277 49 5MB
English Pages 391 [392] Year 2011
ME T H O D S
IN
MO L E C U L A R BI O L O G Y
Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK
For other titles published in this series, go to www.springer.com/series/7651
TM
Metabolic Profiling Methods and Protocols
Edited by
Thomas O. Metz Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA, USA
Editor Thomas O. Metz Pacific Northwest National Laboratory Biological Sciences Division P.O. Box 999, MS K8-98 Richland, WA 99352 USA [email protected]
ISSN 1064-3745 e-ISSN 1940-6029 ISBN 978-1-61737-984-0 e-ISBN 978-1-61737-985-7 DOI 10.1007/978-1-61737-985-7 Springer New York Dordrecht Heidelberg London © Springer Science+Business Media, LLC 2011 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper Humana Press is part of Springer Science+Business Media (www.springer.com)
Preface After accepting the task to edit a volume of Methods in Molecular Biology devoted to metabolic profiling, I began to contemplate the definition of the term. Fiehn referred to “metabolic profiling” as the identification and quantification of a select number of metabolites in an entire metabolic pathway or intersecting pathways (1). Closely related disciplines were targeted metabolite analysis, metabolic fingerprinting, and metabolomics, the latter of which was defined as the quantitative measurement of perturbations in the metabolite complement of a biological system (2). These four terms are often used interchangeably; indeed, in reviewing the literature over the past 40 years, it is evident that these various disciplines of metabolite analysis are related via an evolution of methods and technology. For example, while the field of metabolomics is now 10 years old, the protocols and instrumentation that form the foundation for the myriad approaches of this discipline are based on those originally established for the diagnosis of inborn errors of metabolism and drug metabolite analysis. Thus, in compiling this volume, I have made an attempt to incorporate protocols that are illustrative of the evolution of metabolic profiling from single molecule analysis to global metabolome profiling. The constraints of this volume necessitate that its contents will be perspective based, rather than comprehensive. However, it is my hope that the methods contained herein will be a resource for both established and new investigators in the field of metabolic profiling. Thomas O. Metz
References metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 29, 1181–1189.
1. Fiehn, O. (2002) Metabolomics – the link between genotypes and phenotypes. Plant Mol Biol 48, 155–171. 2. Nicholson, J. K., Lindon, J. C., Holmes, E. (1999) ‘Metabonomics’: understanding the
v
Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
1.
Origins of Metabolic Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . Arthur B. Robinson and Noah E. Robinson
1
2.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism . . . . . Monique Piraud, Séverine Ruet, Sylvie Boyer, Cécile Acquaviva, Pascale Clerc-Renaud, David Cheillan, and Christine Vianey-Saban
25
3.
Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . David S. Millington and Robert D. Stevens
55
Analysis of Organic Acids and Acylglycines for the Diagnosis of Related Inborn Errors of Metabolism by GC- and HPLC-MS . . . . . . . . . . . . . . Giancarlo la Marca and Cristiano Rizzo
73
4.
5.
HPLC Analysis for the Clinical–Biochemical Diagnosis of Inborn Errors of Metabolism of Purines and Pyrimidines . . . . . . . . . . . . . . . . . . . . Giuseppe Lazzarino, Angela Maria Amorini, Valentina Di Pietro, and Barbara Tavazzi
99
6.
Bile Acid Analysis in Various Biological Samples Using Ultra Performance Liquid Chromatography/Electrospray Ionization-Mass Spectrometry (UPLC/ESI-MS) . . . . . . . . . . . . . . . . . . . . . . . . . . 119 Masahito Hagio, Megumi Matsumoto, and Satoshi Ishizuka
7.
Analysis of Glycolytic Intermediates with Ion Chromatography- and Gas Chromatography-Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . 131 Jan C. van Dam, Cor Ras, and Angela ten Pierick
8.
Analysis of the Citric Acid Cycle Intermediates Using Gas Chromatography-Mass Spectrometry . . . . . . . . . . . . . . . . . . . . . . . 147 Rajan S. Kombu, Henri Brunengraber, and Michelle A. Puchowicz
9.
Quantification of Pentose Phosphate Pathway (PPP) Metabolites by Liquid Chromatography-Mass Spectrometry (LC-MS) . . . . . . . . . . . . 159 Amber Jannasch, Miroslav Sedlak, and Jiri Adamec
10.
High-Performance Liquid Chromatography-Mass Spectrometry (HPLC-MS)-Based Drug Metabolite Profiling . . . . . . . . . . . . . . . . . . 173 Ian D. Wilson
vii
viii
Contents
11.
Gas Chromatography-Mass Spectrometry (GC-MS)-Based Metabolomics . . . . 191 Antonia Garcia and Coral Barbas
12.
The Use of Two-Dimensional Gas Chromatography–Time-of-Flight Mass Spectrometry (GC×GC–TOF-MS) for Metabolomic Analysis of Polar Metabolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Kimberly Ralston-Hooper, Amber Jannasch, Jiri Adamec, and Maria Sepúlveda
13.
LC-MS-Based Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Sunil Bajad and Vladimir Shulaev
14.
Capillary Electrophoresis–Electrospray Ionization-Mass Spectrometry (CE–ESI-MS)-Based Metabolomics . . . . . . . . . . . . . . . . . . . . . . . . 229 Philip Britz-McKibbin
15.
Liquid Chromatography-Mass Spectrometry (LC-MS)-Based Lipidomics for Studies of Body Fluids and Tissues . . . . . . . . . . . . . . . . . . . . . . 247 Heli Nygren, Tuulikki Seppänen-Laakso, Sandra Castillo, Tuulia Hyötyläinen, and Matej Orešiˇc
16.
Electrospray Ionization Tandem Mass Spectrometry (ESI-MS/MS)Based Shotgun Lipidomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 Giorgis Isaac
17.
Processing and Analysis of GC/LC-MS-Based Metabolomics Data . . . . . . . . 277 Elizabeth Want and Perrine Masson
18.
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling . . . . . 299 Eva M. Lenz
19.
Nuclear Magnetic Resonance (NMR)-Based Metabolomics . . . . . . . . . . . . 321 Hector C. Keun and Toby J. Athersuch
20.
Slow Magic Angle Sample Spinning: A Non- or Minimally Invasive Method for High-Resolution 1 H Nuclear Magnetic Resonance (NMR) Metabolic Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 Jian Zhi Hu
21.
Processing and Modeling of Nuclear Magnetic Resonance (NMR) Metabolic Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 Timothy M.D. Ebbels, John C. Lindon, and Muireann Coen
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Contributors CÉCILE ACQUAVIVA • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France JIRI ADAMEC • Department of Biochemistry, University of Nebraska-Lincoln, Lincoln, NE, USA ANGELA MARIA AMORINI • Institute of Biochemistry and Clinical Biochemistry, Catholic University of Rome, Rome, Italy TOBY J. ATHERSUCH • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, South Kensington, London, UK SUNIL BAJAD • Sutro Biopharma Inc., South San Francisco, CA, USA CORAL BARBAS • Faculty of Pharmacy, San Pablo-CEU, Campus Monteprincipe, Madrid, Spain SYLVIE BOYER • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France PHILIP BRITZ-MCKIBBIN • Department of Chemistry and Chemical Biology, McMaster University, Hamilton, ON, Canada HENRI BRUNENGRABER • Department of Nutrition, Mouse Metabolic Phenotyping Center, Case Western Reserve University, Cleveland, OH, USA SANDRA CASTILLO • VTT Technical Research Centre of Finland, Espoo, Finland DAVID CHEILLAN • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France PASCALE CLERC-RENAUD • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France MUIREANN COEN • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK VALENTINA DI PIETRO • Institute of Biochemistry and Clinical Biochemistry, Catholic University of Rome, Rome, Italy TIMOTHY M.D. EBBELS • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK ANTONIA GARCIA • Faculty of Pharmacy, San Pablo-CEU, Campus Monteprincipe, Madrid, Spain MASAHITO HAGIO • Division of Applied Bioscience, Research Faculty of Agriculture, Hokkaido University, Sapporo, Japan JIAN ZHI HU • Pacific Northwest National Laboratory, Richland, WA, USA TUULIA HYÖTYLÄINEN • VTT Technical Research Centre of Finland, Espoo, Finland GIORGIS ISAAC • Bio Separation and Mass Spectrometry, Pacific Northwest National Laboratory, Richland, WA, USA; Water corporation, Mulford, MA SATOSHI ISHIZUKA • Division of Applied Bioscience, Research Faculty of Agriculture, Hokkaido University, Sapporo, Japan AMBER JANNASCH • Bindley Bioscience Center, Purdue University, West Lafayette, IN, USA HECTOR C. KEUN • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, South Kensington, London, UK
ix
x
Contributors
RAJAN S. KOMBU • Department of Nutrition, Mouse Metabolic Phenotyping Center, Case Western Reserve University, Cleveland, OH, USA GIUSEPPE LAZZARINO • Division of Biochemistry and Molecular Biology, Department of Chemical Sciences, University of Catania, Catania, Italy EVA M. LENZ • AstraZeneca Pharmaceuticals, Mereside, Macclesfield, UK JOHN C. LINDON • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK GIANCARLO LA MARCA • Mass Spectrometry, Clinical Chemistry and Pharmacology Laboratory, Department of Pharmacology, University of Florence, Meyer Children’s Hospital, Florence, Italy PERRINE MASSON • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK MEGUMI MATSUMOTO • Meiji Dairies Research Chair, Creative Research Institution Sousei (CRIS), Hokkaido University, Sapporo, Japan DAVID S. MILLINGTON • DUMC Biochemical Genetics Laboratory, Department of Pediatrics, Duke University Medical Center, Durham, NC, USA HELI NYGREN • VTT Technical Research Centre of Finland, Espoo, Finland MATEJ OREŠI Cˇ • VTT Technical Research Centre of Finland, Espoo, Finland MONIQUE PIRAUD • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France MICHELLE A. PUCHOWICZ • Department of Nutrition, Mouse Metabolic Phenotyping Center, Case Western Reserve University, Cleveland, OH, USA KIMBERLY RALSTON-HOOPER • Ecosystem Research Division, National Research Council Post-Doctoral Fellow, United States Environmental Protection Agency, Athens, GA, USA COR RAS • Department of Biotechnology, Delft University of Technology, Delft, The Netherlands CRISTIANO RIZZO • Metabolic Unit and Laboratories, Bambino Gesù Children’s Hospital, Rome, Italy ARTHUR B. ROBINSON • Oregon Institute of Science and Medicine, Oregon, OR, USA NOAH E. ROBINSON • Oregon Institute of Science and Medicine, Oregon, OR, USA SÉVERINE RUET • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France MIROSLAV SEDLAK • Laboratory of Renewable Resources Engineering, Purdue University, West Lafayette, IN, USA; Department of Agricultural and Biological Engineering, Purdue University, West Lafayette, IN, USA TUULIKKI SEPPÄNEN-LAAKSO • VTT Technical Research Centre of Finland, Espoo, Finland MARIA SEPÚLVEDA • Department of Natural Resources, Purdue University, West Lafayette, IN, USA VLADIMIR SHULAEV • Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, Blacksburg, VA, USA ROBERT D. STEVENS • Sarah W. Stedman Nutrition and Metabolism Center, Duke University Medical Center, Durham, NC, USA BARBARA TAVAZZI • Institute of Biochemistry and Clinical Biochemistry, Catholic University of Rome, Rome, Italy
Contributors
xi
ANGELA TEN PIERICK • Department of Biotechnology, Delft University of Technology, Delft, The Netherlands JAN C. VAN DAM • Department of Biotechnology, Delft University of Technology, Delft, The Netherlands CHRISTINE VIANEY-SABAN • Laboratoire des Maladies Héréditaires du Métabolisme et Dépistage Néonatal, Hospices Civils de Lyon, Centre de Biologie Est, Bron, France ELIZABETH WANT • Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College, London, UK IAN D. WILSON • AstraZeneca, Macclesfield, UK
Chapter 1 Origins of Metabolic Profiling Arthur B. Robinson and Noah E. Robinson Abstract Quantitative metabolic profiling originated as a 10-year project carried out between 1968 and 1978 in California. It was hypothesized and then demonstrated that quantitative analysis of a large number of metabolites – selected by analytical convenience and evaluated by computerized pattern recognition – could serve as a useful method for the quantitative measurement of human health. Using chromatographic and mass spectrometric methods to measure between 50 and 200 metabolites in more than 15,000 human specimens, statistically significant and diagnostically useful profiles for several human diseases and for other systematic variables including age, diet, fasting, sex, and other variables were demonstrated. It was also shown that genetically distinct metabolic profiles for each individual are present in both newborn infants and adults. In the course of this work, the many practical and conceptual problems involved in sampling, analysis, evaluation of results, and medical use of quantitative metabolic profiling were considered and, for the most part, solved. This article is an account of that research project. Key words: Metabolic profiling, metabolomics, urine, breath, chromatography, mass spectrometry, aging, diagnostic medicine, preventive medicine.
1. Introduction Since the dawn of the age of modern chemistry, biochemistry has been of great interest. When molecular structure became established as an exact discipline, the minds of scientists naturally turned toward those molecules of which they themselves are made. Extensive cataloging and structure determination of these substances followed. As the role of proteins in catalyzing the chemical reactions of metabolism was revealed, progress was made in understanding the metabolites – the smaller molecules required for life that protein
T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_1, © Springer Science+Business Media, LLC 2011
1
2
Robinson and Robinson
catalysts select from the many atomic combinations available and produce to make life possible. Detailed understanding of metabolism was not, however, possible until the discovery of carbon 14 (1) and the development of tracer methodology (2), which now includes both radioactive and stable isotopes. When it became possible to label the atoms of metabolites and trace their paths through living systems, a thorough understanding of metabolism was achievable. This understanding and the rapid advance of protein chemistry then led to explanations for some of the simplest metabolic diseases – genetic errors that cause well-defined inborn errors of metabolism. As analytical technology advanced, the list of known genetic illnesses expanded to include a large number of such diseases which, while individually rare, together cause much suffering. This work was further accelerated by findings that, in some cases such as phenylketonuria, understanding of the disease could lead to effective therapy. Simultaneously, improvements in analytical chemistry led to a search for single metabolites that are diagnostic of more prevalent diseases – including those with non-genetic components. An extensive armament of single-substance measurements entered the inventory of clinical laboratories – tests for both inborn errors and other illnesses. Businesses arose to measure these substances, primarily in blood and urine, which have now grown in the United States alone into a $100 billion industry. This work usually involved the correlation of one substance with a condition of interest in human health. Scientists searched for metabolites and proteins, the quantities of which contained sufficient information about health and disease to warrant their measurement. A few such measurements became standard in health screening of ordinary patients, while a much larger number were made available in clinical laboratories, available upon request by physicians for specific patients. While the many substances measureable in human samples were increasingly evident as analytical methods improved, no practical efforts were made to test the possibility that the simultaneous quantitative analysis of large numbers of metabolites followed by computerized pattern recognition could yield health information of significant value. Forty years ago, however, there arose in California an experimental project with the potential to cause a paradigm shift toward the use of simultaneous measurement of large numbers of metabolites for the quantitative measurement of human health. This effort was ahead of its time and, therefore, faced daunting challenges in the construction of analytical and computational capabilities. This work was known in the 1970s as “quantitative metabolic profiling.” It is now a growing part of “metabolomics.” While
Origins of Metabolic Profiling
3
metabolomics still contains substantial single-substance components, extraordinary advances in analytical and computational technology are rapidly moving this field toward metabolic profiling – a continuation of that 1970s’ effort with greatly superior modern analytical equipment and computers. The California work was funded by private donors, NIH grants, and the personal savings of some of the scientists themselves. This effort proved the enormous analytical power of metabolic profiling as applied to human tissues and developed new analytical and computational tools. It had its origin in a collaboration beginning in 1968 between Linus Pauling and Art Robinson at the University of California at San Diego (UCSD). Later, it continued at Stanford University and the Institute of Orthomolecular Medicine (later renamed the Linus Pauling Institute of Science and Medicine), which Pauling and Robinson cofounded in Menlo Park, California, in 1973.
2. Orthomolecular Psychiatry
Pauling hypothesized (3) that the distribution functions of optimum human nutritional requirements are very wide, leading to nutritional deficiencies and illness, especially mental illness, in many people. He invented the term “orthomolecular psychiatry” – meaning right molecule in the right amount for mental health – to designate the treatment of mental illness by means of megavitamin therapy. Later this was designated “orthomolecular medicine” to include treatment of other illnesses in a similar way. Having worked together at Caltech in 1962–1963 on the chemical basis of general anesthesia (4), both Pauling and Robinson were faculty members at UCSD when Pauling made this proposal. At the time, Pauling was developing a theory of the structure of the atomic nucleus, and Robinson and his students were studying the deamidation of asparginyl and glutaminyl residues in peptides and proteins. In addition to this ongoing work, in 1968 the two men began a collaboration to test Pauling’s ideas about orthomolecular psychiatry, with Robinson directing the experimental work and Pauling extending the theoretical aspects, which led eventually to his widely known hypotheses concerning the role of vitamin C in health and disease. Pauling initially proposed an experimental program using vitamin-loading tests, in which large doses of vitamins were given to experimental subjects – those having mental illnesses and control subjects – and the urinary excretion of the vitamins measured. It was postulated that those individuals with greater needs for the substances would retain more, excreting lesser amounts in
4
Robinson and Robinson
their urine. Robinson assembled a small research group and set up a laboratory for this purpose, while continuing to direct his own laboratory – the size of which was increased by UCSD to accommodate the new work. The initial experiments emphasized loading tests with ascorbic acid, niacin, and pyridoxine, and some interesting results were obtained. It soon became evident, however, that this approach was of less value than hoped. The experiments gave very limited information, and the necessary analytical procedures of that day were laborious, time consuming, and expensive, which diminished their practical value.
3. Origin of the Profiling Hypothesis
4. Scientists Who Tested the Hypothesis
In the course of this work, Robinson utilized a method for measurement of pyridoxine in chemically derivatized urine by means of packed-column gas chromatography, which involved resolution of the pyridoxine peak from the large number of metabolic products that are present in urine. During these experiments, Robinson began to think that the information they needed might be more readily available in the many metabolic constituents evident in the chromatograms rather than in the pyridoxine peak itself. He reasoned as follows. The fundamental need was for a method to measure health vs. the amounts of ingested nutrients, as is illustrated in Fig. 1.1. This required, however, a means of measuring metabolic health quantitatively. He hypothesized that the needed values might be obtained by measuring the amounts of a large sampling of urinary metabolites and statistically correlating the patterns in these profiles with various states of human health and disease. Robinson, therefore, initiated an experimental program, with Pauling’s support, to test the hypothesis that quantitative metabolic profiles contained sufficient information for this purpose. As this work progressed, he assembled a skilled group of co-workers for this project.
These included Roy Teranishi, Dick Mon, and Robert Flath – highly skilled experts in gas chromatography; Martin Turner and Carl Boehme – engineers who built and maintained the PDP-11 vintage computer hardware used for lab automation and data collection; Laurelee Robinson – who wrote the computer software
Origins of Metabolic Profiling
5
Fig. 1.1. Diagrammatic representation of the nutritional problem that led to the development of metabolic profiling. The goal was to measure health as a function of nutrition quantitatively for (a) a single individual and (b) groups of individuals. It was hypothesized – and then demonstrated – that quantitative analyses of a large set of easily measurable metabolites contained sufficient information to determine the numbers needed for the vertical axis in (a) and that the same set could be used for a wide variety of human conditions. Reprinted with permission from the Proceedings of the 8th Annual Conference of the National Society for Autistic Children (15).
used for lab automation, data calculation, and profile evaluation; Fred Westall – Salk Institute chemist who provided samples for multiple sclerosis, muscular dystrophy, and Huntington’s disease work and helped organize the sample bank project; Bill Aberth – physicist who built the molecular ion mass spectrometers; Robert Melville – National Institutes of Health Administrator who arranged for and supervised the NIH support; glassblower Paul
6
Robinson and Robinson
Yeager; chemists Henri Dirren, Kent Matsumoto, and Lowell Brenneman; and biochemist Milton Winitz – who invented the synthetic diet Vivonex 100 and supplied it to the project in large quantities without charge, making possible the diet control studies. These people were aided by several technicians, including Sue Oxley, Maida Bergeson, Janet Tussey, Betsy Dore, Mark Weiss, and Walter Reynolds. Additionally helped by their own students and former students, Pauling’s – John Cheronis and Ian Keaveny – and Robinson’s – Fred Westall, Paul Cary, David Partridge, and Alan Sheets – and by other colleagues from UCSD, Stanford, Kaiser Permanente, and other institutions, a remarkable laboratory was built. Human samples were obtained from many institutions and physicians, including especially Drs. John Mann and George Ellison. Key to the laboratory’s success – from its completely automated, custom-made analytical equipment to its advanced computerized fund-raising systems (donations were obtained from more than 50,000 private individuals) – were four PDP-11 computer systems, utilizing machine language and Fortran programs that Art Robinson’s chemist and systems programmer wife Laurelee wrote in the 8-year period between 1971 and 1978. This group of people gradually built the finest physiological fluid analysis laboratory of its time – fully automated and designed for analysis of large numbers of samples. Ultimately, this laboratory measured, during a period of 8 years, more than 15,000 metabolic profiles – each profile including quantitative measurement of between 50 and 200 substances, primarily in human urine.
5. Accomplishments of the 10-Year Project
Most of the analytical work was by automated gas and ion exchange chromatography, with molecular ion mass spectrometry added during the later years and fragmentation mass spectrometry used for chemical characterizations of chromatographically resolved constituents. The early gas chromatographic work utilized 6-ft-long packed columns. These were replaced by 1000ft-long open-bore stainless steel columns, which were used for analysis of urine vapor and breath. The ion exchange system utilized the ninhydrin-positive physiological fluid analytical procedures developed by Dionex, with added PDP-11 automation. The identification work was carried out with a Finnegan gas chromatograph–fragmentation analysis mass spectrometer.
Origins of Metabolic Profiling
7
Between its beginning in 1969 and the end of the work in 1978, this research effort definitively showed the advantages of metabolic profiling – demonstrating unique statistically significant metabolic profiles for aging, multiple sclerosis, Duchenne dystrophy, Huntington’s disease, breast cancer, fasting, sex, diurnal variation, and chemical birth control use. The numbers of correlating metabolites found in these profiles ranged between 10 and 60, depending upon analytical technique and sample type (5–24). Reference 5 provides a review of much of this work, and References 25 and 26 provide more recent perspective. They also demonstrated, in experiments on 1,000 newborn infants, that the ninhydrin-positive metabolites in human urine are not unimodally distributed at birth. About half of these compounds are bimodally and trimodally distributed, apparently reflecting distinct genetic variations. In addition, they showed that the urinary profiles of adult humans are individually unique so that people can be identified by their urinary profiles just as with fingerprints. In experiments on themselves, they discovered the simplification of urinary profiles that occurs with strict chemical diet control but additionally showed that useful metabolic profiles can be obtained in most instances without diet control. Also demonstrated were metabolic profiles for physiological age in fruit flies (22), mice (21), and men (25), with the finding that about one-third of the metabolites in human urine are age correlated. This discovery opened the way toward quantitative evaluation of dietary and other measures that may decrease the degenerative diseases of aging. Some human conditions show no profiles. For example, their experiments on Stanford students failed to find a urinary metabolic pattern correlating with grade point average (5) – a useful comparison to the patterns that were found in other groups. The work of these scientists and engineers included the following: 1. Building the first fully automated metabolic profiling laboratory. Using PDP-11 computer technology, all of the chromatography and the later molecular ion mass spectrometry equipment that they built were entirely computer controlled and all of the data produced were computer collected and analyzed with very little manual intervention. 2. Building the first chromatographic breath and urine vapor metabolic analyzer (9). This machine contained four 1000ft-long stainless steel open-bore capillary columns. Each column was capable of resolution and quantitative analysis of about 200 volatile substances in a urine or a breath sample with a 6-h cycle time. In routine operation, it could analyze 16 samples per day.
8
Robinson and Robinson
3. The conceptual development and computerized implementation of mathematical tools that produced baselinecorrected integration of the chromatograms; corrected for chromatographic drift and automatically matched all metabolites within sets of chromatograms; normalized the experimental values to correct for variable physiological dilution and other systematic variables; computed nonparametric probabilities of significance for all metabolites in an experiment; corrected these probabilities for random correlation; tabulated and plotted cumulative probability distribution functions to determine the numbers of significant correlating substances; and, by means of a conceptually unique method (13, 25), computed the “diagnostic power” of any discovered metabolic pattern. 4. Origination and implementation of the concept of metabolic profiling by means of molecular ion mass spectrometric separation and quantitative analysis, without prior sample preparation or subdivision. This was done before the invention of the electrospray or laser ionization methods commonly in use today. In 1971, Stanford Research Institute experimental physicists William Aberth and Cap Spindt suggested a new way of producing ions for mass spectrometry without fragmentation, which made use of an array of hollow volcano-shaped structures with a grid registered above it, and Robinson suggested that this be used as a profiling device. With the help of Robert Melville at NIH, Aberth, Robinson, and Pauling received funding to build this device. Aberth ultimately built two such mass spectrometers – one at SRI with this NIH funding and one, later, at the Institute of Orthomolecular Medicine, with the second machine fully automated by Walter Reynolds, Carl Boehme, and Laurelee Robinson. These spectrometers had an initial resolution of 1 mass unit, with a design potential of 0.1 mass units in the mass range from about 50 to 1000. 5. Use of metabolic profiling in urine, breath, and cultured human cells. Similar profiling techniques can, of course, be applied to blood, saliva, and other sources of metabolites. The cultured cell work was undertaken to test the hypothesis that cell cultures from single individuals, monitored by metabolic profiles, might serve as individualized experimental systems in which therapeutic procedures for single individuals such as dietary requirements could be tested.
Origins of Metabolic Profiling
6. Goals and Requirements for Quantitative Metabolic Profiling
9
While the advance of technology during the past 30 years now makes a wonderful amount of fundamental biochemical knowledge available through metabolic profiling, these early workers were motivated almost entirely by a practical, empirical goal – in Robinson’s words, “improvement of the quality, length, and quantity of human life,” or, in Pauling’s, “diminution of the amount of human suffering.” Figure 1.2 summarizes these goals. In order to achieve these objectives, practical empirical metabolic profiling requires several things: 1. Suitable reproducible, quantitative, and low-cost analytical methods. 2. Control and experimental sample sets for test and calibration that are sufficiently free of uncontrolled systematic variables to allow reliable characterization of the differences under consideration. 3. Computerized mathematical methods that allow objective evaluation of the experimental results, without, through unnecessary complexity, separating the experimenters and their scientific intuitions from their data and results. During the 10-year duration of this project, these three problems were successfully addressed. The emphasis of the single-substance-orientated clinical chemistry industry of the 1970s was then, as it is today, primarily upon diagnosis of overt disease by technologically obsolete methods. Physicians are offered the amounts of single substances in human samples and comparisons with so-called normal values, typically two-standard-deviation ranges for the general population. Measurements of a couple of dozen such substances are included in ordinary analyses, and single substances beyond the normal range are noted and considered in patient evaluation. A large suite of additional single-substance measurements is available in industrial laboratories, which the physician can order to extend or confirm his diagnosis. This paradigm is expensive, so the number of substances measured is low and the application is limited to patients already exhibiting disease symptoms. Moreover, it entirely misses the metabolic patterns available from groups of substances that have values within the normal ranges – patterns that require computer analysis to discover. Quantitative metabolic profiling of analytically convenient metabolites allows a single analytical procedure, measuring a single large set of metabolites, to diagnose essentially all disease
10
Robinson and Robinson
Fig. 1.2. The primary health goal of “metabolic profiling” or, now, “metabolomics.” The gradually sloping curve of human survival, beginning to decrease at early ages, represents great amounts of suffering and lost years of human life, as illustrated here by the lifespan distribution of men in the United States in 1974 (a). The first goal of metabolomics should be the “squaring” of this curve (b) so that most people live a long, disease-free life during their intrinsic lifespan of about 90 years. A second goal should be extension of the healthful human lifespan (c). Reprinted with permission from (23).
Origins of Metabolic Profiling
11
conditions with one inexpensive procedure. Moreover, by including computerized pattern recognition, metabolic profiling extracts far more complete and valuable medical information than does the traditional method. The low cost and much greater information content of metabolic profiling permits its use in preventive medicine, allowing the individual to combat the probability of disease rather than overt disease itself. It also provides a convenient and inexpensive means of quantitative measurement of illness so that therapeutic procedures can be evaluated in real time – a capability almost entirely absent from current therapeutic medicine. Moreover, the finding that physiological age can be quantitatively measured by metabolic profiling opens the way toward the conduct of objective research for the evaluation of the effects of various adjustable nutritional and other parameters on aging and, when refined in the future, will allow single individuals to monitor their own rate of aging and probabilities of disease as a function of time and their own habits. The low cost and therefore increased availability of health evaluation that metabolic profiling makes possible can save the lives of many people that are now lost because current methods – imbedded in an expensive, inconvenient health system and providing inferior information – fail to diagnose their illnesses in time. “Health” is a concept that varies with individual objectives. Optimum health means different things to an athlete, to an artist, to a mathematician, or to a soldier. Each seeks to optimize different aspects of his abilities. The quantitative measurement of health that quantitative health profiling should eventually make possible will allow each person to optimize those abilities that he considers most valuable. It was these possibilities and other similar objectives that motivated Art Robinson and his colleagues in the 1970s to develop the techniques of quantitative metabolic profiling.
7. Analytical Methods Requirement 1, the choice of analytical methods is simplified by the fact that the many substances in a living metabolism are interlinked in synthesis and function, with each substance providing information about some of the others. When a large subset of these substances are quantitatively analyzed, the metabolites measured can be chosen by analytical convenience and economy rather than maximum information content per metabolite. Instead of seeking, as had been historically customary, one or two
12
Robinson and Robinson
high-information-content substances that are diagnostic of a disease or useful for some other practical health purpose, metabolic profiling combines the information in a large number of substances with lesser information content per substance. The profiler asks, “How many metabolic substances can be measured for a practical cost?” instead of “How can a single diagnostic substance be found?” This markedly reduces overall cost, an especially important goal for preventive medicine. A single cost-selected multi-substance profile is much less expensive than separate analyses of hundreds of singly selected substances. Thus, the same set of substances – between 50 and 200 during the original work in the 1970s and potentially thousands with current technology – can be used for many purposes, such as diagnosing disease, monitoring disease therapy, measuring optimum health as a function of diet, and other health goals. Quantitative measurement of a large number of metabolic substances, appropriate normalization, and computerized pattern recognition combine to make many new things possible. For example, quantitative measurement of individual human physiological age allows the graphing of physiological age vs. time. With this capability, the effect on physiological age of diet, exercise, and other adjustable parameters can be objectively monitored – both for groups and for single individuals. In the case of disease, metabolic profiling makes possible the quantitative measurement of the probability of a specific illness by comparing the profile of an individual with that of diseased individuals and those who will later become ill. During disease therapy, an individual can be measured as he moves along this probability axis, allowing the objective evaluation of therapy. Moreover, assessment of disease probabilities rather than overt symptoms opens the option of fighting the probability of disease rather than disease itself. While these and other similar health goals have become a part of our custom and culture, they have remained largely in the realm of qualitative discussion and guesswork, without the benefit of objective science and technology. The reason for this has been the lack of a means for measuring health quantitatively. Metabolic profiling provides this means. In the 1970s, chromatography was a slow but inexpensive method, while mass spectrometry was fast but expensive. Chromatography linked to mass spectrometry was, therefore, slow and expensive. Robinson imposed a condition upon the analytical methods chosen by that research group that no profiling tool would be used that could not, in industrial application, be offered to the general public at a cost of $5–$10 or less. This was necessary to the health goals of the project and is still a useful limitation.
Origins of Metabolic Profiling
13
8. Sampling Requirement 2, sampling has always been and still remains the weak link in the development of metabolic profiling. Even the simplest experiment – comparison of profiles of individuals known to be sick with a specific illness to those of well individuals – suffers from the danger of systematic error. There is always a possibility that the two groups also differ in ways not fundamental to the illness such as lifestyle, diet, therapeutic drugs, and other factors – differences that may exhibit profiles that are mistaken for those of interest. Moreover, comparison of profiles of individuals to those of groups of other individuals markedly decreases the information content of the profiles because it introduces biological variation into the observations. Longitudinal measurements wherein an individual serves as his own control and differences in his own profile with time are measured with reference to known population profiles contain far more information. Metabolic profiling – as a result of the enormous amount of information that it makes available – provides, for the first time, an objective means of fighting the probability of disease rather than disease itself. By reducing the data to a linear probability axis extending between sick and well, the probability of illness can be measured and the individual’s position on the axis determined – and moved empirically toward the well end of the axis by means of preventive measures before disease symptoms are evident. This cannot be done, however, unless the techniques can be calibrated by means of longitudinal samples obtained from people before they became ill. Realizing this need for sample sets from people that include serial samples from the same individuals before illness was evident, Robinson, Westall, and Pauling, in collaboration with colleagues at Kaiser Permanente, proposed in 1976 that a sample bank be created in which blood and urine samples were collected at 6-month intervals from 50,000 people in ordinary health and stored at –80◦ C. After 5 years, with current disease incidences, this bank would have provided statistically significant prospective sample sets for all diseases with incidences equal to or greater than multiple sclerosis. If collection continued beyond 5 years, suitable sample sets from more rare conditions would be available. Kaiser Permanente agreed to pay the costs of this project if their contribution were matched by $5 million from the National Institutes of Health. After 2 years of lobbying by Robinson, NIH decided in 1978 to fund an initial project with 10,000 subjects, with the possibility of later expansion to 50,000. Unfortunately, just after
14
Robinson and Robinson
this approval, work on profiling at the Pauling Institute collapsed, so this sample bank was not built. The sampling problem is today still the greatest impediment to progress in metabolomics. This sample bank should be built – with urine, breath, saliva, blood, and culturable human tissue collected from at least 100,000 ordinary people at regular intervals and stored at –80◦ C. After a few years, with such a bank in hand, a great increase in progress in the application of metabolic profiling for the improvement of human health would become possible.
9. Computation Requirement 3, computation was more difficult with the PDP11s of the 1970s, which were primitive computers by today’s standards. They were slow, expensive, and required continual repair. They were, however, a great advance over the mainframe computer systems that then existed at most research centers. The PDP-11s permitted dedicated computer systems for specific purposes, and they were not subject to the frequent outages and other limitations that rendered institutional mainframe systems unsuitable for metabolic profiling work. For real-time automation and data collection, the PDP-11s required programming in machine language in order to meet time requirements. For data analysis, Fortran programming was sufficient because time was not a constraint. Most of the multivariate computational methods in use today were available in the 1970s. These, however, proved less than ideal for the profiling work. First, although designed to give attractive displays such as pseudo two-dimensional plots of metabolically similar groupings, the outputs of these systems did not lend themselves to practical, medically useful decision making. Second, especially during research and development, computer procedures that move beyond the intuition of the experimentalists pose great risks of error. The danger of false positives in profile detection through the use of too many adjustable parameters, improper, uncorrected statistical analysis, introduction of systematic errors, and other factors must always be avoided – avoidance that becomes more difficult if “trust” is placed in computer calculations that are beyond checking by simple manual methods or mental calculations. Since there is a large amount of data to be collected and much repetitive calculation required, computers are essential for metabolic work, but they must not be allowed to become experimental variables in the research itself. For this reason, Laurelee and Art Robinson developed specialized
Origins of Metabolic Profiling
15
calculation procedures for this work (Fig. 1.3). These included the following: 1. Fully automatic baseline fitting, integration, and peak matching software to extract the necessary data from the chromatograms. While today, online mass spectrometry can do this, in the 1970s, automatic peak matching in highresolution chromatograms was an especially difficult problem, which they successfully solved. 2. After normalization to weighted averages of the measured values – a procedure they introduced to eliminate systematic errors that affect all of the values such as urine volume – probabilities of correlation with the phenomenon under study were calculated for each metabolite with nonparametric statistics. Since the distribution functions of most metabolites are not Gaussian and many are multi-modally distributed, parametric statistics is unsuitable. So, a method based on modified Wilcoxon statistics was employed. Cumulative distribution functions of these probabilities were then constructed to correct for random correlations and to test for the existence of metabolic profiles in the experiments, as illustrated in Fig. 1.4. 3. After measuring a set of n substances that correlate with a condition of interest, the values from this set were compressed to one dimension because most actions that would be based upon the analysis – for example, medical therapy – are one dimensional. So, simple mathematical means for making this compression were devised as diagrammatically represented in Fig. 1.5. 4. Since experiments to discover metabolic profiles do not inherently contain information about how these profiles will be used, a general method for evaluating profile strength is also needed. In medical practice, the detection of a deadly disease that will be treated by a safe therapy requires that false positives be risked at the expense of avoiding false negatives. Conversely, a mild disease with a relatively dangerous therapy requires the opposite bias. These and other similar interpretive biases are not, however, inherently a part of a metabolic profile. Therefore, a pattern strength methodology was invented, which averages all possible uses, as illustrated in Fig. 1.6. With the correlation indices of individual subjects for the n correlating metabolites compressed on a linear axis extending between the averaged profiles for the two groups being compared, that axis is divided at all possible points and the results graphed as shown. If no diagnostic power is present in the profiles, the values will lie along the diagonal line. If the diagnostic power of the metabolic profile is perfect, the values comprise
16
Robinson and Robinson
Fig. 1.3. The highest resolution metabolic profiler utilized in the 1970s (a). In 6 h, this device performed completely automated quantitative measurement of about 200 volatile compounds (b) in four different urine or breath samples. The equipment was completely automated with PDP-11 computers. All data analysis was also automated in another PDP-11 system, pictured here with Laurelee Robinson, who wrote the software for the 10-year project. Laurelee died in 1988 at the age of 43 – a death that could have been prevented had the metabolic profiling that she helped to develop been in routine medical use. Personal photographs – Art Robinson.
