture after boiling [27–29]. It is therefore surprising that, despite the importance of potato, there are so few published studies where NMR/MRI has been used to detect internal quality defects in raw tubers. Such internal defects include brown core and hollow heart. Using multi-slice spin-echo imaging, Wang et al. [30] have demonstrated that both brown core and hollow heart can be detected by MRI. Brown core, a physiological disorder of potato characterized by medulla tissues in the center of the tuber turning from light- to dark-brown in color, is associated with loss of membrane structures, cytoplasm, and organelles and a thickening of cell walls that results in the affected tissue having a longer T2 relaxation time. It is therefore recognizable by high signal intensity in scans with long echo times. The “hollow heart” condition results from cells in the pith area of the tuber splitting apart to form star- or lens-shaped air-filled cavities near the center of the tuber. The hollow heart condition is therefore distinguished by the absence of signal in images or one-dimensional profiles.

Another quality defect which is adversely affecting the marketing of potato products has been termed the hardening syndrome. Here, internal lumps form inside the potato during cooking. While a number of biochemical and physiological investigations are underway to try to understand the origins of this syndrome it remains to be seen whether NMR studies on the raw potato can identify regions likely to form internal lumps.

Conclusions Although Table 2 shows that many diverse quality factors in fruit and vegetables can be detected with NMR it is also clear that a great deal of fundamental NMR remains to be done. Apple is one of the most intensively studied fruit, yet we find no reports on field cycling relaxometry, no attempts to measure cross-relaxation rates or T1ρ , and no systematic investigation of water diffusivity in relation to physiological defects such as mealiness. Little work has been done to correlate PGSE measurements of the restriction of water diffusivity to quality factors. The gaps in Table 2 also emphasize the extent of the more applied work that remains to be done before NMR sensors can be applied to all these diverse types of fruit and vegetable. Very few of the references in Table 2 attempt any quantitative modeling of their NMR data in terms of cell microstructure or composition. Such models would be extremely useful in choosing the optimum acquisition pulse sequences and for rationalizing differences between sample varieties and the effects of harvesting times and storage conditions. The numerical cell model is a first step in this direction but more realistic cell morphologies could be tackled with finite element and Monte Carlo numerical methods. As mentioned in the introduction, many of the reports cited in this review have been based on measurements made with superconducting magnets at relatively high spectrometer fields. Unless there is a major breakthrough such as high temperature superconductivity, these high field magnets are unlikely to be used in NMR sensors for sorting and grading. There is therefore a need to repeat many of the measurements at much lower fields (1m), the contribution of field inhomogeneity and other mechanisms will be significant and measurements from FID are no longer adequate. In such cases, T2 has to be measured using CPMG method [41], in which the short T2 component will be “filtered-off” by choosing an appropriate spin-echo delay tSE (Figure 1), leaving only the long T2 components. Examples of this situation include water in starch systems and amorphous polysaccharides in the water-rich environments, which will be dealt with in later sections. We shall deal with the short T2 scale first.

1 T2



= obs

1 T2



+ D

1 T2



+ IH

1 T2

(10) others

In solids, the 1 H NMR line shape is dominated by the dipole–dipole interactions (HD ) between the neighboring protons. In the absence of molecular motions, the FID

Magnetic Relaxation in Starch Systems

Proton Spin–Spin Relaxation and Second Moment of Solid Starch Polysaccharides 1749

By using the Jeener-Broekaert sequence [57] (90◦x -t1 45◦y -t2 -45◦y -t3 -acq, Figure 1d), some Zeeman order can be transferred into the dipolar order, allowing the measurement of spin–lattice relaxation times along local dipolar fields (T1D ). For a homogenous solid biopolymer, when t1 ≈ T2 , and t2 ≥ t3 , a single Jeener echo is expected, whose shape and amplitude depends on the time derivative of the FID [57]. However, if an inhomogeneous distributions of second rank spin interactions are present, three echoes can be observed [58] respectively at t3 = t1 and t3 = t2 ± t1 and, for long t1 , superposition of echoes are present together with the changes in the shapes of the echoes as a function of t2 variations [59]. In addition, one can use the Goldman-Shen sequence [56] (90◦x -τ 1 -90◦−x -τ 2 -90◦x -acq, Figure1C) to prepare the spin system in a non-equilibrium state with distinct domains at different spin temperatures. The recovery to a quasi equilibrium state can then be monitored as domains achieve a common spin temperature through spin diffusion. When τ1 ≥ 5 T2short , where T2short is the T2 of the fast decay component, the macroscopic magnetisation of the rigid domain is selectively filtered-off. The residual magnetisation of the slow decay component, characterized by a T2long , is then realigned along the magnetic field and the transfer of the magnetisation between domains (or the changes of spin temperature) via spin diffusion can be monitored by varying τ 2 . If spin-diffusion or spin translation occurs between some distinct phases of a heterogeneous sample, the return and growth of the fast component of the FID will be observed when τ 2 ≥ T2long . In such cases, the size and dimension of domains can be estimated [60–63] from the diffusion coefficient, D, and relaxation times. This will be described in later sections. On the other hand, however, in the event of a broad distribution of correlation times or fast spin translation between spins intermingled at short distances or crosscorrelation effects of a segment moving anisotropically, only one spin population is expected corresponding to a single homogenous phase. The recovery process of the T2short component will be rapid (τ2 < T2long ). In such cases, the domains are absent.

Domains and Spin-Diffusion in the Solid Starch Polysaccharides Assuming the spin diffusion as a random walk process, the mean diffusion distance d over the interval T1 is [60–63] d =



2DT1

(11)

where D is the diffusion coefficient. For biopolymers in the solid state with a Gaussian line shape, the diffusion