Origins of Metabolic Profiling
17
Fig. 1.4. Cumulative distribution functions of nonparametric probabilities of correlation. In this example of 72 ninhydrin-positive urinary metabolites as a function of sex in 71 male and 77 female Stanford students (a), the diagonal line corrects for random correlation. About 30 of the 72 substances correlate with sex, as is shown by deviation from the diagonal line of random correlation. A similar experiment with 57 male Stanford students (b) shows that, since no profile is present for grade point average, the measured values fall along the diagonal line. Alternatively, a tabular presentation of number of substances below each probability vs. the number expected below that probability can be used. Reprinted with permission from Clinical Chemistry (5). In ordinary singlesubstance clinical chemistry, it is more difficult to rigorously evaluate the significance of a reported correlation because there is no suitable way to correct for the unknown number of unreported experiments that have failed to detect a correlation. Metabolic profiling, in which each experiment evaluates a statistically significant number of potential correlations, has the advantage that correction for random correlation can be made from the experimental data itself.
18
Robinson and Robinson
Fig. 1.5. Diagrammatic representation of three important uses of metabolic profiling. The n-dimensional profiles of substances that correlate with physiological age, probability of illness, and severity of illness are computationally compressed onto one-dimensional axes suitable for empirical evaluation of interventions for preventive and diagnostic medicine and for therapeutic monitoring. This one dimensional compression is also used in the calculation of profile “diagnostic power” as illustrated in Fig. 1.6. Reprinted with permission from Mechanisms of Ageing and Development (23).
Fig. 1.6. Diagnostic power of the sex-correlated ninhydrin-positive compounds illustrated in Fig. 1.4a. The n correlates are computationally compressed to one dimension. The one-dimensional axis extending between the male profile and the female profile as diagrammatically illustrated in Fig. 1.5 has one value for each of the 148 students. In the computation of these values, the profile of the student being calculated is always omitted from the calculation of the profiles to which he is being compared. Reprinted with permission from Clinical Chemistry (5). The resulting linear axis is divided at all of the 149 possible places between and beyond the 148 values and the errors plotted as shown. If there were no diagnostic power, the values would lie along the diagonal line. If the diagnosis is perfect, the graph is reduced to a point in the origin. The fraction of the area between the diagonal and perfect correlation that has been successfully eliminated is the “diagnostic power,” in this case 0.93. This quantitative experimental endpoint is useful in evaluating the strength of a profile, in comparing the relative values of alternative profiling techniques, and in optimizing experimental and computational parameters, such as the coefficients used to normalize the data.
Origins of Metabolic Profiling
19
a point at the origin. The “diagnostic power” is defined as the percentage of the area between the diagonal line and the origin that has been successfully eliminated by the technique. If the data are from more than one replicated experiment or from groups of subjects that are fundamentally impossible to completely separate into unique groups, the probability and diagnostic power distributions are non-linear, so appropriately curved expected distributions are used (13, 24) rather than the linear ones illustrated in Figs. 1.4 and 1.6. The invention of this method (13) permitted further optimization of the metabolic profiling experiments and calculation methods because it provided a quantitative score for the entire experiment and calculation. For example, data normalization coefficients could be optimized by iterative calculations to maximize diagnostic power. This “diagnostic power” method has proved to be generally useful in many unexpected ways. For example, it was recently used to optimize the parameters involved in predicting the deamidation rates in asparginyl residues for proteins of known threedimensional structure (27–30).
10. Summary of the Project In 1978, after 10 years of effort, the initial hypothesis about metabolic profiling had been verified. In every case in which the profiling laboratory completed an experiment intended to detect a profile distinctive of human disease and in many other experiments on variables of interest, a distinctive, diagnostically useful profile had been found. A set of 1000 humans had been profiled at birth with the goal of correlating their genetically unique urinary profiles with their health in later life. Analytical methods using chromatography and direct injection molecular ion mass spectrometry had been refined and automated, and suitable computational tools developed. Moreover, funding had been obtained for the regular sampling of 10,000 humans and permanent storage of the samples at –80◦ C so that profiling technologies could be tested and calibrated with longitudinal controls. Also, the various metabolic profiles discovered over the 10 years of work had been correlated with a single database.
11. Termination of the Project During the 10 years, all of the conceptual and computational questions that arose had been answered. So, the discipline of
20
Robinson and Robinson
metabolic profiling was firmly established. One metabolic profiling laboratory was well funded, equipped, and staffed with welltrained and experienced scientists and engineers and was ready to move forward. Robinson, 36 years old, was then research professor and president of the Pauling Institute and Pauling, 77 years old, was research professor and chairman of the board. Unfortunately, a disagreement arose between these two men regarding a series of experiments on nutrition and the growth rate of squamous cell carcinoma in hairless mice that had been carried out under Robinson’s direction. Involving experiments with about 2,000 mice, these experiments showed that sub-optimum nutrition decreased the cancer growth rate, while better nutrition increased it. The growth rate of squamous cell carcinoma in these mice was varied over a range of 20-fold by means of diet alone, an important finding that may have importance in the management of human cancer (30). General dietary restriction or provision of single nutritional components such as protein in amounts above or below those providing optimum overall health decreased the cancer growth rate, a finding that included vitamin C. With a supplement of the human equivalent of about 3 g/d (a dose Pauling was recommending to the public), vitamin C increased the cancer growth rate by twofold as compared with no supplement – a result that, unfortunately, Pauling believed harmful to his personal campaign to promote vitamin C. Exacerbated by several individuals who sought to gain from this disagreement, this controversy caused the Institute’s research to collapse, and all further work on metabolic profiling ended. Pauling and his attorneys who gained control of the Institute succeeded in prohibiting Robinson and his colleagues access to their equipment and research data, and the very high cost of rebuilding the necessary technology elsewhere prevented them from continuing the work. Accounts of a large part of their work in the 1970s have, however, been published (refs. 5–24 are a complete listing) and should prove useful to those in the field of metabolomics who are now rapidly advancing by means of the wonderful new analytical technology that is available today.
12. Some Thoughts About the Future
In the 1970s, the goal of metabolic profiling was to obtain as much quantitative health information as conveniently and economically as possible. At that time, urine analysis in central laboratories was the best path to this goal. Recognizing that other tissues and fluids were potentially complimentary or superior
Origins of Metabolic Profiling
21
sources, small forays into breath, saliva, blood, and cultured tissue were also undertaken. With regard to the substances analyzed, volatile compounds were easiest with the technology of that day, but non-volatile compounds appeared to contain more information per substance. While nucleic acids, proteins, and metabolites are all now of interest as a result of technological advance, only metabolites were within realistic analytical reach in the 1970s. Even today, it is likely that metabolites are still superior for these purposes. Nucleic acids and proteins are the blueprints for the biological structure and the machines that comprise and operate it. Quantitative health determination based on these molecules is, however, still in its infancy. Metabolites provide immediate real-time access to biochemical information reflected in the chemical output of basic processes in the living system. This profiling is becoming especially valuable, since the computer revolution now makes possible the transfer of diagnostic technology away from centralized laboratories and into each individual’s personal environment (25, 26). This transfer will markedly lower cost and increase convenience and, thereby, make possible a continuous flow of very useful information to each individual. We are not far from a time when each person’s personal computer will constantly acquire data about his health and submit that data for interpretation to the free-enterprise entrepreneurs of his choice – anywhere in the world. Which technologies will be ascendant in the future? All relevant health measurement technologies as applied to a wide variety of samples will obviously compete among research laboratories and in the marketplace. Only this competitive environment can sort out which approaches will provide the greatest health benefits for the greatest numbers of people. Many human factors as well as scientific factors will be important, such as the crucial issue of human compliance. Miraculous machines may arise for the analysis of drops of urine or blood, for example, but how many people would routinely utilize them? While we surely do not know the answers to these questions, breath metabolite analysis seems especially promising. On the basis of information content and ease of analysis, current techniques of breath analysis are much inferior to, for example, urine or blood analysis, but the human compliance problem is fundamentally much easier. Miniaturized mass spectrometers or other devices, for example, purchased as computer peripherals or perhaps routinely included on ordinary personal computer motherboards could continuously sample the breath of a single computer user or the air in a room frequented by a few people, without distraction of the people being monitored. During its birth in the 1970s, metabolic profiling required custom-built, slow, and expensive analytical machines coupled to
22
Robinson and Robinson
primitive, costly, and temperamental computer systems. Still, even then, it could have been incorporated as a very useful tool in centralized clinical laboratories. It is tragic that this was not done. Now, the advance of technology has removed these limitations. Analytical technology is greatly improved and is becoming less expensive and more reliable, while powerful interlinked computers have brought the computational requirements of metabolic profiling within reach of virtually every individual in the developed world. The applications of this new technology to increase the quality, quantity, and length of human life during the coming decades should be spectacular indeed. References 1. Kamen, M. D., Ruben, S. (1940) Production and properties of carbon 14. Phys Rev 58, 194. 2. Kamen, M. D. (1957) Isotopic Tracers in Biology: An Introduction to Tracer Methodology, Academic Press, New York, NY. 3. Pauling, L. (1968) Orthomolecular psychiatry. Science 160, 265–271. 4. Robinson, A. B., Manly, K. F., Anthony, M. P., Catchpool, J. F., Pauling, L. C. (1965) Anesthesia of artemia larvae: method for quantitative study. Science 149, 1255–1258. 5. Robinson, A. B., Pauling, L. C. (1974) Techniques of orthomolecular diagnosis. Clin Chem 20, 961–965. 6. Teranishi, R., Mon, T. R., Robinson, A. B., Cary, P., Pauling, L. C. (1972) Gas chromatography of volatiles from breath and urine. Anal Chem 44, 18–20. 7. Pauling, L. C., Robinson, A. B., Teranishi, R., Cary, P. (1971) Quantitative analysis of urine vapor and breath by gas– liquid partition chromatography. Proc Natl Acad Sci USA 68, 2374–2376. 8. Robinson, A. B., Pauling, L. C. (1973) Quantitative chromatographic analysis in orthomolecular medicine, in W H Freeman & Co (Hawkins, D., ed.), Orthomolecular Psychiatry, pp 35–53. 9. Robinson, A. B., Partridge, D., Turner, M., Teranishi, R., Pauling, L. C. (1973) An apparatus for the quantitative analysis of volatile compounds in urine. J Chromatogr 85, 19–29. 10. Matsumoto, K. E., Partridge, D. H., Robinson, A. B., Pauling, L. C., Flath, R. A., Mon, T. R., Teranishi, R. (1973) The identification of volatile compounds in human urine. J Chromatogr 85, 31–34. 11. Pauling, L. C., Robinson, A. B. (1973) Techniques of orthomolecular medicine. In First
12.
13.
14.
15.
16.
17.
18.
19.
20.
Conference on the Analysis of Multicomponent Mixtures and Their Application to HealthRelated Problems 1, 1–7. Robinson, A. B., Cary, P., Dore, B., Keaveny, I., Brenneman, L., Turner, M., Pauling, L. (1973) Orthomolecular diagnosis of mental retardation and diurnal variation in normal subjects by low-resolution gas–liquid chromatography of urine. Int Res Commun Syst 70, 3. Robinson, A. B., Westall, F. C. (1974) The use of urinary amine measurement for orthomolecular diagnosis of multiple sclerosis. J Orth Psychol 3, 1–10. Robinson, A. B., Westall, F. C., Ellison, G. W. (1974) Multiple sclerosis: urinary amine measurement for orthomolecular diagnosis. Life Sci 14, 1747–1753. Robinson, A. B. (1974) Orthomolecular medicine – diagnosis and therapy. In Proceedings of the 8th Annual Conference National Society for Autistic Children, 1–8. Robinson, A. B. (1975) Looking for optimum health: A guided tour through the Linus Pauling Institute. Prevention 89–96. Robinson, A. B., Weiss, M., Reynolds, W. E., Robinson, L. R. (1975) Use of mass spectrometry for Orthomolecular diagnosis. In Proceedings Twenty-Third Annual Conference on Mass Spectrometry and Allied Topics, 182–184. Dirren, H., Robinson, A. B., Pauling, L. C. (1975) Sex-related patterns in the profiles of human urinary amino acids. Clin Chem 21, 1970–1975. Rosenberg, R. N., Robinson, A. B., Partridge, D. (1975) Urine vapor pattern for olivopontocerebellar degeneration. Clin Biochem 8, 365–368. Robinson, A. B., Dirren, H., And Sheets, A., Miquel, J., Lundgren, P. R. (1976) Quanti-
Origins of Metabolic Profiling
21.
22. 23.
24.
25.
tative aging pattern in mouse urine vapor as measured by gas–liquid chromatography. Exp Gerontol 11, 11–16. Robinson, A. B., Willoughby, R., Robinson, L. R. (1976) Age dependent amines, amides, and amino acid residues in Drosophila melanogaster. Exp Gerontol 11, 113–120. Robinson, A. B., Pauling, L. C., Aberth, W. (1977) A controversy: diagnosis of infectious hepatitis. Clin Chem 23, 908–910. Robinson, A. B. (1979) Molecular clocks, molecular profiles, and optimum diets: three approaches to the problem of aging. Mech Ageing Dev 9, 225–236. Robinson, A. B., Robinson, L. R. (1991) Quantitative measurement of human physiological age by profiling of body fluids and pattern recognition. Mech Ageing Dev 59, 47–67. Robinson, A. B. (2007) Revolutionizing 21st century medicine with consumer-based
26. 27.
28.
29. 30.
23
diagnostics and the internet. J Am Phys Surg 12(1), 14–21. Robinson, A. B. (2007) Human health in the Telecosm. Gilder/Forbes Telecosm 10. Available at www.oism.org/health Robinson, N. E., Robinson, A. B. (2001) Prediction of protein deamidation rates from primary structure and three-dimensional structure. Proc Natl Acad Sci USA 98, 4367–4372. Robinson, N. E., Robinson, A. B. (2001) Deamidation of human proteins. Proc Natl Acad Sci USA 98, 12409–12413. Robinson, N. E. (2002) Protein deamidation. Proc Natl Acad Sci USA 99, 5283–5288. Robinson, A. B., Hunsberger, A., Westall., F. C. (1994) Suppression of squamous cell carcinoma in hairless mice by dietary nutrient variation. Mech Ageing Dev 76, 210–214.
Chapter 2 Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism Monique Piraud, Séverine Ruet, Sylvie Boyer, Cécile Acquaviva, Pascale Clerc-Renaud, David Cheillan, and Christine Vianey-Saban Abstract The diagnosis of inherited metabolic disorders of amino acid (AA) metabolism is based on the qualitative and/or the quantitative analysis of AAs, mainly in blood and urine. For years, the most widespread technique in use was ion-exchange chromatography followed by post-column derivatization with ninhydrin, a method which is the basis of numerous automated AA analyzers with a throughput of about eight samples/day. The emergence of tandem mass spectrometry (MS/MS) coupled to liquid chromatography (LC) has made possible the measurement of many metabolites for the diagnosis of inborn errors of metabolism. The LC-MS/MS method described here allows the clinical diagnosis of AA disorders by analysis of underivatized AAs and derivative molecules in various biological samples prepared by methanol precipitation. AAs are separated by ion-pairing reversed-phase LC, using perfluorocarboxylic acid as an ion-pairing agent. Each AA is detected in MS/MS-positive ionization mode by its specific transition. The method allows the analysis of about 40 biological samples/day. Key words: Amino acids, inborn errors of metabolism, tandem mass spectrometry, liquid chromatography.
1. Introduction The diagnosis of inherited metabolic disorders of amino acid (AA) metabolism is based on the qualitative and/or the quantitative investigation of about 80 AAs or derivative molecules in biological fluids, mainly blood and urine, but also cerebrospinal fluid (CSF), leukocytes, and amniotic fluid (AF). For years, the most T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_2, © Springer Science+Business Media, LLC 2011
25
26
Piraud et al.
widespread technique for this purpose was ion-exchange chromatography followed by post-column derivatization with ninhydrin, a method which is the basis of numerous automated AA analyzers with a throughput of about eight samples/day (1). Tandem mass spectrometry (MS/MS) coupled to liquid chromatography (LC) is still a developing technology capable of replacing classical methods in clinical biology (2) for the diagnosis of inborn errors of metabolism (3–5). Indeed, LC-MS/MS has been applied to newborn screening and other numerous metabolic conditions (6). An LC-MS/MS method has been developed allowing for the analysis of underivatized AAs in various biological samples for clinical diagnosis (7–9). The samples (plasma, urine, CSF, leukocytes, AF) are prepared by methanol precipitation. AAs are separated by ion-pairing reversed-phase liquid chromatography, using perfluorocarboxylic acid as an ion-pairing agent. Each AA is detected in MS/MS-positive ionization mode by its specific transition. The method allows a much higher throughput (about 40 biological samples/day) compared to traditional methods, and the technique can be adapted to other kinds of samples.
2. Materials Reagents used for daily analysis of AAs are underlined. All of them are made ready for analysis. 2.1. Common Reagents
1. Acetonitrile. 2. HPLC-grade methanol. 3. Sterile deionized water.
2.2. LC
1. Solvent A: 0.5 mM Tridecafluoroheptanoic acid (TDFHA). Store at room temperature. For preparation of the stock solution, see Note 1. 2. Solvent B: Acetonitrile. Store at room temperature. 3. Octadecyl-bonded silica gel LC column: Modulo-cart QS Uptisphere, 120 Å, 3 µm BP2, 50 mm × 2 mm (Interchrom; Interchim, Montluçon, France) (see Note 2). 4. Pre-columns of the same stationary phase (Interchim): 10 mm × 2 mm. 5. On-line filters (Interchim): 2 µm.
6. Direct connectors for 10-mm pre-column (Interchim). 7. Two Series 200 micropumps (Perkin-Elmer, Norwalk, CT, USA) (see Note 3).
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
27
8. Series 200 autosampler (Perkin–Elmer) (see Note 3). 9. The temperature of the LC column is controlled with a Croco-Cil (Cluzeau Info Labo, Sainte-Foy-la-Grande, France). 2.3. MS/MS
1. API 2000 triple-quadrupole mass spectrometer (AB Sciex, Toronto, Canada) equipped with a TurboIonSpray source. 2. Balston generator (Parker Hannifin International Ltd, Milan, Italy) to produce nitrogen from compressed (8 bars) air for desolvation and collision. 3. Analyst 1.3.1 software (AB Sciex).
2.4. Amino Acids and AA Qualitative Standards
Beside the individual stock solution for each molecule, a solution containing 26 common AAs is prepared for daily checking of the chromatographic separation. 1. Investigated AAs and other derivative molecules and their manufacturers are listed in Table 2.1, with abbreviations used. 2. Stock solutions of each molecule (5 mM in water (see Note 4)). Store in aliquots at –20◦ C for a maximum of 1 year. 3. L-Cysteine-L-homocysteine disulfide (Cys-Hcy) is not commercially available (see Note 5). 4. Anhydrides of L-argininosuccinic acid (ASA anhydrides) are not commercially available (see Note 6). 5. 50 µM stock solution (26 AAs): Combine 1 mL of each 5 mM AA solution (see list in Table 2.2) and complete to 100 mL with water. Store in 1 mL aliquots at –20◦ C for a maximum of 1 year. 6. 25 µM solution (26 AAs): Dilute 1 mL of the 50 µM stock solution with 1 mL of 1.25 mM TDFHA in a 2-mL injection vial. This solution should be prepared every other week. 7. S-2-Aminoethylcysteine (S2AE, see Note 7) 1 mM stock solution in methanol: prepare in a glass flask with grinded cap by dissolving 20.1 mg of S2AE in 100 mL methanol. Store in a glass flask with grinded cap tightly closed at 4◦ C (maximum 6 months, see Note 8). Do not let the flask open longer than necessary when preparing the deproteinization reagent (see Note 8).
28
Piraud et al.
Table 2.1 List of amino acids and derivative molecules of biological interest for the diagnosis of inherited disorders of amino acids metabolism Name
Abbreviation MM
Manufacturer Biological interest
Sample
Ethanolamine, HCl
EA
61
Calbiochem
Hypoxia
P,U
Glycine
Gly
75
Merck
Nonketotic hyperglycinemia, organic acidurias
P,U,CSF
L -Alanine
Ala
89
Calbiochem
Hyperlactacidemia
P,U
Sarcosine
Sar
89
Sigma
MAD (ETF or ETF-QO deficiency)
P,U
β-Alanine
β-Ala
89
Merck
Hyper-β-alaninemia
U
L -α-Amino-n-
Abu
103
Calbiochem
Protein intake
P
N,NDimethylglycine
(Me)2 -Gly
103
Sigma
MAD (ETF or ETF-QO deficiency)
P,U
β-Aminoisobutyric acid
β-AIB
103
Sigma
Neoplasia, excessive breakdown
γ-Aminobutyric acid L -Serine
GABA
103
Calbiochem
Ser
105
Calbiochem
CNS disorders, neurotrans- P,U,CSF mitter disorders Vitamin B6 deficiency, B6 P,U, CSF synthesis deficiency, serine deficiency disorders
1 -Pyrroline5-carboxylic acid
PC
113
Sigma
Hyperprolinemia (type II)
L -Proline
Pro
115
Calbiochem
Hyperprolinemia (types I, II) P,U
butyric acid
tissue U
U
L -Valine
Val
117
Merck
Maple syrup urine disease
L -Homoserine
Hse
119
Sigma
Interference (neuroblastoma) P, U
L -Threonine
Thr
119
Calbiochem
Hepatic failure
Taurine
Tau
125
Sigma
Sulfite oxidase deficiency
P,U
Pip
129
Sigma
Peroxisomal disorders
P,U
Pyro-L-glutamic acid
pGlu
129
Fluka
Pyroglutamic aciduria
P,U
ε-Aminocaproic acid
eCap
131
Calbiochem
Interference (therapeutic agent)
δ-Aminolevulinic acid
dALA
131
Sigma
Tyrosinemia type I, porphyrias
U
4-Hydroxy-Lproline
Hyp
131
Calbiochem
Hydroxyprolinemia
P,U
L -Leucine
Leu
131
Calbiochem
Maple syrup urine disease
P,U
L -allo-Isoleucine
aIle
131
Sigma
Maple syrup urine disease
P,U
L -Pipecolic
acid
P,U P,U
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
Table 2.1 (continued) Name
Abbreviation MM
Manufacturer Biological interest
Sample
L -Isoleucine
Ile
131
Calbiochem
Maple syrup urine disease
P,U
Orn
132
Merck
Urea cycle disorders, HHH, P,U hyperornithinemia
L -Asparagine
Asn
132
Fluka
Asparaginase treatment (leukemia)
P,U
L -Aspartic
Asp
133
Merck
Asparaginase treatment (leukemia)
P,U
L -Homocysteine
Hcy
135
Sigma
Homocystinuria, Cbl C/D P,U deficiency
L -Glutamine
Gln
146
Merck
Hyperammonemia, urea cycle disorders
L -Lysine,
Lys
146
Calbiochem
Hyperlysinemia, cystinuria– P,U lysinuria
Glu
147
Calbiochem
Urea cycle disorders
P,U
L -Methionine
Met
149
Calbiochem
Homocystinuria, hypermethioninemia
P,U
L -Histidine,
HCl
His
155
Calbiochem
Histidinemia
P,U
L -α-Aminoadipic
Aad
161
Sigma
α-Aminoadipic aciduria
P,U
5-Hydroxy-L-lysine, HCl
Hyl
162
Sigma
Hydroxylysinuria
U
N-Acetyl-L-cysteine
NAcCys
163
Merck
Interference (therapeutic agent)
P,U
L -Phenylalanine
Phe
165
Merck
Phenylketonuria, hyperphenylalaninemia
P,U
N3 -Methyl-Lhistidine
His-1Me
169
Calbiochem
Renal insufficiency
P,U
N1 -Methyl-Lhistidine
His-3Me
169
Calbiochem
Renal insufficiency, denutri- P,U tion, starvation
Glycyl-L-proline
Gly-Pro
171
Sigma
Iminodipeptiduria
L -Ornithine,
HCl
acid
HCl
L -Glutamic
acid
P,U
acid
U
L -Arginine
Arg
174
Merck
Urea cycle disorders
P,U
Formimino-Lglutamic acid
FIGLU
174
Sigma
FIGLU aciduria
U
N-Acetyl-Lornithine
NAcOrn
174
Sigma
Interference (hyperornithinemia)
L -Citrulline
Cit
175
Merck
Urea cycle disorders
P,U
L -Tyrosine◦
Tyr
181
Calbiochem
Tyrosinemia type I or II
P,U
L -Homocitrulline
Hci
189
ICN
HHH
P,U
L -3-(3,4-
Dopa
197
Sigma
Tyrosinemia, neurotrans- U mitter disorders
Dihydroxyphenyl)alanine◦
29
30
Piraud et al.
Table 2.1 (continued) Name
Abbreviation MM
Manufacturer Biological interest
Sample
S-Sulfo-L-cysteine
S-Cys
201
Sigma
Sulfite oxidase deficiency
P,U
N G ,N G Dimethylarginine, HCl
asym(Me)2 Arg
202
Sigma
Physiological
U
L -Tryptophan◦
Trp
204
Calbiochem
Hartnup disease
P,U
L -Kynurenine
Kyn
208
Sigma
Vitamin B6 deficiency
U
5-Hydroxy-Ltryptophan
Hyt
220
Sigma
Therapeutic agent, neuro- U transmitter disorders
L -Cystathionine◦
Hcy(Ala)
222
Sigma
Cystathioninuria, neuroblastoma
P,U
3-Hydroxy-Lkynurenine
Hyk
224
Sigma
Vitamin B6 deficiency
U
L -Carnosine
Car
226
Calbiochem
Carnosinemia, physiological U (meat)
L -Anserine
Ans
240
Sigma
Carnosinemia, physiological U (meat)
L -Cystine
(Cys)2
240
Merck
Cystinuria, cystinuria– P,U lysinuria, homocystinuria, sulfite oxidase deficiency
L -Cysteine- L -
Cys-Hcy
254
∗
Cystinosis Homocystinuria, cystinuria
L P,U
(Hcy)2
268
Sigma
Homocystinuria
P,U
ASA anh
272
∗
Urea cycle disorders (AS aciduria)
P,U
L -Saccharopine
Sac
276
Sigma
Saccharopinuria
P,U
L -Argininosuccinic
ASA
290
Sigma
Urea cycle disorders (AS aciduria)
P,U
Glutathione (reduced form)
GSH
307
Sigma
γGT deficiency, GSH synthase deficiency
P,U
β-Aspartylglucosamine
GlcNAcAsn
335
Sigma
Aspartylglucosaminuria
U
homocysteine disulfide L -Homocystine L -Argininosuccinic
anhydrides
acid
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
31
Table 2.1 (continued) Name
Abbreviation MM
Manufacturer Biological interest
Sample
S-Adenosyl-Lhomocysteine
Ad-Hcy
384
Sigma
Homocystinuria, hyperme- P,U thioninemia
S-Adenosyl-Lmethionine
Ad-Met
398
Sigma
Homocystinuria, hyperme- P,U thioninemia
Glutathione (oxidized form)
GSSG
612
Sigma
γGT deficiency, GSH syn- P,U thase deficiency
∗ : for preparation, see text. ◦ : molecules for which 2 drops of 37% HCl are added per 10 mL of 5 mM stock solution
for dissolution. MM, molecular mass; MAD, multiple acyl-CoA dehydrogenase deficiency; ETF, electron transfer flavoprotein; ETF-QO, ETF-ubiquinone oxidoreductase; CNS, central nervous system; Cbl, cobalamine; HHH, hyperammonemia, hyperornithinemia, homocitrullinuria; P, plasma; U, urine; CSF, cerebrospinal fluid; L, leukocytes. Reproduced with permission from ref. (7).
2.5. Stable Isotope-Labeled AAs (Internal Standards, AAs∗ )
1. Stable isotope-labeled standards for MS/MS are listed in Table 2.3, along with their abbreviations (see Note 9). 2. Stock solutions of each molecule (25 mM for Gly∗ , 15 mM for Gln∗ , 10 mM for Ala∗ and Pro∗ ; 5 mM for others) in water (see Note 4). Store in aliquots at –20◦ C for a maximum of 1 year. 3. Working solutions: 17 AAs∗ at 200 µM (except Pro∗ and Ala∗ at 400 µM, Gln∗ at 600 µM, and Gly∗ at 1 mM). In a 50-mL flask, add 2 mL of each AA∗ stock solution. Complete to 50 mL with water, mix, and store in aliquots at – 20◦ C for a maximum of 1 year.
2.6. AA Quantification Standards
Two quantitative standards are used daily for quantification: one is prepared from a commercially available AA standard supplemented in order to obtain a standard with all the most commonly measured physiological AAs for the diagnosis of inborn errors of metabolism (100 and 50 µM AAs standards) and one is homemade and contains all the molecules to be quantified that are not present in the other standard (supplemental AAs standard). 1. “Amino acid standards, physiological acidics, neutrals and basics for calibrating amino acid analyzers” (Sigma-Aldrich) containing several AAs at 500 µM. 2. AA additional stock solution: Gln 6.25 mM, Gly 5 mM, Asn 1.25 mM. Store in single-use aliquots at –20◦ C for a maximum of 6 months.
β-Ala
Abu
β-Alanine
L -α-Amino-n-
20 20 20 20
104 > 58
104 > 86
104 > 86
104 > 87
(Me)2 Gly
β-AIB
GABA
β-Aminoisobutyric acid (1)
γ-Aminobutyric acid
20
20
90 > 72
104 > 58
20
20
10
20
DP
90 > 44
90 > 44
76 > 30
62 > 44
Monitored transition
N,NDimethylglycine
butyric acid
Ala
Sar
L -Alanine
Gly
Glycine
Sarcosine
EA
Ethanolamine
Abbreviation
14
14
14
18
18
12
20
20
16
16
CE
x
9.5 9.5
Lys∗ Lys∗
3.7
Pro∗
1.3
8.0
Val∗
Gly∗
1.6
Ala∗ x
1.8
x
Ala∗ 2.4
8.8
Expected RT
Gly∗
26 AAs
Leu∗
AA∗
100
100
100
100
100
100
500
100
200
200
Conc. in Conc. in 100 Supplemental µM AA stan- AA standard dard (in µM) (in µM)
3
3
3
1
1–2
2–3
1
1
1
3
Monitoring period
Table 2.2 List of unlabeled AAs and derivative molecules of biological interest for the diagnosis of inherited disorders of amino acid metabolism. Corresponding specific transitions used for the identification of AAs in MS/MS positive ionization mode are given with their specific declustering potential (DP, in V) and collision energy (CE, in eV). Concentrations (in µM) of analytes in the 100 µM AA standard solution and in the supplemental standard solution are indicated, as well as AAs present in the 25 µM solution (26 AAs). Transitions are monitored either in period 1 (0–3.5 min of the chromatogram), in period 2 (3.5–9 min), and/or in period 3 (9–20 min). #: transition 132 > 86 is common to Hyp, Leu, Ile, aIle and dALA; transition 175 > 130 is common to Arg and FIGLU; RT: retention time; In italics: molecules for which quantitation could not be validated. Reproduced with permission from ref. (9)
32 Piraud et al.
114 > 68
PC
Pro
Val
Hse
1-Pyrroline-5carboxylic acid
L -Proline
L -Valine
L -Homoserine
130 > 84 130 > 84
132 > 68 132 > 43
Pip
pGlu
eCap
dALA
Hyp
Leu
aIle
Ile
Pyroglutamic acid
ε-Aminocaproic acid
δ-Aminolevulinic acid
4-Hydroxy-L-proline
L -Leucine
L -allo-isoleucine
L -Isoleucine
acid
L -Pipecolic
120 > 74
20
30
36
16
132 > 86# (2) 25
26 26
25
30
16
16
24
20
20
16
32
32
32
16
20
20
14
CE
25
132 > 69
132 > 69
30
20
20
132 > 114 132 > 114
30
30
30
40
132 > 79
126 > 108
Thr
Tau
L -Threonine
20
120 > 74
Taurine
20
25
20
25
10
DP
120 > 44
118 > 72
116 > 70
106 > 60
Monitored transition
Ser
Abbreviation
L -Serine
Table 2.2 (continued)
1.1
Glu∗
x x x
Leu∗ Leu∗
9.8
9.2
10.4
1.3
x
Asp∗ Leu∗
9.6
Leu∗
11.2
5.2
Val∗
1.9 0.77
x
1.9
Ala∗
Ala∗
x
Val∗ 6.0
2.3
x
Pro∗
1.6 2.0
x
Ser∗
Expected RT
Val∗
26 AAs
AA∗
100
100
100
100
100
100
100
100
200
200
200
200
Conc. in Conc. in 100 Supplemental µM AA stan- AA standard dard (in µM) (in µM)
2–3
2–3
3
1
3
3
1
2
1
1
1
1
2
1
1
1
Monitoring period
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism 33
Asn
Asp
L -Asparagine
L -Aspartic
Gln
Lys
Glu
L -Glutamine
L -Lysine
L -Glutamic
His
Aad
L -Histidine
L -α-Aminoadipic
Hyl
Phe
His-3Me
His-1Me
Gly-Pro
Arg
FIGLU
5-Hydroxy-L-lysine
L -Phenylalanine
N1 -Methylhistidine
N3 -Methylhistidine
Glycyl-L-proline
L -Arginine
Formimino-Lglutamic acid
acid
Met
L -Methionine
acid
Hcy
L -Homocysteine
acid
Orn
Abbreviation
L -Ornithine
Table 2.2 (continued)
22
175 > 130# 20
32
16
20
20
20
18
24
18
14
32
20
15
30
30
30
5
20
20
10
24
36
24
24
16
18
18
22
CE
25
175 > 84
175 > 70
173 > 116
170 > 126
170 > 124
166 > 120
163 > 128
162 > 98
156 > 110
150 > 104
15
20
147 > 67
148 > 84
20
20
10
25
25
10
DP
147 > 84
147 > 84
136 > 90
134 > 74
133 > 74
133 > 70
Monitored transition
x
Lys∗
x
Glu∗
3.5
9.7 12.7
x
12.0 11.9
Arg∗
x
Leu∗
Lys∗
Lys∗
12.4 11.2
Phe∗
3.9
12.0
6.5
2.0
12.6
1.74
4.3
1.3
1.5
12.1
Expected RT
Lys∗ x
x
Met∗ Pro∗
x
Glu∗
x
x
Asp∗
Lys∗
x
Asp∗
x
x
Orn∗
Gln∗
26 AAs
AA∗
100
100
100
100
100
100
100
100
100
500
100
100
100
200
200
Conc. in Conc. in 100 Supplemental µM AA stan- AA standard dard (in µM) (in µM)
2
2
3
2–3
3
3
3
3
1–2
3
2
1
3
3
1
2
1
1
3
Monitoring period
34 Piraud et al.
Tyr
Hci
Dopa
L -Tyrosine
L -Homocitrulline
L -3-(3,4-
asym(Me)2 Arg 203 > 70
Trp
Kyn
Hyt
Hcy(Ala)
Hyk
Car
Ans
NG ,NG Dimethylarginine
L -Tryptophan
L -Kynurenine
5-Hydroxy-Ltryptophan
L -Cystathionine
3-Hydroxy-Lkynurenine
L -Carnosine
L -Anserine
241 > 109
227 > 110
225 > 208
223 > 88
221 > 204
209 > 192
205 > 188
S-Cys
202 > 120
198 > 152
190 > 173
182 > 165
176 > 159
Monitored transition
S-Sulfo-l-cysteine
Dihydroxyphenyl)alanine
Cit
Abbreviation
L -Citrulline
Table 2.2 (continued)
30
30
20
15
20
20
10
20
15
10
20
25
10
DP
35
30
14
40
10
14
18
42
18
20
16
12
14
CE
12.8 12.4 11.9
Lys∗ Lys∗ Lys∗
3.5 10.0 14.0 13.8
Ala∗ Lys∗ Lys∗ Lys∗
10.4
0.6
Ala∗
7.0
8.5 4.5
x
Tyr∗
2.6
Met∗
x
Ala∗ , Pro∗
Expected RT
Val∗
26 AAs
AA∗
100
100
100
100
100
200
200
200
200
200
200
200
Conc. in Conc. in 100 Supplemental µM AA stan- AA standard dard (in µM) (in µM)
3
3
3
1–2
3
3
3
3
1
2
2
2–3
1–2
Monitoring period
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism 35
acid
336 > 126
613 > 355 165 > 120
GlcNAc-Asn
Ad-Hcy
Ad-Met
GSSG
S2AE
β-Aspartylglucosamine
S-Adenosyl-lhomocysteine
S-Adenosyl-Lmethionine
Glutathione (oxidized form)
S-2-Aminoethylcysteine
8
20
15
30
15
25
30
25
20
10
25 100
DP
18
30
20
64
24
18
48
36
55
12
40 18
CE
1.6 12.7 15.3 8.7 11.1
Lys∗ Val∗ Lys∗
4.1
Lys∗
11.5
Val∗ Asp∗
5.3
Lys∗
12.4
Val∗
11.8
2.1 5.0
Lys∗
x
(Cys)2 ∗ (Cys)2 ∗
Expected RT
(Hcy)2 ∗
26 AAs
AA∗
100
100
200
200
200
200
Conc. in Conc. in 100 Supplemental µM AA stan- AA standard dard (in µM) (in µM)
(1) BAIB is not measurable when GABA is detectable. (2) Common transition to Leu, Ile, aIle, dALA, and Hyp. It cannot be used for Ile quantification when dALA is present.
399 > 250
385 > 88
308 > 179
291 > 70
GSH
ASA
277 > 84
273 > 70
269 > 136
241 > 74 255 > 134
Monitored transition
Glutathione (reduced form)
L -Argininosuccinic
L -Saccharopine
Sac
ASA anh
L -Argininosuccinic
anhydrides
(Hcy)2
(Cys)2 Cys-Hcy
Abbreviation
L -Homocystine
homocysteine disulfide
L -Cysteine- L -
L -Cystine
Table 2.2 (continued)
2
3
3
1
2
3
2
3
3
1 2
Monitoring period
36 Piraud et al.
20
(5,5-D2 ), 2HCl
acid (2,4,4-D3 )
177 > 70 184 > 167 245 > 74 277 > 140
Arg∗ Tyr∗ (Cys)2 ∗ (Hcy)2 ∗
(guanido-15 N2 ), HCl
(ring 3,5-D2 )
(3,3,3’,3’-D4 )
L -Tyrosine
DL -Cystine
10
25
25
20
30
10
20
20
15
25
10
12
40
12
32
20
14
24
24
24
18
22
36
16
16
20
14
20
16
CE
CIL
CIL
CIL
CIL
CIL
Euriso-top
CIL
Euriso-top
CIL
Euriso-top
CIL
Euriso-top
CIL
CIL
CIL
CIL
CIL
Manufacturer
5
5
5
5
5
5
15
5
5
5
5
5
5
5
10
5
10
25
Stock solution (mM)
Corresponding specific transitions used for their identification in MS/MS positive ionization mode are given with their specific declustering potential (DP, in V) and collision energy (CE, in eV). CIL, Cambridge Isotope Laboratories. Reproduced with permission from ref. (8).
DL -Homocystine
(3,3,3’,3’,4,4,4’,4’-D8 )
(methyl-D3 )
L -Phenylalanine
L -Methionine
L -Arginine
152 > 88
Gln∗ 171 > 125
151 > 88
Lys∗ 153 > 107
151 > 87
Glu∗
Phe∗
137 > 75
(ring-D5 )
135 > 72
Met∗
(2,3,3,4,4-D5 )
(4,4,5,5-D4 ), 2HCl
L -Glutamine
DL -Lysine
DL -Glutamic
acid (2,3,3-D3 )
L -Ornithine
L -Aspartic
Asp∗
30
135 > 46 Orn∗
25
25
126 > 80 135 > 89
Leu∗
(5,5,5-D3 )
20
L -Leucine
(2,3,3,4,4,5,5-D7 )
DL -Valine-D8
DL -Proline
123 > 77
10
Val∗
94 > 48 109 > 63
Ser∗
10
DP
Pro∗
(2,3,3- D3 )
(2,3,3,3-D4 )
DL -Alanine
DL -Serine
78 > 32
Gly∗
Glycine (2,2-D2 ) Ala∗
Monitored transition
Abbreviation
Table 2.3 List of the 17 stable isotope-labeled standards (AA∗ ) Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism 37
38
Piraud et al.
3. AA diluted solution: Gln 625 µM, Gly 500 µM, Asn 125 µM. Add 1 mL AA additional stock solution to 9 mL of water. Use this solution within the day of preparation. 4. 100 µM AA standard (see list in Table 2.2): all AAs at 100 µM (except Gly and Gln at 500 µM). Add 300 µL of the Sigma standard (“amino acid standards, physiological acidics, neutrals, and basics for calibrating amino acid analyzers”) to 1200 µL of AA diluted solution. Store in aliquots at –20◦ C for a maximum of 1 month. When thawed, store at 4◦ C for a maximum of 1 week. 5. 50 µM AA standard: Dilute the 100 µM AA solution. Prepare weekly and use for a maximum of 1 week with storage at 4◦ C. 6. Supplemental AA standard (see list in Table 2.2): several molecules at 200 µM, which are not present in the 100 µM standard. Combine 1 mL of each 5 mM stock solution and complete to 25 mL. Store in 1 mL aliquots at –20◦ C for a maximum of 2 years. When thawed, store at 4◦ C for a maximum of 1 week. 7. 0 µM standard is a water blank. 2.7. Materials for Preparation of Standards and Samples
1. 2-mL injection vials with screw-top caps (Interchim). 2. 9-mm silicone/PTFE slit screw caps (Interchim). 3. Microtest tube rack (Brand, Wertheim, Germany). 4. Microtubes PP 1.5 mL with attached PP safety cap (Sarstedt France, Marnay, France). 5. Deproteinization reagent: 20 µM S2AE in methanol (see Note 7). Dilute the S2AE stock solution 1:49 with methanol in a 50-mL glass flask with grinded cap. Store tightly closed at 4◦ C for a maximum of 1 month. This reagent can be aliquoted to small glass flasks with grinded caps. Open only when used and close immediately after use (see Note 8). 6. 1.25 mM TDFHA working solution (for preparation of the stock solution, see Note 1) 7. Plasma: Sample blood with heparin, then centrifuge 10 min at 5000×g at 4◦ C. Transfer plasma (200 µL minimum) in a tube. Deproteinize immediately or store at –80◦ C. 8. Urine: Determine creatinine concentration for each urine sample with any classical creatinine determination method for this purpose (e.g., Jaffé reaction) and freeze one aliquot at –20◦ C for AA determination (1 mL minimum). 9. Cerebrospinal fluid (CSF) and amniotic fluid (AF): Freeze the sample (300 µL minimum) at –20◦ C immediately.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
39
10. Leukocytes: Isolate the leukocytes pellet from 15–20 mL of blood as soon as possible (maximum 48 h) after sampling with citric acid–citrate–dextrose as anticlotting agent, according to usual procedures currently in use in the laboratory for leukocytes isolation. For preparation of the leukocyte mixture and analysis, see Note 10. 2.8. Quality Control Samples
1. External QAP (Quality Assurance Program): ERNDIM amino acid scheme (www.erndimqa.nl). 2. External QAP: ERNDIM cystine in white blood cell scheme (www.erndimqa.nl). 3. Internal QAP: SKML control material amino acids (www.erndimqa.nl).
3. Methods Samples for AA analysis are first deproteinized with methanol containing a known quantity of S2AE for control of the correct dilution/precipitation of the sample. Ion-pairing reversed-phase liquid chromatography is used for separation of AAs, and the use of volatile mobile phases (ACN, TDFHA, water) allows MS/MS detection, particularly with the TurboIonSpray source adapted to evaporation of water containing mobile phases. This chromatographic system must be used in stable conditions as described (with respect to rinsing periods) in order to obtain repeatability of the amino acid retention times and of the quantification results. As necessary for MS/MS quantitative analysis, 17 stable isotope-labeled AAs are used as IS. Quantification of the 17 corresponding AAs has been validated in these conditions. Validation of quantification has been obtained for many other AAs but not all. Several AAs for which quantitative validation has not been obtained are described anyway as qualitative markers useful for the diagnosis of some inherited disorders of AA metabolism. 3.1. Sample Preparation
1. Push 1.5-mL microtubes firmly onto the microtube rack. 2. AA standards (0, 50, and 100 µM, AA Suppl. STD 200 µM), plasma, CSF, AF, and quality control samples are used undiluted. For leukocytes, see Note 10. 3. Do not dilute urine samples if the creatinine concentration is below 4 mM. If the creatinine concentration exceeds 4 mM, then dilute urine with water as follows: a. 4–8 mM creatinine, dilute 1:1. b. 8–12 mM creatinine, dilute 1:2.
40
Piraud et al.
c. 12–16 mM creatinine, dilute 1:3. d. 16–20 mM creatinine, dilute 1:4. e. ≥20 mM creatinine, dilute 1:9.
4. Mix 150 µL of each sample (standards, plasma, CSF, AF, controls or adequately diluted urine) with 600 µL of deproteinization reagent, close the cap, and mix thoroughly for 2 min. 5. Store the samples at room temperature for 5 min. 6. Centrifuge the samples at 17,500×g for 5 min at +4◦ C, then transfer the supernatants to new 1.5-mL microtubes, and store at –80◦ C until used. 7. Prior to analysis, thaw and homogenize the deproteinized samples and standards. 8. Combine 200 µL of each supernatant with 40 µL of AA∗ working solution and 160 µL of 1.25 mM TDFHA in an injection vial and mix. 3.2. Liquid Chromatography
1. Assemble the column and the pre-column using the connectors and add a pre-filter. 2. Maintain the column at 26◦ C. 3. Before the first use, rinse the column with ACN at 200 µL/min for 2 h. 4. Separations are carried out at a flow rate of 200 µL/min. 5. Inject 5 µL of sample or standard. 6. Gradient elution is as follows: from 0 to 15% B in 1 min; maintain 15% B for 5 min; from 15 to 25% B in 3 min; maintain 25% B for 6 min; from 25 to 0% B in 1 min; maintain 0% B for 15 min to re-equilibrate the column before a new injection. 7. The sample throughput of the method is 2 samples/h. The LC effluent is directed to the mass spectrometer for the first 20 min of the separation, after which it can be directed to waste.
3.3. Tandem Mass Spectrometry
1. Set the ionization source to positive ionization mode. 2. Set the TurboIonSpray source to 450◦ C. 3. Set the curtain gas (CUR) to 20. 4. Set the collision gas to 2. 5. Set the ion source nebulization gas (GS1) to 25 (arbitrary units are given by the manufacturer). 6. Set the auxiliary gas (GS2) to 40. 7. Set Q1 and Q3 to unit resolution. 8. Set the ion spray voltage (IS) to 5000 V.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
41
9. Set the focusing potential (FP) to 200 V. 10. Set the entrance potential (EP) to –10 V. 11. Specific transitions optimized for each unlabeled (Table 2.2) or labeled (Table 2.3) compound are listed (see Note 11) with their specific declustering potential (DP, in V) and collision energy (CE, in eV) parameters (see Note 12). 12. For each sample, acquisition of data is performed during the first 20 min of the gradient analysis, according to the chosen transitions in positive ionization mode and specific parameters (Tables 2.2 and 2.3). During the re-equilibration time, the flow can be diverted to waste. 13. The predicted retention time for each molecule is indicated (see Table 2.2). 14. For qualitative analysis (25 µM solution (26 AAs)), the acquisition method contains all the involved transitions in a single period with a dwell time of 25 ms each. 15. For quantification using the Analyst 1.3.1 software, the acquisition time is divided into three periods (period 1 is from 0 to 3.5 min; period 2 is from 3.5 to 9 min; period 3 is from 9 to 20 min), during which only the relevant molecules are managed with their own transition (for the corresponding period, see Table 2.2). For each transition, the dwell time is 25 ms, except for Gly and Gly∗ (100 ms). 16. Some molecules are monitored over two periods in case there is a slight shift in RT (see Table 2.2). 3.4. LC-MS/MS Analysis
1. After rinsing the LC column, inject samples in the following order (see Note 13): a. One 0.5 mM TDFHA sample (not usable). b. One 25 µM solution (26 AAs) (for checking the quality of the analysis). c. 12–14 standards or samples (for quantification). d. Rinse the column on line 1 h with ACN at 200 µL/min. 2. The same sequence (from a to d) can be repeated two more times. Modifying or avoiding the rinsing periods may alter the chromatogram and cause shift in retention times. 3. The 0 and 100 µM AA standards are analyzed in each sequence. 4. The 50 µM and supplemental AA standards are analyzed at least two times every 24 h. 5. After each use (maximum 24 h, three sequences, 36–42 samples), rinse the column (out-line, reverse flow) with methanol/water (85:15) for 16 h at 100 µL/min.
42
Piraud et al.
Fig. 2.1. Overall chromatogram obtained with the 26 AA sample, with an MS/MS method integrating the transitions of each molecule during all the analysis time (20 min). Column: Uptisphere 50 mm × 2 mm ID 3 µm. Elution gradient: solvent A: 0.5 mM TDFHA in water and solvent B: acetonitrile; from 0 to 15% B in 1 min, 1–6 min 15% B maintained, 6–9 min linear gradient to 25% B, 9–15 min 25% B maintained, 15–16 min gradient back to 0% B, then 16–31 min 0% B to equilibrate the column before a new injection. Injection volume, 5 µL; concentration of molecules in mixtures, 25 µM each.
6. Check the quality of the chromatography based on the 25 µM solution (26 AAs) (see Fig. 2.1). 7. Identify each AA based on its transition and its retention time. 8. The overall aspect of the chromatogram and the elution order of the AA must remain the same. The observed retention time for each AA must correspond to the expected value (Table 2.2 and Fig. 2.2; see Note 14). 9. Visually inspect the separation of the critical pairs of AAs (see Fig. 2.2), i.e., mainly Glu and Gln (see Fig. 2.2c), Asp and Asn (see Fig. 2.2d), aIle, Ile, and Leu (see Fig. 2.2g; see Note 15). 10. Verify the sensitivity of the system daily by comparing peak intensities of AAs with those of the previous days.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
(a)
Fig. 2.2. (continued)
(b)
Fig. 2.2. (continued)
43
44
Piraud et al.
(c)
Fig. 2.2. (continued)
(d)
Fig. 2.2. (continued)
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
(e)
Fig. 2.2. (continued)
(f)
Fig. 2.2. (continued)
45
46
Piraud et al.
(g)
Fig. 2.2. Extracted ion current (XIC) chromatograms of the amino acid pairs that have similar MS/MS characteristics. Column: Uptisphere 50 mm × 2 mm ID, 3 µm. Elution gradient: solvent A: 0.5 mM TDFHA in water and solvent B: acetonitrile; from 0 to 15% B in 1 min, 1–6 min 15% B maintained, 6–9 min linear gradient to 25% B, 9–15 min 25% B maintained, 15–16 min gradient back to 0% B, then 16–31 min 0% B to equilibrate the column before a new injection. Injection volume, 5 µL; concentration of molecules in mixtures, 25 µM each. (a) Ala, Sar (90 > 44); N-acetyl-L-ornithine (NAcOrn), Arg (175 > 70). (b) Abu, (Me)2 Gly (104 > 58). (c) Gln (147 > 84), Glu, Lys (148 > 84). (d) Asn (133 > 74), Asp (134 > 74). (e) Pip, pGlu (130 > 84). (f) Cys (122 > 76), Hcy (136 > 90), (Cys)2 (241 > 74), (Hcy)2 (269 > 136), Cys-Hcy (255 > 134). (g) Isobar molecules giving a parent ion at m/z 132: Leu, Ile, aIle, dALA, Hyp, and εCap. Reproduced with permission from ref. (8).
3.5. Quantification
1. Perform quantification of AA after acquisition of all the samples in the three sequences using Analyst 1.3.1 software. 2. Use the corresponding AAs∗ as internal standards (see Table 2.2, Note 16) (peak area ratio), and the 0 and 100 µM AA standards (or 200 µM supplemental AA standard) as external standards according to a linear calibration curve. 3. The 50 µM AA standard is used as a control sample. 4. The retention times of AAs and AAs∗ are noted, and the AA retention time/AA∗ retention time ratio (ratio of retention times) is calculated. 5. Confirm the quality of the peak integrations manually (see Note 17). 6. Quantitative validation has been obtained for most AAs measured by this method (see Note 17). However, quantita-
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
47
tive validation could not be obtained for some of them (Table 2.2). Therefore, their analysis is only qualitative (see Note 18). 7. In several cases, qualitative identification of the disease marker is sufficient for a diagnosis (e.g., ASA, pGlu, FIGLU, S-Cys, Cys-Hcy). Using the stable isotope-labeled IS analogues of the compounds when available will improve the results. 8. Direct quantification of Cys-Hcy and ASA is not possible because of the lack of quantitative standards (see Notes 19 and 20). 9. Validation of the series is achieved if quantitative results match with the calculated consensus values of the internal control SKML amino acids.
4. Notes 1. Preparation of the 13.73 mM Tridecafluoroheptanoic acid (TDFHA) stock solution: Prepare by dissolving 5 g of TDFHA in 1 L water and store at room temperature (maximum 6 months). 2. Other ultrapure octadecyl-bonded silica gel columns of the same characteristics could be used with no or slight modifications to the gradient shape, provided that the separation of critical pairs of AAs is obtained (see 7, Section 3.4, Fig. 2.2 and Notes 14 and 15). 3. Any other HPLC pumps/autosampler can be used for solvent delivery and automated sample introduction, if adapted to the given flow/sample introduction volume. 4. When necessary, stock solutions are acidified with 2 drops of 37% HCl per 10 mL, in order to enhance the solubility of the compound (see Table 2.1). 5. A qualitative standard solution containing Cys-Hcy can be obtained by heating a solution containing 0.5 M Cys and 0.5 M Hcy at 50◦ C for 2 h. 6. A 0.5 mM ASA anhydrides solution can be obtained by heating a 0.5 M argininosuccinic acid solution at 100◦ C for 60 min. 7. S2AE is used as an internal standard for the control of the correct dilution/precipitation of samples.
48
Piraud et al.
8. Methanol is an efficient reagent for protein precipitation. Its use limits ion suppression that can occur with other precipitating reagents. This solution is used as a standard for deproteinization. Methanol is highly volatile. If evaporation occurs, the concentration may be altered. If the solution stays open longer than necessary, then it will be best to discard the flask and use a new one. 9. The use of stable isotope-labeled ISs of AAs to be quantified corrects the ion suppression phenomenon occurring when several molecules co-elute, assuming that the signal modification present in each sample is the same for the analyte and its corresponding IS. Due to their cost, the number of AAs∗ has been limited to 17, and they have been chosen according to the necessity of a precise quantification of the AAs for the diagnosis of metabolic disorders. For the remaining AAs, one of the 17 nonhomologous AAs∗ has been used as a surrogate IS after careful validation of their quantification in these conditions. In some cases, the quantification could not be validated, but the method can be used for qualitative analysis of the marker. If necessary, other AAs∗ can be easily added to the method (e.g., Cit∗ ), in order to improve quantification results and/or to extend the linear dynamic range of the measurement. 10. Preparation and analysis of leukocyte samples for cystine measurements are slightly modified when compared to those for other samples, in order to enhance the sensitivity of the method for this purpose. The method is modified as follows: a. A leukocyte mixture is prepared by mixing the leukocyte pellet with 250 µL water and sonicating according to usual procedures currently in use in the laboratory for this purpose. b. Protein concentration is determined. c. The 10, 5, and 2.5 µM AA standards for cystine measurement in leukocytes are prepared by diluting adequately the 100 µM AA standard (see Step 4 of Section 2.6) in water. d. Leukocytes extracts, standards, and controls (ERNDIM external QAP for cystine in white blood cells and SKML internal QAP) are immediately deproteinized in 1.5-mL microtubes, as described in Step 4 of Section 3.1, with the exception that 200 µL of leukocyte extracts, standard (0, 2.5, 5, and 10 µM), or controls are mixed with 600 µL of deproteinization reagent.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
49
e. After deproteinization, the supernatant is kept frozen at –80◦ C. f. Samples for injection are prepared by mixing 200 µL of each supernatant with 50 µL of AA∗ working solution in an injection vial. g. 5 µL is injected. h. The injection sequence is similar to that described in Step 1 of Section 3.4, with the exception that adequate standards and control samples are used. As leukocyte samples are usually very few, these samples can be included in another sample series, provided the leukocyte standards are also included in the same series. i. Quantification (see Section 3.5) involves only the corresponding transitions for Cys and Cys∗ and can be done with a specific quantification method. 11. Transitions are pairs of [M+H]+ precursor ion > fragment ion obtained before (precursor) and after (fragment) fragmentation of the AA and specific of it, e.g., 132 > 86 corresponds to the precursor ion (m/z = 132) and fragment ion (m/z = 86) produced by leucine. For each AA, the most sensitive/specific transition has been chosen. In some cases, the monitoring of two transitions for one AA is necessary to ensure the confident identification/quantification of the AA to be measured. Every molecule that can give the same fragmentation as another is a potential interference when these two molecules are present in the same mixture. This interference phenomenon is mainly due either to isobaric molecules, arising from in-source, collision-induced fragmentation, or to natural isotopic contributions (i.e., mainly 1.1% 13 C and 4.5% 34 S). Interference occurring between molecules analyzed in this method is reported in (7), but most of them have been separated from the AA of interest in the described chromatographic system. 12. Specific transitions have been determined with the API 2000 triple-quadrupole mass spectrometer. However, these transitions can be used in any triple-quadrupole mass spectrometer. DP and CE reported here are specific to the API 2000. The use of another triple-quadrupole mass spectrometer will necessitate the preliminary study of the specific parameters of the new apparatus for each AA. Various parameters (e.g., dwell time) for the data processing should be adapted to the software. Contact the manufacturer for any questions concerning this technology transfer. 13. This method uses ion-pairing, reversed-phase liquid chromatography, a technique which must be run in strictly standardized conditions. Repeatability of the retention time is obtained only when the system is equilibrated with a few
50
Piraud et al.
gradients. Frequent rinsing is necessary in order to limit the fouling of the column (by samples and ion-pairing agent) and thus the peak drift. In a series, the shape of the gradient must not be modified nor the duration of the L re-equilibration time. The first sample injected after rinsing the column cannot be used for analysis. The second sample (25 µM solution (26 AAs) is used to check the quality of the chromatography in the sequence. No more than three sequences must be analyzed in a series. When the column has to be maintained on-line for more than 24 h, it is possible to rinse the column on-line (10 h with 100 µL/min ACN) and use it again for a 24-h period (three sequences, 36–42 samples). 14. Retention times can be slightly different if the conditions of the chromatography are slightly different than those described (HPLC pumps, column, etc.). Calculating the ratios of the retention time of the analyte to that of the IS is helpful. For example, in instances of slight shifts of the retention times after several runs or on some samples, this ratio has been found to be very stable and allows correct identification of the peaks (see Note 15). 15. Gln/Glu (see Fig. 2.2c) and Asn/Asp (see Fig. 2.2d) present interference due to the 13 C isotope contribution and must present as separated peaks. Asp and Glu may be overestimated if they are not baseline separated from Asn and Gln, respectively (mainly Glu, due to the important concentration of Gln in biological fluids). aIle, Ile and Leu share a common 132 > 86 transition and must give three clear peaks with baseline separation (see Figs. 2.1 and 2.2 g). dALA coelutes with Ile. Thus, the common 132 > 86 transition cannot be used for the measurement of Ile when dALA is detectable. GABA and β-AIB cannot be separated by this method. β-AIB is measurable only if GABA is undetectable in the sample. Hse and Thr cannot be separated by this method. Thr is measurable only if Hse is undetectable in the sample. However, in our experience, no Hse has been detected among more than 10,000 analyzed biological samples. 16. When the homologous labeled AA is not available, another AA∗ , as close as possible to the target AA in terms of retention time and structure, is used (see Table 2.2). 17. For every series, the following parameters must be systematically checked: a. Each peak of each AA must be correctly integrated by the software. If not, manually correct the integration. b. The AA retention time/AA∗ retention time ratio must be 100% for each AA for which the corresponding AA∗ is
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism
51
present. For other molecules, the AA retention time/IS retention time ratio is the same in the standards and in the samples. c. The surface of the S2AE peak should be stable among all the standards and samples. If not, then repeat the preparation for injection and analysis. d. The identification of Leu, Ile, and aIle peaks must be correct: Leu/Leu∗ retention time ratio is 100%; Ile/Leu∗ retention time ratio is about 95%, and aIle/Leu∗ retention time ratio is about 90%. All of these can be quantified using the 132 > 86 common transition (135 > 89 for Leu∗ ) except for Ile when dALA is present. The 132 > 43 specific transition can be used for Leu quantification (135 > 46 for Leu∗ ). The 132 > 69 specific transition can be used for Ile and aIle quantification (135 > 46 for Leu∗ ). e. Hyp can be quantified either with the common 132 > 86 or with the 132 > 68 specific transition. f. The Asn/Asp∗ retention time ratio must be about 123%. g. The Gln/Gln∗ retention time ratio must be 100%. h. The Gln/Glu retention time ratio must be about 88%. Because of the prominent concentration of Gln in biological fluids (mainly plasma), Glu may be present as a double peak, with one peak at 100% of Glu∗ (corresponding to Glu) and one peak at 88% of Glu∗ (corresponding to Gln interference). Check that only the Glu part (as well as the Glu∗ part) is integrated for Glu quantification. i. The Sar/Ala∗ retention time ratio must be about 68%. j. The 76 > 30 transition used for Gly is the most sensitive of those characteristic for this AA, but it is not as sensitive relative to transitions for other AA. However, Gly concentrations are elevated in biological fluids, allowing for its accurate quantification. In CSF, the concentration of Gly is reduced, but elevated concentrations found in nonketotic hyperglycinemia can be detected. k. The His-3Me/Lys∗ retention time ratio is about 93%. An interfering peak is present at 88%, which may not be considered as His-3Me. l. The basic AA Arg, His, Lys, Orn elute late in the chromatographic system; a small interfering peak is present in the blank (0 µM AA standard) which must be taken into account in blanks for their measurement.
52
Piraud et al.
18. Quantitative validation, including the determination of the limit of quantification (LOQ), linear dynamic range, intra-day and inter-day precision, and accuracy (correlation and/or recovery experiments) have been obtained for all the AAs for which the corresponding AAs∗ are available (8). It has also been obtained for many AAs for which the corresponding AAs∗ are not used (9) (see Table 2.2). LOQs given in (7) are those determined with the API 2000. LOQs are highly instrument dependent, and a linearity limit has been found up to 3 mM for AAs for which the corresponding AAs∗ are used in the method. For other AAs for which AAs∗ were not used, the linearity limit was found up to 1 mM (9). For some AAs, the quantitative validation could not be obtained (9) (Table 2.2). 19. Cys-Hcy can be qualitatively evaluated by the presence/absence of the peak. A semi-quantitative estimation is possible using (Cys)2 as external standard and (Cys)2 ∗ as IS for the management of patients affected with homocystinuria. 20. The dual form of ASA (acid and anhydrides) is the cause of poor quantitative results for ASA and ASA anhydride by this method. Complete transformation of ASA to ASA anhydride is obtained by heating standards and samples at 100◦ C for 60 min. Total ASA anhydride can thus be measured and allow for the management of patients affected with argininosuccinic aciduria.
Acknowledgments The authors would like to thank Luc Anselmini for reading over this manuscript. The authors thank the journal Rapid Communication in Mass Spectrometry and Wiley Publishing for authorization to reproduce tables and figures from refs.(7–9). References 1. Moore, S., Spackman, D. H., Stein, W. H. (1958) Automatic recording apparatus for use in the chromatography of amino acids. Anal Chem 30, 1190–1206. 2. Vogeser, M., Seger, C. (2008) A decade of HPLC-MS/MS in the routine clinical laboratory – goals for further developments. Clin Biochem 9, 649–662.
3. Dooley, K. C. (2003) Tandem mass spectrometry in the clinical chemistry laboratory. Clin Biochem 36, 471–481. 4. Rashed, M. S. (2001) Clinical applications of tandem mass spectrometry: ten years of diagnosis and screening for inherited metabolic diseases. J Chromatogr B Biomed Sci Appl 758, 27–48.
Amino Acid Profiling for the Diagnosis of Inborn Errors of Metabolism 5. Gelb, M. H., Turecek, F., Scott, C. R., Chamoles, N. A. (2006) Direct multiplex assay of enzymes in dried blood spots by tandem mass spectrometry for the newborn screening of lysosomal storage disorders. J Inherit Metab Dis 29, 397–404. 6. Dietzen, D. J., Rinaldo, P., Whitley, R. J., Rhead, W. J., Hannon, W. H., Garg, U. C., et al. (2009) National academy of clinical biochemistry laboratory medicine practice guidelines: follow-up testing for metabolic disease identified by expanded newborn screening using tandem mass spectrometry executive summary. Clin Chem 55, 1615–1626. 7. Piraud, M., Vianey-Saban, C., Petritis, K., Elfakir, C., Steghens, J. P., Morla, A., Bouchu, D. (2003) ESI-MS/MS analysis of underivatised amino acids: a new tool for the diagnosis of inherited disorders of amino acid metabolism. Fragmentation study of 79
53
molecules of biological interest in positive and negative ionisation mode. Rapid Commun Mass Spectrom 17, 1297–1311. 8. Piraud, M., Vianey-Saban, C., Petritis, K., Elfakir, C., Steghens, J. P., Bouchu, D.. (2005) Ion-pairing reversed-phase liquid chromatography/electrospray ionization mass spectrometric analysis of 76 underivatized amino acids of biological interest: a new tool for the diagnosis of inherited disorders of amino acid metabolism. Rapid Commun Mass Spectrom 19, 1587–1602. 9. Piraud, M., Vianey-Saban, C., Bourdin, C., Acquaviva-Bourdain, C., Boyer, S., Elfakir, C., Bouchu, D. (2005) A new reversed-phase liquid chromatographic/tandem mass spectrometric method for analysis of underivatised amino acids: evaluation for the diagnosis and the management of inherited disorders of amino acid metabolism. Rapid Commun Mass Spectrom 19, 3287–3297.
Chapter 3 Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry David S. Millington and Robert D. Stevens Abstract The acylcarnitine profile is a diagnostic test for inherited disorders of fatty acid and branched-chain amino acid catabolism. Patients with this type of metabolic disorder accumulate disease-specific acylcarnitines that correlate with the acyl coenzyme A compounds in the affected mitochondrial metabolic pathways. For example, propionylcarnitine accumulates in patients with both propionic and methylmalonic acidemias. The test identifies and quantifies the species of acylcarnitines in the whole blood or blood plasma of patients at risk for or suspected of having such a disorder. The acylcarnitines are analyzed using electrospray ionization–tandem mass spectrometry. The instrument is used in the precursor ion scan mode to record the molecular species giving rise to fragment ions at m/z 99, derived specifically from the methylated acylcarnitines within the specimen. Quantification is based on the principle of stable isotope dilution, whereby concentrations are derived from the response ratio of each acylcarnitine species to that of a deuterium-labeled acylcarnitine standard. Interpretation of the acylcarnitine profile requires recognition of abnormal concentrations of specific analytes or patterns of analytes and knowledge of their metabolic origin. Key words: Acylcarnitines, tandem mass spectrometry, inherited metabolic disease, metabolomics, fatty acid oxidation.
1. Introduction Acylcarnitines are the catabolic end products of fatty acids and several branched-chain amino acids that are utilized to generate cellular energy. They are derived from their corresponding acyl coenzyme A (acyl-CoA) analogs through exchange of acyl groups between coenzyme A and L-carnitine by the action of a series of carnitine acyl-transferases. These transferases have overlapping T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_3, © Springer Science+Business Media, LLC 2011
55
56
Millington and Stevens
chain length specificity for the various acyl groups, ranging from 2 carbons (acetyl) to over 18 carbons (stearoyl) in length. Unlike their corresponding acyl-CoA analogs, acylcarnitines can cross mitochondrial and cell membranes, and are readily detectable in plasma. The plasma acylcarnitine profile is thus a fair reflection of the intramitochondrial acyl-CoA status at the time of analysis (1). Under normal metabolic circumstances, the acylcarnitine pattern in fasting plasma or serum is relatively stable and consists primarily of acetylcarnitine, plus relatively minor amounts of species derived from the branched-chain amino acids (C3–C5 acylcarnitines) and still lower concentrations of fatty acid intermediates (2) (Fig. 3.1a). Whole blood also exhibits significant concentrations of long-chain acylcarnitines in proportion to their corresponding dietary precursors (2) (C16, C18:1, C18:2, etc.; Fig. 3.1b). The presence of a block in the catabolic pathway of either the fatty acids or branched-chain amino acids results in the accumulation of one or more acyl-CoA intermediates. This in turn elevates the corresponding acylcarnitine concentrations. Analysis of plasma acylcarnitines can detect more than 25 metabolic disorders of fatty acid and branched-chain catabolism (3, 4), and it has become a frontline diagnostic test for these
100
C2
(a)
%
*
* *
*
*
* C14:1 C16
C18:1
0 m/z 200 220 240 260 280 300 320 340 360 380 400 420 440 460 480 500
100
(b)
C2
%
* *
*
*
*
C16
*
C18:1
C14
0 m/z 200 220 240 260 280 300 320 340 360 380 400 420 440 460 480 500
Fig. 3.1. (a) Mass spectrum of plasma acylcarnitines derivatized as their methyl esters, generated from a precursor ion scan of m/z 99 (sum of approximately 50 individual scans) after baseline subtraction, peak smoothing, and centroiding. Internal standard signals are indicated by (∗ ). Note the unit mass resolution at all points in the mass scale. (b) Mass spectrum of blood acylcarnitines derivatized as their methyl esters, generated from a precursor ion scan of m/z 99 (sum of approximately 50 individual scans).
Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry
57
types of disorders. The value of this test has gained international recognition as the basis for the so-called expanded newborn screening panel (5). All babies born in the United States, most European, and several other countries are now routinely screened by analysis of acylcarnitines and amino acids in dried blood spots on filter paper. Those detected as at risk for an inherited metabolic disorder are referred for confirmatory metabolic testing that includes analysis of plasma acylcarnitines (6, 7). Thus the acylcarnitine profile is widely used as a follow-up test to an abnormal newborn screen, as well as for the evaluation and monitoring of patients suspected of having a metabolic disorder (4). Results from the test are usually correlated with urine organic acid analysis performed at the same time. In vitro tests based on the investigation of defects in pathways using metabolic probes with cultured cell lines, particularly fibroblasts, also employ acylcarnitine analysis by tandem mass spectrometry (8, 9). More recently, acylcarnitine analysis has become integrated into targeted metabolomics platforms that have aided the discovery of animal models of human disease and the discovery of new mechanisms of insulin resistance (10, 11). Here we describe a popular method for the analysis of acylcarnitines as their methyl esters using tandem mass spectrometry. Although the preferred specimen type is plasma (or serum), satisfactory (though less accurate) results can also be achieved from dried blood or plasma spots on filter paper. The method is straightforward, robust, highly sensitive to changes in acylcarnitine concentration outside the control range and has a rapid turnaround time. It relies on the molecular specificity of the tandem mass spectrometer and the reproducibility of the analytical procedure. Six isotope-labeled internal standards are used to enable pseudo-quantitative analysis of the acylcarnitines. It should be noted that free carnitine cannot be reliably quantified by this method because the derivatization step partially hydrolyzes acylcarnitines to free carnitine. A separate procedure is required for accurate analysis of free and total carnitine by tandem mass spectrometry (12).