Part III

of such a system decays rapidly; the corresponding proton NMR spectrum appeared as a broad featureless peak. Whilst many solid polymers featured a Gaussian decay for their proton FIDs, the FID of solid starch as in the cases of most solid sugars can be characterized neither by an exponential function nor by a Gaussian function. Therefore, it is often not practical to analyse the “rigid” starch by the spin–spin relaxation time T2 . Similar to the other carbohydrate systems such as glassy maltose [42], plant cell walls [43–45] and model systems [45–49], however, the proton FIDs can be approximated adequately by a Gaussian decay modified by a sinc function [43–50], from which the second moment of the NMR spectra, M2r , can be evaluated quantitatively. The rigid lattice second moment of a biopolymer can be calculated with lattice summation described by van der Vleck equation [51] if reasonable structure information is available. Proton positions in X-ray structures are often not accurate and the neutron structures give much more accurate results. The true second moment, M2 , is independent of molecular motions. However, the measured second moment also known as residual second moment, M2r , can be used to investigate the motions. In general, when the correlation time, τ c , is greater √ compared to the inverse of linewidth, i.e. τ c (γ M2r )−1 , the hydrogen nuclei (protons) are said to undergo slow motions and give rise to √ so-called “rigid lattice” NMR spectra. When τ c ≈ (γ M2r )−1 , the fluctuation of the dipolar interactions becomes fast enough to average HD to some extent √ and the M2r is expected to start decreasing. If τc (γ M2r )−1 , the hydrogen nuclei are said to undergo fast motions. The motion modulated reduction of the apparent proton second moment is observed by an amount M2 = M2 − M2r . The FIDs of starch consist of a fast decay and a slow decay parts. After removal of water (and OHs) by exhaustive D2 O exchange, the FID remains as two parts though the slow decay part is often reduced [11] in intensity due to the removal of the contributions of H2 O. The interpretation for such composite line shapes is often simplified as two different structure domains, “rigid” and “mobile” ones. Whilst such interpretation is true in many cases, it can sometimes suffer from oversimplification [52]. In fact, the multiple-decays in T2 process can also arise in other situations, for example, when there are broad distributions of correlation times [53,54], cross-correlation effects in the anisotropic motion of molecular segments [55], and cross-relaxation between two spin species intermingled at short distances from each other [56]. It is therefore always desirable to ascertain the nature of the multiple-component decays experimentally. This can be achieved by inspecting the shapes of the dipolar echoes and monitoring the behaviour of the magnetisation transfers via the Goldman-Shen experiment [56] (Figure 1c).

1750 Part III

Food Science

Part III

coefficient is given [64] approximately by D=

r  60π T2

reaches a minima or R1 reaches a peak where

2

where r 2 is the mean-squared inter-nuclear spacing and T2 is the measured FID decay constant. Since the local dipolar field is given by

H HL  = √ 12

R1 Max = 1.425C/ω0

(12)

C in the Equation (14) and (16) is the relaxation constant and is related to the spatial arrangement of protons in the sample, thus, a useful structural parameter. For a powder, C is given by Equation (17)

(13)

where H is the line width (in Gauss) at the half height [65]. By measuring H and working out HL , an average r  for the proton separation can be obtained from Equation (2). This together with the measured T1 and T2 values allows estimation of the domain dimensions according to the Equation (11–12). However, one has to be aware that in the dipolar coupled systems, spatial inhomogeneity measured by 1 H NMR can be achieved only for a certain range of length scales. Spatial fluctuations within the correlation length of the local dipolar fields cannot be resolved. This sets a lower bound resolution to a few near neighbour distances, ˚ i.e. about 5–10 A.

Proton Spin–Lattice Relaxation of Solid Starch Polysaccharides in the Laboratory Frame 1

H relaxation times are most widely employed for studying the dynamics due to high sensitivity and wellestablished theories. For biopolymers, the proton T1 can be conveniently measured using inversion-recovery or saturation-recovery sequences either by directly detecting the 1 H NMR signals or indirectly via 13 C (e.g. CPMAS). In the dipolar interaction dominated system, the proton relaxation time, T1 , obeys the relaxation theory described by Bloembergen, Purcell and Pound [66]. Assuming exponential correlation function and single correlation time, the T1 can be evaluated quantitatively with the well-known Kubo-Tomita expression (Equation 14) [67–70],    1 4τci τci = Ci + (14) R1 = T1 1 + ω02 τci2 1 + 4ω02 τci2 i≥1 where ω0 is the proton Larmor frequency and τ c the rotational correlation time of the motions responsible for the spin–lattice relaxation, which normally follows the Arrhenius activation law

Ea (15) τc = τ0 exp RT where τ o is the pre-exponential factor corresponding to the rotational correlation time at infinite temperature, E a the activation energy for the motion and R the gas constant. For a motional process, when ω0 τc = 0.6154, T1

(16)

C=

    9 −6 · γ 4 -h2 r −6 + r , jk jl 20N j k l= j,k

(17)

where N is the number of protons to be relaxed and, r jk is the inter-proton distances between a pair of protons j and k, γ is the proton magnetogyric ratio. The first term in parenthesis describes the intra-pair spin properties whereas the second term describes the inter-pair ones. It is thus apparent that, by measuring T1 as a function of temperature or the Larmor frequency, one can access parameters describing the molecular motions (E a , τ o , C) and the geometry of the groups (r ) involved in the motions if no phase transition or chemical decompositions were intermingled into the motional processes. However, even in most of single-phase homopolymer systems, the assumption of a single τ c is not always valid [71–76]. A distribution of τ c is more realistic [71–76] although the reasons for such τ c distribution are not well studied. The distribution of activation energy due to the distribution of molecular conformers is one possibility (P.S. Belton, personal communication). In biopolymer systems such as starch, such distribution is often complex. The dynamics of starch polysaccharides in much slow frequency region (several kHz) is also extremely important especially for the mechanical properties. Such dynamics cannot be easily probed with T1 but can be investigated by measuring the proton T2 or M2r , T1D and T1ρ relaxation times.

Proton Spin–Lattice Relaxation of Solid Starch Polysaccharides in the Rotating Frame To study slow motions in several kHz regions, 1 H T1ρ can be conveniently measured by a single pulse followed with a variable spin-lock scheme. T1ρ can also be used to quantitatively analyse the molecular motions by fitting the experimental R1ρ data to an expression similar to that used in R1 analysis [46,47,49,67–69]:  1 3 5 τci = Ci + R1ρ = T1ρ 2 i≥1 1 + 4ωei 2 τci 2 3  τci 2 τci × + (18) × 1 + ωo 2 τci 2 3 1 + 4ωo 2 τci 2

Nuclear Magnetic Relaxation in Starch Systems

ωe =

 ωSP 2 + ωL 2

(19)

In practice, ωL is temperature dependent and has to be modeled by an empirical relationship, such as the Sigmoid function [43–49,67–69], to give the closest values to that of the experimental ones. For a motional process, when ωe τc = 0.6101, T1ρ reaches a minimum. In most of starch systems, multiple components are frequently detected [5,38] for 1 H T1ρ of starch systems due to the heterogeneity of such systems. The magnetization transfers between domains via spin diffusion over the whole sample can be used to obtain information on the dimension of the domains. In the meantime, the multiple components of relaxation times provide a great opportunity to access information on the structure and interactions of the different domains via relaxation-edited spectroscopy. In theory (see Equation 18), T1ρ measurements as a function of temperature or relaxation frequency can also be used to characterize the molecular dynamics quantitatively (E a , τ o , C, r ) in the similar way to that for T1 . However, in real starch systems, measuring the relaxation time as a function of sample temperature is not always possible without causing samples to undergo physical changes, such as phase transition, and chemical decomposition. The changes often occur “inconveniently” in the temperature range one want to investigate. Nevertheless, such problems can be avoided by measuring T1 as a function of the frequency [79], ωo , and analyse the motional processes accordingly. The commercial fastfield cycling spectrometers with field switching speed of 0.1 ms/MHz are now available, covering the frequency range of 10 kHz to 80 MHz. This makes it possible to study molecular motions in the range of several tens of kilohertz to hundred of megahertz at one given temperature [31,79,80]. 13