2. Materials 2.1. Standards and Internal Standards
1. Acetyl-L-carnitine and palmitoyl-L-carnitine hydrochloride salts (Sigma Chemical Co, St. Louis, MO) and octanoyl-Lcarnitine (Larodan, Malmo, Sweden). These standards are stored in airtight screw-cap vials within a sealed container with self-indicating desiccant at a temperature of –70◦ C or below. Storage time is 5 years.
58
Millington and Stevens
2. D3-L-Acetylcarnitine, D3-L-propionylcarnitine, D3-Lbutyrylcarnitine, D9-L-isovalerylcarnitine, D3-Loctanoylcarnitine, and D3-L-palmitoylcarnitine hydrochloride salts (Cambridge Isotope Laboratories, Andover, MA). These standards are stored in airtight screw-cap vials within a sealed container with selfindicating desiccant at a temperature of –70◦ C (see Note 1). Storage time is 5 years. 3. Stock solutions of each standard and internal standard (except for C16) are prepared in deionized water by rapidly weighing approximately 25–50 mg of each and making up to a concentration of 0.1 M by addition of the exact volume of deionized water (DI-H2 O) required (see Note 1). The palmitoylcarnitine (C16) standard and internal standard solutions are prepared in MeOH:DI-H2 O (75:25 v/v). Stock solutions are stored at –20◦ C for up to 1 year. 4. Quality control stock solutions for the plasma assay are prepared by first mixing the standard solutions in correct proportion and diluting the mixture in MeOH:DI-H2 O (1:1, v/v) to provide a stock solution (5 mL) with final concentrations of 3 mM acetylcarnitine, 1 mM octanoylcarnitine, and 2 mM of palmitoylcarnitine. This quality control stock solution is stored at –20◦ C for up to 1 year. 5. Working quality control solution for the plasma assay is made first by a 1:5 dilution of the stock solution with MeOH:DI-H2 O (1:1, v/v) to make the working mixture (0.1 mL). Then 0.06 mL of the working mixture is equilibrated with 12 mL bovine adult serum (BAS) in a-15 mL centrifuge tube by agitating for 30 min on a sample rocker at ambient temperature. 6. 100 µL aliquots of the working quality control solution, containing final concentrations of 3 µM acetylcarnitine, 1 µM octanoylcarnitine, and 2 µM palmitoylcarnitine, are stored at –20◦ C. QC samples are analyzed at least in duplicate within each batch of samples. A “patient QC” is also selected to be analyzed within each sample run. This is a sample selected at random from a previously analyzed batch of patient samples. 7. Quality controls for the blood spot assay are prepared from a fresh specimen of whole blood (10 mL, heparinized). This whole blood specimen can be spiked with octanoyl-Lcarnitine at a concentration of 1 µM for comparison with the plasma QC standard. Then 50 µL aliquots are pipetted onto the pre-defined circles on newborn screening specimen cards. The spots are allowed to air-dry for at least 6 h and are then stored in paper envelopes within a sealed
Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry
59
plastic bag in the presence of self-indicating desiccant at –70◦ C. 8. Stock internal standard mixture is prepared by first mixing the internal standard solutions and diluting the mixture in MeOH:DI-H2 O (1:1, v/v) to provide final concentrations of 10 mM D3-acetylcarnitine, 4 mM D3palmitoylcarnitine, and 2 mM of each of the other internal standards. The stock internal standard solution is stored at –20◦ C for up to 1 year. 9. Working internal standard solution for the plasma assay is prepared from stock by serial dilution in MeOH:DIH2 O (1:1, v/v) to provide final concentrations of 0.1 mM D3-acetylcarnitine, 0.04 mM D3-palmitoylcarnitine, and 0.02 mM of each of the other internal standards (see Note 2). The working internal standard solution is stored for up to 1 month at 0–4◦ C. 10. Working internal standard solution for the dried blood spot assay is prepared from stock by serial dilution in MeOH:H2 O (1:1, v/v) to provide final concentrations of 12 µM D3-acetylcarnitine, 4.8 µM D3-palmitoylcarnitine, and 2.4 µM of each of the other internal standards. The working internal standard solution is stored for up to 1 month at 0–4◦ C. 11. A mass spectrometry tuning solution is prepared from the internal standard stock solution as follows. 30 µL of stock (containing 1 mM D3-acetyl-, 0.2 mM D3-octanoyl-, and 0.4 mM D3-palmitoyl-carnitines) is pipetted into a 5-mL glass vial. Evaporate to dryness under nitrogen. Add 100 µL of 3 M HCl in MeOH, cap and vortex mix, then incubate for 15 min at 50◦ C. Uncap the vial, evaporate to dryness, and reconstitute in 2 mL of final matrix (MeOH:H2 O, 85:15, v/v). Store at 0–4◦ C for up to 2 weeks. 2.2. Solvents and Reagents
1. Anhydrous methanolic hydrogen chloride (Supelco, Bellefonte, PA): 3 M. It is advisable to aliquot this reagent as rapidly as possible upon receipt into small screw-cap glass vials intended for single use (1–2 mL), wrap with parafilm, and store with desiccant in a sealed plastic bag at –20◦ C (see Note 3). 2. Anhydrous butanolic hydrogen chloride (Supelco): 3 M. It is advisable to aliquot this reagent as rapidly as possible upon receipt into small screw-cap glass vials intended for single use (1–2 mL), wrap with parafilm, and store with desiccant in a sealed plastic bag at –20◦ C (see Note 3). 3. HPLC-grade methanol.
60
Millington and Stevens
4. Deionized water (resistivity >17 M/cm). 5. Mobile phase: MeOH:DI-H2 O (15:85, v/v). This should be filtered prior to use through a 0.2-µm nylon membrane (Grace, Deerfield, IL). 2.3. Equipment and Other Materials
1. Tandem quadrupole mass spectrometer (Quattro Micro, Waters Associates, Milford, MA) equipped with ESI source, solvent delivery, and autosampler systems (Acquity, Waters). Equivalent systems from other vendors are acceptable. 2. Plate dryer, 96-well (Biotage, CA), preferably equipped R with Teflon -coated stainless steel head. 3. Harvard pump model HA11 (Instech Labs, Inc., Plymouth Meeting, MA). 4. A set of high-quality variable pipettes (0.5–10; 5–50, 20–200; 100–1000 µL; e.g., from Finnpipette or Eppendorf) are required, plus appropriate disposable pipette tips. 5. Microtiter plates (96-well; Evergreen, Los Angeles, CA). 6. A good quality steel hole puncher, 3/16 in. diameter. 7. Aluminum foil (7.5-cm-wide roll; Fisher Scientific, Pittsburgh, PA) and adhesive film (microadhesive film; USA Scientific, Ocala, FL), used to cover 96-well plates. 8. Plastic microcentrifuge and centrifuge tubes of 1.5, 2, and 15 mL capacity. 9. Laboratory benchtop centrifuge (e.g., Fisher model 235C from Fisher Scientific, Pittsburgh, PA or equivalent). 10. Vortex mixer. 11. Orbital shaker. 12. Incubator/oven.
3. Methods The general objective of this method is to generate a report showing the concentrations of up to 30 acylcarnitine species relative to the upper limits, and for a few analytes the lower limits, of a control range for each of them. It is not a rigorous quantitative analysis. The general principles are as follows. Isotope-labeled internal standards are added to the specimen as early as possible during the sample preparation procedure to establish a fixed ratio of their concentrations to those of the target analytes. When the sample is a dried blood spot (DBS), the internal standards are added to an extract from the DBS. Thus, the accuracy of the
Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry
61
method is inherently lower for DBSs than for plasma, which is compounded by the fact that the volume of blood in a DBS is estimated from the diameter of the punch and by the variable hematocrit. Similarly, dried plasma spots on filter paper, although facilitating specimen collection and transfer from remote or challenging locations, do not provide the most accurate results. For these reasons, liquid plasma (or serum) is regarded as the best and preferred specimen for this test. DBSs collected by the heelstick method are widely used to screen newborns using a method similar to that described here (6). The errors inherent in this approach are minimized in the diagnostic laboratory by using a larger sample quantity and a more rigorous analytical technique. The amounts of the added internal standards are designed to provide high sensitivity at the upper limits of the normal analyte concentration ranges, thus enhancing the discrimination between normal individuals and those at risk for a metabolic disorder. After precipitation of proteins, the acylcarnitines are converted to methyl esters to enhance sensitivity, which is especially important for dicarboxylic species (see Note 4). Their analysis is performed by flow injection of the derivatized sample directly into the electrospray ion source of a tandem quadrupole mass spectrometer operated at unit mass resolution, using a precursor ion scan of m/z 99. The precursor ions are recorded as a single accumulated spectrum from m/z 200 to 500 (Fig. 3.1a). After baseline subtraction, peak smoothing, and centroiding, the signal ratios for the analytes and their respective internal standards (see Table 3.1) are generated, converted to an approximate concentration by multiplying by the final concentration of internal standard added to the aliquot of sample, and inserted into a final report (see examples in Tables 3.2 and 3.3). In some cases, in order to distinguish between isomers (Section 3.4), samples may be reanalyzed as their butyl esters, in which case the precursor ion scan of m/z 85 is used and the precursor ions are recorded from m/z 250 to 550 (6).
Table 3.1 Pseudo-molecular ion masses of the target analytes and their respective internal standards (in bold characters) as methyl esters and butyl esters Species
Methyl ester
Butyl ester
C2
218
260
C2–IS
221
263
C3
232
274
C3–IS
235
277
C4
246
288
62
Millington and Stevens
Table 3.1 (continued) Species
Methyl ester
Butyl ester
C4–IS
249
291
C5:1
258
300
C5
260
302
C4–OH
262
304
C5–IS
269
311
C6
274
316
C5–OH
276
318
C3–DC
276
360
C4–DC
290
374
C8:1
300
342
C8
302
344
C5–DC
304
388
C8–IS
305
347
C6–DC
318
402
C10:2
326
368
C10:1
328
370
C10
330
372
C8–DC
346
430
C12:1
356
398
C12
358
400
C14:2
382
424
C14:1
384
426
C14
386
428
C14:1–OH
400
442
C14–OH
402
444
C16
414
456
C16–IS
417
459
C16–OH
430
472
C18:2
438
480
C18:1
440
482
C18
442
484
C18:2–OH
454
496
C18:1–OH
456
498
C16–DC
458
542
C18:1–DC
484
568
Acylcarnitines: Analysis in Plasma and Whole Blood Using Tandem Mass Spectrometry
63
Table 3.2 Example of a final report from the acylcarnitine profile analysis of a plasma specimen, showing comparison of patient results (bold characters) with control values and interpretation Nanoliter range (nmol/mL)
Result (nmol/mL)
Status
21.48
NL
2.41
NL
57
Succinic acid
117 > 73
Methylmalonic acid
117 > 73
Glutaric acid
131 > 87
Ethylmalonic acid
131 > 87
Methylsuccinic acid
131 > 87
Butyrylglycine
144 > 74
Isobutyrylglycine
144 > 74
2-Ketoglutaric acid
145 > 101
3-Methylglutaric acid
145 > 101
Tiglylglycine
156 > 74
3-Methylcrotonylglycine
156 > 74
Isovalerylglycine
158 > 74
Methylbutyrylglycine
158 > 74
suppression effect is minimized using several labeled internal standards. Apart from routine ion source cleaning performed at approximately weekly intervals, no other precautions are taken to minimize ion source contamination. 12. A careful selection of appropriate MRM transitions makes it possible to measure several isomeric metabolites. In some cases, no suitable transitions to solve isomeric compounds are found (Table 4.5) (5). This condition can be solved in the future by coupling the tandem mass spectrometer with HPLC separation. 13. The primary application of the LC tandem mass spectrometer urine screening method is as a first test for suspected inborn errors of metabolism. While the tandem mass spectrometry method can be used to test all urine samples, GCMS-based organic acid analysis is typically used to confirm abnormal LC tandem mass spectrometer results or to distinguish isomers.
4. Notes 1. The 24-h collected urine should be considered a first choice for the analysis of organic acids. The majority of patients with suspicion of IEM are in neonatal age; thus, it is important to
96
la Marca and Rizzo
perform a rapid test, preferably on the first morning voiding. When possible, urine samples must be stored in a freezer at –20◦ C if a –70◦ C system is not available. In fact, poor preservation of samples can lead to (1) bacterial conversion of some organic acids to artifacts such as the conversion of all keto acids to their respective hydroxy acids, (2) increased levels of succinic acid from bacterial degradation of glutamine, and (3) an abnormal concentration of pyroglutamic acid due to conversion from glutamine. The concentration of such volatile compounds as small molecule organic acids (e.g. short-chain organic acids) can be shortened if the sample is exposed to high temperature (especially in summer). The administration of drugs should be suspended in patients undergoing organic acids analysis to avoid artifacts due to drugs or drug metabolism. The following are some examples of the interference of drugs on the analysis of organic acids. Valproic acid administration causes an abnormal excretion of 3-hydroxy-isovaleric acid, tiglylglycine, dicarboxylic acids (even saturated, e.g., suberic, adipic, and sebacid acids), 2methylbutyrylglycine, 7-hydroxy-octanoic acid, 5-hydroxyhexanoic acid, and hexanoylglycine. An increase of pivalic acid can be due to the administration of the antibiotics pivampicillin and pivmecillinam. Fluvoxamine maleate, a selective serotonin reuptake inhibitor (SSRI) used to treat obsessive–compulsive disorder and depression, can cause an increment of maleic acid. A therapy with the antihyperuricemic drug allopurinol can increase orotate urinary excretion. Artifactual formation of new compounds during sample preparation can also occur. For example, during extraction with ethyl acetate, decarboxylation of keto acids can occur (14). Additionally, it is essential to pay attention to solvent contaminants, plasticizers (phthalate, adipate, and sebacate esters), glassware-cleaning agents, lubricants, resins, and bleeding from the stationary phase of the chromatographic columns. 2. IS work solution stability is only at 4◦ C or below. For example, the peak intensity of PDA-TMS is considerably larger when analyzed 1 week after preparation and storage at room temperature than when immediately analyzed. This is due to solvent evaporation at room temperature (12). 3. Bilirubine, a potential interferent, is oxidized by using potassium ferricyanide K3 Fe(CN)6 at 0.13 mmol/L. 4. Do not mix organic and aqueous phases during liquid/liquid extraction of urine with solvents. 5. Clean the injector syringe everyday to avoid encrustation, purging, and bleeding.
Analysis of Organic Acids and Acylglycines
97
6. Incorrect mass spectrometer calibration can result in abnormal qualitative and quantitative results. 7. Clean the mass spectrometer source and change the liner every 500 samples. Contamination is usually identified by an abnormal background. In a GC-MS instrument, there are sources of contamination within the gas chromatograph (bleeding septa, dirty injection port liners, air leaks, etc.) and the mass spectrometer (ion source leaks, fluid/oil, manifold dirt). Unfortunately, not all contamination can be removed through running clean carrier gas overnight. When this is the case, the instrument may need extensive cleaning. Cleaning the mass spectrometer ion source to remove contamination is critical to restore the electrostatic properties of the ion source lens. The authors suggest that the ion source be cleaned every 500 runs. Another common source of mass spectrometer contamination is from inadequate venting and maintenance of the diffusion pump. Preventive maintenance ensures that diffusion pump oil/fluids are topped to the correct levels. It is important to maintain proper fluid levels to avoid pump failure and ensure optimum performance of the vacuum system. Air leaks can occur if a seal becomes damaged or is not correctly fastened. This is another common problem for any instruments that use a vacuum and can be identified by a higher than normal vacuum manifold pressure, low relative ion abundance, and poor sensitivity. 8. The use of labeled internal standards and isotope dilution allows one to obtain a correct and absolute quantitation of organic acids, but it can be very expensive. References 1. Björkman, L., McLean, C., Steen, G. (1976) Organic acids in urine from human newborns. Clin Chem 22, 49–52. 2. Chalmers, R. A., Healy, M. J., Lawson, A. M., Watts, R. W. (1976) Urinary organic acids in man. II. Effects of individual variation and diet on the urinary excretion of acidic metabolites. Clin Chem 22, 1288–1291. 3. Kumps, A., Duez, P., Mardens, Y. (2002) Metabolic, nutritional, iatrogenic, and artifactual sources of urinary organic acids: a comprehensive table. Clin Chem 48, 708–717. 4. Pitt, J. J., Eggington, M., Kahler, S. G. (2002) Comprehensive screening of urine samples for inborn errors of metabolism by electrospray tandem mass spectrometry. Clin Chem 48, 1970–1980.
5. Rebollido, M., Cocho, J. A., Castiñeiras, D. E., Bóveda, M. D., Fraga, J. M. (2006) Aplicación de la espectrometría de masas en tándem al análisis de aminoácidos, acilcarnitinas, acilglicinas y ácidos orgánicos en muestras de orina en papel. Quím Clín 25, 64–74. 6. Chalmers, R. A., Healy, M. J., Lawson, A. M., Hart, J. T., Watts, R. W. (1976) Urinary organic acids in man. III. Quantitative ranges and patterns of excretion in a normal population. Clin Chem 22, 1292–1298. 7. Perry, T. L., Hansen, S. (1974) Artifacts and pitfalls in interpretation of gas chromatograms in Application of Gas Chromatography-Mass Spectrometry to the Investigation of Human Disease. In Proceedings of a workshop, Montreal, May 1973. McGill University-Montreal Children’s
98
8. 9. 10.
11. 12.
la Marca and Rizzo Hospital Research Institute, Montreal, Quebec, Canada, 89–102. Paterson, N. (1967) Relative constancy of 24-hour urine volume and 24-hour creatinine output. Clin Chim Acta 18, 57–58. Zorab, P. A. (1969) Normal creatinine and hydroxyproline excretion in young persons. Lancet 29, 1164–1165. Applegarth, D. A., Hardwick, D. F., Ross, P. M. (1968) Creatinine excretion in children and the usefulness of creatinine equivalents in amino acid chromatography. Clin Chim Acta 22, 131–134. Larsen, K. (1972) Creatinine assay by a reaction-kinetic principle. Clin Chim Acta 41, 209–217. Tanaka, K., West-Dull, A., Hine., D. G., Lynn, T. B., Lowe, T. (1980) Gas-
chromatographic method of analysis for urinary organic acids. II. Description of the procedure, and its application to diagnosis of patients with organic acidurias. Clin Chem 26, 847–853. 13. Tanaka, K., Hine, D. G., West-Dull, A., Nad Lynn, T. B. (1980) Gas-chromatographic method of analysis for urinary organic acids. I. Retention indices of 155 metabolically important compounds. Clin Chem 26, 1839–1846. 14. Thompson, R. M., Belanger, B. G., Wappner, R. S., Brandt, I. K. (1975) An artifact in the gas chromatographic analysis of urinary organic acids from phenylketonuric children: decarboxylation of the phenylpyruvic acid during extraction. Clin Chim Acta 61, 367–374.
Chapter 5 HPLC Analysis for the Clinical–Biochemical Diagnosis of Inborn Errors of Metabolism of Purines and Pyrimidines Giuseppe Lazzarino, Angela Maria Amorini, Valentina Di Pietro, and Barbara Tavazzi Abstract The determination of purines and pyrimidines in biofluids is useful for the clinical–biochemical characterization of acute and chronic pathological states that induce transient or permanent alterations of metabolism. In particular, the diagnosis of several inborn errors of metabolism (IEMs) is accomplished by the analysis of circulating and excreted purines and pyrimidines. It is certainly advantageous to simultaneously determine the full purine and pyrimidine profile, as well as to quantify other compounds of relevance (e.g., organic acids, amino acids, sugars) in various metabolic hereditary diseases, in order to screen for a large number of IEMs using a reliable and sensitive analytical method characterized by mild to moderate costs. Toward this end, we have developed an ion-pairing HPLC method with diode array detection for the synchronous separation of several purines and pyrimidines. This method also allows the quantification of additional compounds such as N-acetylated amino acids and dicarboxylic acids, the concentrations of which are profoundly altered in different IEMs. The application of the method in the analysis of biological samples from patients with suspected purine and pyrimidine disorders is presented to illustrate its applicability for the clinical–biochemical diagnosis of IEM. Key words: Purines, pyrimidines, HPLC, body fluids, clinical–biochemical diagnosis, inborn errors of metabolism (IEMs), head injury, neurodegeneration, ischemia and reperfusion, multiple sclerosis.
1. Introduction In the clinical biochemistry setting, the terms “purines” and “pyrimidines” identify several low molecular weight compounds deriving from nucleic acid and nucleotide catabolism. Most of the T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_5, © Springer Science+Business Media, LLC 2011
99
100
Lazzarino et al.
circulating and excreted purines and pyrimidines are uncharged, with various degrees of water solubility, and have excellent absorption coefficients in the UV region (230–300 nm). Depending on the compound considered, purines and pyrimidines have broad ranges of concentrations (from 0 to the millimolar levels) in bodily fluids of normal subjects; these levels are influenced by various factors such as age, sex, diet. Very often, those absent or present physiologically in trace amounts are subject to tremendous concentration fluctuations due to several pathologic states for which the analysis of purines and pyrimidines in urine and plasma provides a strong indication of an altered metabolic state. In fact, the determination of purines and pyrimidines could be of particular clinical relevance in a number of acute and chronic diseases, such as myocardial ischemia (1, 2), traumatic brain injury (3, 4), and neurodegenerative disorders (multiple sclerosis, amyotrophic lateral sclerosis, Alzheimer’s disease, and Parkinson’s disease) (5–7). Under these conditions, the determination of certain compounds of these two classes is useful to monitor changes in the metabolic status of the patient, thus giving information on the progression of the disease and the effectiveness of possible pharmacological treatments. Purine and pyrimidine concentrations in bodily fluids are also dramatically altered in various inborn errors of metabolism (IEMs). In fact, a large number of IEMs, characterized by various degrees of disability and neurological disturbances, include several enzymatic defects involving pathways of purine and pyrimidine catabolism, recovery, and de novo biosynthesis. For many of these IEMs, an early diagnosis may greatly reduce consequences for the patients but, since clinical signs are not always discriminatory, the final diagnosis is not simple and is frequently delayed. It is therefore evident that purine and pyrimidine profiling in biological fluids plays a critical role for both clinicians and IEM-affected patients (8–10). Based on the physical–chemical characteristics of those purines and pyrimidines with clinical–biochemical significance, various methods have been proposed during the last decade based on the use of high-performance liquid chromatography (HPLC) coupled with either UV-VIS (11–18) or mass spectrometric (19– 23) detectors, capillary electrophoresis (24, 25), thin-layer chromatography (TLC) (26), and gas chromatography–mass spectrometry (GC–MS) (27). We have recently described a highly sensitive, reproducible, easy to use, and low-cost ion-pairing HPLC method coupled with a diode array detector (DAD) for the determination of these compounds in biological fluids (28), suitable for the clinical– biochemical diagnosis of IEMs of purines and pyrimidines (29). This method also allows for the separation of compounds other than purines and pyrimidines and is therefore useful for the diagnosis of additional IEMs of clinical relevance (30). The method is
HPLC Analysis for the Clinical–Biochemical Diagnosis
101
characterized by minimal manipulation of biological fluids (urine, plasma/serum, cerebrospinal fluid, amniotic fluid) and allows the separation of purines (hypoxanthine, xanthine, uric acid, inosine, guanosine, adenosine, adenine, S-adenosylhomocysteine, S-adenosylmethionine, adenylosuccinate, succinylaminoimidazole carboxamide riboside, aminoimidazole carboxamide riboside) and pyrimidines (cytosine, cytidine, thymine, thyimidine, uracil, beta-pseudouridine, uridine, orotic acid), simultaneously with N-acetylated amino acids (N-acetylaspartate, N-acetylglutamate, N-acetylaspartylglutamate), dicarboxylic acids (propionic acid, malonic acid, methylmalonic acid), creatinine, reduced and oxidized glutathione, and ascorbic acid. Herein, we describe in detail how to carry out the analysis of the aforementioned compounds in biological fluids (serum and urine) of healthy subjects and of two selected patients each suffering from an IEM of purine metabolism, i.e., hypoxanthine phosphoribosyltransferase (HPRT) deficiency and adenylosuccinate lyase (ADSL) deficiency. A comprehensive list of the genetic disorders based on metabolic defects of the aforementioned compounds, for which the application of this HPLC method might be helpful in clinical diagnoses is reported in Table 5.1.
2. Materials 2.1. Chemicals and Standards
1. Tetrabutylammonium hydroxide is supplied as a 55% water solution (Nova Chimica, Milan, Italy), stored at room temperature, and light protected (see Note 1). 2. HPLC-grade methanol. 3. HPLC-grade water. 4. Monobasic potassium phosphate (KH2 PO4 ). A 1 M stock solution is prepared by dissolving the proper amount of the compound in HPLC-grade water, with stirring and moderate heating; this solution is stable for up to 1 month at 4◦ C. 5. Ultrapure standards for HPLC (Sigma, St. Louis, MO, USA). Stock solutions (1 mM final concentration) of the following compounds are prepared by dissolving the proper amount of each substance with HPLC-grade water. All solutions, with the exceptions of ascorbic acid and GSH, can be stored in aliquots and maintained at –80◦ C with no significant loss (decrease in concentration 95% purity; Sigma-Aldrich, Inc.). 2a. O-Acetyl-L-carnitine (C2) 2b. Adenosine (A) 2c.
L -Alanyl- L -alanine
(diAla)
232
Britz-McKibbin
2d.
L -Arginine
(Arg)
2e. Atenolol (At) 2f.
L -Carnitine
2g.
L -Carnosine
(C0) (Carn)
2h.
L -Citrulline
(Cit)
2i. 3-Choloro-L-tyrosine (ClTyr) 2j. Cystathionine (Cyst) 2k. Dopamine (DopN) 2l. Estradiol-3-glucuronide (E2-3G) 2m. Estradiol-3-sulfate (E2-3S) 2n. Estriol-3-glucronide (E3-3G) 2o. Estriol-3-sulfate (E3-3S) 2p. Estrone-3-glucuronide (E1-3G) 2q. Estrone-3-sulfate (E1-3S) 2r.
L -Glutamine
2s.
L -Glutamic
(Gln)
acid (Glu)
2t. Glutathione – oxidized (GSSG) 2u. Glutathione – reduced (GSH) 2v. Guanosine (G) 2w.
L -Histidine
(His)
2x. 3-Hydroxy-L-tryptophan (OHTrp) 2y.
L -Leucine
(Leu)
2z.
L -Isoleucine
(Ile)
2aa.
L -Allo-isoleucine
2ab.
L -Lysine
2ac.
L -Methionine
(allo-Ile)
(Lys) (Met)
2ad. 3-Methyl adenosine (MeA) 2ae. N-Methyl-L-aspartic acid (MeAsp) 2af. 3-Methyl-L-histidine (MeHis) 2ag. O-Myristoyl-L-carnitine (C14) 2ah. 3-Nitro-L-tyrosine (NTyr) 2ai.
L -Ornithine
(Orn)
2aj. O-Palmitoylcarnitine (C16) 2ak. p-Aminobenzoic acid (PABA) 2al. 2am.
L -phenylalanine L -Proline
(Phe)
(Pro)
2an. 5-Oxo-L-proline or pyroglutamic acid (5-oxo-Pro) 2ao. Serotonin (Sero)
CE–ESI-MS-Based Metabolomics
233
2ap. Trans-4-hydoxy-L-proline (4-OH-Pro) 2aq. Tryptamine (TyrN) 2ar.
L -Tryptophan
2as.
L -Tyrosine
2at.
L -Valine
(Trp)
(Tyr)
(Val)
3. Acylcarnitine standards (Larodan Fine Chemicals Inc., Malmö, Sweden). 3a. Propionyl-L-carnitine HCl (C3) 3b. Butyryl-L-carnitine HCl (C4) 3c. Octanoyl-L-carnitine HCl (C8) 4. Dilute individual stock solutions of metabolites (10 mM) in 1:1 methanol/water and prepare all reduced thiol standards (e.g., GSH) daily due to their intrinsic lability. 5. Degas all solutions and metabolite standards by sonication, then store at 4◦ C prior to use. 6. Quality control mixture: 20 µM of Ala, Leu, Phe, Tyr, Glu, Asp, Lys, His, and Arg; 50 µM diAla. 2.3. Equipment
1. Barnstead EASYpure II LF ultrapure water system (Dubuque, USA). 2. Polypropylene centrifuge tubes (0.5 and 1.5 mL; VWR International, Inc., Toronto, Canada). R centrifugal filters (Pall Life Sciences, Inc., 3. 3-kDa Nanosep Michigan, USA).
4. Disposable lancets (Unistik 3; Owen Munford Ltd, Georgia, USA). 5. Grade 903 Protein Saver Card filter paper and 1/8th-in.diameter hole puncher (Whatman, Inc., New Jersey, USA). 6. Data processing and statistical analysis software: Excel 2007 (Microsoft, Inc., Redmond, WA), MATLAB 2008 (The Mathworks, Inc., Natick, MA), and Igor 5.0 (Wavemetrics, Inc., Lake Oswego, OR).
3. Methods Metabolite profiling by CE–ESI-MS is typically performed under either acidic (pH 1.8) or alkaline (pH > 8.5) background electrolyte (BGE) conditions in conjunction with positive-ionor negative-ion-mode ESI-MS in order to resolve and detect complex mixtures of cationic and anionic metabolites, respectively. This not only ensures that weakly ionic or zwitterionic metabolites are adequately ionized but also avoids the pH range at
234
Britz-McKibbin
which the EOF tends to be more variable (pH ≈ 4–8). Thus, neutral metabolites (e.g., androgens) are not analyzed by CE–ESIMS unless using a surface-modified capillary suitable for capillary electrochromatography (CEC) (16) or partial-filling micellar electrokinetic chromatography (17), which are hybrid separation techniques more difficult to implement routinely. In some cases, complementary ionization sources can be coupled to CE in order to improve the detection of certain classes of metabolites, such as atmospheric photoionization (18). Nevertheless, a common constraint of all liquid-infused MS platforms is the lack of a universal ion source for gas-phase desorption (19), since ionization efficiency in ESI is highly dependent on the intrinsic physicochemical properties of metabolites with relative response factors that can vary up to three orders of magnitude (5). Given the selectivity constraints of the separation technique and ion source, CE–ESI-MS is primarily amenable to the analysis of polar and weakly ionic metabolites, which comprise a majority (>60%) of known metabolites or degradation products associated with primary metabolism (20). The sample pretreatment procedure for processing a variety of biological specimens is described in detail below, where only 10 µL of sample is typically required for injection (≈ 0.1 µL injected) using commercial CE–ESI-MS instruments. However, special precaution is needed when analyzing labile metabolites (e.g., thiols) in complex biological samples due to oxidation artifacts that can occur during sample collection, pretreatment, storage, and/or analysis (7). In most cases, internal standards are included in all samples in order to improve assay reproducibility in terms of quantification (i.e., integrated relative peak area) and migration time (i.e., RMT) performance. In addition, recovery standards can also be used to assess method accuracy after sample pretreatment, as well as identify sources of bias during large-scale batch analyses. Data pre-processing of large-scale metabolomic datasets requires adequate noise filtering, data alignment/normalization, and data exploration methods to reveal biologically significant features relevant to the experimental design, which can be facilitated with integrated data analysis software. However, a major challenge in untargeted metabolomics remains the identification of a large fraction of unknown metabolites that do not correspond to known candidates listed within public databases (e.g., KEGG, Human Metabolome Database, MetLin) despite access to accurate mass, isotopic composition, and fragmentation information by ESI–TOF-MS/MS (21). Indeed, this dilemma is considerable when less than 10% of total metabolite peaks (i.e., features) detected by hyphenated separation MS techniques can be quantified due to limited access to authentic standards (22). In this context, CE separations offer orthogonal qualitative information for unambiguous identification among
CE–ESI-MS-Based Metabolomics
235
putative candidates based on their characteristic migration behavior that can be predicted from the intrinsic properties of metabolites (9). Several examples of targeted analyses of cationic or anionic analytes relevant to metabolite profiling (e.g., expanded newborn screening), as well as a comprehensive metabolomic strategy for biomarker discovery (e.g., oxidative stress), are illustrated in the subsequent figures. 3.1. Sample Pretreatment of Dried Blood Spots
1. Collect human blood samples using a finger-prick method via disposable lancets and spot them on a Grade 903 Protein Saver Card and dry overnight. 2. Punch out a 3.2-mm (1/8 in.) disk (≈ 3.4 µL) manually from each dried blood spot (DBS) with a hole puncher into a 0.5-mL centrifuge tube that contained 100 µL of icecold 1:1 MeOH:H2 O with the internal standard dialanine (DiAla, 100 µM). 3. Extract the DBS under sonication for 10 min. 4. Filter the resulting extract solution through a 3-kDa R Nanosep centrifugal filter at 150×g at 4◦ C for 10 min prior to analysis (see Note 3). 5. Dilute the filtrate 1:1 with an aqueous ammonium acetate solution (400 mM, pH 7.0) to produce the final sample solution used for analysis (200 mM ammonium acetate, 25% MeOH, 50 µM dialanine). 6. In most cases, filtered DBS samples are analyzed directly by CE–ESI-MS without sample pretreatment steps such as chemical derivatization or solvent/reagent evaporation (see Note 4).
3.2. Sample Pretreatment of Human Plasma and Red Blood Cell Lysates
1. Collect human blood samples using a finger-prick method (see above) or via a venous catheter inserted in the antecubital vein, kept patent using a saline (0.9%, w/v) solution. 2. Immediately place each blood sample on ice and subsequently centrifuge at 20×g at 4◦ C for 5 min to fractionate plasma from erythrocytes. 3. Transfer the plasma supernatant and remove plasma proR centrifugal filter at 150×g teins using a 3-kDa Nanosep for 10 min prior to dilution and subsequent analysis (see Note 3). 4. Wash the red blood cells (RBCs) with phosphate buffered saline (10 mM NaH2 PO4 , 150 mM NaCl, pH 7.4, 4◦ C), then vortex and centrifuge at 70×g at 4◦ C for 1 min to isolate RBCs. 5. Repeat the washing steps three times until the supernatant is clear without any evidence of premature hemolysis.
236
Britz-McKibbin
6. Hemolyze 200 µL of RBCs by adding 600 µL of pre-chilled deionized water, then centrifuge at 70×g at 4◦ C for 1 min to sediment cell debris. 7. Next, filter 100 µL of RBC lysate with a 3-kDa filter at 150×g at 4◦ C for 10 min to remove excess hemoglobin. 8. Dilute all filtered protein-free RBC lysates twofold in BGE containing 50 µM diAla and 50 µM HEPES as internal standards for positive and negative ESI, respectively, and store frozen at –80◦ C. Thaw only once prior to analysis (see Note 5). 3.3. Sample Pretreatment of Human Urine
1. Collect human urine mid-stream as a first void morning sample and store at 4◦ C and normalize to creatinine to correct for urine dilution effects (see Note 6). 2. Centrifuge the urine samples at 70×g at 4◦ C for 2 min to remove particulate matter, then dilute the samples 10fold using deionized water containing 50 µM diAla and 50 µM HEPES as internal standards when using CE–ESIMS with positive (acidic BGE) and negative (alkaline BGE) ion modes, respectively.
3.4. Capillary Conditioning, Buffer Preparation, and CE–ESI-MS Operation
1. Prior to first use, condition fused silica capillaries installed in the coaxial sheath liquid interface (see Note 7) for 15 min each with MeOH, 1 M NaOH, 1 M formic acid, de-ionized H2 O and then 60 min with acidic or alkaline BGE. 2. Perform several trial runs with the newly conditioned capillary using a quality control test mixture as the sample and the desired BGE until stable currents and reliable ion signals are attained (see Note 8). 3. Once conditioning is complete, perform a 10-min prerinse/flush of the capillary with BGE prior to each separation. 4. High-purity (>98%) volatile buffer reagents (SigmaAldrich, Inc.) used for CE separations were prepared on a weekly basis, filtered, and sonicated prior to use. 5. In general, 1.0 M formic acid, pH 1.8 (diluted from concentrated formic acid), can be used as the acidic BGE for cationic metabolites under positive-ion-mode ESI-MS, whereas 50 mM ammonium acetate or ammonium bicarbonate, pH 8.5–9.5 (adjusted with ammonium hydroxide), can be used as the alkaline BGE for anionic metabolites under negative-ion-mode ESI-MS (see Note 9). 6. Perform CE separations at 20◦ C with an applied voltage of 30 kV unless otherwise indicated (see Note 10).