C Spin–Lattice Relaxation for Starch Polysaccharides The T1 and T1ρ of 13 C for starch polysaccharides can be measured directly, providing information on the

spectroscopically resolved sites. Using appropriate pulse sequences [81–84], the chemical shift anisotropy (CSA) can also be measured to gain dynamics information. The advantage of using 13 C method is that the NMR signals contain only these of polysaccharides and there will be no water signals directly intermingled. So far, there is still a lack of systematic 13 C relaxation study of starch though some data have been reported for maltose [85] and starch gels [33,86]. Some early efforts were concentrated on the measurements of 13 C T1 for the prepared crystalline starches [87], which was only possible after the invention of CPMAS experiment. For both A- and B-type dry crystalline amylose [87], 13 C T1 values were about 20 s and 5 s for the C-1 and C-6 respectively. The C-6 has a much shorter 13 C T1 presumably due to the trans-gauche conformational motions of CH2 OH group [43–46,88–91]. However, 13 C T1 values are much smaller than these for crystalline cellulose I of cotton [87] (140–210 s), indicating more motional mobility for starch polysaccharides than cellulose. Full hydration of starch led to only a small decrease [87] of T1 value for C-1 (13–16 s) and C-6 (3–4 s). Similar observations were made [87] for C-4 (dry∼14, hydrated∼11–13 s). In contrast, glycogen, a similar polysaccharides, showed dramatic decrease of 13 C T1 values upon hydration [92]. It is suggested that the hydration-induced enhancement for molecular dynamics is only limited for the ordered starch polysaccharides. For the V-type amylose as a single helical amylose-lipids complex, T1 of C-1 showed significant increase upon hydration [87], from 7 s (anhydrous) to 11 s (fully hydrated). Similar observations were made for other pyran ring carbons, indicating that hydration promoted the formation of the ordered structure in this case. The 13 C T1 of amorphous starch are considerably shorter than the crystalline ones. For example, the amorphous wheat starch containing 33% water showed 13 C T1 of 300ms for CH2 OH and 600 ms for the ring carbons [38]. The difference of 13 C T1 has been employed in some spectral editing of 13 C NMR spectra, termed as mobility-resolved spectroscopy [93]. It is conceivable that a systematic 13 C relaxation studies ought to provide some significant insights into the dynamics of the starch polysaccharides. Nevertheless, for systems as complex as starch polysaccharides, some background knowledge is always helpful on the single-phase starch models such as the crystalline sperulites and short chain amorphous polysaccharides.

Water in Starch Systems In most starch systems, water is an essential constituent thus plays important roles in starch functionality. NMR techniques have already been employed in studying water molecules in starch systems for decades. The first such

Part III

where ω0 , τc and C have the same meaning as in the case of T1 . ωe is the effective field for relaxation in frequency units, which is often dependent on both the spin-locking frequency (ωSP ) and the local dipolar field (in frequency unit), ωL . When τc < T2 and ωSP ωD , T1ρ can be interpreted through the theory of Bloembergen, Purcell and Pound [66], therefore, ωe can be replaced by ωSP . When τc

T2 , the so-called Slichter-Ailion conditions apply [77]. However, if ωSP and ωD are in the same order of magnitude and τc ≤ T2 stands, the effects of the local dipolar field has to be taken into consideration as follows [78]:

Water in Starch Systems 1751

1752 Part III

Food Science

Part III 0.0001

0.001

0.01

0.1

1

10

T2 (s)

Fig. 4. A TRESY spectrum of water in wet potato starch granules showing four water populations detectable in the system as water in the crystalline region (0.001 s), water in the amorphous region (0.01 s), water trapped between granules (0.1 s) and bulk water (10 s).

study was published in Nature [94] in 1960 on the starch polysaccharide–water interactions, showing the broadening water proton lines in high resolution NMR spectra of starch gels. In 1970s, some low-resolution 1 H NMR studies of the “bound water” of corn starch [95] were also published and such studies were continuously reported from 1980s to 2000s [10–15,32,96–98] with the gradual improvement of NMR hardware. Some progress has also been made in understanding of the water dynamics in acid hydrolysis and gelatinisation. This area of research will remain active for sometime. Since most of water in starch systems is in the long T2 regime, the CPMG spin echo experiment [41] will be most suitable. The CPMG echo trains are better analysed as a continuous distribution of exponentials, yielding a spectrum [3,4,16,99–103] of T2 components with distributive T2 values (Figure 4). This method was called transverse relaxation spectroscopy (TRESY) [3] and has found widespread application especially in the porous systems. The use of TRESY methods has led to new insights into the state of water in starch systems. For example, TRESY methods enabled most populations [3,11,23,104] of water in the native starch granules to be identified in all three types of starch granules; they are bulk water, water trapped between granules, water in the amorphous growth rings, water in the semi-crystalline regions and structural water in the hexagon channels formed by the double helices. Amongst them, the “channel water” in the granules and hydrolysed starch spherulites, which remained unfrozen until temperature reached as low as −50◦ C [3,4,11], was also observable from the D2 O saturated B-type [3,4,11,14,15] and C-type [3,4,11] starches using 2 H NMR as an 1kHz splitting. 1 H TRESY methods have also been employed to investigate the effects of acid hydrolysis (Linterization),

thermal heating and gelatinisation on the dynamics of starch and microscopic distribution of water within the starch systems [3,11]. By monitoring the changes of proton T2 spectra of a starch/water system, it has been possible to monitoring the changes of the dynamic behaviour of water such as swelling induced water absorption [4,11]. By replacing H2 O with D2 O, the proton TRESY method were employed to monitor the molecular dynamic changes of starch polysaccharides [3,11], especially near the gelatinisation temperature. Below the gelatinisation temperature, the proton NMR signals are mostly from polysaccharides and 1 H NMR signals from FID offers a direct observation of the polysaccharides [3,11]. On the other hand, 1 H CPMG signals near and above gelatinisation temperature provide an unambiguous monitoring handle to the mobilized starch [3,11]. To the similar systems, 2 H and 17 O NMR is useful to monitor the changes of water dynamics in response to the processing such as gelatinisation.

Future Perspective Although studies of starch using NMR relaxation have extended to various extent to both starch polysaccharides and water, more systematic studies are still in need to understand the systems more comprehensively. Whilst most efforts were devoted to the crystalline starch, limited attentions were paid to the amorphous starches even though most of the processed starch is in the amorphous form and hence is important for the rubbery properties. Furthermore, only few studies have been directed towards the interactions between components in the complex. NMR diffusion is not extensively employed even though it is expected to provide essential information on dynamics and interactions of starch systems. It is thus conceivable that the future studies will be focused more on the dynamics and interactions of polysaccharides. The other important questions, which remain unanswered, include: what are the relationships between the physical properties and the domain distribution and dynamic state of the starch polysaccharides (if any); how are these properties relevant in establishing and determining the response of starch granules to multi-parametric processing and ultimately functionality.