CE–ESI-MS-Based Metabolomics
237
7. The sheath liquid consisting of 1:1 MeOH:H2 O with 0.1% (v/v) formic acid (used with acidic BGE) or 5 mM ammonium acetate (used with alkaline BGE) is supplied by the 1100 series isocratic pump at a flow rate of 10 µL/min. 8. Nitrogen is used as both a nebulizing and a drying gas supplied at 6 psi and 10 L/min, respectively, whereas helium at 6 × 10–6 mbar is used as a damping gas for the ion trap.
9. Perform all ESI-MS analyses using a ± 4 kV cone voltage in positive- or negative-ion mode at a temperature of 300◦ C unless otherwise stated. 10. Record MS data within a range of 50–750 m/z using an ultrascan mode of 26,000 m/z per second (see Note 11). 3.5. Targeted Profiling of Cationic Metabolites
1. Introduce all samples to the capillary after a two- or tenfold dilution in 100 mM ammonium acetate, pH 7.0, using a low-pressure hydrodynamic injection by first injecting a sample for 75 s at 50 mbar (see Note 12) followed by a 60 s injection of the acidic BGE (1 M formic acid, pH 1.8) at 50 mbar prior to voltage application (see Note 13). 2. Perform all separations for cationic metabolites using unmodified fused-silica capillaries under normal polarity by CE with positive-ion-mode ESI-MS. 3. In most cases, all analyses should be performed in triplicate using a fresh BGE reservoir (inlet/anode) for each run. Data should be processed (e.g., peak integration) using extracted ion electropherograms (EIE) based on the m/z for each target metabolite (M+H+ ). 4. Perform a pre-rinse with acidic BGE for 10 min after each run in order to re-condition the capillary. 5. Figure 14.1a shows a schematic that depicts the coaxial sheath liquid interface configuration used in CE with an acidic BGE under positive-ion-mode ESI-MS conditions for the analysis of cationic metabolites, which elute in order of their apparent charge density (i.e., charge–size ratio) (see Note 14). 6. Figure 14.1b depicts the simultaneous analysis of polar amino acids (e.g., Arg) and surface-active acylcarnitines (e.g., palmitoyl-L-carnitine, C16) under a single elution condition by CE–ESI-MS relevant to expanded newborn screening of inborn errors of metabolism as derived from filtered DBS extracts (6), which also allows for the resolution of isomeric (e.g., Ile, allo-Ile) and isobaric (e.g., 4-OH-Pro) co-ion interferences (see Note 15). 7. Figure 14.2a highlights the interference-free time window (≈ 7–12 min) for quantification of low-abundance cationic
238
Britz-McKibbin
(a)
(b)
N+ H
C16
OO
O
O
6
H2N
6
Isomeric/Isobaric Ions
0.8
C14
0.6 0.4
Ile 0.2
4 2
Leu IS
Arg
NH3+
C0
8
x106
Ion Response
C2
O-
N H
1.2 1.0
O H
1.4x10
NH2+
C4
4 - OH - Pro Pro Cit allo - Ile Met
C16
C3
Arg
C8
Phe Tyr Orn
0 10
Val
11
12
13
14
15
16
5 - oxo - Pro
0.0 12
13
14
15
16
17
18
Time (min)
Fig. 14.1. (a) High-efficiency metabolite separations by CE in free solution and multivariate ion desorption model in ESIMS, which allows for (b) isomeric resolution of a wide variety of polar metabolites in filtered DBS extracts with minimal sample handling (reproduced from ref. 5 (a) and Ref. 6 (b) with permission from ACS).
metabolites from filtered RBC lysates under acidic conditions (5), where excess sodium ions (Na+ ) and reduced glutathione (GSH) represent major co-ion interferences that are fully resolved (see Note 16). 3.6. Targeted Profiling of Anionic Metabolites
1. Introduce all samples to the capillary using a low-pressure hydrodynamic injection by first injecting a sample prepared in 10 mM HCl/5 mM ammonium acetate, pH 2.0, for 75 s
CE–ESI-MS-Based Metabolomics
239
(a) O
C4-DC-M
Gln
C5-3-OH
H2N
allo-Ile
O O
2.0x107
OH
GSH
Isomeric Ions: 232 m/z
Salt Front
IS
C4-Butyryl
3
Reduced Glutathione
4
1.5
Na
x10
Apparent Ion Response
O O
2
At
C4-Isobutryl
1
1.0 0 12.0
12.5
13.0
MeA
0.5
Thea Orn
Carn MeHis
Sero DopN Cyst TyrN PABA
ClTyr
OHTrp Cit
NTyr
G MeAsp
0.0 6
8
10
12
14
Migration Time (min)
(b)
O
17-OH-Preg-S
HO
OH H H
Gal
H
C6-Gly H
OH O OH
3-OH-PA
N
HO
O
OH
Glucuronides
Apparent Ion Response
50x103
Sulfates
E23G
Positional Isomers
40
E13S
E13G
30
E217G
E33S
E33G
20
E23S
10
6.0
6.2
6.4
6.6
6.8
7.0
7.2
7.4
Migration Time (min)
Fig. 14.2. Representative 2D chemical structures of diverse classes of (a) cationic and (b) anionic metabolites and their isomers amenable to CE–ESI-MS and a series of extracted ion electropherograms showing direct analysis of cationic metabolites derived from red blood cell lysates without ion suppression effects (Na+, GSH)∗ 21 and anionic estrogen glucuronide/sulfate conjugates in urine (reproduced from ref. (5) (a) with permission from ACS, whereas (b) represents unpublished data).
at 50 mbar followed by a 60 s injection of the alkaline BGE at 50 mbar prior to voltage (see Notes 12 and 13). 2. Perform all separations for anionic metabolites using unmodified fused-silica capillaries under normal polarity by CE with negative-ion-mode ESI-MS detection (see Note 17).
240
Britz-McKibbin
3. Figure 14.2b highlights the direct analysis of several estrogen (e.g., E1, E2, E3) sulfate (S) and glucuronide (G) conjugates spiked in urine without chemical derivatization or desalting steps, which also allows for resolution of positional isomers, such as E2 -3-G and E2 -17-G (see Note 18). 4. Various classes of weakly and strongly acidic metabolites can also be analyzed by CE–ESI-MS, including organic acids, nucleotides, acylglycines, steroid conjugates, and sugar derivatives. However, the ionization efficiency of small polar metabolites tends to be poor in ESI-MS that compromises concentration sensitivity (5) notably under negativeion-mode detection due to ionization suppression effects caused by co-ion electrolytes used in the BGE (e.g., acetate) (23). 5. Due to the strong EOF under alkaline conditions in CE, both cations and anions can be simultaneously analyzed; however, in practice, the migration time window is too narrow for adequate resolution of cationic metabolites from background interferences to allow for reliable quantification (see Section 3.5). 3.7. Comprehensive Metabolomics for Biomarker Discovery
1. Perform sample pretreatment and CE–ESI-MS analysis for cationic and anionic metabolites as described in previous sections. 2. Neutral metabolites and poorly responsive/low-abundance polar metabolites are undetected by this method without selective chemical derivatization (23) or the use of a complementary hyphenated separation platform, such as LC–ESIMS or GC–EI-MS. 3. Perform data analysis by first manually processing each EIE at nominal mass resolution (∆m/z = 1) over a total range of 50–750 m/z under both positive and negative-ion modes in CE–ESI-MS (see Note 19). 4. Perform noise filtering and data normalization (see Note 20) prior to multivariate analysis by processing each putative metabolite signal as a single-paired variable denoted as m/z:RMT, which facilitates data alignment among replicates and experimental conditions. 5. Most of the data pre-processing procedures described above can be more efficiently performed with new advances in integrated software for CE–ESI-MS (13), although manual inspection of data quality is still strongly advised. 6. Perform unsupervised data exploration and pattern recognition of pre-processed metabolomic datasets (see Note 21) by principal component analysis (PCA) and hierarchal cluster analysis (HCA) using commercial multivariate software, such
CE–ESI-MS-Based Metabolomics
241
as MATLAB. These methods are particularly useful for classification and differentiation of biologically relevant metabolites in complex datasets. 7. Figure 14.3a depicts a 3D heat map of dynamic changes in intra-cellar metabolism of erythrocytes as a result of a 52-min acute bout (time interval: 45–82 min) of exhaustive exercise for a healthy volunteer (e.g., standardized ergometer cycling), where each segment represents a unique metabolite defined by its characteristic m/z:RMT, which are sequenced using a HCA algorithm (24). Various groupings of metabolites can be deduced from their characteristic time-dependent normalized peak area pattern from 0 h (preexercise) until 6 h (post-exercise/recovery). 8. Figure 14.3b illustrates a PCA (i.e., control and trial) and differential PCA (i.e., control–trial, refer to inset) scores plot, which allows for clear identification of putative biomarkers of oxidative stress associated with the study (24), two of which were subsequently identified (see Note 22) as oxidized glutathione (GSSG) and L-carnitine (C0).
Heat Map:Control Pr e- exercise Exhaustive exercise
(b)
Post-exercise (Recovery)
241 m/z: 4 132 m/z: 0 .953 RMT 1.054 RMT
204 m/z: 0.983 RMT
*
–10
–15
0 –5
* *
* *
307 m/z: 1.321 RMT
-2
-6
5
-8
307 m/z: 1.321 RMT
4
-10 -12
2
241 m/z: 1.054 RMT
204 m/z: 0.983 RMT
0
–3
2 –2
162 m/z: 0 .929 RMT
–4
Time (min)
(c)
0
-4
6 PC2
*
Metabolite ( m/z : RMT)
Hierarchical Cluster Analysis
PC1 (68.3%)
8
308 m/z: 1.447 RMT
2
162 m/z: 0 .929 RMT
10
PC2 (24.5%)
(a)
12
7 132 m/z: 0 .953 RMT
17
22
PC1
Control
NAC
308 m/z: 1.447 RMT
180 150 120
X2
Concentration (μM)
90 60
GSSG
30 0 100
0
200
300
C0
80 60
X6
40 20 0 0
100
200
300
Time (min)
Fig. 14.3. Differential metabolomics strategy using CE–ESI-MS for the discovery of biomarkers of exercise-induced oxidative stress and the assessment of the efficacy of nutritional intervention (e.g., NAC pretreatment) to delay the onset of fatigue, where (a) 3D heat maps of global perturbations in erythrocyte metabolism, (b) differential PCA for detection of putative biomarkers of oxidative stress, and (c) unknown metabolite identification and quantification, such as oxidized glutathione (GSSG) and L-carnitine (C0).
242
Britz-McKibbin
9. Figure 14.3c shows that GSSG and C0 can serve as early and late-stage biomarkers of oxidative stress, respectively, based on their unique time-dependent concentration profiles (24).
4. Notes 1. Alternative CE systems can also be coupled with other MS instruments using various sheath and sheathless interface designs; however, analyses of large sample sizes can be challenging without full system integration/automation. Indeed, there is increasing use of CE in conjunction with time of flight (TOF)/MS in metabolite profiling applications due to the improved robustness, mass resolution, and cost effectiveness of recent commercial instruments (25). 2. The polyimide coating at the distal end of the capillary (≈ 2–3 cm) that is directed toward the ion source was removed by burning it with a match and subsequent cleaning using a Kimwipe wetted with a few drops of methanol (MeOH). This capillary was then installed into the coaxial sheath liquid interface to generate a bare fused-silica tip that served as the effective emitter in the ion source. 3. Ultrafiltration is the deproteinization method of choice when processing whole blood or RBC lysate samples without oxidation artifacts, which are generated with protein denaturation (e.g., oxygenases, hemoglobin) with acid or organic solvent precipitation (7). 4. The solvent of the dried blood spot extract can also be evaporated under a gentle stream of N2 and reconstituted in 20 µL of sample solution in order to improve concentration sensitivity, which can provide an additional 10-fold sample enrichment prior to CE–ESI-MS. 5. Blood collection and sample workup should be processed within about 2 h, while blood specimens are kept on ice (4◦ C) at all times using degassed/pre-chilled solutions in order to avoid premature hemolysis and oxidation artifacts. 6. Creatinine is determined under acidic BGE conditions with positive-ion mode by CE–ESI-MS. The creatinine concentration is then used to normalize all other reported urinary metabolite concentration levels (i.e., µmol/mmol creatinine), which corrects for dilution effects when using singlespot morning urine samples. 7. Reproducible alignment of the capillary emitter into the ion source was facilitated by the Agilent coaxial sheath liquid ESI-MS interface, where the distal end of the bare
CE–ESI-MS-Based Metabolomics
243
fused-silica capillary was allowed to protrude from the sprayer by about 0.1 mm in order to minimize postcapillary dilution effects. 8. The average capillary lifespan is about 3 weeks when analyzing complex biological samples by CE–ESI-MS. In most cases, capillary replacement is made evident by an unstable capillary current and/or increasing noise in the total ion electropherogram that cannot be improved with subsequent conditioning or rinsing with BGE. The incorporation of a daily quality control test run is strongly recommended to ascertain capillary condition. Also, periodic cleaning of sprayer assembly and ion source is recommended to improve long-term performance. 9. In some cases, a small fraction of organic solvent may be added to the BGE (e.g., 15% (v/v) acetonitrile or methanol) in order to improve the solubility of hydrophobic analytes or suppress micellar formation of surface-active analytes, such as long-chain acylcarnitines. 10. In all cases, the current generated during voltage application in CE should be kept under 50 µA in order to minimize Joule heating effects and improve ion signal stability. 11. These conditions have been found to be appropriate for comprehensive metabolite profiling of various classes of metabolites when using full-scan ion monitoring in CE– ESI-MS. However, optimization of coaxial sheath liquid, ion source, and/or MS scanning conditions is recommended for targeted metabolite profiling to lower detection limits since ionization efficiency in ESI-MS is highly dependent on solute physicochemical properties. 12. The first long sample injection plug (≈ 10% of capillary length) allows for online sample preconcentration of lowabundance metabolites directly in capillary during electromigration prior to ionization based on differences in co-ion electrolyte mobility and pH at the sample and BGE interface. Up to a 50-fold improvement in concentration sensitivity can be realized by this method without compromising separation efficiency or resolution when using conventional instrumentation (9). 13. The second injection sequence is used to displace the original sample plug within the capillary past the electrode interface at the inlet (anode), which is required to avoid CEinduced oxidation artifacts when analyzing low micromolar levels of oxidized glutathione (GSSG) in the presence of excess reduced glutathione (GSH) (7). 14. Under these conditions, neutral metabolites (e.g., urea) co-migrate with the suppressed EOF (>20 min), whereas strongly acidic anions (e.g., chloride) migrate out of the
244
Britz-McKibbin
capillary at the inlet upon voltage application. The RMT of an ion can be accurately predicted in CE based on its characteristic absolute mobility (µo ) and thermodynamic pKa as derived from its putative chemical structure (9), which provides a novel strategy for identification of unknown metabolites complementary to ESI-MS. 15. Stereoselective analysis of allo-Ile from dried blood spot extracts allows for specific diagnosis of maple syrup urine disease without false positives associated with total Leu, whereas resolution of 4-OH-Pro prevents misdiagnosis of hydroxyprolinemia, a benign disorder. 16. A variety of classes of metabolites can be analyzed by this method, including amino acids, (biogenic) amines, peptides, acylcarnitines, and nucleosides, where the relative response factor of recently identified/novel metabolites can be predicted based on their fundamental physicochemical properties in CE–ESI-MS (5). 17. Note that separations performed under reverse polarity (inlet/cathode) conditions for anionic metabolites with cationic polymer-coated capillaries can lead to corrosion of the stainless steel electrode (i.e., ESI spray needle) in CE–ESI-MS with poor long-term robustness (26). Although platinum-based electrodes have been suggested as a solution to this problem, our protocol uses inexpensive, unmodified, bare fused-silica capillaries under normal polarity with a conventional stainless steel spray needle. 18. In this case, E2 -17-G migrates with a longer migration time (i.e., higher negative mobility) under alkaline conditions (pH > 9) relative to E2 -3-G due to its weakly acidic (unmodified) phenol moiety. Similarly, native estrogens can also be analyzed by CE–ESI-MS due to their weakly acidic phenol (pKa ≈ 10.2) functionality.
19. Each EIE was manually integrated with a minimum signal threshold over background noise of S/N ≈ 10 in order to avoid signal artifacts. Also, subsequent data pre-processing was performed to eliminate redundant signals originating from the same ion, including isotope contributions, fragment ions, salt adducts, as well as common ions detected under both ± ESI-MS conditions for unequivocal quantification of unique metabolite signatures. 20. Due to potential signal artifacts when analyzing complex biological samples, quantification of metabolite features in the total ion electropherogram was performed for peaks above a signal/noise threshold of >10. All integrated peak areas were then normalized to the internal standard. Data alignment was performed manually by
CE–ESI-MS-Based Metabolomics
245
labeling each metabolite feature as a paired variable, namely mass/charge:relative migration time (m/z:RMT) that had a corresponding normalized peak area, which allowed for easy comparison of datasets among different replicates and/or samples despite differences in apparent migration times caused by EOF. 21. All multivariate analyses were pre-processed as standardized datasets (X-matrix), where mean-centered responses for all metabolites were normalized to the inverse of the standard deviation determined at each experimental condition (e.g., time variable), which provided an effective way to scale different ion responses having different absolute values and ranges. 22. Unambiguous metabolite identification was realized by a series of experiments involving the acquisition/interpretation of multi-stage mass spectra when using a 3D ion trap mass analyzer, which was followed by an extensive search for candidate metabolites on public databases. Lead metabolite candidates were rationally selected based on several criteria, including equivalent nominal mass, characteristic fragments/isotope patterns as well as relevant intrinsic charge states deduced by their putative chemical structure. Once a series of lead candidates were selected, a comparison of their measured (or predicted) RMTs by CE and/or spiking of filtered RBC lysate samples with authentic standards (if available) was performed for unambiguous identification.
Acknowledgments The author wishes to acknowledge funding support from National Science and Engineering Research Council of Canada, Premier’s Research Excellence Award, as well as a Japan Society for Promotion of Science – Invitation Fellowship.
References 1. Smith, R. D., Barinaga, C. J., Udseth, H. R. (1988) Improved electrospray ionization interface for CE-MS. Anal Chem 60, 1948–1952. 2. Maxwell, J. E., Chen, D. D. Y. (2008) Twenty years of interface development for capillary electrophoresis–electrospray
ionization–mass spectrometry. Anal Chim Acta 627, 25–33. 3. Nilsson, S. L., Bylund, D., Jornten-Karlsson, M., Petersson, P., Markides, K. W. (2004) A chemometric study of active parameters and their interaction effects in a nebulized sheath–liquid electrospray interface for
246
4.
5.
6.
7.
8. 9.
10. 11. 12.
13.
14.
15.
Britz-McKibbin capillary electrophoresis–mass spectrometry. Electrophoresis 25, 2100–2107. Mokaddem, M., Gareil, P., Belgaied, J. -E., Varenne, A. (2009) New insight into suction and dilution effects in CE coupled to MS via an ESI interface II – dilution effect. Electrophoresis 30, 1692–1697. Chalcraft, K. R., Lee, R., Mills, C., BritzMcKibbin, P. (2009) Virtual quantification of metabolites by CE–ESI-MS: predicting ionization efficiency without chemical standards. Anal Chem 81, 2506–2515. Chalcraft, K. R., Britz-McKibbin, P. (2009) Newborn screening of inborn errors of metabolism by CE–ESI-MS: a second-tier method with improved specificity and sensitivity. Anal Chem 81, 307–314. Lee, R., Britz-McKibbin, P. (2009) Differential rates of glutathione oxidation for assessment of cellular redox status and antioxidant capacity by CE-MS: an elusive biomarker of oxidative stress. Anal Chem 81, 7047–7056. Hruska, V., Janos, M., Gas, B. (2006) Simul 5 – free dynamic simulator of electrophoresis. Electrophoresis 27, 984–991. Lee, R., Niewzcas, L., Britz-McKibbin, P. (2007) Integrative metabolomics for characterizing unknown low-abundance metabolites by CE-MS with computer simulations. Anal Chem 79, 403–415. Ramautar, R., Somsen, G. W., Jong, G. Jd. (2009) CE-MS in metabolomics. Electrophoresis 30, 276–291. Monton, M. R. N., Soga, T. (2007) Metabolome analysis by CE-MS. J Chromatogr A 1168, 237–246. Huhn, C., Ramautar, R., Wuhrer, W., Somsen, G. W. (2010) Relevance and use of capillary coatings in capillary electrophoresis– mass spectrometry. Anal Bioanal Chem 396, 297–314. Sugimoto, M., Hirayama, A., Ishikawa, T., Robert, M., Baran, R., Uehara, K., Kawai, K., Soga, T., Tomita, M. (2010) Differential metabolomics software for CEMS data analysis. Metabolomics 6, 27–41. 10.1007/s11306-009-0175–1. Williams, B. J., Cameron, C. J., Workman, R., Broeckling, C. D., Sumner, L. W., Smith, J. T. (2007) Amino acid profiling in plant cell cultures: an inter-laboratory comparison of CE-MS and GC-MS. Electrophoresis 28, 1371–1379. Buscher, J. M., Czernik, D., Ewald, J. C., Sauer, W., Zamboni, N. (2009) Crossplatform comparison of methods for quantitative metabolomics of primary metabolism. Anal Chem 81, 2135–2143.
16. Kato, M., Onda, Y., Sekimoto, M., Degawa, M., Toyo’oka, T. (2009) A CEC–ESI-MS method for simultaneous analysis of charged and neutral constituents of a hepatocarcinoma cell metabolome. J Chromatogr A 1216, 8277–8282. 17. Amundsen, L. K., Kokkonen, J. T., Soren, H. (2008) Comparison of partial filling MEKC analyses of steroids with use of ESI-MS and UV spectrophotometry. J Sep Sci 31, 803–813. 18. Mol, R., Jong, G. D., Somsen, G. W. (2005) On-line capillary electrophoresis–mass spectrometry using dopant-assisted atmospheric pressure photoionization: setup and system performance. Electrophoresis 26, 146–154. 19. Nordstrom, A., Want, E., Northen, T., Lehtio, J., Siuzdak, G. (2008) Multiple ionization mass spectrometry strategy used to reveal the complexity of metabolomics. Anal Chem 80, 421–429. 20. Ohashi, Y., Hirayama, A., Ishikawa, T., Nakamura, S., Shimizu, K., Ueno, Y., Tomita, M., Soga, T. (2008) Depiction of metabolome changes in histidine-starved Escherichia coli by CE–TOFMS. Mol Biosyst 4, 135–147. 21. Kind, T., Fiehn, O. (2006) Metabolomic database annotations via query of elemental compositions: mass accuracy is insufficient even at less than 1 ppm. BMC Bioinform 7, 234–244. 22. Soga, T., Ohashi, Y., Ueno, Y., Naraoka, H., Tomita, M., Nishioka, T. (2003) Quantitative metabolome analysis using capillary electrophoresis mass spectrometry. J Proteome Res 2, 488–494. 23. Yang, W. C., Regnier, F. E., Adamec, J. (2008) Comparative metabolite profiling of carboxylic acids in rat urine by CE– ESI MS/MS through positively pre-charged and (2)H-coded derivatization. Electrophoresis 29, 4549–4560. 24. Lee, R., West, D., Phillips, S. M., Britz-McKibbin, P. (2010) Differential metabolomics for quantitative assessment of oxidative stress with strenuous exercise and nutritional intervention: thiol-specific regulation of cellular metabolism with N-acetylL-cysteine. Anal Chem 82, 2959–2968. 25. Staub, A., Schappler, J., Rudaz, S., Veuthey, J. -L. (2009) CE–TOF/MS: fundamental concepts, instrumental considerations and applications. Electrophoresis 30, 1610–1623. 26. Soga, T., Igarashi, K., Ito, C., Mizobuchi, K., Zimmermann, H.-P., Tomita, M. (2009) Metabolomic profiling of anionic metabolites by CE-MS. (2009) Anal Chem 81, 6165–6174.
Chapter 15 Liquid Chromatography-Mass Spectrometry (LC-MS)-Based Lipidomics for Studies of Body Fluids and Tissues Heli Nygren, Tuulikki Seppänen-Laakso, Sandra Castillo, Tuulia Hyötyläinen, and Matej Orešicˇ Abstract In this paper, analytical methodologies for the global profiling of lipids in serum and tissue samples are reported. The sample preparation is based on a modified Folch extraction, and the analysis is carried out with ultrahigh-performance liquid chromatography combined with mass spectrometry (UPLC-MS). For further identification, MSn mass spectrometry is carried out utilizing an LTQ-Orbitrap mass spectrometry as the detector. Such a system affords determination of accurate masses and is thus a highly useful tool for lipid identification. The repeatability of the analysis proved to be good, with relative standard errors for spiked samples being between 4.51 and 10.44%. The throughput of the methodology described here is over 100 samples a day. Key words: Lipidomics, liquid chromatography, mass spectrometry.
1. Introduction Lipids are a broad group of naturally occurring molecules, which include fats, waxes, sterols, monoglycerides, diglycerides, triglycerides, phospholipids, and other hydrophobic or amphiphilic small molecules. Lipids originate entirely or in part from two distinct types of building blocks: ketoacyl and isoprene groups (1). They are a functionally as well as a structurally very diverse group of compounds in part due to the many possible variations of lipid building blocks and the different ways of noncovalent linkage (2).
T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_15, © Springer Science+Business Media, LLC 2011
247
248
Nygren et al.
Lipids play important roles in biological systems as signaling or energy storage molecules and as major constituents of cellular membranes (3). Tight control of membrane lipid composition is of central importance for the maintenance of normal cellular physiology, and its dysregulation may affect, e.g., membrane fluidity as well as topology, mobility, or activity of membranebound proteins. Even minor changes of lipid composition may affect the membrane properties (4) and thus the physiology of the cell. To be able to detect such changes, sensitive and accurate platforms are needed to measure the lipids in biological systems. Global characterization of lipids in biological samples, i.e., lipidomics, remains a challenging task. In recent years, lipidomics has emerged owing to the advances in mass spectrometry-based analytical technologies, which allow detection and quantitation of hundreds of intact lipid molecular species in parallel (5, 6). These global lipidomics approaches include shotgun lipidomics, which uses direct infusion of lipid extracts into the mass spectrometer (7, 8), as well as liquid chromatography coupled to mass spectrometry (LC-MS) (9, 10). Both approaches have their own advantages and disadvantages. In both methodologies, sample preparation is a critical step and should be fast, repeatable, and nonselective. The shotgun approach is advantageous because it is a relatively simple and rapid way to profile the lipids in crude lipid extracts. However, ion suppression is the major complication in this methodology (11, 12). To avoid ion suppression, more careful sample pre-treatment, using, e.g., fractionation, is often needed, increasing the total analysis time. The advantage of LCMS-based methodologies, on the other hand, is potentially higher sensitivity as well as the ability, by using a non-targeted strategy, to detect and identify novel lipids. However, with conventional LC-MS methods, the analysis time can be relatively long. Moreover, the LC-MS methods can suffer from carryover effects, which must be taken into account in the method development. It should also be noted that due to the extremely large number of different lipids, coelution cannot be totally avoided. Coelution of lipids can be reduced by using novel ultrahigh-pressure liquid chromatography such as UPLCTM , which allows highly efficient separation in a very short analysis time (13). Here we describe a UPLC-MS-based global lipidomics platform, with applications to serum and tissue sample profiling. As an integral component of the platform, identification of lipids using UPLC-MSn is also described. In the identification stage, an LTQOrbitrap is used as the detector because this system provides outstanding mass accuracy and mass resolution. The ability to detect accurate masses allows the unequivocal compositional and structural elucidation of the compounds.
LC-MS-Based Lipidomics
249
2. Materials 2.1. Samples
1. Samples: 10 µL serum/plasma or cell concentrate (centrifugation residue, e.g. macrophages, islets, adipocytes); 5 mg of heart, liver, muscle or fat tissue (e.g. white or brown adipose tissue).
2.2. Standards and Chemicals
The internal standard mixture contains compounds from several different lipid classes: phosphatidylcholines (PC), a phosphatidylethanolamine (PE), a ceramide (Cer), a phosphatidylserine (PS), a phosphatidic acid (PA) as well as mono-, di-, and triacylglycerols (MG, DG, and TG, respectively). Lipids are denoted by their molecular composition as follows: :/< Number of carbon atoms in the second fatty acid moiety>:. For example, the abbreviation PC(17:0/17:0) indicates a phosphatidylcholine comprising two C17 fatty acids with no double bonds. 1. Standard mixture 1A added to plasma and cell samples before extraction: PC(17:0/0:0), PC(17:0/17:0), PE (17:0/17:0), PG(17:0/17:0)[rac], Cer(d18:1/17:0), PS(17:0/17:0), and PA(17:0/17:0) (Avanti Polar Lipids, Inc., Alabaster, AL, USA) and MG(17:0/0:0/0:0)[rac], DG(17:0/17:0/0:0)[rac] and TG(17:0/17:0/17:0) (Larodan Fine Chemicals, AB, Malmö, Sweden). The concentration of each standard is approximately 0.2 µg/sample. 2. Standard mixture 1B added to fat tissue samples (after homogenization) or other tissue samples (before homogenization): PC(17:0/0:0), PC(17:0/17:0), PE(17:0/17:0), and Cer(d18:1/17:0) (Avanti Polar Lipids, Inc.) and TG(17:0/17:0/17:0) (Larodan Fine Chemicals). The concentration of each standard is 0.5–1 µg/sample. 3. Standard mixture 2 added after extraction: PC (16:1/0:0D3 ), PC(16:1/16:1-D6 ), and TG(16:0/16:0/16:0-13 C3) (Larodan Fine Chemicals). The concentration of each standard is 1 µg/tissue sample and 0.1–0.2 µg/serum or cell sample. 4. Extraction solvents: mixture of HPLC-grade chloroform/methanol (2:1, v/v).
2.3. UPLC-MS Systems
1. Acquity Ultra Performance LCTM (UPLC; Waters Corporation, Wexford, Ireland). 2. Sample organizer for automatic sampling.
250
Nygren et al. R 3. LC column: Acquity UPLC BEH C18 1.7 µm, 2.1 mm × 100 mm column (Waters Corporation).
4. Solvent A: 1% 1 M NH4 Ac and 0.1% HCOOH in water. 5. Solvent B: AcCN/2-propanol (1:1, v/v), 1% 1 M NH4 Ac, and 0.1% HCOOH. 6. Quadrupole-time-of-flight (Q-Tof) Premier mass spectrometer (Waters Corporation). 7. LTQ-Orbitrap mass spectrometer (Thermo Scientific Corp., San Jose, CA). 2.4. Other Instrumentation
1. Ultrasonication bath (Fritsch, Laborette 17, Idar-Oberstein, Germany). 2. Retsch Mixer Mill MM400 homogenizer (Retsch GmbH, Haan, Germany). 3. Multiskan EX instrument (Thermo Scientific, Inc., Waltham, MA, USA) for spectrophotometric determination. 4. Micro BCATM Protein Assay Kit (Pierce, Rockford, IL, USA). 5. Vortexer. 6. Centrifuge.
2.5. Data Processing Software
1. MZmine 2 software for peak detection, alignment, and normalization is used to process raw MS data (http://mzmine.sourceforge.net/). Supported data formats are mzML (version 1.0 and 1.1), mzXML (versions 2.0, 2.1, and 3.0), mzData (versions 1.04 and 1.05), NetCDF (no MSn support), and Thermo RAW (only on Windows with Thermo Xcalibur version 2).
3. Methods The UPLC-MS-based lipidomics platform consists of multiple steps (Fig. 15.1), including sample preparation, separation and detection, and finally data analysis. Total lipid extracts are obtained using a modified Folch extraction and the extracts are analyzed by UPLC-MS in positive- (ESI+) and/or negative (ESI–)-ion mode. The choice of ionization may depend on the biological questions asked. To avoid potential bias, it is important that the samples are randomized prior to sample preparation and analysis. In large-scale analyses, it is common that the samples are analyzed in multiple batches of several hundred samples (14, 15).
LC-MS-Based Lipidomics
Tissue sample 5 mg
Cell concentrate 10 µl
Addition of PBS-buffer/NaCl Homogenisation
251
Serum sample 10 µl Addition of NaCl
Addition of internal standards Liquid extraction CHCl3:MeOH (2:1) Centrifugation Collection of lower phase Addition of external standards
Protein content determination
UPLC-QTof-MS
UPLC-LTQ-Orbitrap: MS/MS
Data analysis: MZmine integration, alignment of the data, identification, normalization
Fig. 15.1. Typical UPLC-MS lipidomics sample analysis workflow.
Peaks are detected, aligned, and calibrated with internal standards in specialized software packages such as MZmine 2. The identification of lipids may be performed automatically based on an internal database of mass-to-charge ratio (m/z) and retention time values (16). However, both identified and unidentified lipids are usually included in further data analyses. The most important peaks, either for validation purposes if known or for de novo identification if unknown, need to be characterized further by mass spectrometry (UPLC-MS/MS or UPLC-MSn ). To minimize carryover effects, relatively strong solvent is used as the UPLC eluent. No significant carryover was noticed, proved by blank analyses after real samples. The repeatability of the analysis proved to be good, with RSD of the spiked standards varying from 4.51% (TG standard) to 10.44% (PE standard) (n = 10). 3.1. Sample Preparation
1. Serum/plasma: a. Dilute 10 µL of serum/plasma with 10 µL of 0.15 M (0.9%) sodium chloride and add internal standard mixture 1A such that the concentration of internal standard is 0.2 µg/10 µL serum/plasma. b. Extract serum/plasma lipids using 100 µL of HPLCgrade chloroform/methanol (2:1, v/v). c. Vortex the mixture for 2 min, then allow to stand for 30 min.
252
Nygren et al.
d. Centrifuge the sample for 3 min at 7800×g. e. Collect the lower phase and add 10–20 µL of internal standard mixture 2 such that the concentration of internal standard is 0.1–0.2 µg/10 µL serum/plasma. 2. Cells: a. Dilute 10 µL of cell concentrate with 100 µL of PBS buffer or 0.15 M NaCl (0.9%) and add standard mixture 1A such that the concentration of internal standard is 0.2 µg/10 µL serum/plasma. b. Homogenize in an ultrasonicating bath. c. Remove 5 µL of homogenized cell suspension for determination of protein content. Dilute this aliquot with PBS buffer for use in the Micro BCATM Protein Assay Kit and determine protein concentration on a Multiskan EX instrument or equivalent. The lipid concentrations in cells will be normalized with the measured protein content. d. Extract cellular lipids by combining 20 µL of the homogenate with 100 µL HPLC-grade chloroform/ methanol (2:1, v/v). e. Vortex the mixture for 2 min and allow to stand for 30 min. f. Centrifuge the sample for 3 min at 7800×g. g. Collect the lower phase and add 10–20 µL of internal standard mixture 2 such that the concentration of internal standard is 0.1–0.2 µg/10 µL cell concentrate. 3. Tissues: a. Add 10–20 µL of internal standard mixture 1B and to 5 mg of heart, liver, or muscle sample such that the concentration of internal standard is 0.5–1 µg/5 mg of sample and homogenize with the extraction solvent (HPLC-grade chloroform/methanol, 2:1, v/v) and three grinding balls (Ø 3 mm) for 2 min (20 Hz) with mixer mill homogenizer. b. Add 50 µL of 0.15 M NaCl (0.9%) to the homogenate, vortex the mixture for 2 min, and allow to stand for 30 min. c. Centrifuge the sample for 3 min at 7800×g. d. Collect the lower phase and add 10–20 µL of internal standard mixture 2 such that the concentration of internal standard is 1 µg/5 mg tissue sample. 4. Adipose tissue: a. Add 100 µL of 0.15 M (0.9%) sodium chloride or PBS buffer to 5 mg of white or brown adipose tissue and homogenize with three grinding balls (Ø 3 mm) for 2 min (20 Hz) with mixer mill homogenizer.
LC-MS-Based Lipidomics
253
b. Dilute 10 µL of the homogenate with 10–40 µL of 0.15 M NaCl (0.9%), add 10–20 µL of internal standard mixture 1B such that the concentration of internal standard is 0.5–1 µg/5 mg of sample, and combine with 100–400 µL of HPLC-grade chloroform/methanol (2:1, v/v), vortex the mixture for 2 min, and allow to stand for 30 min. c. Centrifuge the sample for 3 min at 7800×g. d. Collect the lower phase and add 10–20 µL of internal standard mixture 2 such that the concentration of internal standard is 1 µg/5 mg tissue sample. 1. The UPLC gradient is as follows: set to start at 35% B to reach 80% B in 2 min, 100% B in 7 min and remain there for 7 min. The total run time including a 4-min reequilibration step at initial conditions is 18 min. An example chromatogram is shown in Fig. 15.2.
3.2. Lipidomics Profiling Platform (UPLC-MS)
2. Set the temperature of the column to 50◦ C. 3. Set the temperature of the sample organizer to 10◦ C. 4. Set the flow rate to 0.400 mL/min. 5. Inject 2 µL of lipid extract (chloroform phase). 6. Lipid compounds are detected by a Q-Tof mass spectrometer using electrospray ionization in positive- or negative-ion mode. 7. Collect data in continuum mode using extended dynamic range over a mass range of m/z 300–1200 with a scan duration of 0.2 s (see Note 1). 8. Process data using the MZmine 2 software (17, 18) (see Note 2) or equivalent: Phospholipids
Triacylglycerols
%
100
Lysophospholipids
0 1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
Time (min)
Fig. 15.2. Total ion chromatogram (positive ion mode) from UPLC-MS-based lipidomics analysis of serum (Waters QTof Premier as mass analyzer) showing major lipid fractions.
254
Nygren et al.
a. Data processing includes alignment of peaks, peak integration, normalization, and identification. b. Lipids are identified using an internal spectral library or with tandem mass spectrometry (see below and Fig. 15.3). A
866.81738
100 90
[M+NH4]+
Δ M = +0.3 ppm
Relative Abundance
80 70 867.82080
60 50 40 30
868.82434
20 10 0 865
B
866
867
868 m/z
870
871
579.50
100
Neutral loss of RCOOH+NH3
90
Relative Abundance
869
+
[M+H-270]
80 70 60 50 40 30 20 10 0
Relative Abundance
C
160 150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0
250
300
350
400
450
500
550 600 m/z
650
700
750
800
850
[C17H32OH]+ Δ M = –1.6ppm
253.25217
354.30007 235.24130 220 240 260 280 300 320 340 360 380 400 420 440 460 m/z
Fig. 15.3. UPLC LTQ-Orbitrap lipidomics analysis of human serum extract in positive-ion mode. (a) Orbitrap highresolution mass spectrum at retention time 9.2 min, showing the peak of internal standard TG (17:0/17:0/17:0). (b) MS2 spectrum of the peak with m/z 866.8, showing the product ion with m/z 579.6. (c) Orbitrap high-resolution MS3 spectrum, showing the product ion of protonated aldehyde.
LC-MS-Based Lipidomics
255
c. The data is normalized using one or more internal standards representative of each class of lipid present in the samples; the intensity of each identified lipid is normalized by dividing it with the intensity of its corresponding standard and multiplying it by the concentration of the standard. All monoacyl lipids except cholesterol esters, such as monoacylglycerols and monoacylglycerophospholipids, are normalized with PC (17:0/0:0), all diacyl lipids except ethanolamine phospholipids are normalized with PC (17:0/17:0), all ceramides with Cer (d18:1/17:0), all diacyl ethanolamine phospholipids with PE (17:0/17:0), and TG and cholesterol esters with TG (17:0/17:0/17:0). Other (unidentified) molecular species are calibrated with PC (17:0/0:0) for retention time 224,000 Kovats retention index values for ∼22,000 compounds, while the Wiley database contains ∼400,000 EI mass spectra and >180,000 chemical structures. Together these two databases form an important resource for metabolite identification using GC-MS. The GOLM open access database (27) at the Max Planck Institute of Molecular Plant Physiology focuses on EI data and also acts as a GC-MS data repository. Well-characterized metabolites may be identified through database searches, and if samples are analyzed using highresolution MS, many candidates can be excluded at this stage. For unambiguous metabolite identification, co-chromatography and comparison of MS/MS data with the authentic compound are necessary (28, 29). In the case of an unknown molecule, de novo
Processing and Analysis of GC/LC-MS-Based Metabolomics Data
295
identification is required – a significant challenge given the often trace quantities of some metabolites and limited sample amount. Elemental composition can be obtained using a combination of the following: 1. Accurate mass determinations using high-resolution FTMS or Orbitrap instrumentation 2. High accurate tandem mass measurements (e.g. on a Q-TOF, LTQ-FT or LTQ-Orbitrap) for structural characterization However, despite recent advances, the lack of comprehensive MS libraries and databases often hinders metabolite identification based solely on MS information. Ultimately, a combination of technologies will be required for metabolite identification. These include high-sensitivity capillary NMR, chemical modification for functional group identification and finally independent compound synthesis for verification.
4. Notes 1. Parameter optimization. It is crucial to optimize the data preprocessing parameters for each metabolomics study, as every dataset will differ. Parameter optimization is not straightforward and can be time consuming, as some input parameters that need to be selected for peak alignment are not always simple to determine. Key peak picking parameters include the signal-to-noise (S/N) threshold, peak width parameters and peak shape parameters. The optimum values for settings such as peak width should be easy to determine and in general are constant for all chromatograms within a dataset. Conversely, parameters that define real chromatographic peaks vs. noise or window sizes in which peaks in two chromatograms are considered the same are more difficult to determine. Criteria have to be set for determining the alignment quality, which can be subjective. Some studies use control samples that are analyzed both spiked with known compounds and unspiked to check the analytical system and the method, as well as evaluating peak alignment quality. 2. Baseline correction, noise reduction and smoothing. Correct use of baseline correction and noise reduction tools is essential. Different GC-MS and LC-MS technologies may require specific parameter settings and algorithms, which should be optimized by each vendor, or perhaps the instrument operator. Therefore, as a rule, smoothing and baseline correction are best performed by vendor and system-specific software applications.
296
Want and Masson
3. Scores and loadings plots. Scores and loadings plots are 2D representations of the scores and loadings space. It is important to keep in mind how much variation is explained by the plotted PCs. When many principal components are necessary to explain all the variations in the data, precaution is needed when interpreting the plots. 4. Predictive performance. It is important to note that OPLS and PLS have equivalent prediction performance: an OPLS model that uses 1 predictive component and n orthogonal components is equivalent in terms of prediction quality to a PLS model using n+1 components. However, the model interpretation is easier with OPLS. 5. Cross-validation. The basic idea of cross-validation is to exclude part of the data, construct a new model on remaining data, predict the omitted data with this model and compare the predictions with the actual values, this being done until all data have been removed once. The sum of the squared differences between predicted and observed values can be used as a measure of the predictive power of the model. 6. Model validation. A good way of validating a model, if the number of samples is sufficient, is to split the dataset into two parts: a training set and a test set (the latter containing typically at least one-third of the samples). A model is constructed and optimised using only the training set. Then, the developed model is used to predict the samples in the test set, which have not influenced the model. The goodness of fit on the test data is an indicator of the predictive ability of the model.
Acknowledgements The authors would like to acknowledge Dr. Timothy Ebbels for valuable discussions during the preparation of this chapter. EW would like to acknowledge Waters Corporation for funding. Perrine Masson would like to acknowledge Servier Laboratories Ltd. for funding. References 1. Nicholson, J. K., Connelly, J., Lindon, J. C., Holmes, E. (2002) Metabonomics: a platform for studying drug toxicity and gene function. Nat Rev Drug Discov 1, 153–161. 2. Fiehn, O. (2002) Metabolomics – the link
between genotypes and phenotypes. Plant Mol Biol 48, 155–171. 3. Nicholson, J. K., Lindon, J. C. (2008) Systems biology: metabonomics. Nature 455, 1054–1056.
Processing and Analysis of GC/LC-MS-Based Metabolomics Data 4. Trygg, J., Holmes, E., Lundstedt, T. (2007) Chemometrics in metabonomics. J Proteome Res 6, 469–479. 5. Smith, C. A., Want, E. J., O’Maille, G., Abagyan, R., Siuzdak, G. (2006) XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal Chem 78, 779–787. 6. Burton, L., Ivosev, G., Tate, S., Impey, G., Wingate, J., Bonner, R. (2008) Instrumental and experimental effects in LC-MS-based metabolomics. J Chromatogr B Anal Technol Biomed Life Sci 871, 227–235. 7. Scholz, M., Gatzek, S., Sterling, A., Fiehn, O., Selbig, J. (2004) Metabolite fingerprinting: detecting biological features by independent component analysis. Bioinformatics 20, 2447–2454. 8. Wang, W., Zhou, H., Lin, H., Roy, S., Shaler, T. A., Hill, L. R., Norton, S., Kumar, P., Anderle, M., Becker, C. H. (2003) Quantification of proteins and metabolites by mass spectrometry without isotopic labeling or spiked standards. Anal Chem 75, 4818– 4826. 9. Oresic, M., Clish, C. B., Davidov, E. J., Verheij, E., Vogels, J., Havekes, L. M., Neumann, E., Adourian, A., Naylor, S., van der Greef, J., Plasterer, T. (2004) Phenotype characterization using integrated gene, protein and metabolite profiling. Appl Bioinform 3, 205–217. 10. Yeung, K. Y., Ruzzo, W. L. (2001) Principal components analysis for clustering gene expression data. Bioinformatics 17, 763–774. 11. Jolliffe, I. T. (2002) Principal Components Analysis, 2nd edn, Springer, New York, NY. 12. Ivosev, G., Burton, L., Bonner, R. (2008) Dimensionality reduction and visualization in principal components analysis. Anal Chem 80, 4933–4944. 13. Barker, M., Rayens, W. (2003) Partial least squares for discrimination. J Chemom 17, 166–173. 14. Bylesjo, M., Rantalainen, M., Cloarec, O., Nicholson, J. K. (2006) OPLS discriminant analysis: combining the strengths of PLSDA and SIMCA classification. J Chemom 20, 341–351. 15. Trygg, J., Wold, S. (2002) Orthogonal projections to latent structures (O-PLS). J Chemom 16, 119–128. 16. Trygg, J. (2002) O2-PLS for qualitative and quantitative analysis in multivariate calibration. J Chemom 16, 283–293. 17. Jackson, J. E. (2003) A User’s Guide to Principal Components, Wiley–Interscience, New York, NY.
297
18. Zelena, E., Dunn, W. B., Broadhurst, D., Francis-McIntyre, S., Carroll, K. M., Begley, P., O‘Hagan, S., Knowles, J. D., Halsall, A., HUSERMET Consortium, Wilson, I. D., Kell, D. B. (2009) Development of a robust and repeatable UPLC-MS method for the long-term metabolomics study of human serum. Anal Chem 81, 1357–1364. 19. Gika, H. G., Macpherson, E., Theodoridis, G. A., Wilson, I. D. (2008) Evaluation of the repeatability of ultra-performance liquid chromatography–TOF-MS for global metabolic profiling of human urine samples. J Chromatogr B Anal Technol Biomed Life Sci 871, 299–305. 20. Brereton, R. G. (2006) Consequences of sample sizes, variable selection, model validation and optimisation for predicting classification ability from analytical data. Trends Anal Chem 25, 1103–1111. 21. Anderssen, E., Dyrstad, K., Westad, F., Martens, H. (2006) Reducing over-optimism in variable selection by cross-model validation. Chemom Intell Lab Syst 84, 69–74. 22. Romero, P., Wagg, J., Green, M. L., Kaiser, D., Krummenacker, M., Karp, P. D. (2005) Computational prediction of human metabolic pathways from the complete human genome. Genome Biol 6, R2. 23. Wishart, D. S., Tzur, D., Knox, C., Eisner, R., Guo, A. C., Young, N., Cheng, D., Jewell, K., Arndt, D., Sawhney, S., Fung, C., Nikolai, L., Lewis, M., Coutouly, M. A., Forsythe, I., Tang, P., Shrivastava, S., Jeroncic, K., Stothard, P., Amegbey, G., Block, D., Hau, D. D., Wagner, J., Miniaci, J., Clements, M., Gebremedhin, M., Guo, N., Zhang, Y., Duggan, G. E., Macinnis, G. D., Weljie, A. M., Dowlatabadi, R., Bamforth, F., Clive, D., Greiner, R., Li, L., Marrie, T., Sykes, B. D., Vogel, H. J., Querengesser, L. (2007) HMDB: the human metabolome database. Nucleic Acids Res 35, D521–D526. 24. Ogata, H., Goto, S., Sato, K., Fujibuchi, W., Bono, H., Kanehisa, M. (1999) KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res 27, 29–34. 25. Smith, C. A., O’Maille, G., Want, E. J., Qin, C., Trauger, S. A., Brandon, T. R., Custodio, D. E., Abagyan, R., Siuzdak, G. (2005) METLIN: a metabolite mass spectral database. Ther Drug Monit 27, 747–751. 26. Babushok, V. I., Linstrom, P. J., Reed, J. J., Zenkevich, I. G., Brown, R. L., Mallard, W. G., Stein, S. E. (2007) Development of a database of gas chromatographic retention properties of organic compounds. J Chromatogr A 1157, 414–421.
298
Want and Masson
27. Kopka, J., Schauer, N., Krueger, S., Birkemeyer, C., Usadel, B., Bergmüller, E., Dörmann, P., Weckwerth, W., Gibon, Y., Stitt, M., Willmitzer, L., Fernie, A. R., Steinhauser, D. (2005) [email protected]: the Golm Metabolome Database. Bioinformatics 21, 1635–1638. 28. Weckwerth, W., Morgenthal, K. (2005) Metabolomics: from pattern recognition to
biological interpretation. Drug Discov Today 10, 1551–1558. 29. Saghatelian, A., Trauger, S. A., Want, E. J., Hawkins, E. G., Siuzdak, G., Cravatt, B. F. (2004) Assignment of endogenous substrates to enzymes by global metabolite profiling. Biochemistry 43, 14332–14339.
Chapter 18 Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling Eva M. Lenz Abstract The identification of drug metabolites in biofluids such as urine, plasma and bile is an important step in drug discovery and development. Proton nuclear magnetic resonance (1 H-NMR) spectroscopy can provide detailed information regarding the structural transformation of a compound as a consequence of metabolism. However, successful identification of drug metabolites by 1 H-NMR spectroscopy is generally compromised by the presence of endogenous metabolites, which can obscure the signals of the drug metabolites in question. Hence, sample clean-up and separation of the metabolites from the biofluid matrix is crucial. This is generally achieved by extraction of the biofluid, solid-phase extraction (SPE), high-performance liquid chromatography (HPLC) or any combination of these. Apart from 1 H, other NMR-active nuclei, such as 19 F, can provide a useful handle for metabolite profiling, provided they are not naturally present in the biofluid. Successful studies have shown that the presence of a fluorine-handle on the drug and its metabolites can provide additional qualitative and quantitative data by 19 F-NMR spectroscopy. This chapter provides guidelines and examples of NMR-based drug metabolite profiling. Key words: 1 H- and 19 F-NMR spectroscopy, SPE, HPLC, drug metabolites.
1. Introduction Proton nuclear magnetic resonance (1 H-NMR) spectroscopy is a very useful structural tool to assess the biotransformations of drugs, provided the drug metabolites are isolated from the biological matrix and present in concentrations of >50 µM when using conventional probes (although with new CryoProbe technology, detection limits are in the nanomolar region). A prerequisite of successful structural determination of metabolites by 1 H-NMR spectroscopy is the full structural characterisation of T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_18, © Springer Science+Business Media, LLC 2011
299
300
Lenz
G G
A)
G
B) S
S
S
G C
hippurate
P urea 7.5
7.0
6.5
6.0
5.5
ppm
2.15
2.10 ppm
creatinine
C)
9.5
9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
ppm
Fig. 18.1. The 1 H-NMR spectrum of urine from a healthy human volunteer 3–6 h after administration of paracetamol (dose = 13.7 mg/kg, i.e. 1 g in total) without any sample preparation. The 1 H-NMR spectrum was acquired with water
suppression (with noesypr1d), showing the characteristic resonances from the major urinary paracetamol metabolites, the paracetamol glucuronide (G) and the sulphate (S). The regions of interest are shown in the insets: (a) representing the aromatic region, as well as the diagnostic anomeric proton at δ1 H 5.15 of the paracetamol glucuronide moiety, and (b) representing the N-acetyl groups. The whole 1 H-NMR spectrum of the urine collected 3–6 h post-dose is shown in (c). Key: G, glucuronide conjugate; S, sulphate conjugate; C, cysteinyl conjugate and P, parent.
the parent molecule itself (see Note 1). According to metabolite concentration or amount extracted, several NMR experiments can be carried out to aid the structural identification of the metabolite (1). In the simplest case, drug metabolites can be detected and partially identified directly by 1 H-NMR spectroscopy of the biofluid without further sample workup, as in the example of paracetamol (APAP), for which a body of literature evidence exists (e.g. 2). The diagnostic resonances of APAP are visible in the 1 H-NMR spectra of urine and plasma, despite the presence of endogenous signals (Fig. 18.1). However, in most cases, drug metabolite identification by 1 H-NMR spectroscopy is obscured by the presence of endogenous metabolites (as shown in Fig. 18.2), as well as exogenous contaminants, such as solvent or dosing vehicles. Exogenous contaminants comprise solvent signals (e.g. ACN, MeOH) and pH modifiers (e.g. acetic and formic acid). This can be a challenge in hyphenated HPLC–NMR, as solvent concentrations are present in vast excess, resulting in large 1 H-NMR signals that require suppression to enable the drug metabolite to be digitised and detected, as shown in Fig. 18.3. Other typical contaminants are polyethylene glycol (PEG, from dosing solutions) or glycerol (contained in microsomal solutions), which can be present in high concentrations (Fig. 18.4; ref. 3).
301
water
TSP
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
A
B
C
9.5
9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
ppm
Fig. 18.2. 1 H-NMR spectra of typical biofluids, showing the wealth of signals from the endogenous metabolites contained. Biofluid profiles comprise complex mixtures of glucose, amino acids, organic acids, lipids and proteins. Whilst the urine spectrum is characterised by sharp peaks from small molecular weight components such as creatinine, urea, citrate, amino acids and glucose (amongst many other components), the plasma profile is characterised by a rolling baseline deriving from the protein macromolecules contained, embedding amino and organic acids, and additional broad signals from lipids. Amino acids, bile acids and lipids are the major contributors to the bile profile. Successful metabolite identification by 1 H-NMR spectroscopy relies on the removal of the interfering biological matrices. Key: (a) a human urine sample, (b) human plasma and (c) rat bile. All acquired with water suppression (with noesypr1d) and containing TSP as internal reference standard.
Unfortunately, these agents are UV silent, which can make sample isolation quite difficult (see Note 2). Endogenous contaminants are generally co-eluting biofluid constituents, which have similar polarity as the analyte in question. Again, some are UV silent, such as bile acids, hence during the HPLC-UV separation, these can co-elute unnoticed, as shown in Fig. 18.5. Therefore, every effort must be made to extract the metabolites of interest from the biological matrix in order to minimise the problem of signal overlap with either endogenous or interfering material such that the structures of the metabolites can
302
Lenz
ACN 13C
* *
satellites*
0.55% of ACN–peak
A
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
2.5
2.0
1.5
1.0
13C
ppm
ppm
satellites
B water
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
ppm
Fig. 18.3. These 1 H-NMR spectra represent a typical HPLC–NMR separation, using D2 O and ACN as mobile phase, and
highlight the problem solvent signals can pose. The spectra are acquired in stop-flow mode, without solvent suppression. (a) showing the dominant ACN peak and its 13 C satellites. Double-solvent suppression (b) achieves attenuation of the ACN peak (to below the 13 C satellite level) and, hence, enables digitisation and detection of the drug metabolite signals. The drug metabolite was derived from a 48-h pooled urine sample from a human volunteer, which was extracted first by step-gradient SPE, where it eluted in the 50% MeOH fraction. The dried SPE fraction was then further separated by hyphenated HPLC–NMR (the experimental conditions are detailed in ref. (11)).
be fully elucidated. However, there is no single way to achieve drug metabolite separation and identification by 1 H-NMR spectroscopy due to the differences in biofluid matrices and the concentration and chemical properties of the metabolites (4, 5). Hence, metabolite isolation typically involves extraction, protein precipitation, solid-phase extraction or HPLC separation to obtain a pure enough drug metabolite sample to result in successful identification by NMR spectroscopy. In addition, several extraction steps might be required in order to isolate the metabolite in question from the biofluid matrix. Perhaps the least used extraction method is liquid–liquid extraction, as the more polar metabolites might not partition into the organic phase (see Note 3). Hence, generally, chromatographic separations such as SPE or HPLC are used for the isolation of metabolites from the biological matrix. HPLC is a high-resolution chromatographic method enabling the separation and isolation of drug metabolites from endogenous metabolites. However, co-elution can still be an issue with metabolites of similar polarity and, hence, elution characteristics. Therefore, retention characteristics have to be assessed
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
303
A
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
2.8
ppm
B
3.8
4.4
4.2
4.0
3.8
3.6
3.4
3.7
3.2
3.0
3.6
2.8
ppm
ppm
Fig. 18.4. 1 H-NMR spectra of the typical exogenous UV-silent contaminants (a) glycerol, often contained in the medium of microsomal preparations, and (b) PEG, often used as a dosing vehicle. These components co-extract or co-elute undetected but can obscure large sections of the NMR spectrum and even pose a dynamic range problem when present in high concentrations compared to the drug metabolite in question.
with the parent compound, prior to loading of the biofluid or the biofluid extract, in order to ascertain maximum retention of the drug metabolites. In general, HPLC is coupled with UV or DAD detection; however, when dealing with a radiolabelled compound, radiodetection (HPLC–RAD) greatly facilitates the identification of drug metabolites, provided the label is preserved in the metabolites. Drug metabolite isolation and identification can be achieved off-line (e.g. by SPE or HPLC; see Notes 4 and 5) or, if the concentrations of the drug metabolites are sufficiently high (e.g. >50 µM), on-line by HPLC–NMR or HPLC–SPE–NMR spectroscopy, which enables the separation and structural characterisation by 1 H-NMR spectroscopy in a single step. Alternatively, 19 F-NMR spectroscopy has been successfully utilised (6–8), as it provides a sensitive and specific NMR ‘handle’ and does not
acetate
Lenz
MeOH–d4
304
taurine
C18
C19
C21
*
*
* * taurocholic acid 4.0
3.5 21
OH 18 19
HO
H
3.0
2.5
2.0
1.5
1.0
0.5
ppm
1
ppm
CONHCH2CH2HSO3
taurocholic acid
H
OH
biliary drug metabolite*
9
8
7
6
5
4
3
2
Fig. 18.5. 1 H-NMR spectrum of an isolated biliary drug metabolite contaminated with taurocholic acid, which co-eluted undetected during the HPLC separation as it is UV-silent. In this case, the presence of the taurocholic acid signals did not perturb metabolite identification, although the cholesterol-backbone occupies a large section of the aliphatic region (0.7–2.2 ppm). Note: this HPLC separation was carried out with acetic acid buffer, hence the additional signal in the spectrum from acetate.
suffer from biological contamination and solvent interferences. 19 F-NMR spectroscopy is 83% as sensitive as 1 H-NMR spectroscopy and, with its large spectral width (∼200 ppm), the signals are generally nicely dispersed. Assuming that no fluorinated solvent or buffer is used, the observed signals solely derive from the drug and its metabolites, providing a metabolite profile and information on the relative amounts and concentrations of the metabolites present, prior to their isolation. The metabolite profile can also be assessed quantitatively against an internal standard, if required, as shown in Fig. 18.6. For HPLC–NMR spectroscopy, due to the spectral simplicity, it can be used as an additional or even sole detector in order to identify metabolites during the HPLC separation (e.g. Fig. 18.7). When used in conjunction with off-line fractionation, each fraction can be examined for 19 F content, giving additional evidence as to whether the fractions contain a drug metabolite (Fig. 18.8, see Note 6). This chapter provides guidelines and examples of NMR-based drug metabolite profiling.
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
305
Is 24–48h
C Is 8–24h
B 0–8h
Is
A
Fig. 18.6. 19 F-NMR profile of neat rat urine samples collected at (a) 0–8 h, (b) 8–24 h and (c) 24–48 h intervals following a single i.p. dose of 50 mg/kg of 2-trifluoromethylacetanilide. Addition of an internal standard (Is) enabled the quantification of the metabolites contained (experimental details described in ref. (8)). Reprinted from (8), with permission from Elsevier.
2. Materials 2.1. Sample Preparation, Extraction and Isolation 2.1.1. Freeze-drying of Samples or Large Volumes of Sample Extracts
1. Freeze-drier (e.g. Edwards Modulya 4K or VirTis Sentry 2.0). 2. Glass vials with screw-top neck. 3. Liquid nitrogen (–196◦ C) or solid carbon dioxide (Cardice, –140◦ C) to freeze the sample.
2.1.2. Liquid–Liquid Extraction
1. Biological fluid (i.e. aqueous samples, generally acidified to suppress ionisation). 2. Water-immiscible solvents (e.g. HPLC-grade chloroform; Fisher Scientific, Loughborough, UK).
306
Lenz
A)
C)
B) UV-chromatogram rows
A time
* transfer-time
* *
F2 retention time/min.
Fig. 18.7. The pseudo-2D contour plot of an HPLC separation of a rat urine concentrate monitored by 19 F-NMR spectroscopy. The pseudo-2D contour plot shows the individual metabolites of 2-trifluormethylacetanilide (dosed at 50 mg/kg) represented by their CF3 -handles (a). The relationship between the pseudo-2D experiment and the chromatogram is schematically depicted (b), enabling the assessment of their exact chromatographic retention times. The HPLC separation was then repeated in stop-flow mode, where the chromatographic separation is halted at the retention times of interest. Once the metabolite fractions are trapped inside the NMR flow-cell, 1 H-NMR spectra can be acquired for structural identification of the metabolites (c) (the experimental details described in ref. (8)). Reprinted from (8), with permission from Elsevier.
3. Separating funnel. 4. Solvent evaporator or water bath. 2.1.3. Solid–Liquid Extraction
1. Biological fluid (e.g. urine or bile), dry or freeze-dried, or faeces. 2. Organic solvent (e.g. HPLC-grade methanol (MeOH) or acetonitrile (ACN); Fisher Scientific). 3. Centrifuge (e.g. Megafuge 1.0R; Heraeus Instruments, Newport Pugnell, UK). 4. Solvent evaporator, water bath and/or freeze-drier.
2.1.4. Plasma or Serum Protein Precipitation
1. Organic solvent (e.g. HPLC-grade ACN; Fisher Scientific). 2. Centrifuge. 3. Solvent evaporator, water bath and/or freeze-drier.
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
A0013 100
47.59
Channel 1 An1 4.09e5
A0146 3.94
100
Parent drug
A
307
7.02 9.48 11.11 13.39 13.86 17.60
An1 1.00e6
46.49 47.45
31.61 37.33
B 21.95
43.97
32.69 23.89 25.42
45.94
19.47
27.56
39.20
%
%
18.69
3.94
19.80
55.18
42.76 28.12
42.06
49.04
54.72
33.57
59.83
29.92
34.25
50.06
52.92 50.55
41.45
59.86
0
Time
10.00
20.00
30.00
40.00
50.00
0
60.00
5.00
Time
10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00 50.00 55.00 60.00
**
C
*
FA Metabolite*
water impurities
9.5
9.0
8.5
8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
D
2.0
1.5
1.0
ppm
19F
111
112
113
114
115
116
117
118
119
120
121
122
123
124
ppm
Fig. 18.8. An example of an off-line HPLC separation of a drug metabolite from 400 mL of human urine. The fluorinecontaining drug was metabolised extensively (28 metabolites eluting over 55 min, as determined by HPLC-MS (data not shown)). The isolated HPLC fractions were subsequently analysed by 19 F-NMR for ‘ metabolite content’ and identified by 1 H-NMR for structural identification. Shown are (a) HPLC method development with parent compound (eluting at 48 min at 30% ACN), (b) the HPLC-UV trace of an aliquot (80 µL) of urine concentrate collected into 96-well plates, and the 1 Hand 19 F-NMR spectra (c and d) of a metabolite isolated by HPLC. Key: FA, formic acid, contained in the mobile phase.
2.1.5. Precipitation of Bile Salts
1. Acid or acidified water (e.g. formic acid, HPLC-grade acidified water at pH 2; Fisher Scientific). 2. Centrifuge (for small volumes, such as 2 mL aliquots, e.g. Eppendorf Centrifuge 5417C, Hamburg, Germany). 3. Organic solvent (HPLC-grade MeOH or ACN; Fisher Scientific) for extraction of the precipitate, in case metabolites or parent drug have also precipitated. 4. Solvent evaporator, water bath and/or freeze-drier. 5. Alternatively, samples may be freeze-dried and extracted with organic solvent as described in Section 3.1.3.
2.1.6. Solid-Phase Extraction (SPE) Chromatography
1. 1–3 mL SPE columns of suitable packing material (e.g. R Waters Oasis HLB or Whatman C18 packing). 2. Organic solvent (e.g. HPLC-grade MeOH; Fisher Scientific).
308
Lenz
3. Aqueous phase (e.g. HPLC-grade water at pH 2, acidified with HCL or formic acid; Fisher Scientific). 4. Solvent mixtures to enable step-gradient elution of the analytes (e.g. MeOH:aqueous, 0:100, 20:80, 40:60, 60:40, 80:20, 100:0% (vol/vol); or even finer steps, e.g. 10% MeOH). 5. Vacuum manifold to aid elution of the chromatographic fractions. 2.1.7. High-Performance Liquid Chromatography (HPLC)
1. HPLC column (appropriate for analyte, e.g. Prodigy 5 µm ODS; Phenomenex, UK; BDS Pursuit 3 µm PFP; Varian, Netherlands). 2. HPLC pump and UV detector (e.g. Agilent Technologies, West Lothian, UK) or radiodetector if compound is radiolabelled (e.g. 14 C, Berthold radiodetector, Harpenden, UK, equipped with solid cell). 3. Fraction-collector or glass vials for manual collection. 4. Mobile phases: HPLC-grade water and ACN (Fisher Scientific). 5. Buffers of choice: e.g. HPLC-grade ACN and aqueous phase, pH adjusted with analytical-grade formic acid (e.g. 0.1% formic acid, v/v; Fisher Scientific) and/or buffered with analytical-grade ammonium formate (e.g. 10 mM; Fisher Scientific) (see Note 7).
2.2. NMR Spectroscopy 2.2.1. NMR Spectroscopy
1. NMR tubes (ideally of small volume not to dilute sample unnecessarily, e.g. 1-, 2.5- or 3-mm tubes, or the use of Shigemi tubes) and corresponding NMR probes. 2. Suitable deuterated NMR solvents (e.g. D2 O, MeOH-d4 , ACN-d3 , dimethyl sulphoxide (DMSO-d6 ); Goss Scientific, Nantwich, UK).
2.2.2. Hyphenated HPLC–NMR
1. HPLC–NMR system comprised of an integrated HPLC pump allowing stop-flow analysis (e.g. Bruker Agilent HPLC system, operated via HystarTM software) and an LC– NMR flow probe (e.g. 1 H/19 F dual tunable (LCDXI), with 60 µL cell volume). 2. HPLC solvents: D2 O (Eurisotop, Giv-sur-Yvette, France) and ACN (Pestanal Grade, Riedel-De-Haën, Sigma-Aldrich, Dorset, UK), or ACN-d3 (99.8 atom %D(it says on the bottle); Sigma-Aldrich). The most suitable acid to adjust the pH of the mobile phase is formic acid-d2 (Goss Scientific, Nantwich, UK).
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
2.2.3. Hyphenated HPLC–SPE–NMR
309
1. Dedicated HPLC–SPE system (e.g. Bruker Agilent (HPLC) Spark Prospekt 2 (SPE) system) with Hystar SPE cartridges (Spark, Netherlands). 2. Dedicated LC–SPE flow probe (e.g. 1 H/19 F dual tunable (LCDXI), with 30 µL cell volume). 3. Non-deuterated mobile-phase constituents for initial HPLC separation: water and ACN (Pestanal Grade, Riedel-DeHaën, Sigma-Aldrich); pH adjusted with formic acid. 4. Deuterated ACN-d3 (99.8 at.% D; Sigma-Aldrich) for elution of SPE cartridge.
3. Methods 3.1. Sample Preparation, Extraction and Isolation 3.1.1. Freeze-drying
Freeze-drying is an ideal way to concentrate biofluid samples and to reduce large volumes of biofluid extracts and off-line collected SPE or HPLC fractions (see Note 8). 1. Ensure complete evaporation of organic solvents from sample prior to freeze-drying. This applies to biofluids extracted with organic solvents such as MeOH. 2. Freeze small aliquots (∼1–5 mL) of biofluids or biofluid extracts in liquid nitrogen (–196◦ C) or solid carbon dioxide (Cardice, –140◦ C) in screw-neck glass vials placed at a slight angle. 3. Operate freeze-dryer according to instruction manual.
3.1.2. Liquid–Liquid Extraction
1. Liquid–liquid extraction is based on the partition coefficient of an analyte between two immiscible phases. This is probably the least used extraction method in drug metabolite studies, as metabolites generally exhibit a range of polarities; hence, the extraction efficiency can be inferior. 2. Where appropriate, biofluids are pH adjusted (e.g. acidified prior to extraction in order to suppress ionisation of the metabolites by drop-wise addition of an acid while measuring the pH of the biofluid). 3. Add equal volumes of the acidified biofluid and chloroform into a separating funnel. Mix vigorously, then gently release pressure. To avoid inhalation of chloroform, this procedure must be carried out in a fume hood.
310
Lenz
4. Allow the phases to separate and carefully decant and collect the chloroform phase, then the aqueous phase. The chloroform phase ought to contain the drug and its metabolites. 3.1.3. Solid–Liquid Extraction
1. This refers to solid material (e.g. rodent faeces) or freezedried biofluid samples (e.g. urine or bile). 2. If large quantities of human urine are to be extracted, subaliquot the samples into smaller volumes (e.g. 5–10 mL) and freeze-dry. 3. Extract the material successively in aliquots of organic solvent (e.g. 1–5 mL MeOH), mix thoroughly, centrifuge at +4◦ C for 10 min at 1800×g, then carefully remove the supernatant. Repeat the process several times and combine the supernatant fractions. 4. Evaporate the supernatant solvent.
3.1.4. Plasma Protein Precipitation
1. This can be achieved twofold, either via protein crashing or ultrafiltration (see Note 9). 2. Protein crashing requires aliquots of plasma to be combined with ACN (1:3, v/v), mixed and centrifuged at +4◦ C for 10 min at 1800×g. 3. Collect and combine the supernatants and evaporate the ACN.