References 1. Gallant DJ, Bouchet B, Baldwin PM, Carbohydr. Polym. 1997;32:177. 2. Buleon A, Colonna P, Planchot V, Ball S. Int. J. Biol. Macromol. 1998;23:85. 3. Tang HR, Godward J, Hills B. Carbohydr. Polym. 2000;43:375.

Nuclear Magnetic Relaxation in Starch Systems

37. Kojima T, Sekine M, Sugiyama J, Ishida N, Nagata T. J. Jap. Soc. Food Sci. Technol.– Nippon Shokuhin Kagaku Kogaku Kaishi. 1996;43:1098. 38. Morgan KR, Furneaux RH, Larsen NG. Carbohydr. Res. 1995;276:387. 39. Powles JG, Mansfield P. Phys. Lett. 1962;2:58. 40. Mansfield P. Phys. Revi. 1965;137:a961–a974. 41. Meiboom S, Gill D. Rev. Sci. Instrum. 1958;29:688. 42. Hills BP, Pardoe K. J. Mol. Liq. 1995;63:229. 43. Tang HR, Belton PS, Ng A, Waldron KW, Ryden P. Spectrochim. Acta. Pt. A. 1999;55:883. 44. Tang HR, Zhao BL, Belton PS, Sutcliffe LH, Ng A. Magn. Reson. Chem. 2000;38:765. 45. Tang HR, Belton PS. In: PS Belton, BP Hills, GA Webb (Eds). Advances of Magnetic Resonance in Food Science. Royal Society of Chemistry: Cambridge, 1999, p 166. 46. Tang HR, Belton PS. Solid State NMR. 2002;21:117. 47. Wang YL, Tang HR, Belton PS. J. Phys. Chem. B.0 2002;106:12834. 48. Tang HR, Wang YL, Belton PS. Phys. Chem. Chem. Phys. 2004;6:3694. 49. Tang HR, Belton PS. Solid State NMR. 1998;12:21. 50. Abragam A. The Principles of Nuclear Magnetism. Oxford University Press: Oxford, 1961. 51. van Vleck JH. Phys. Rev. 1948;74:1168. 52. Lacelle S, Gerstein BC. Biopolymers. 1987;26:849. 53. Resing HA. Adv. Mol. Relaxation Processes. 1972;3:199. 54. Resing HA. J. Chem. Phys. 1967;43:669. 55. Werbelow LG, Marshall AG. J. Magn. Reson. 1973;11:299. 56. Goldman M, Shen L. Phys. Rev. 1966;144:321. 57. Jeener J, Broekaert P. Phys. Rev. 1967;157:232. 58. Bloom M, Burnell EE, Roeder SBW, Valic MI. J. Chem. Phys. 1977;66:3012. 59. Cheung TTP. Phys. Rev. B. 1981;23:1404. 60. Cheung TTP. J. Chem. Phys. 1982;76:1248. 61. Cheung TTP. Appl. Spectr. 1997;51:1703. 62. Cheung TTP. Gerstein BC. J. Appl. Phys. 1981;52:5517. 63. Goldman M. Spin Temperature and Nuclear Magnetic Resonance in Solids. Clarendon Press: Oxford, 1970. 64. Cheung TTP, Yaris R. Carbohydr. Polym. 1980;72:3604–16. 65. Bloembergen N, Purcell EM, Pound RV. Phys. Rev. 1948;73:679. 66. Belton P, Wang YL. Mol. Phys. 1997;90:119. 67. Wang YL, Belton PS, Tang HR. Chem. Phys. Letts. 1997;268:387. 68. Wang YL, Belton PS, Tang HR. Solid State NMR. 1999;14:19. 69. Kubo R, Tomita K. J. Phys. Soc. Jpn. 1954;9:888. 70. Andrew ER, Gaspar R, Vennart W. Biopolymers. 1978;17:1913. 71. Andrew ER, Bryant DJ, Cashell EM, Gaspar R, Meng QA. Polymer. 1981;22:715. 72. Andrew ER, Bryant DJ, Cashell EM, Meng QA. Febs Letters. 1981;126:208. 73. Andrew ER, Bryant DJ, Cashell EM, Meng QA, Phys. Lett. A. 1982;88:487. 74. Andrew ER, Bone DN, Bryant DJ, Cashell EM, Gaspar R, Meng QA. Pure. Appl. Chem. 1982;54:585. 75. Andrew ER. Polymer. 1985;26:190. 76. Slichter CP, Ailion D. Phys. Rev. 1964;135:A1099– A1100.

Part III

4. Tang HR, Hills BP. In: GA Webb, PS Belton, AM Gil, I Delgadillo, (Eds). Magnetic Resonance in Food Science— A View to the Future. Royal Society of Chemistry: Cambridge, 2001, p 155. 5. Tang HR, Hills BP. Biomacromol. 2003;4:1269. 6. Shogren RL, Lawton JW, Doane WM, Tiefenbacher KF. Polymer. 1998;39:6649. 7. Simonsen L, Hovgaard L, Mortensen PB, Brondsted H. Euro. J. Pharm. Sci. 1995;3:329. 8. Wu HCH, Sarko A. Carbohydr. Res. 1978;61:7. 9. Wu HCH, Sarko A. Carbohydr. Res. 1978;61:27. 10. Baianu IC. Cereal Foods World. 1986;31:584. 11. Tang HR, Brun A, Hills B. Carbohydr. Polym. 2001;46:7. 12. Tanner SF, Hills BP, Parker R. J. Chem. Soc. Farad. Trans. 1991;87:2613. 13. Kulik AS, Decosta JR, Haverkamp J. J. Agri. Food Chem. 1994;42:2803. 14. Yakubu PI, Ozu EM, Baianu IC, Orr PH. J. Agri. Food Chem. 1993;41:162. 15. Yakubu PI, Baianu IC, Orr PH. J. Food Sci. 1990;55: 458. 16. Chatakanonda P, Chinachoti P, Sriroth K, Piyachomkwan K, Chotineeranat S, Tang HR, Hills B. Carbohydr. Polym. 2003;53:233. 17. Chen PL, Long Z, Ruan R, Labuza TP. Food Sci. Techn. 1997;30:178. 18. Ruan RR, Zou C, Wadhawan C, Martinez B, Chen PL, Addis P. J. Food Proc. Preserv. 1997;21:91. 19. Ruan RR, Long ZZ, Song AJ, Chen PL. Food Sci. Techn. 1998;31:516. 20. Kimshin MS, Mari F, Rao PA, Stengle TR, Chinachoti P. J. Agri. Food Chem. 1991;39:1915. 21. Tang HR, Wang YL, Belton PS. Solid State NMR. 2000;15:239. 22. Li S, Dickinson LC, Chinachoti P. Cereal Chem. 1996;73:736. 23. Kou Y, Dickinson LC, Chinachoti P. J. Agri. Food Chem. 2000;48:5489. 24. Colquhoun IJ, Parker R, Ring SG, Sun L, Tang HR, Carbohydr. Polym. 1995;27:255. 25. Dybowski C, Lichter RL. NMR Spectros. Tech. Dekker: New York, 1987. 26. Mehring M. Principles of High Resolution NMR in Solids. Springer-Verlag: Berlin, 1983. 27. Fyfe CA. Solid Sate NMR for Chemists. C.F.C. ress: Guelph, Ontario, Canada, 1983 28. Spiess HW. Chem. Rev. 1991;91:1321. 29. Tang HR, Belton PS, Davies SC, Hughes DL. Carbohydr. Res. 2001;330:391. 30. Schmidtrohr K, Spiess HW. Multidimensional solid state NMR and Polymers. Academic Press: London, 1994. 31. Hills BP, Wang YL, Tang HR. Mol. Phys. 2001;99:1679. 32. Baianu IC, Yakubu PH, Ozu E. Abstr. Papers Am. Chem. Soc. 215, 239-OLY (1998). 33. Chinachoti P, White VA, Lo L, Stengle TR. Cereal Chem. 1991;68:238. 34. Chinachoti P, Kimshin MS, Mari F, Lo L. Cereal Chem. 1991;68:245. 35. Lim H, Setser CS, Paukstelis JV, Sobczynska D. Cereal Chem. 1992;69:382. 36. Cheetham NH, Tao LP. Carbohydr. Polym. 1998;35:279.