3.1.5. Precipitation of Bile Salts
1. This is a useful approach to remove excess salt in bile samples, as precipitated bile salts can block the SPE or HPLC columns when large quantities (e.g. dog bile or pooled samples) are to be loaded. Mix aliquots of bile with acid or acidified water (∼1:1, v/v) (see Note 10). 2. Centrifuge the precipitate for 5 min at 8000×g and collect the supernatant. Repeat process several times and collect and combine the supernatants. 3. Freeze-dry supernatant extracts prior to separation by SPE or HPLC, if required, and reconstitute in the initial mobile phase. 4. To avoid loss of metabolites, additionally extract the precipitate–pellet with organic solvent (e.g. MeOH), evaporate the solvent and combine with the supernatant. 5. As an alternative approach, especially when dealing with large volumes of bile (e.g. dog bile), the sample can be freeze-dried and then extracted with an organic solvent (see Section 3.1.3)
3.1.6. SPE Chromatography
1. This section assumes a reversed-phase column (e.g. C8 or C18 ).
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
311
2. Prepare elution solvents, such as acidified water (e.g. pH 2, ∼100 mL), MeOH (∼100 mL) and stock solutions of MeOH:acidified water mixtures with increasing MeOHcontent, to allow step-gradient elution (e.g. 20% MeOH or finer, e.g. 10% MeOH (v/v), ∼20 mL total volume each). 3. Acidify sample to pH 2 and centrifuge at +4◦ C for 10 min at 1800×g, if necessary (in case of precipitation). 4. Condition the column/cartridge before loading the sample, according to the manufacturer’s requirements. For a 3-mL C18 column, typical sorbent equilibration conditions are detailed below: (a) Elute column with MeOH (∼3 mL). (b) Follow by elution with acidified water (e.g. pH 2) (∼3 mL). 5. Make up the parent drug in an aqueous solvent acidified to pH 2. Prior to loading of the biofluid sample, ensure that the parent drug is retained on the column and elutes in the higher MeOH-fractions (e.g. 80% MeOH). 6. Once retention of parent compound is ensured, load acidified sample (onto the freshly equilibrated column). 7. Collect unretained material (known as eluate). 8. Elute with acidified water and collect fraction (known as acid wash). 9. Elute with a step gradient (e.g. 10–100% or 20–100% MeOH, v/v) and collect every fraction separately. 10. Evaporate MeOH, reconstitute in water, readjust pH of the fractions back to pH 7 and freeze-dry fractions prior to NMR analyses. 3.1.7. HPLC
1. This section assumes description of HPLC in one of the earlier chapters (here, the procedure using a reversed-phase column is described as an example). For reversed-phase columns, conditioning of the HPLC column-packing material with the mobile phase (e.g. to ensure elution of potential contaminants from previous analyses) is required prior to loading of the biofluid sample or extract. This generally means elution with, e.g. ACN and aqueous phase (both buffered with, e.g. 0.1% formic acid, v/v, or 10 mM ammonium formate), followed by a steep gradient, e.g. 10–80% ACN over 20 min. 2. Develop a suitable HPLC method to maximise retention of the parent compound and to ensure good peak shape. 3. Equilibrate the HPLC column well with the initial mobile phase, prior to loading of the sample.
312
Lenz
4. Inject small volumes of biofluid or biofluid extract and collect fractions of interest into individual glass vials. 5. Combine fractions and evaporate solvent prior to freezedrying and NMR analysis. 3.2. NMR Spectroscopy
3.2.1. 1 H-NMR Spectroscopy
These instructions assume the use of a Bruker 600-MHz NMR spectrometer; however, they should be easily adapted to other instruments. As for every NMR experiment, acquisition of spectra involves locking to the solvent and tuning and matching of each sample, followed by shimming to achieve optimum line-shape. 1. A prerequisite of structural determination of metabolites is the full structural characterisation of the parent molecule itself. Hence, ideally, the parent drug should be dissolved in the same solvent or a similar solvent system as the metabolites in order to avoid differences in chemical shifts. Once the parent spectrum has been fully assigned, spectral modifications as a consequence of metabolism can be observed and interpreted on the isolated metabolites of the chromatographic fractions with more confidence (see ref. 1). 2. Depending on the amount of metabolite extracted, several NMR experiments can be carried out to aid the structural identification of the metabolite. However, owing to the typical low concentration of the metabolites, the choice of NMR experiments can be limited. Generally, 1 H-NMR experiments involve solvent suppression. Typical 1D 1 HNMR experiments include zgpr, noesypr1d (for singlesolvent suppression) and lc1pnf2 (for double-solvent suppression), whilst typical 2D NMR experiments comprise COSY (cosyphpr) and TOCSY (mlevphpr). Heteronuclear experiments include 1 H–13 C HSQC (hsqcetgpsisp2) and HMBC (hmbcgplpndqf). 3. Dissolve the isolated metabolite in the smallest amount of suitable solvent possible (e.g. 0.2–0.5 mL). Ideally, this would be the same solvent as the parent itself, in order to compare chemical shift changes as a result of biotransformation (see Notes 1 and 11). 4. If solvent suppression is required, define the exact frequency of the solvent peak to be suppressed and set as the offset frequency (O1). Ensure that the spectral width is sufficiently large to capture all the signals. Typical values are sw = 20 ppm and td = 64 k, resulting in an acquisition time of 2.7 s. Allow a further 2.3 s for d1 between successive scans, i.e. ensuring a pulse repletion time of ∼5 s. Typically, the solvent signal itself (e.g. the water contained in D2 O) or the water peak contained in the organic
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
313
solvent (e.g. MeOH-d4 or ACN-d3 ) might require suppression; hence, set the O1 onto the solvent peak and apply a pulse sequence such as zgpr or noesypr1d. For suppression with zgpr, set the power level for suppression (pl9) to approximately 70 dB. For noesypr1d, a mixing time of 100 ms (delay d8) is recommended and pl9 of ∼60– 70 dB for the suppression during the relaxation delay. For noesypr1d, the measurement of the 90◦ pulse width is required. 5. Small amounts of isolated metabolite (e.g. 10 µg and below) can require long acquisition times (e.g. 1–4 k scans); hence, temperature control is advisable. 3.2.2. 19 F-NMR Spectroscopy
1. This assumes that the compound dosed or administered contains a fluorine-handle, which is unlikely to be lost as a consequence of metabolism. 2. Initially, the whole biofluid can be examined to obtain a 19 F-NMR metabolite profile, revealing the number and relative concentrations of the metabolites. For this purpose, the biofluid sample may require concentrating by freeze-drying and reconstituting in a smaller volume of a suitable solvent (e.g. D2 O or MeOH-d4 ). 3. Fluorine atoms are quite sensitive to structural modifications due to metabolism (approx. up to eight bonds from the fluorine atom), causing changes in the chemical shift compared to the parent peak (Fig. 18.6; ref. 8). 4. Biofluid fractions (isolated metabolites) can be ‘screened’ for fluorine in addition to the 1 H-NMR spectrum. Although 19 F-NMR spectroscopy does not provide structural information, it gives additional evidence that the metabolite is contained in the sample (e.g. Fig. 18.8). 5. The acquisition of a 19 F-NMR spectrum involves setting up an experiment for 19 F detection (e.g. rpar 19 F or 19 FCPD) and tune and match the sample on the 19 F channel. Fluorine spectra are run with either pulse programs zg (if 1 H decoupling is not required, i.e. for a CF3 group) or zgfhigqn (for 1 H decoupling). 6. Again, acquire a spectrum of the parent compound first. This allows optimisation of the correct O1 and pulse width measurements. Typically, set O1 to –100 ppm (for monofluoro) or –60 ppm (for CF3 groups), sw = 100 ppm and td = 128 k. For 19 F-NMR experiments, pulse widths smaller than 90◦ can be used (e.g. zg30). For 1 H decoupling, the decoupling power pl12 needs to be measured or calculated (in edprosol) to give a 90◦ decoupling pulse of 80 µs (pcpd2). Cpdpg2 should be set to waltz 16.
314
Lenz
7. Depending on metabolite concentrations, lengthy acquisition times can be required; hence temperature control is advisable. 3.2.3. HPLC–NMR Spectroscopy
1. These instructions assume the use of a Bruker NMR spectrometer equipped with a dedicated HPLC system (e.g. Bruker Agilent HPLC system) operated via Bruker’s HystarTM software. This set-up also includes a dedicated LC flow probe (e.g. 3-mm 1 H/19 F dual-tunable LCDXI, with 60 µL cell volume) and a DAD-UV detector. 2. This description assumes that the transfer time between the UV detector and the flow probe is calibrated so that the peak of interest is reliably eluted into the NMR flow-probe. 3. Again, the HPLC method has to be developed and optimised with the parent compound to ensure retention of the metabolites. 4. The mobile phase typically consists of D2 O (instead of H2 O) and ACN (not necessarily deuterated), both buffered with 0.1% formic acid-d2 (v/v). 5. It is not atypical to load the entire biofluid extract to capture all the metabolites contained therein (see ref. 9 and Note 12). Hence, the biofluid extract or freeze-dried biofluid requires reconstitution in a very small volume (e.g. 100 µL) of the initial mobile phase (see Note 13). 6. During the HPLC gradient, the metabolites can be collected either in time-slice mode (in timed intervals, e.g. every 10 s) or in stop-flow mode, halting the pump at given UV peaks. 7. Once the peak is captured in the NMR flow cell, the ‘captured fraction’ requires tuning and matching, as well as reshimming and defining of the frequencies for suppression of the mobile-phase peaks (i.e. the residual water in the D2 O and the ACN peak). Double-solvent suppression is achieved with the lc1pnf2 pulse sequence set-up via rpar LC1D12. Generally, O1 is set on the larger peak (generally the ACN peak) and O2 on the residual water peak. Typical values are mixing time (d8) of 100 ms, 70 dB for pl9 (for suppression of the larger peak, defined by O1) and 80 dB for pl21 (for the smaller solvent peak, defined by O2). This is followed by ‘rga’ (automatic receiver gain adjustment) prior to acquisition of the spectra (as described in Section 3.2.1). Alternatively, this can be run in automation, rpar LC1D12, xaua (au-prog au_lc1d). 8. Again, as the acquisitions can be lengthy (1–8 k scans), depending on metabolite concentration, temperature control is advisable.
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
315
9. Post-acquisition, the individual peaks can be collected straight from the LC probe outlet if further purification or analysis (e.g. MS) is required. 10. If the metabolites contain a fluorine handle, each chromatographic fraction can be additionally screened by 19 FNMR spectroscopy once captured inside the LC-probe. This gives supporting evidence that the metabolite in question is indeed drug related. Here, the solvent signals are invisible; hence, no solvent suppression is required. The acquisition is as in Section 3.2.2. 11. Alternatively, if the metabolites are sufficiently concentrated in the biofluid sample (e.g. rodent urine) and enough biofluid sample is available, then 19 F-NMR can be utilised in the on-flow mode in order to identify the exact retention times of the metabolites during the HPLC separation, as shown in Fig. 18.7. This takes advantage of the specificity of fluorine and the lack of interfering solvent signals. 12. For on-flow HPLC–NMR, a pseudo-2D experiment is acquired (see Note 14). This comprises the acquisition of 1D experiments (representing the F2 axis) over the period of the chromatographic separation (represented by F1). In order to mirror the chromatographic separation, the residence times of the fractions in the flow cell are limited; hence, the quality of corresponding NMR spectra is typically compromised. Typically, the flow rate is reduced (e.g. 0.5 mL/min); however, the data acquisition by NMR has to be accelerated to collect sufficient data of the fractions flowing through the flow cell. Typical NMR parameters, therefore, include a reduced number of scans (e.g. 16–24 scans per increment, a spectral width of ∼60 ppm, 32 k data points, an acquisition time of ∼1 s and a very short relaxation delay, e.g., of 1 s). The required pulse program is lc2 (without decoupling, ideal for CF3 groups) or lc2pg (if 1 H decoupling is required). The duration of the separation has to be ‘covered’ by collecting sufficient increments (in eda). Hence, calculate the time required to acquire the 1D spectrum and enter the required number of increments in F1 for the full chromatographic separation to complete. When the pseudo-2D data set is acquired (or even during the acquisition), the pseudo-2D matrix is converted with xf2. This achieves Fourier transformation in F2 only (as F1 represents a time axis). 13. After completion of the on-flow experiment, the exact retention times of the metabolites, based on the presence of the 19 F signals in the pseudo-2D matrix, can be evaluated.
316
Lenz
The separation is then repeated, stopping the separation on the signals of interest (in stop-flow mode), capturing the metabolite inside the flow cell and acquiring 1 H-NMR spectra as described in Section 3.2.1. 3.2.4. HPLC–SPE–NMR Spectroscopy
1. These instructions assume the use of a Bruker NMR spectrometer equipped with a dedicated HPLC–SPE system (Bruker Agilent Spark Prospekt 2-system) operated via Bruker’s HystarTM software. This set-up also includes a dedicated LC flow probe (e.g. 3-mm, 1 H/19 F dual-tunable LCDXI, with 30 µL cell volume) and a DAD-UV detector. 2. Here, the HPLC separation is developed as in conventional HPLC. The mobile phase is not required to be deuterated. Again, method development with the parent compound is required before loading of the sample, in order to ensure optimum retention of the metabolite peaks. 3. In SPE–NMR, it is imperative that the SPE cartridges are pre-equilibrated, prior to loading of the HPLC fractions/peaks, in order to avoid ‘breakthrough’. This requires knowledge of the number of expected metabolites to be captured. 4. Once the HPLC fractions of interest are identified, they are transferred onto the SPE cartridges, a process during which the fractions are diluted with water in order to ensure retention. 5. Several trapping steps can be carried out, i.e. the HPLC separation can be repeated three times and the fractions/peaks are trapped onto the same cartridge every time (see ref. 10). 6. The trapped SPE fractions are dried with nitrogen gas and finally eluted with a single organic solvent, typically ACNd3 , into the LC–SPE–NMR probe head, where 1 H- or 19 FNMR experiments can be acquired.
4. Notes 1. Ideally, the parent drug ought to be dissolved in the same solvent as the metabolites in order to avoid differences in chemical shifts that may occur in different solvents. However, in some cases, this is not possible, and a suitable solvent should be used that dissolves the parent fully. Generally speaking, the metabolites are more polar than the parent and might be soluble in D2 O and MeOH-d4 . Both
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling
317
of these solvents can be easily freeze-dried or evaporated if further experiments with the metabolite are required. 2. Due to the high concentrations of these agents, several repeat chromatographic ‘washing steps’ are recommended, where the analyte is loaded onto the column (SPE or HPLC) and several wash steps (elutions with the aqueous phase) are performed. 3. In order to achieve partitioning into the organic phase, pH adjustment, e.g. acidification of the biofluid prior to liquid– liquid extraction, ensures ion suppression/protonation. 4. SPE is a low-resolution chromatographic method and a range of column-packing materials and sizes are commercially available to suit the properties of the drug and its metabolites. Optimisation of retention has to be carried out with drug parent prior to loading of the biofluid matrix. 5. SPE allows large volumes of biological fluid (e.g. 20 mL) to be eluted, the fractions of which can subsequently be dried and reconstituted in small volumes of suitable solvent prior to NMR analysis or further HPLC separation. In many cases, owing to the low concentration of the metabolites excreted (urine, bile) or contained (plasma), it might be necessary to extract the total volume of the biofluid. This can be a problem in human studies when dealing with large quantities, e.g. several litres of urine collected over 24–48 h, containing the metabolites in microgram quantities. Here, samples can be initially cleaned up and separated by SPE chromatography. This allows, provided the metabolites have a range of polarities, separation of the metabolites from the majority of endogenous metabolites and also enables concentration of the metabolites. 6.
19 F-NMR
signals can be prone to chemical shift differences, mainly due to solvent or pH differences. Thus, the exact chemical shift values are not always compatible between the overall metabolite profile and the isolated metabolites.
7. Avoid buffers that give rise to NMR signals, such as acetic acid (acetate signal). Opt for ammonium formate or formic acid instead of the acetic acid equivalents. Formate has the advantage of evaporating during the freeze-drying process. For hyphenated HPLC–NMR, deuterated formic acid is generally used (see Section 2.2.2). 8. Freeze-drying and solvent evaporation are generally carried out several times during the sample preparation stages, in order to produce a dry isolated extract (ideally containing only the analyte/metabolite in question) which can then be dissolved in a suitable deuterated solvent for NMR analysis.
318
Lenz
9. Protein crashing has the advantage that the sample will be mixed with a single solvent (ACN), rather than being exposed to a filter that contains glycerol. For ultrafiltration, the filter has to be cleaned several times with water to get rid of the glycerol. Hence, due to the higher contamination risk, the ultrafiltration process is not discussed here. 10. Bile salt precipitation can lead to some loss of metabolites, which can also co-precipitate in the process. Hence, it might be advisable to dilute the sample and inject in smaller aliquots, with plenty of aqueous wash cycles prior to reloading. For SPE, the frit sitting on the surface of the sorbent bed can be punctured. 11. Ensure that the solvent signals do not obscure important signals in the spectrum. This can be assessed with the parent compound prior to the analysis of the metabolites. 12. As a precaution, the entire HPLC separation should be collected (post-NMR flow probe), in case the sample elutes in the void volume. 13. Make sure the injection loop is sufficiently large for the volume injected. Alternatively, the sample can be injected in small volumes (e.g. in 20 µL aliquots) in successive intervals, whilst keeping the mobile phase at 0% organic for ∼5–10 min, to ensure refocusing of the sample on the column head (see ref. 9). 14. Optimise the spectral parameters first with parent compound, i.e. the definition of the correct centre of spectrum (O1) and the spectral width and the correct pulse sequence (i.e. the need for proton decoupling or not). In addition, measure (or generously estimate) the total chromatographic run-time with the parent compound, as sufficient increments for F1 in eda will have to be defined to cover the separation of the biofluid sample. References 1. Braun, S., Kalinowski, H.-O., Berger, S. (1998) 150 and More Basic NMR Experiments, Wiley-VCH Verlag GMBH, 69469 Weinheim, second expanded edition, Germany. 2. Bales, J. R., Nicholson, J. K., Sadler, P. J. (1985) Two-dimensional proton nuclear magnetic resonance “maps” of acetaminophen metabolites in human urine. Clin Chem 31, 757–762. 3. McCormick, A. D., Slamon, D. L., Lenz, E. M., Phillips, P. J., King, C. D., McKillop, D., Roberts, D. W. (2007) In vitro metabolism of a tricyclic alkaloid (M445526) in human
liver microsomes and hepatocytes. Xenobiotica 37, 972–985. 4. Moffat, A. C., Jackson, J. V., Moss, M. S., Widdop, B. (eds.) (1986) Clarke’s Isolation and Identification of Drugs in Pharmaceuticals, Body Fluids, and Post-mortem Material, 2nd edn, The Pharmaceutical Press (publications division of The Pharmaceutical Society of Great Britain, 1 Lambeth High Street, London SE1 7JN). 5. Holzgrabe, U., Wawer, I., Diehl, B. (1999) NMR Spectroscopy in Drug Development and Analysis, Wiley-VCH Verlag GmbH, 69469 Weinheim, Germany.
Nuclear Magnetic Resonance (NMR)-Based Drug Metabolite Profiling 6. Lindon, J. C., Nicholson, J. K., Wilson, I. D. (1995) The development and application of coupled HPLC–NMR spectroscopy. Adv Chromatogr 36, 315–382. 7. Lindon, J. C., Nicholson, J. K., Wilson, I. D. (1996) Direct coupling of chromatographic separations to NMR spectroscopy. Prog Nucl Magn Res 29, 1–49. 8. Tugnait, M., Lenz, E. M., Hofmann, M., Spraul, M., Wilson, I. D., Lindon, J. C., Nicholson, J. K. (2003) The metabolism of 2-trifluoromethyl aniline and its acetanilide in the rat by 19f NMR monitored enzyme hydrolysis and 1 H/19 F HPLC–NMR spectroscopy. J Pharm Biomed Anal 30, 1561– 1574. 9. Lenz, E. M., D’Souza, R. A., Jordan, A. C., King, C. D., Smith, S. M., Phillips, P. J.,
319
McCormick, A. D., Roberts, D. W. (2007) HPLC–NMR with severe column overloading: fast-track metabolite identification of urine and bile samples from rat and dog treated with [14 C]-ZD6126. J Pharm Biomed Anal 43, 1065–1077. 10. Exarchou, V., Fiamegos, Y. C., van Beek, T. A., Nanos, C., Vervoort, J. (2006) Hyphenated chromatographic techniques for the rapid screening and identification of antioxidants in methanolic extracts of pharmaceutically used plants. J Chromatogr A 1112, 293– 302. 11. Martin, P. D., Warwick, M. J., Dane, A. L., Hill, S. J., Giles, P. B., Lenz, E. (2003) Metabolism, excretion, and pharmacokinetics of rosuvastatin in healthy adult male volunteers. Clin Ther 25, 2822–2834.
Chapter 19 Nuclear Magnetic Resonance (NMR)-Based Metabolomics Hector C. Keun and Toby J. Athersuch Abstract Biofluids are by far the most commonly studied sample type in metabolic profiling studies, encompassing blood, urine, cerebrospinal fluid, cell culture media and many others. A number of these fluids can be obtained at a high sampling frequency with minimal invasion, permitting detailed characterisation of dynamic metabolic events. One of the attractive properties of solution-state metabolomics is the ability to generate profiles from these fluids following simple preparation, allowing the analyst to gain a naturalistic, largely unbiased view of their composition that is highly representative of the in vivo situation. Solutionstate samples can also be generated from the extraction of tissue or cellular samples that can be tailored to target metabolites with particular properties. Nuclear magnetic resonance (NMR) provides an excellent technique for profiling these fluids and is especially adept at characterising complex solutions. Profiling biofluid samples by NMR requires appropriate preparation and experimental conditions to overcome the demands of varied sample matrices, including those with high protein, lipid or saline content, as well as the presence of water in aqueous samples. Key words: NMR spectroscopy, urine, plasma, tissue, cellular extracts, culture media.
1. Introduction Solution-state nuclear magnetic resonance (NMR) spectroscopy is an efficient profiling tool that can generate information-rich spectra that have proven success in the study of metabolism and related research and clinical areas (1–3). NMR spectroscopy possesses many general features that make it highly fit for purpose as a metabolic profiling technique. It can operate in a largely untargeted fashion, i.e. any sufficiently abundant molecule containing the nuclei of interest will be detected. As for biological NMR in general, 1 H-NMR spectroscopy provides in principle access to T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_19, © Springer Science+Business Media, LLC 2011
321
322
Keun and Athersuch
many metabolites, given the ubiquitous presence of these nuclei in biochemistry. Other NMR-visible nuclei such as 13 C, 31 P and 15 N can also be exploited to a lesser extent and provide complementary information. The resonances produced are very sensitive to structure, so many compounds can be identified in spectral data simultaneously and the fine structure of the signals in the frequency domain provides strong structural clues for the identification of unknown metabolites. In addition, routinely available technical extensions, such as multidimensional NMR, give further detailed structural information. Where stable isotopes are used to target one or several metabolic pathways, it is possible to distinguish isotopomers and isotopologues and thus gain information on biotransformation and pathway flux. Crucially, the data generated are amenable to quantitative interpretation across a wide dynamic range. Despite being relatively insensitive compared to other forms of spectroscopy and mass spectrometry, NMR remains highly competitive in many applications, due in part to the quality of the structural information it can provide. However, the most unique feature of NMR spectroscopy is that it can analyse the ‘native’ tissue or biofluid with minimal preparation and in an essentially non-destructive manner. This has many important consequences: NMR spectroscopy tends to be more analytically reproducible and robust than other platforms (4, 5); it also allows the measurement of metabolites non-invasively in vivo. These characteristics in turn make NMR spectroscopy a useful tool in translational research, i.e. taking results from the bench into practical use in the field or clinic (1, 6). One of the key advantages of metabolic profiling as a means of biomarker discovery (and metabolic biomarkers themselves) over genomic and proteomic counterparts is that metabolites are a defined chemical entity irrespective of species, genotype, localisation and biological matrix. In principle therefore, analytical procedures should be more translatable and the particular properties of NMR spectroscopy enhance that advantage. There are two general problems faced in biofluid NMR that have led to a preferential use of two pulse sequences for metabolic profiling, namely the 1D nuclear Overhauser effect spectroscopy with presaturation (NOESYpresat) (7) and the 1D Carr–Purcell– Meiboom–Gill (CPMG) (8) sequences. The first issue is adequate suppression of the solvent resonance. Although very effective solutions using excitation sculpting (9) exist, presaturation continues to be the most simple and prevalent option, applied during the relaxation delay. Field inhomogeneity decreases suppression efficiency, and so a single increment of the standard NOESY pulse sequence is typically used, rather than the single pulse experiment, to reduce contributions from regions of the active volume that
Nuclear Magnetic Resonance (NMR)-Based Metabolomics
323
experience an incomplete 90◦ pulse, thus reducing the residual water resonance. The second common experimental issue in biofluid NMR is the discrimination between metabolites of low molecular weight (typically 10,000×g. d. Transfer 550 µL of sample to a 5-mm NMR tube, taking care not to disturb any pelleted material (see Note 2). 3. General instrument set-up (Steps a–f are readily automated, see Note 3). a. Set probe temperature (e.g. 300 K), insert sample and wait for temperature equilibration (∼5 min). b. ‘Lock’ the instrument to the D2 O resonance. c. Tune and match the probe. d. Adjust shims to optimise spectral lineshape. Half-height linewidth of ∼1 Hz should be readily obtainable on samples with low protein content. e. Using a single pulse experiment with presaturation, determine the 90◦ pulse length and optimise the spectrometer frequency offset to minimise the residual solvent resonance. f. Optimise the receiver gain to remove digitisation errors while not exceeding the dynamic range of the receiver. g. Select suitable recycle delay (RD) for total recycle time to be of the order of 5∗T1 and for suitable water suppression. Typical parameters at 600 MHz 1 H frequency might be an RD of 2 s and a total acquisition time (AQ) of 2.73 s recorded into 64 K complex data points to give a spectral width of ∼12 kHz.
h. Presaturation pulse power should be the minimum required to achieve the necessary reduction in the water resonance, e.g. equivalent of 25 Hz bandwidth.
4. Specific pulse sequence optimisation. a. Change to ‘NOESYpresat’ pulse sequence (RD90◦ -3 µs-90◦ -tm -90◦ -AQ) for more effective solvent
Nuclear Magnetic Resonance (NMR)-Based Metabolomics
327
suppression. A typical mixing time (tm ) is 100 ms (see Note 4). b. Reoptimise the receiver gain if required. c. Accumulation of 128 scans should provide an adequate signal-to-noise (s/n) ratio in ∼11 min.
5. Spectral processing. a. Typically, application of 0.3 Hz exponential linebroadening transformation and zero filling to at least 64 K data points if fewer are recorded is beneficial prior to Fourier transformation (FT).
b. After FT, reference the chemical shift scale to TSP (see Note 5) and apply zero- and first-order phase correction and a global linear baseline correction. Examples of typical urine NMR spectra are shown in Fig. 19.2. 3.2. Method B: Plasma/Serum (Also Suitable for Other High-Protein Samples, e.g. Cyst Fluid)
1. Sample collection (also see (10)). a. Collect blood into heparinised (plasma) or plain (serum) collection tubes. b. Centrifuge within a defined time interval and note clot contact time (ideally 10,000×g at 4◦ C to remove suspended debris. d. Transfer 550 µL of the supernatant to a 5-mm NMR tube taking care not to disturb any pelleted material. 3. General instrument set-up. a. See Method A, Step 3. 4. Specific pulse sequence optimisation. In addition to the 1D ‘NOESYpresat’ (Method A, Step 4), the CPMG pulse sequence (RD-90◦ -(τ -180◦ -τ ) n -AQ) with presaturation can be used for suppression of macromolecular signals on the basis of T2 editing. Optimise τ and n to achieve desired suppression of macromolecular signals; τ is often kept fixed and n varied. Typical parameters for blood serum/plasma are τ =400 µs and n=300 giving a total mixing time of ∼240 ms (see Note 7). Other possible profiling experiments include J-resolved and diffusion-edited pulse sequences (see Beckonert et al. (13)). 5. Spectral processing. As for Step 5 of Method A, with the exception that spectral calibration is often via the glucose
328
Keun and Athersuch TMA
(a) taurine
succinate
creatine + creatinine
citrate
creatine
3-hydroxypropionic acid creatinine DMA acetate
lysine
(b) creatinine
2-OG
2-OG
TMAO
creatinine
hippurate trans-aconitate
glycine N-methylnicotinamide + N-methylnicotinic acid
lactate alanine
4
3
2
1
0
Fig. 19.2. Typical 600-MHz 1 H NOESYpresat NMR spectra of urine from (a) mouse and (b) rat. Clear species differences can be seen in the relative concentrations of metabolites in these urinary profiles. Key: DMA, dimethylamine; DMG, dimethylglycine; MA, methylamine; TMA, trimethylamine; TMAO, trimethylamine-N-oxide. Reproduced from Bollard et al. (12).
alpha-anomeric doublet resonance at 5.233 ppm, a process that can be automated (14). 3.3. Method C: Tissue Extracts
1. Sample collection. a. Tissues should be snap frozen using liquid nitrogen and the time to freezing controlled to minimise the effects of ischemia, etc.
Nuclear Magnetic Resonance (NMR)-Based Metabolomics
329
b. Store at –80◦ C or below and transport on dry ice. 2. Sample extraction (see Note 8). a. Weigh frozen tissue into an Eppendorf or glass vial (a minimum of 20 mg). b. Add cold CHCl3 /MeOH (2:1) solution (300 µL assuming a 20–30 mg sample) and manually grind the tissue or preferably use a tissue homogeniser. c. Add an equivalent volume (300 µL) of HPLC-grade water and vortex mix. Keep the homogenate on ice. d. Centrifuge the homogenate for 10 min at >10,000×g. e. The lower organic (CHCl3 ) and upper aqueous (methanol/water) phases should be clearly separated by an insoluble interface. Carefully isolate the two layers by pipetting the aqueous (upper) phase and organic (lower) phase into separate clean glass vials. f. Repeating the extraction with the same volume of solvents and pooling the second set of fractions with the first will increase recovery. g. Remove the organic solvents from the samples using a speed vacuum concentrator/rotary evaporator or a stream of nitrogen. h. Freeze-dry the aqueous phase to remove residual water. i. Perform a ‘blank’ extraction procedure on an empty sample tube to allow a baseline spectrum to be obtained (allows identification of contamination during preparation). 3. NMR sample preparation. a. Reconstitute the aqueous extract in 600 µL of reconstitution buffer (0.2 M Na2 HPO4 , 0.043 M NaH2 PO4 , 1 mM TSP, 3 mM NaN3 and 100% D2 O). b. Transfer to a sample tube and centrifuge for 5 min at >10,000×g). c. Pipette 550 µL into an NMR tube. d. Reconstitute the organic extract in 700 µL of CDCl3 (containing 0.03% (v/v) TMS). e. Transfer to a sample tube and centrifuge for 5 min at >10,000×g. f. Pipette 600 µL into an NMR tube. 4. General instrument set-up. a. See Method A, Step 3. 5. Specific pulse sequences. a. For the aqueous sample, the NMR acquisition can proceed as for urine or blood serum (Method B, Step 4) with
330
Keun and Athersuch
either NOESYpresat or CPMG sequences used. Residual protein signal can be suppressed by the use of the CPMG sequence with a short T2 relaxation time (e.g. 64 ms). For either experiment, an increased number of scans (e.g. 256) is recommended. b. For the organic sample the residual solvent resonance is small and hence a single pulse experiment can be used. 3.4. Method D: Cell Extracts and Culture Media
1. Sample generation. a. For a monolayer culture of mammalian cells, typically >106 cells are required, with 4–5 × 106 giving adequate concentrations of metabolites. Cells can be cultured in 75-cm2 flasks with 12 mL of media, with each flask yielding a single biological replicate (see Notes 9 and 10). 2. Sample collection (also see (15)). a. Remove cultures from the incubator and aspirate the media. b. If desired, retain the media in a sterilised tube, centrifuge (4◦ C, 4 min, 150×g) to pellet dead cells and freeze supernatant for later analyses. c. Wash the cells (add–distribute–remove) using 1 mL of cold (4◦ C) PBS to remove residual media. Discard, wash and repeat. d. Lyse cells and quench metabolism by adding 1 mL cold methanol (4◦ C) to the culture vessel. e. After 2 min, detach cellular material from the culture vessel using a cell scraper. f. Transfer the resulting suspension of cellular material to an Eppendorf or glass tube and dry down the sample using a solvent evaporator. 3. Sample preparation. a. The resulting cell pellet can be extracted using the same procedure as for whole tissue (Method C, Step 2) without the need for grinding or sonication. b. The residual pellet after extraction can be used for sample normalisation (see Note 11). 4. NMR sample preparation. a. Aqueous and organic cell extracts can be prepared as for tissue extracts, (Method C, Step 3) but for mass-limited samples, it may be advisable to dilute the reconstitution buffer two to fivefold. b. For culture media, mix 550 µL of sample with 50 µL of 0.2 % (w/v) TSP in 100% D2 O before transferring to a 5-mm NMR tube.
Nuclear Magnetic Resonance (NMR)-Based Metabolomics
331
Fig. 19.3. Typical 600-MHz 1 H CPMG spin-echo NMR spectrum of an aqueous intracellular extract obtained from cultured primary rat hepatocytes. Key: GPC, glycerophosphocholine; PC, phosphocholine; GLY, glycogen; GLU, glucose; GSH, glutathione. Adapted from Ellis et al. (16).
5. General instrument set-up. a. As for Method A, Step 3. 6. Specific pulse sequences. a. As for Method C, Step 5. The culture media can be treated similarly to the aqueous cell extract (see Note 12). An example of a typical cell extract spectrum is shown in Fig. 19.3.
4. Notes 1. The pH of samples can be finely adjusted by the addition of small volumes (5–10 µL) of 1 M HCl or 1 M NaOH. Probes for measuring pH directly in NMR tubes can also be obtained. 2. Many sample types including urine can be conveniently stored frozen in NMR tubes down to –40◦ C prior to analysis at the expense of one freeze/thaw cycle. Breakages can be avoided by ensuring the NMR tube is tightly capped, inverted so that the sample is at the head of the tube against the cap and the tube placed horizontally during freezing. 3. Where resources allow, the use of cooled sample holders will reduce the opportunity for changes in sample composition/degradation during prolonged wait times in well plate racks or in a sample carousel. 4. In standard implementations of NOESYpresat acquisitions, presaturation occurs during the mixing time (tm ) in addition to the prescan delay (RD). Mixing time presaturation
332
Keun and Athersuch
does not have a significant impact on spectral quality, but should be kept constant across an analytical batch. 5. Several chemical shift reference substances are used in solution-state NMR of aqueous samples. Most commonly used is 3-(trimethylsilyl)propionic-2,2,3,3-d4 acid sodium salt (TSP) (17), although 4,4-dimethyl-4-silapentane-1sulfonic acid (DSS) and 4,4-dimethyl-4-silapentane-1ammonium trifluoroacetate (DSA) (18) are also used. An additional shift reference that can be used in combination with one invariant with pH can be used to accurately determine sample pH (1,1-difluoro-1trimethylsilanyl methylphosphonic acid, DFTMP) (19). 6. Broad background resonances attributable to protein can also be reduced by deproteinisation via methanol precipitation or MW filtering (e.g. 10-kDa filter, see Section 2.2). 7. In NMR experiments using spin echoes, the optimal τ value will vary with field strength. Typically, at 600 MHz (14.1 T), a τ value between 300 and 400 µs will be optimal. This value can be determined on a single representative sample. Long CPMG sequences can induce heating in the sample and cause broadening of peaks. Increasing the number of dummy scans (e.g. 16–32) can remedy this problem by improving thermal equilibration at the start of the acquisition. 8. Solvents used for sample extraction and reconstitution should be checked for purity prior to use to minimise the inadvertent introduction of contaminants. For organic extraction buffers, a larger aliquot should be dried down under N2 and then any remaining material reconstituted in the NMR solvent (e.g. CDCl3 ). 9. Cell cultures should be harvested at a similar level confluence. 10. Profiles of media, media with growth supplements added and a dummy incubation (no cells) should accompany cell culture experiments to aid spectral assignment, indicate any changes in media independent of cellular activity and help identify sources of unwanted contamination (e.g. nonsterile media). 11. The insoluble interface of the liquid–liquid extraction procedure can be used for estimating sample protein. Most commonly, the bicinchoninic acid (BCA) assay is used – a colorimetric method assay with a wide linear range and detection limit of around 5 µg/mL that can be conducted easily in 96-well plate format. Calibrated protein measurements are a useful measure for the normalisation of cell culture materials and can be used to approximate cell number.
Nuclear Magnetic Resonance (NMR)-Based Metabolomics
333
12. Organic solvents are commonly found at relatively high concentrations in cell culture media samples when used as the solvent for chemical exposure experiments. Solvent signals in NMR spectra can often be successfully suppressed by the use of multiple presaturation or excitation sculpting pulse sequences. Alternatively, the sample may be dried down using a solvent evaporator, but this may present problems where the solvent has a high boiling point (e.g. DMSO), requiring severe conditions that may compromise sample quality.