References 1753

1754 Part III

Food Science

Part III

77. Mccall DW, Douglass DC. Appl. Phys. Lett. 1965;7:12. 78. Belton PS, Wang Y. Magnetic Resonance in Food Science—A View to the Future. 2001;262:145. 79. Wang YL, Belton PS. Chem. Phys. Letts. 2000;325: 33. 80. Gan ZH. J. Am. Chem. Soc. 1992;114:8307. 81. Gan ZH, Ernst RR. J. Magn. Reson. A. 1996;123:140. 82. Gan ZH, Grant DM, Ernst RR. Chem. Phys. Letts. 1996;254:349. 83. Hu JZ, Wang W, Liu F, Solum MS, Alderman DW, Pugmire RJ, Grant DM. J. Magn. Reson. A. 1995;113:210. 84. Tromp RH, vanDusschoten D, Parker R, Ring SG. Phys. Chem. Chem. Phys. 1999;1:1927. 85. Vodovotz Y, Chinachoti P. In: Belton PS, Hills BP, Webb GA (eds). Advances of Magnetic Resonances in Food Science. Royal Society of Chemistry: Cambridge, 1999, p. 185. 86. Horii F, Yamamoto H, Hirai A, Kitamaru R. Carbohydr. Res. 1987;160:29. 87. Reynhardt EC. Mol. Phys. 1990;69:1083. 88. Latanowicz L, Reynhardt EC. Ber. Bunsen Gesell. Phys. Chem. Int. J. Phys. Chem. 1994;98:818. 89. Latanowicz L, Reynhardt EC, Utrecht R, Medycki W. Ber. Bunsen-Gesell. Phys. Chem. Chem. Phys. 1995;99:152.

90. Reynhardt EC, Latanowicz L. Chem. Phys. Letts. 1996;251:235. 91. Jackson CL, Bryant RG, Biochem. 1989;28:5024. 92. Foster TJ, Ablett S, Mccann MC, Gidley MJ. Biopolymers. 1996;39:51. 93. Collison R, McDonald MP. Nature. 1960;186:548. 94. Mousseri J, Steinberg JMP, Nelson AI, Wei LS. J. Food Sci. 1974;39:114. 95. Kuhn K, Kruschak W, Lechert H. Starch-Starke. 1987;39:1. 96. Rugraff YL, Desbois P, LeBotlan DJ. Carbohydr. Res. 1996;295:185. 97. Chinachoti P, Stengle TR. J. Food Sci. 1990;55:1732. 98. Hills BP, Lefloch G. Food Chem. 1994;51:331. 99. Hills BP, Wright KM, Snaar JM. Magn. Reson. Imag. 1996;14:715. 100. Hills BP, Wright KM, Snaar JM. Magn. Reson. Imag. 1996;14:305. 101. Hills BP, Remigereau B. Int. J. Food Sci. Technol. 1997;32:51. 102. Hills BP, Godward J, Manning CE, Biechlin JL, Wright KM. Magn. Reson. Imag. 1998;16:557. 103. Li S, Dickinson LC, Chinachoti P. J. Agri. Food Chem. 1998;46:62.

Part III

High Resolution Solid State Methods

1757

Maria Antonietta Brescia and Antonio Sacco Dipartimento di Chimica, Universit`a di Bari, via Orabona 4, 70126 Bari, Italy

Introduction The study of the properties of wheat flours is necessary to understand the mechanical and rheological properties of the resulting dough and therefore of the finished baked products. In the specific case of bread, recent works have concerned low resolution NMR for studying water and lipid mobility [1–3]. The mobility of water in dough is an important parameter being related to the quality of gluten network. To gain more insight in the structural characteristics of dough and of its principal components, gluten and starch, solid-state Magic Angle Spinning (MAS) NMR was applied. The interactions of each dough component with water and effects of heating during bread making were also studied. MAS methods improve resolution by averaging dipolar interactions, along with chemical shift anisotropy and magnetic susceptibility. This permits to detect easily and quantify signals due to various components of dough and flours. In this chapter applications of 13 C and 1 H MAS NMR spectroscopy to flours and dough and to their major components, starch and proteins, will be reviewed and the potential of these techniques in exploring the correlation between molecular level information and macroscopic properties of dough and flours will be examined.

13

C Cross Polarization MAS NMR of Flours

The first 13 C cross polarization (CP-MAS) NMR study of flours was performed by Baianu and F¨orster [4]. They found that the broad features of spectra of wheat grains were similar to those of the corresponding flours. The subtle differences were supposed to be due to starch damage and mechanical breaking of protein chains. The major signals present in the CP-MAS 13 C NMR spectrum of flour are due to sugar carbons in the starch components of the flours, resonating at 96–102 ppm (C1 ), 70–73 ppm (C2 , C3, and C5 ), 77–83 ppm (C4 ), and 59–62 ppm (C6 ). The tail of the 62 ppm starch resonance obscures signals due to Cα’s of the proteins, while signals due to protein side-chain aliphatic carbon resonate at 20–35 ppm, and do not overlap with other signals. The relative intensities of the protein and starch resonances allow calculating the protein content of durum and tender wheat flours. Graham A. Webb (ed.), Modern Magnetic Resonance, 1757–1763.
C 2008 Springer.

The resolution of the signals permitted the application of relaxation experiments and to determine the average rotating-frame proton relaxation time T1ρ(H) of each carbon [5]. From these measurements it emerged that all the starch protons have a similar T1ρ(H) while for proteins T1ρ(H) is 30% lower indicating that starch and proteins are separated into distinct domains. 1

H High Resolution MAS NMR of Flours

Even if 13 C CP-MAS NMR has the advantage of avoiding sample manipulation, the information that can be inferred from these spectra is strongly limited by the low resolution due to the presence of very broad signals generated by high molecular weight (HMW) components. 1 H High Resolution (HR-MAS) NMR technique permitted to overcome these difficulties combining the typical advantages of solid- and liquid-state NMR techniques. Samples for 1 H HR-MAS measurements were prepared in the rotors mixing 33 mg of D2 O with 40 mg of flour to obtain homogeneous samples with similar degrees of swelling. A typical 1 H HR-MAS spectrum of durum wheat flour, obtained using a presaturation sequence for water suppression, is reported in Figure 1. More than 80 peaks can be distinguished resulting mainly from lipids and polysaccharides. Lipids assignment was performed [6] on the basis of correlations observed in TOCSY and 1 H-13 C HMQC spectra (Table 1). A detailed assignment of saccharide moieties was not possible due to overlap among TOCSY correlations. The heights of the resonances showing neither overlapping nor correlation with other signals were evaluated in the 1 H HR-MAS spectra of 12 durum wheat flours coming from southern Italy to be used as variables for Principal Component Analysis. The results permitted to detect differences among the samples mainly due to their geographical origin. This procedure demonstrated its reliability also when it was applied on an enlarged data set formed by 26 samples of two cultivars at 13 different locations in Italy [7] and was able to differentiate samples coming from North, Central, and South Italy according to the geographical origin and cultivar. 1 H HR-MAS was also employed to study the kinetic release of oligosaccharides due to hydrolases in the amylolytic starch degradation of wheat flours [8]. The increasing intensities of β-glucose and anomeric protons of the

Part III

Magic Angle Spinning NMR of Flours and Doughs

1758 Part III

Food Science

Part III 6.0

5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5 ppm

Fig. 1. 1 H HR-MAS spectrum of a durum wheat flour (From Ref. [7]).

reducing ends of β-maltose were determined as function of time for several durum and soft wheat samples and were fit with a Michaelis Menten kinetic equation to obtain kinetic parameters that are associated to the transformation of starch into maltose and maltose into glucose. A relationship between the kinetic degradation of starch and the botanical origin of the flour was suggested. 1

H and 13 C MAS NMR of Doughs

The hydration of flour is an important step in the baking process. Differences in the mechanical properties of dough are related to the interactions of water and flour. Hydrated durum and tender wheat flours were analyzed with CP-MAS and showed a decreased intensity of signals due to protein carbons in comparison with the spectra of the dry flours, being nearly absent in the spectrum of the tender wheat flours [5] (Figure 2). This was explained supposing a reduced effect of water on protein

components in durum wheat flours, because proteins are linked to starch–protein particles and remain solid-like. Hydration of the tender flour proteins generates a liquidlike component that can act like a compatibilizer rendering tender flour more processable than durum wheat flour. Some properties of the flours were investigated, in order to assess its suitability for technological purposes. In particular, the peculiar viscosity and elasticity of hydrated flours are associated to the quantity and quality of gluten. These properties were extensively investigated [9] to understand their origin at a molecular level and the behavior of specific groups or protein moieties on the formation of the hydrated network. These studies were firstly performed on molecules that could be used as simple models [10]. For this reason, the major constituents of gluten, gliadins, and glutenins, were separately analyzed. In particular, it was interesting to gain information about the water-insoluble network obtained by hydration of these proteins to low water content, since this is closer to the protein during dough making.

MAS NMR of Flours and Doughs

δ (ppm)

Moiety

0.83 1.22 1.50 1.98 2.18 2.69 3.26 3.40 3.41 4.64 4.65 5.23 5.23 5.24 5.41 5.41 5.35–5.50

CH3 –(CH2 )n–CH2 – –(CH2 )n–CH2 – –(CH2 )–CH2 –CO–O– –CH2 –CH=CH– –CH2 –CH2 –CO—O– –CH=CH–CH2 –CH=CH– C2 C4 C4 C1 C1 reducing end C1 C1 reducing end –CH=CH– C1 non reducing end C1 non reducing end Non reducing ends

Assignment Lipids Lipids Lipids Lipids Lipids Lipids β-glucose β-glucose α-glucose β-glucose β-maltose α-glucose α-maltose Lipids α-maltose β-maltose carbohydrates

Gliadins The first 13 C CP-MAS NMR studies on gliadins were performed by Baianu [11]. The wheat proteins ω-gliadins, that are associated with dough viscosity, have a relatively simple structure; therefore they were used as a model

to study hydration with 13 C CP-MAS NMR [12,13]. The changes in resolution and intensity of the signals present in the spectrum were analyzed at different hydration percentages and were correlated with the protein conformational changes (Figure 3). An improvement of spectral resolution with hydration reflected a change in the protein conformation. Moreover, the loss of the proline Cδ signal was detected, indicating an increased mobility of this group. These observations were enriched with results obtained with the T2 relaxation time measurements, and were consistent with the formation of mobile protein loops together with residual regions of strong interprotein interactions with hydration. These evidences support the “loop and train” model, common in some polymers, in which “loops” of solvated mobile polymer coexist with “trains” polymer bonded to the surface [9]. The effect of hydration was also studied by 1 H HRMAS NMR on ω-gliadins. Proton T1 measurements were made under MAS conditions for 19% hydrated samples showing that the most mobile protons were leucine methyl groups [12]. Anyway, great care was required to interpret T1 relaxation times, since they were found to be dependent on the sample spinning rates [13]. In the spectra of 48% hydrated ω-gliadins, it was possible to observe resonances of backbone NH groups (8.2 ppm), arising from glutamine and phenylalanine residues, as well as side-chains NH groups (6.9 ppm) and aromatic protons (7.3 ppm), indicating that mobility is not restricted to protein side-chains, but includes also protein backbone regions [12].

HARD FLOUR (12% protein)

SOFT FLOUR (9% protein)

×10

×10

200

100 dc

Fig. 2.

13 C

0

PPM

200

100

0

dc

CP-MAS NMR spectrum of a wheat flours: (left) durum wheat flour, (right) soft wheat flour (From Ref. [5]).

PPM

1759

Part III

Table 1: Proton assignment of durum wheat flour lipid moieties (From ref. [6]).

1 H and 13 C MAS NMR of Doughs

Pro Cβ Gln Cγ Gln Cβ Pro Cγ

Pro Cα Gln Cα Pro Cδ

Gln Cδ CO e

d

c

b

a

200

150

100 δ(13C)/ppm

50

0

Fig. 3. 13 C CP-MAS spectra of ω-gliadins at different hydrations: (a) 0% water, 464 scans; (b) 7% water, 11,720 scans; (c) 19% water, 32,305 scans; (d) 35% water, 5500 scans; (e) 50% water, 17,732 scans (From Ref. [13]).

MAS NMR of Flours and Doughs

1 H and 13 C MAS NMR of Doughs

Moreover, evidences were found about the existence of two main environments for glutamine residues: one situated in protein segments adopting the β-sheet conformation, remaining relatively hindered, and the other associated to a more mobile section, possibly adopting a β-turn configuration.

High Molecular Weight Fractions

Dry and hydrated gluten was directly analyzed by 13 C CP-MAS NMR and Single Pulse Excitation (SPE)-MAS experiment [19,20]. While cross polarization detects the solid-like parts of the samples, because the present signals are due to carbon nuclei strongly coupled to protons via dipolar interactions, only mobile nuclei generate signal in SPE experiment (Figure 4). Therefore these two approaches are different and give complementary information about local chemical environments and mobility. It was shown that resonances of residual starch and protein backbone carbons are enhanced in CP spectra of dry gluten, suggesting that they are located in rigid domains. Lipid carbons and glutamine side-chain carbonyl groups are clearly visible only in SPE spectra, indicating that they are present in a more mobile environment. With hydration, starch and proteins gave signals with a lower signal to noise ratio, due to a higher mobility of these fractions, more pronounced for proteins than for starch. The T1ρ(H)values of the hydrated gluten were similar for all the carbon resonances. This result is indicative of a mixing among the different domains of the samples (protein backbone, side-chains and starch), induced by hydration. 13 C CP-MAS NMR gluten spectra did not differ when various baking quality wheat samples were compared; therefore no correlation was found between protein environment sampled by the technique and the baking quality [19]. 1 H MAS spectra of hydrated gluten contain wellresolved signals due to lipids, while protein aliphatic chains give broad resonances. As in the spectra of hydrated ω-gliadins, resonances of backbone and side-chain NH groups were observed, supporting the loop and train model. The variations of gluten due to heating/cooling processes were checked to understand its behavior, at a molecular level, during the baking process. From the study of the relaxation times it was shown that gluten protein aliphatic peaks have an increase of T2 with heating and a decrease with cooling, which indicates the reversibility of the dynamic changes. The increase of line width at half height of water in gluten samples during heating and cooling was interpreted as reflecting an increasing hindrance of the system.

The gluten HMW fraction has been extensively studied due to its high contribution in determining the viscoelastic properties of gluten. HMW subunits are most important in the structure of the wheat proteins glutenins, since they form the backbone of glutenin being linked together by means of disulphide bonds. 13 C CP-MAS studies performed on the hydrated proteins helped to individuate some relationships between structural properties and the characteristics of the hydrated network [15]. The decreasing of the signal to noise ratio with hydration, due to higher protein mobility, was analyzed in order to find those structural characteristics that were related to higher or lower mobility. In particular, it was demonstrated that the presence of disulphide bonds was responsible of higher mobility, while the irregular sequences present in chain ends cause hindrance of the network due to the formation of interchain bonds. In absence of irregular chain ends long chain peptides are insoluble in water and form a rather mobile network. The influence of the protein primary structure on the type of network formed was also studied on model systems [16]. 13 C and 1 H MAS studies on HMW subunits helped the elucidation of the structure of the protein segments involved in the loops and trains sections. Trains sections contain glutamine residues close to hydrophobic residues forming hydrogen bonds that hold protein segments together, while loops involve hydrophilic glutamine-rich segments that may also contain glycine. The network formed by the hydrated protein could be further hindered by hydrophobic interactions established by proteins terminal domains [17]. 1 H MAS spectroscopy was a suitable method to study the effect of hydration on wheat proteins, since the weakened dipolar interactions that characterize these systems enable the acquisition of well-resolved 1 H NMR spectra. The resolution of protein spectra was significantly improved using HR-MAS spectroscopy [18] that allowed the application of 2D experiments to obtain the assignment of the protein structures. Considerable shifts were observed for some resonances, relative to the chemical shifts of amino acids in solution, indicating that specific interactions occur in the hydrated protein network.

Gluten

Part III

13 C CP-MAS NMR was also employed to the conformational and dynamical study of the hydration of the barley storage protein, C-hordein that is structurally homologous to ω-gliadins [14]. It was found that two types of domains with different mobility were present. It appeared that the relative proportion of rigid and mobile domains was related to the extent of hydration, which was dependent on the proportion of hydrophilic amino acid residues.

1761

1762 Part III

Food Science

Part III ∗

∗

ppm

150

100

50

0

ppm

150

100

50

0

Fig. 4. 13 C NMR spectra recorded by means of SPE and CP experiments on dry (sample A, left) and H2 O hydrated (sample B, right) gluten: (top) SPE; (bottom) CP. All experiments were performed under MAS conditions at a spinning rate of 6 kHz; asterisks denote spinning sidebands (Adapted from Ref. [20]).

Starch The behavior of starch upon heating was investigated in order to understand the processes that occur to this dough component during baking. Starch contains both crystalline and amorphous structures. Its crystalline nature is revealed by the pattern shown by the C1 glucose peak, resonating at 101 ppm. In particular, in the A-type crystalline form, typical of cereal starches, it appears as a triplet, while in B form, typical of tuber starches, it is a doublet. The existence of a distinct amorphous region was suggested [21] by the presence of peaks resonating at 103.3 and 82.0 ppm having different rotating-frame relaxation rates compared to peaks of others starch protons. These peaks were assigned to random single helices present in the (1→6) linked branching regions. The major noticeable change in the 13 C CP-MAS NMRspectra of heated starch was the loss of signal intensity due to the increased mobility associated to its gelatinization. Anyway, the increase in relative intensity of peaks originated from the amorphous region was also

observed. This means that upon gel formation the amorphous form of starch is the dominant component. The effect of heating and cooling process was studied on starch fractions extracted by wheat flours [22]. From relaxation measurement it emerged that T2 of water decreases during heating, due to the formation of a starch gel at high temperatures. This parameter decreases during cooling, probably for separation of water from starch, due to starch precipitation. This explanation is consistent with the broadening of starch peak upon cooling and from T2 values. The addition of the flour oligosaccharides pentosans to the isolated starch resulted in a more mobile network and in a starch higher organization upon cooling.

Bread Spectra of bread are similar to those obtained for starch gels. During storage of some breads, an increase of C1 signal intensity revealed a decreased molecular mobility due to starch crystallization or retrogradation. Traces of Btype configuration were revealed in aged breads through

MAS NMR of Flours and Doughs

References 1. Chen PL, Long Z, Ruan R, Labuza TP. Food Sci. Technol. 1997;30:178. 2. Roudaut G, van Dusschoten D, Van Has H, Hemminga MA, Le Meste M. J. Cereal Sci. 1998;28:147. 3. Roudaut G, Maglione M, van Dusschoten D, Le Meste M. Cereal Chem. 1999;76:70. 4. Baianu IC, F¨orster H. J. Appl. Biochem. 1980;2:347. 5. Garbow JR, Schaefer J. J. Agric. Food. Chem. 1991;39:877. 6. Sacco A, Bolsi IN, Massini R, Spraul M, Humpfer E, Ghelli S. J. Agric. Food. Chem. 1998;46:4242. 7. Brescia MA, Di Martino G, Fares C, Di Fonzo N, Platani C, Ghelli S, Reniero F, Sacco A. Cereal Chem. 2002;79:238. 8. Amato ME, Ansanelli G, Fisichella S, Lamanna R, Scarlata G, Sobolev AP, Segre A. J. Agric. Food. Chem. 2004;52:823. 9. Belton PS. J. Cereal Sci. 1999;29:103. 10. Gil AM, Masui K, Naito A, Tatham AS, Belton PS, Saitˆo H. Biopolymers. 1997;41:289. 11. Baianu IC. J. Sci. Food Agric. 1981;32:309. 12. Belton PS, Gil AM, Grant A, Alberti E, Tatham AS. Spectrochim. Acta A. 1998;54:955. 13. Gil AM, Alberti E, Tatham AS, Belton PS, Humpfer E, Spraul M. Magn. Reson. Chem. 1997;35:S101. 14. Gil AM, Masui K, Naito A, Tatham AS, Belton PS, Saitˆo H. Biopolymers. 1996;41:289. 15. Gil AM, Alberti E, Naitˆo A, Okuda K, Saitˆo H, Tatham AS, Gilbert S In: PS Belton, BP Hills, GA Webb (Eds). Advances in Magnetic Resonance in Food Science. The Royal Society of Chemistry: Cambridge, 1999, p 126. 16. Alberti E, Gilbert SM, Tatham AS, Shewry PR, Naito A, Okuda K, Saitˆo H, Gil AM. Biopolymers. 2002;65:158. 17. Alberti E, Gilbert SM, Tatham AS, Shewry PS, Gil AM. Biopolymers. 2002;67:487. 18. Alberti E, Humpfer E, Spraul M, Gilbert SM, Tatham AS, Shewry PS, Gil AM. Biopolymers. 2001;58:33. 19. Belton PS, Duce SL, Tatham AS. Int. J. Biol. Macromol. 1987;9:357. 20. Calucci L, Forte C, Galleschi L, Geppi M, Ghiringhelli S. Int. J. Biol. Macromol. 2003;32:179. 21. Morgan KR, Furneaux RH, Stanley RA. Carbohydr. Res. 1992;235:15. 22. Gil AM, Alberti E, Santos D. In: GA Webb, PS Belton, AM Gil, I Delgadillo (Eds). Magnetic Resonance in Food Science— A View to the Next Century. The Royal Society of Chemistry: Cambridge, 2001, p 43. 23. Baik M, Dickinson LC, Chinachoti P. J. Agric. Food. Chem. 2003;51:1242. 24. Brescia MA, Sgaramella A, Ghelli S, Sacco A. J. Sci. Food Agric. 2003;83:1463. 25. Fan TWM. Prog. Magn. Reson. Spectrosc. 1996;2:161.

Part III

the inspection of C1 resonance patterns. These observations were supported by X-ray diffraction data [23]. Some authors studied the effect of glycerol addition on starch rigidity of bread samples stored with and without crust [23]. The addition of glycerol determined the lowering of signal intensities, revealing a decrease in the rigidity of starch. Breads stored with crust gave higher signal intensities, probably due to higher level of rigid domains. 1 H HR-MAS NMR spectroscopy was applied to the analysis of typical bread samples of southern Italy [24]. Apart from the difference in the total lipid content (higher in flours than in bread) substantial differences between the 1 H HR-MAS NMR spectra of a bread crumb and of the relative flour were noticed. Two considerably intense signals appear in the crumb spectrum (respectively at 1.91 and at 2.43 ppm). Their assignment was made on the basis of the HSQC 1 H-13 C spectrum (with 13 C signals resonating respectively at 23.12 and 33.09 ppm) and the aid of chemical shifts found in literature [25]. The observed signals were assigned to acetic and succinic acids, formed as a consequence of secondary fermentation processes. These compounds were determined in other analyzed crumb samples and were also detected in the dough before cooking. Nevertheless the amount of these acids was variable in different crumb samples and, in some cases, succinic acid could not be easily detected. The dependence of these differences on the type of yeast, fermentation, and cooking conditions is under investigation. In the sugar region (3– 5.5 ppm) of the crumb spectrum it was possible to point out the absence of some peaks present in the flour spectrum. This was due to the scission of some polysaccharides into simple sugars that undergo fermentation. In the crumb TOCSY spectrum a correlation peak was found between resonances at 4.15 and 1.39 ppm, which was absent in the flour spectrum. Neither of these proton resonances could be identified in the monodimensional spectrum due to overlapping with more intense signals. This peak was attributed to lactic acid, formed during fermentation processes, according to the 13 C chemical shift values in the heterocorrelated spectrum. There were no salient differences between the crumb and the crust spectrum of the same bread except for the 3–5.5 ppm range of the latter due to the effects of partial crust carbonization.

References 1763

1765

Ana M. Gil and Iola F. Duarte Department of Chemistry, CICECO, University of Aveiro, 3810-193 Aveiro, Portugal

Nuclear magnetic resonance (NMR) spectroscopy has long been applied to the detection and quantification of plant and fruit extracts, often after some kind of separation/purification treatment for simplification and concentration of the system. The challenge remaining regards the application of NMR to heterogeneous and intact systems, so that full use of the non-invasiveness of NMR is made. Tackling physically inhomogeneous materials such as intact fruits and vegetables or some of their components (e.g. skin, pulp, pips) must take into account that different physical phases are characterized by distinct magnetic susceptibilities and that anisotropic effects such as chemical shift anisotropy and internuclear dipolar interactions play increasingly important roles in semisolid and solid materials. The NMR lines in the spectrum are broadened to an extent related to the strength of the effect. Magnetic susceptibility effects may lead to hundreds of Hz to kHz line broadening, depending on the nature of the phases, whereas dipolar interactions may broaden NMR signals up to tens or even hundreds of kHz, in the extreme case of rigid solids. In order to recover spectral resolution, magic angle spinning (MAS) is employed. This consists of spinning the rotor containing the sample around an axis making an angle of 54◦ 44 with the direction of the external magnetic field B0 . Detailed descriptions of this procedure and its principles may be found elsewhere [1,2]. A necessary condition for MAS to result in satisfactory line narrowing is, however, that the spinning speed is at least of the same order of magnitude as the line broadening interaction. In the case of intact fruits, resolved spectra of tissues are typically achievable at spinning rates