Acknowledgements Development of the protocols described and the preparation of the manuscript was funded in part by the European Union FP6 carcinoGENOMICS project (contract number PL037712) and a Cefic Long-range Research Initiative (LRI) Innovative Science Award. The authors wish to acknowledge Dr. Olaf Beckonert, Dr. James Ellis and Dr. Orla Teahan (Imperial College London) for their useful help and advice. References 1. Keun, H. C., Athersuch, T. J. (2007) Application of metabonomics in drug development. Pharmacogenomics 8, 731–741. 2. Lindon, J. C., Nicholson, J. K. (2008) Spectroscopic and statistical techniques for information recovery in metabonomics and metabolomics. Annu Rev Anal Chem 1, 45–69. 3. Coen, M., Holmes, E., Lindon, J. C., Nicholson, J. K. (2008) NMR-based metabolic profiling and metabonomic approaches to problems in molecular toxicology. Chem Res Toxicol 21, 9–27. 4. Keun, H. C., Ebbels, T. M. D., Antti, H., Bollard, M. E., Beckonert, O., Schlotterbeck, G., Senn, H., Niederhauser, U., Holmes, E., Lindon, J. C., Nicholson, J. K. (2002) Analytical reproducibility in H-1 NMR-based metabonomic urinalysis. Chem Res Toxicol 15, 1380–1386. 5. Dumas, M. E., Maibaum, E. C., Teague, C., Ueshima, H., Zhou, B., Lindon, J. C., Nicholson, J. K., Stamler, J., Elliott, P., Chan, Q., Holmes, E. (2006) Assessment of analytical reproducibility of 1H NMR spec-
6.
7.
8.
9.
10.
troscopy based metabonomics for large-scale epidemiological research: the INTERMAP study. Anal Chem 78, 2199–2208. Bundy, J. G., Iyer, N. G., Gentile, M. S., Hu, D. E., Kettunen, M., Maia, A. T., Thorne, N. P., Brenton, J. D., Caldas, C., Brindle, K. M. (2006) Metabolic consequences of P300 gene deletion in human colon cancer cells. Cancer Res 66, 7606–7614. Neuhaus, D., Ismail, I. M., Chung, C. W. (1996) ‘FLIPSY’ – a new solvent-suppression sequence for nonexchanging solutes offering improved integral accuracy relative to 1d NOESY. J Magn Reson Ser A 118, 256–263. Meiboom, S., Gill, A. (1958) Modified spin-echo method for measuring nuclear relaxation times. Rev Sci Instrum 29, 688–691. Hwang, T. L., Shaka, A. J. (1995) Water suppression that works: excitation sculpting using arbitrary wave-forms and pulsed-field gradients. J Magn Reson 112, 275–279. Teahan, O., Gamble, S., Holmes, E., Waxman, J., Nicholson, J. K., Bevan, C., Keun, H. C. (2006) Impact of analytical bias
334
11.
12.
13.
14.
Keun and Athersuch in metabonomic studies of human blood serum and plasma. Anal Chem 78, 4307– 4318. Griffin, J. L., Nicholls, A. W., Daykin, C. A., Heald, S., Keun, H. C., Schuppe-Koistinen, I., Griffiths, J. R., Cheng, L. L., RoccaSerra, P., Rubtsov, D. V., Robertson, D. (2007) Standard reporting requirements for biological samples in metabolomics experiments: mammalian/in vivo experiments. Metabolomics 3, 179–188. Bollard, M. E., Keun, H. C., Beckonert, O., Ebbels, T. M., Antti, H., Nicholls, A. W., Shockcor, J. P., Cantor, G. H., Stevens, G., Lindon, J. C., Holmes, E., Nicholson, J. K. (2005) Comparative metabonomics of differential hydrazine toxicity in the rat and mouse. Toxicol Appl Pharmacol 204, 135–151. Beckonert, O., Keun, H. C., Ebbels, T. M. D., Bundy, J. G., Holmes, E., Lindon, J. C., Nicholson, J. K. (2007) Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc 2, 2692–2703. Pearce, J. T., Athersuch, T. J., Ebbels, T. M., Lindon, J. C., Nicholson, J. K., Keun, H. C. (2008) Robust algorithms for automated
15.
16.
17.
18.
19.
chemical shift calibration of 1d 1H NMR spectra of blood serum. Anal Chem 80, 7158–7162. Teng, Q., Huang, W. L., Collette, T. W., Ekman, D. R., Tan, C. (2009) A direct cell quenching method for cell-culture based metabolomics. Metabolomics 5, 199–208. Ellis, J. K., Chan, P. H., Doktorova, T., Athersuch, T. J., Cavill, R., Vanhaecke, T., Rogiers, V., Vinken, M., Nicholson, J. K., Ebbels, T. M. D., Keun, H. C. (2010) Effect of the histone deacetylase inhibitor trichostatin a on the metabolome of cultured primary hepatocytes. J Proteome Res 9, 413– 419. Pohl, L., Eckle, M. (1969) Sodium 3trimethylsilyltetradeuteriopropionate, a new water-soluble standard for 1H-NMR. Angew Chem Int Ed Engl 8, 381. Nowick, J. S., Khakshoor, O., Hashemzadeh, M., Brower, J. O. (2003) DSA: a new internal standard for NMR studies in aqueous solution. Org Lett 5, 3511–3513. Reily, M. D., Robosky, L. C., Manning, M. L., Butler, A., Baker, J. D., Winters, R. T. (2006) DFTMP, an NMR reagent for assessing the near-neutral ph of biological samples. J Am Chem Soc 128, 12360–12361.
Chapter 20 Slow Magic Angle Sample Spinning: A Non- or Minimally Invasive Method for High-Resolution 1 H Nuclear Magnetic Resonance (NMR) Metabolic Profiling Jian Zhi Hu Abstract High-resolution 1 H magic angle spinning nuclear magnetic resonance (NMR), using a sample spinning rate of several kilohertz or more (i.e., high-resolution magic angle spinning (hr-MAS)), is a wellestablished method for metabolic profiling in intact tissues without the need for sample extraction. The only shortcoming with hr-MAS is that it is invasive and is thus unusable for non-destructive detections. Recently, a method called slow MAS, using the concept of two-dimensional NMR spectroscopy, has emerged as an alternative method for non- or minimally invasive metabolomics in intact tissues, including live animals, due to the slow or ultra-slow sample spinning used. Although slow MAS is a powerful method, its applications are hindered by experimental challenges. Correctly designing the experiment and choosing the appropriate slow MAS method both require a fundamental understanding of the operation principles, in particular the details of line narrowing due to the presence of molecular diffusion. However, these fundamental principles have not yet been fully disclosed in previous publications. The goal of this chapter is to provide an in-depth evaluation of the principles associated with slow MAS techniques by emphasizing the challenges associated with a phantom sample consisting of glass beads and H2 O, where an unusually large magnetic susceptibility field gradient is obtained. Key words: High-resolution 1 H-NMR metabolomics, tissues, organs, live animals, slow magic angle spinning, magic angle turning, magnetic susceptibility, line broadening, molecular diffusion.
1. Introduction Magnetic resonance imaging (MRI) is widely used for clinical diagnosis of malignancies in various tissues and organs (1, 2). However, MRI is mainly useful for detecting malignancies when tumors have already developed to a relatively large size, e.g., a few T.O. Metz (ed.), Metabolic Profiling, Methods in Molecular Biology 708, DOI 10.1007/978-1-61737-985-7_20, © Springer Science+Business Media, LLC 2011
335
336
Hu
hundreds of micrometers or more depending on the sensitivity of the spectrometer and the methods used. Thus, MRI is not an effective method for early diagnosis. Since biochemical changes in the diseased tissues precede tumor formation, early diagnosis can be achieved if molecular-level information is obtained. These detailed biochemical changes can, in principle, be measured using high-resolution nuclear magnetic resonance (NMR) methods through ex vivo analysis of chemical extracts of excised tissues (3). The ex vivo process usually starts with lysing the cells, followed by extracting the cell lysate with organic solvents. Finally, standard high-resolution, liquid-state NMR can be used to analyze the extracted molecular entities. Although impressive spectral resolution can be obtained, standard ex vivo methods involve extensive sample preparation and are therefore prone to artifacts induced by incomplete sample extraction and fractionation, as well as sample degradation during this lengthy process (4). Like solids, tissues and cells cannot be directly analyzed by standard liquid-state NMR spectroscopy due to the line broadenings induced by the variation of local magnetic field gradients at the compartment boundaries in cells and tissues (5), and to a lesser extent by the residual homonuclear dipolar interactions and residual chemical shift interactions. However, when spinning the sample about an axis at the magic angle (54◦ 44′ ) and using a sample spinning rate of several kilohertz or more, all of the line broadenings can be effectively averaged out, resulting in a high-resolution 1 H-NMR spectrum. High-resolution magic angle spinning (hr-MAS) 1 H-NMR has been successfully applied to analyze intact cells and tissues from organs such as brain, lung, kidney, heart, and muscle (4, 6–12), and the method has been reviewed recently in The Handbook of Metabonomics and Metabolomics (13). With hr-MAS, high-spectral-resolutionapproaching liquid-state NMR has been achieved. The major advantage of hr-MAS 1 H-NMR over other methods for tissue sample analysis is that there is minimal sample preparation and thus less artifacts and better correlation with in vivo techniques. However, the large centrifugal force associated with the fast sample spinning rates destroys the tissue structure and even some cell types (5), which makes the method unusable for nondestructive metabolic profiling or localized spectroscopy in live animals. Hence, it is important to develop alternative methods where the spinning speed can be reduced. However, the spinning rate cannot be arbitrarily reduced. A problem with traditional MAS at slow sample spinning is that it gives rise to numerous, overlapping spinning sidebands (SSBs), which renders analysis of the spectra difficult or impossible. Fortunately, in solid-state NMR, many methods have been developed to overcome this problem. A few of these methods have the potential to be extended to biological fluid samples. These methods include 2D phase-adjusted spinning sidebands (PASS) (14)
Slow Magic Angle Sample Spinning
337
and 2D phase-corrected magic angle turning (PHORMAT) (15). It has been shown (16–18) that nearly sideband-free isotropic spectra could be obtained with PASS in a variety of excised intact organs at spinning speeds as low as 43 Hz and with a spectral resolution comparable to or better than the resolution obtained with fast MAS (19). 1 H PASS using a sample spinning rate of less than 100 Hz has been successfully utilized to study live cells directly from agar growth plate (20) and intact food seeds such as oil seeds. While the frequency used in 1 H PASS is already two orders of magnitude lower than the frequencies employed in hr-MAS, it is still too large to keep larger size soft tissues and organs undamaged. Furthermore, in PASS the magnetization is constantly present in the transverse plane, and the first signal is observed after one rotor period (16). This means that the amplitude of the signal is reduced as a result of the decay of the magnetization during this period, which is governed by the spin–spin relaxation time T2 (16). Therefore, serious signal attenuation occurs when the spinning rate is comparable to or less than (1/T2 ). Moreover, spectral distortions may occur if the T2 values of the different spectral lines are not the same. For example, in brain tissue, where T2 values of the metabolites are of the order of 100–450 ms (21), we found that PASS was limited to spinning speeds of 10–20 Hz or larger. Additionally, the homonuclear J-coupling may complex the spectrum if the spinning frequency becomes comparable to the J-coupling constant because use of the 180 pulses will not refocus the homonuclear J-coupling during the evolution dimension of the PASS experiment. In contrast, T2 attenuation is avoided in a PHORMAT experiment. It has been demonstrated (22) that PHORMAT, applied with a 1 Hz spinning speed, produces spectra of excised rat liver with a resolution approaching that obtained from PASS or hrMAS methods. Because of the ultra-slow sample spinning used, 1 H PHORMAT represents one of the practical ways for obtaining high-resolution 1 H spectroscopy of the metabolites in live whole animals. In fact, a high-resolution metabolite spectrum of whole mouse has been successfully acquired at 2T field using a sample spinning rate of 1.5 Hz (23). Recently, a localized 1 H-PHORMAT, namely the LOCMAT experiment, has been developed for obtaining localized metabolite spectra inside a live mouse (24, 25). The potential application of 1 H PASS and PHORMAT/LOCMAT in metabolomics and the directions for further improving the methods have been extensively discussed recently (26) and will not be duplicated in this article. The focus of the current article is on understanding the operation principles of the methods, in particular the details of line narrowing at the presence of molecular diffusion, that have not yet been reported in detail previously. Understanding these principles is critically important both for correctly running the slow MAS experiment and for choosing the right type of slow MAS method.
338
Hu
2. Methods 2.1. The Basic Principles of a 1 H PHORMAT Experiment
The basic principle of a PHORMAT experiment acting on the magnetic susceptibility field is best illustrated with the prototype magic angle turning (MAT) pulse sequence (15) given in Fig. 20.1. The pulses labeled as r1 , r2 , r3 , and r4 are synchronized to the one-third of the rotor period T. p1 , p2 , and p3 are projection pulses, which project either the cos(ι ) or the sin(ι ), where ι = 1,2,3, component of the procession magnetization during the corresponding t1 /3 period to the z-axis. A free induction decay (FID) is acquired following the last 90◦ read out pulse (r4 ). With proper phase cycling (15, 22) of the projection pulses p1 , p2 , and p3 and the receiver, it is possible to obtain FID(t2 , t1 ) = exp(−i(1 (t1 /3) + 2 (t1 /3) + 3 (t1 /3))FID(t2 ) [1] The precession angles 1, 2 , and 3 can be written in the usual way as time integrals of the resonant angular frequency ω(r, t) = ωiso + γ Bz (r, t), where ωiso is the isotropic frequency of interest and Bz (r, t) is the susceptibility field at the observation point r in the rotor frame, which is a function of time due to sample rotation: 1 = ωiso t1 /3 + 2 = ωiso t1 /3 +
= ωiso t1 /3 +
3 = ωiso t1 /3 +
= ωiso t1 /3 +
r1
p1
r2
t1/3
Φ1 0
γ Bz (r, t)dt
0
[2a]
T /3+t1 /3
γ Bz (r, t)dt T /3
[2b]
t1 /3
γ Bz (r, t 0 2T /3+t1 /3
+ T /3)dt
γ Bz (r, t)dt
2T /3
[2c]
t1 /3
γ Bz (r, t + 2T /3)dt
0
p2
r3
t1/3 L
t1 /3
Φ2 T/3
Fig. 20.1. The prototype MAT experiment.
p3
r4
t1/3 L
Φ3 2T/3
t2/3 L
T
Slow Magic Angle Sample Spinning
339
We obtain 1 + 2 + 3 = ωiso t1 +
t1/3 0
γ [Bz (r, t) + Bz (r, t + T /3)
[3]
+Bz (r, t + 2T /3)] dt The term (Bz (r, t) + Bz (r, t + T /3) + Bz (r, t + 2T /3)) inside the integral of Eq. [3] represents a summation of the susceptibility field at the observation point r located in the rotor frame at three chosen times that are exactly one-third of rotor period apart or that are exactly 120◦ apart around the circle of rotation. If the axis of rotation is at the magic angle, i.e., at an angle of 54◦ 44” relative to the main field B0 , such a rotation will place B0 along three perpendicular axes, i.e., the x, y, and z. Since the effect of sample rotation relative to the magnetic field is equivalent to keeping the sample static while the magnetic field rotates, we found that it is convenient to adapt the concept of rotating the magnetic field to find the value of the summation in the integrant of Eq. [3]. In the following, we will prove that the summation (Bz (r, t) + Bz (r, t + T /3) + Bz (r, t + 2T /3)) is zero for a biological tissue sample of arbitrary shape so that a high-resolution 1 H-NMR metabolite spectrum free from the magnetic susceptibility-induced line broadening is obtained. To reach this goal, the analytical equation Bz (r, t) must be obtained first. 2.2. Analytical Equations of Basic Magnetic Susceptibility Fields
A biological tissue consists mainly of two types of structures, i.e., a cellular matrix, which may be approximated by a combination of many spheres, and a venous system, which can be treated as cylinders. All other shapes can then be constructed using a combination of these two basic geometries. To simplify the discussion, we only treat the susceptibility field created from both a sphere and an infinitely long straight cylinder. The results are then generalized to an arbitrary shape.
2.2.1. Spherical Geometry
Considering a sphere with radius R containing homogenous material with relative permeability µi surrounded by another infinite material with relative permeability of µe (µ = 1 + χ, where χ is the corresponding magnetic susceptibility), it follows from ref. (27). that the field inside the sphere is homogeneous and is parallel to the external magnetic field B0 with the field due to the susceptibility difference given in cgs units (28) by Bi = 8π
µi − µe B0 µi + 2µe
[4]
and the susceptibility field outside the sphere is a dipolar field with the component along B0 , denoted as the z-component, given by
340
Hu
µi − µe 3 3 cos2 θ − 1 R B0 (Be )z = 4π µi + 2µe r3
[5]
where θ is the angle between r, which is a vector joining the origin of the sphere and the point of observation with a length equal to r, and the external field B0 . 2.2.2. Cylindrical Geometry
Considering an infinitely long cylinder with radius R, containing homogenous material with relative permeability µi surrounded by another infinite material with relative permeability of µe , it follows from refs. (27, 29–33) that the field inside the cylinder is homogeneous and is parallel to the external magnetic field B0 with the field due to the susceptibility difference given in cgs units by Bi =
4π µi − µe 3 cos2 θ − 1 B0 3 µi + µe
[6]
and the component of the susceptibility field along B0 outside the cylinder is given by (Be )z = 4π
µi − µe R2 sin2 θ cos 2φB0 , (r ≥ R) µi + µe r 2
[7]
In Eqs. [6] and [7], r is the distance from the point of interest to the cylinder axis, θ is the angle between the cylinder axis and the external field B0 , and ϕ is the angle between the vector r and a plane containing both B0 and the cylinder axis. 2.3. The Effect of Magic Angle Turning on the Susceptibility Field Outside a Magnetized Sphere
All the terms in Eq. [5] are independent of sample rotation except the angle θ, which is the angle between the external main field B0 , i.e., the z-axis, and the vector jointing the observation point and the origin of the laboratory frame. Using the spherical angles θ and φ, the unit vector along r expressed in the Cartesian frame is r = sin(θ)cos(ϕ)i + sin(θ)sin(ϕ)j + cos(θ)k
[8]
where i, j, and k are unit vectors along the x-, y-, and z-directions, respectively. We will adapt the right-hand rotation with the initial field direction along z, i.e., the k-direction with the initial direction cos between the vector r and B0 denoted by cos(θ1 ) = cos(θ)
[9]
After the first 120◦ rotation about the magic angle, B0 is rotated to the x-direction, and the direction cos between the vector r and B0 is now cos(θ2 ) = sin(θ)cos(ϕ)
[10]
Slow Magic Angle Sample Spinning
341
The second successive 120◦ rotation about the magic angle places B0 in the y-direction, and the direction cos between the vector r and B0 becomes cos(θ3 ) = sin(θ)sin(ϕ)
[11]
Bz (r, t) + Bz (r, t + T /3) + Bz (r, t + 2T /3) 3 µi − µe 3 3 cos2 θi − 1 R B0 = 4π µi + 2µe r3 1
= 4π
R3
µi − µe B0 µi + 2µe r 3
3 1
3 cos2 θi − 1
[12]
=0 Equation [3] becomes 1 + 2 + 3 = ω
iso t1
[13]
Equation [13] means that the susceptibility field produced by a magnetized sphere is averaged to zero by the technique of MAT. 2.4. The Effect of Magic Angle Turning on the Susceptibility Field Due to a Cylinder Geometry
Due to the factor of 3 cos2 θ − 1 (see Eq. [6]), the susceptibility field inside the cylinder is averaged to zero by MAT for the same reason as described above, i.e., 1 + 2 + 3 = ωiso t1 . For the field outside the cylinder, we have to find the three sets of (θ ι , ϕ ι ), ι =1,2,3, corresponding to each of the 120◦ rotations about an axis at the magic angle so that we can use Eq. [7] to perform the MAT average. Recall that in Eq. [7], θ is the angle between the main field B0 and the cylinder axis, and ϕ is the angle between the vector r and the plane containing both B0 and the cylinder axis, where r is perpendicular to the cylinder axis (see Fig. 20.2). Since the magic angle forms a cone about the B0 -direction, a general choice is to place the cylinder axis along the z-axis in the laboratory frame, and the initial B0 lies in the z–x plane, i.e., along the z1 -axis. A rotation (ψ) about the z1 -axis will place the magic angle (MA) in a general place in space. In the final frame denoted by x2 –y2 –z2 , each 120◦ rotation about the magic angle axis will rotate the B0 from x2 to y2 then to z2 . The task to find the three sets of (θ ι , ϕ ι ) thus becomes a task to find the three θ ι angles between the cylinder axis and each of the x2 , y2 , z2 axes, respectively, and the ϕ ι angles between the vector r and the plane containing both the cylinder axis and one of the corresponding axes in the frame x2 –y2 –z2 , respectively. Using the right-hand rotation, two successive rotations are needed to rotate the laboratory frame x–y–z to the destination frame x2 –y–z2 . The
342
Hu
z
Magic Angle
B0 Ψ θ
z1
β
z2
y2 y1
φ
y
r
x2 x x1 Fig. 20.2. The coordinate used to perform the susceptibility field average in a cylindrical geometry by MAT. The axis of the cylinder is along the z-direction.
first rotation is θ about the laboratory frame y-axis and the resultant frame is x1 –y1 –z1 . The corresponding rotation matrix is ⎤ cos(θ) 0 sin(θ) ⎥ ⎢ R1 = ⎣ 0 1 0 ⎦ −sin(θ) 0 cos(θ) ⎡
[14]
The second rotation is about the z1 -axis and the resultant frame is x2 –y2 –z2 . The related rotation matrix is ⎤ cos(ψ) −sin(ψ) 0 ⎥ ⎢ R2 = ⎣ sin(ψ) cos(ψ) 0 ⎦ 0 0 1 ⎡
The rotation matrix for the combined rotation is ⎤ ⎡ cos(ψ)cos(θ) −sin(ψ) cos(ψ)sin(θ) ⎥ ⎢ R = R2 R1 = ⎣ sin(ψ)cos(θ) cos(ψ) sin(ψ)sin(θ) ⎦ −sin(θ) 0 cos(θ)
[15]
[16]
The cylinder axis, i.e., the laboratory frame z-axis, expressed in the final frame x2 –y2 –z2 is ⎡ ⎤ 0 ⎢ ⎥ z = R ⎣ 0 ⎦ = cos(ψ) sin(θ)i + sin(ψ) sin(θ)j + cos(θ)k [17] 1 where i, j, and k are unit vectors along x2 -, y2 -, and z2 -axes, respectively.
Slow Magic Angle Sample Spinning
343
It follows from Eq. [17] that the three θ ι (ι =1,2,3) angles have the following relationship: cos(θ1 ) = cos(ψ)sin(θ) cos(θ2 ) = sin(ψ)sin(θ) cos(θ3 ) = cos(θ)
[18]
The vector r, expressed in the final frame x2 –y2 –z2 , is ⎤ cos(ϕ) ⎥ ⎢ r =R ⎣ sin(ϕ) ⎦ = (cos(ψ) cos(θ) cos(φ) − sin(ψ) sin(φ))i 0 ⎡
+ (sin(ψ) cos(θ)cos(φ)+ cos(ψ) sin(φ))j + (− sin(θ) cos(φ))k [19] In order to find the three ϕ ι angles, we need to obtain the three normals ni , i = 1,2,3, where n1 is the normal to the plan formed by x2 and the cylinder axis and n2 is the normal to the plan formed by y2 and the cylinder axis and so on. We will define [20] n1 = x1 i + y1 j + z1 k Since n1 • i = 0 and n1 • z = 0, we obtain x1 = 0, sin(ψ) sin(θ)y1 + cos(θ)z1
= 0,
[21]
y12 + z12 = 1 We obtain
cos(θ) n1 = ± 1/2 j cos2 (θ) + sin2 (ψ)sin2 (θ) sin(ψ)sin(θ) − 1/2 k cos2 (θ) + sin2 (ψ)sin2 (θ)
Similarly, we obtain
cos(θ) n2 = ± 1/2 i cos2 (θ) + cos2 (ψ)sin2 (θ) cos(ψ)sin(θ) − 1/2 k cos2 (θ) + cos2 (ψ)sin2 (θ) n3 = ± [−sin(ψ)i + cos(ψ)j ]
[22]
[23]
[24]
344
Hu
The ± sign in Eqs. [22]–[24] is arbitrary because the normal to a plane can differ by 180◦ . i, j, and k are unit vectors along the corresponding axis in the x2 –y2 –z2 frame. The equation to determine the three φ ι , ι =1,2,3, angles become
It follows that
cos( π2 − φ1 ) = r • n1 cos( π2 − φ2 ) = r • n1 cos( π2 − φ3 ) = r • n1
sin(φ1 ) = ±
(sin ψ cos ϕ+cos ψ sin ϕ cos θ) (cos2 (θ)+sin2 (ψ)sin2 (θ))1/2
sin(φ2 ) = ±
(cos ψ cos ϕ−sin ψ sin ϕ cos θ) (cos2 (θ)+cos2 (ψ)sin2 (θ))1/2
[25]
[26]
sin(φ2 ) = ±sin(ϕ) Finally, with Eqs. [18] and [26] we are able to perform the time average in Eq. [3] for the susceptibility outside a cylindrical geometry: Bz (r, t) + Bz (r, t + T /3) + Bz (r, t + 2T /3) 3
= 4π = 4π
µi − µe R2 2 B0 sin θi cos 2φi µi + µe r 2 1
R2
µi − µe B0 µi + µe r 2
3 1
[27]
(1 − cos2 θi )(1 − 2 sin2 φi )
=0 Once again Eq. [3] becomes 1 + 2 + 3 = ωiso t1
[28]
The results given in Eqs. [3], [27], and [28] indicate that MAT is also able to average out the susceptibility field outside a cylindrical geometry. Since all other shapes can be constructed using a combination of the basic spherical and cylindrical geometries, based on the distribution rules of integration, we conclude that MAT is able to average the susceptibility field of an arbitrary sample shape to zero. This result indicates that in the absence of translational molecular diffusion, a high-resolution isotropic metabolite NMR spectrum should be obtained along the evolution dimension of the PHORMAT experiment regardless of the sample spinning rate used. However, translational molecular diffusion is present in a biological fluid object. In the following, we will investigate how molecular diffusion affects the spectral resolution along the isotropic dimension of the PHORMAT experiment.
Slow Magic Angle Sample Spinning
2.5. The Theory of Translational Molecular Diffusion at a Magnetic Gradient Field
345
It is well known (34–36) that the time dependence of magnetization in the presence of isotropic molecular diffusion for an isolated, one-half spin is described by the Block–Terrey equation (37): Mx i + My j ∂M (r, t) =γ M (r, t) × B(r, t) − ∂t T2 Mz − M0 − k + D0 ∇ 2 M (r, t) T1
[29]
where i, j, and k are unit vectors along the x-, y-, and z-directions in the laboratory reference frame, respectively, γ is the gyromagnetic ratio of the nuclei, T1 and T2 are the NMR spin-lattice and spin–spin relaxation times, respectively, and D0 is the isotropic diffusion coefficient. B(r,t) = B0 k + b(r,t) is the vector sum of the magnetic field at the observation point (r,t), where B0 k is the main external magnetic field and b(r,t) is an additional magnetic field generated by the susceptibility variations in the sample. Since it is generally true that B0 >> |b(r,t)|, only the component of b(r,t) along the main field direction, i.e., bz (r,t), needs to be considered. Equation [29] may be simplified and expanded into its Cartesian terms to yield Mx ∂Mx (r, t) = γ (B0 + bz (r, t))My (r, t) − + D0 ∇ 2 Mx (r, t) ∂t T2 [30] ∂My (r, t) My = −γ (B0 + bz (r, t))Mx (r, t) − + D0 ∇ 2 My (r, t) ∂t T2 [31] Mz − M0 ∂Mz (r, t) =− + D0 ∇ 2 Mz (r, t) ∂t T1
[32]
Equations [30] and [31] are recombined to yield m(r, t) ∂m(r, t) = −iγ (B0 + bz (r, t))m(r, t) − + D0 ∇ 2 m(r, t) ∂t T2 [33] where m(r, t) = Mx (r, t) + iMy (r, t) is the transverse magneti√ zation and i = −1 is the complex unit. Defining m(r, t) = m(r, t) × exp(−(iω0 + T12 )t) with ω0 = γ B0 yields ∂m(r, t) = −iγ bz (r, t)m(r, t) + D0 ∇ 2 m(r, t) ∂t
[34a]
Mz − M0 ∂Mz (r, t) =− + D0 ∇ 2 Mz (r, t) ∂t T1
[34b]
346
Hu
Solving Eqs. [34a] and [34b] under an arbitrary field Bz (r,t) is difficult. Instead, the method of cumulative expansion (35) is used. Using the fact that ∇ 2 M (r, 0) = 0, we obtain ∇ 2 Mz (r, 0) = 0 and ∇ 2 m(r, 0) = 0. Using ∇ 2 m(r, 0) = 0 to replace ∇ 2 m(r, t) in Eq. [34a] and integrating with respect to t, the first cumulative solution of Eq. [34a] is
m(r, t) = m(r, 0) exp −iγ
t
0
′
bz (r, t )dt
′
[35]
Substituting Eq. [35] into the right-hand side of Eq. [34a] and again integrating with respect to t, the second cumulative solution of Eq. [34a] is
m(r, t) =m(r, 0) exp −i
t 0
′
γ bz (r, t )dt
′
⎤ ′ 2 ′ γ t ∇b (r, t ′′ )dt ′′ } dt z 0 0 ⎦ ×⎣ t ′ t′ 2 ′′ ′′ {−iD } + exp 0 0 dt γ 0 ∇ bz (r, t )dt =m(r, 0)exp(−iψ(r, t)) × exp (−bD0 ) + exp −iϕ(r, t) [36] t where ψ(r, t) = 0 γ bz (r, t ′ )dt ′ is the usual phase accumulation of the magnetization in the presence of the field bz (r, t ′ ) and ′ ⎡
exp {−D0
t
t
ϕ(r, t) = {D0
t
dt ′ γ
0
0
∇ 2 bz (r, t ′′ )dt
′′
is a phase accumulation due to the combination of diffusion and the second derivative of the field bz (r, t ′ ). In case ∇ 2 bz (r, t) = 0, which is valid for most pulsed field gradient experiments, Eq. [36] is simplified to m(r, t) = m(r, 0) exp(−iψ(r, t)) × exp (−bD0 ) t k2 (r, t ′ )dt ′ b=
[37a] [37b]
0
k(r, t ′ ) = γ
t′
G(r, t ′′ )dt ′′
[37c]
0
where G(r, t) = ∇bz (r, t) is the gradient of the field. By inserting Eq. [37] into Eq. [34a] to carry out the third cumulative expansion, it is trivial to prove that Eq. [37] is the exact solution of Eq. [34a] provided that ∇ 2 bz (r, t) = 0, i.e., a constant gradient field. Under a constant field gradient G(r, t) = G0 , it follows from
Slow Magic Angle Sample Spinning
347
Eqs. [37a], [37b], and [37c] that the magnetization following a single pulse 900 − acq(t) is
1 2 2 3 m(r, t) = m(r, 0) exp(−iψ(r, t)) × exp − D0 γ G0 t [38] 3
and 0 the magnetization following a CPMG pulse sequence 90 − τ − 1800 − τ − n − acq(t) is
1 2 2 2 m(r, t) = m(r, 0) exp(−iψ(r, t)) × exp − D0 γ G0 (2τ ) (nτ ) 12 [39]
where the effective gradient field defined in ref. (36), i.e., the π pulse changes the sign of the gradient prior to it, is used to perform the integration in Eq. [37c]. These results are consistent with those reported previously (38, 39). Similarly, using the fact that ∇ 2 Mz (r, 0) = 0, it can be seen from Eq. [34b] that ∇ 2 Mz (r, t) = 0 is satisfied when t = 0. Consequently, the longitudinal magnetization Mz (r, t) is independent of molecular diffusion. This means that the T1 values can still be faithfully measured even at the existence of the pulse field or susceptibility field gradients.
2.6. The Effect of Molecular Diffusion to the PHORMAT Experiment
It is known from Eqs. [36] and [37] that if ∇ 2 bz (r, t) = 0, the effect of diffusion on the FID is an attenuation of the signal by a factor of exp (−bD0 ). This means that when working on a pulse sequence, we can treat the phase part and the amplitude part separately. It follows from the principles given above and those given in the original PHORMAT article (15) that the phase part created by the magnetic susceptibility field from both the sphere and the cylindrical geometries is averaged to zero and only the isotropic contribution is left along the evolution dimension (t1 ). The resultant FID observed by the PHORMAT sequence is FID(t2 , t1 ) = exp(j ωiso t1 )F2 (t2 )
[40]
where F2 (t2 ) is the FID along the acquisition dimension that is broadened by the magnetic susceptibility field. The results in Eq. [40] mean that the isotropic dimension of the PHORMAT experiment is free from susceptibility broadening. In the following, we will examine the attenuation factor of the PHORMAT experiment due to diffusion under a constant field gradient. Working backward and ignoring the triple-echo sequence because its length is usually smaller than the evolution
348
Hu t1/3
t1/3
t1/3
t2
RF T/3
T/3
Effect Gradient
Fig. 20.3. The effective field gradient in a PHORMAT experiment at a constant gradient field G0 . The pulses are all π /2 pulses.
increment in a biological application (22), the effective field gradient for both the (+) and the (–) pulse sequences is simplified and the result is pictured in Fig. 20.3. It follows from Eq. [37] that by performing the first-time integration over the effective gradient using Eq. [37c], one finds k(r, t ′ ). b is obtained by performing the second-time integration over the square of k(r, t ′ )(see Eq. [37b]) and the resultant b for the evolution dimension is b=
1 2 2 2 γ G0 t1 T 27
[41]
The b value along the acquisition dimension is the same as that described in Eq. [38], i.e., b=
1 2 2 3 γ G0 t2 3
[42]
With diffusion incorporated, Eq. [40] becomes 1 FID(t2 , t1 ) = exp − D0 γ 2 G02 t12 T exp −jωiso t1 27 [43] 1 2 2 3 exp − D0 γ G0 t2 F2 (t2 ) 3 Including the attenuation due to spin-lattice relaxation time along the evolution dimension, the modified version of Eq. [43] is −t1 1 FID(t2 , t1 ) = exp − 27 exp(−jω D0 γ 2 G02 t12 T exp −2 T3T t ) iso 1 1 exp − 13 D0 γ 2 G02 t23 F2 (t2 ) [44]
The detailed evaluation of Eq. [44] in the presence of large magnetic susceptibility field using the phantom sample of glass beads + H2 O is given in Section 3.
Slow Magic Angle Sample Spinning
2.7. 2D PASS Experiment at a Magnetic Susceptibility Field and at the Existence of Molecular Diffusion
349
It follows from the original 2D PASS report (14) that the powderaveraged signals as a function of pitch θ without molecular diffusion are given by FID(θ, t2 ) =
∞
k =−∞
a(k) exp( − ikθ)exp i(ωiso + kωr )t2
[45]
× exp[−(T + t2 )/T2 ] where a(k) is the powder-averaged sideband amplitude of order k in a conventional MAS experiment, T is the rotor period, and T2 is the spin–spin relaxation time constant. For the 2D PASS experiment, θ is incremented between 0 and 2π and is related to the positions of the five π pulses according to the set of PASS equations given in the original report (14). The PASS experiment is named by the total number of θ steps. For example, PASS-16 means that there are a total of 16 increments along the θ dimension. Equation [45] means that in a 2D PASS experiment, the sidebands are separated according to the order of sideband. T2 produces a uniform attenuation of the FID that is θ independent. In order to simplify the discussion, the concept of constant field gradient, i.e., ∇ 2 Bz (r, t) = 0, is again used. The effective gradient in a 2D PASS experiment is given in Fig. 20.4. It follows from Eqs. [37c] and [37b] that the b factor corresponding to Fig. 20.4b along the evolution dimension θ is 2 b = γ 2 G02 T 3 (1 − t5 )3 + (2t5 − t4 − 1)3 + (t3 − 2t2 + 2t1 )3 3 +(t2 − 2t1 )3 + t13 =γ 2 G02 T 3 f (θ)
[46a] t2
(a) 0
t1
t2
t3
t4
t5
T
t2 (b)
0
t1
t2
t3
t4
t5
T
Fig. 20.4. The effective field gradient in a 2D PASS experiment at a constant gradient field G0 . (a) Pulses sequence, where the initial pulse is a π /2 pulse and the remaining five are π pulses. (b) The effective gradient in the existence of a constant gradient G0 when the effect of the π pulses is considered.
350
Hu
2 (1 − t5 )3 + (2t5 − t4 − 1)3 + (t3 − 2t2 + 2t1 )3 3 +(t2 − 2t1 )3 + t13 [46b] where t1 –t5 are expressed in units of T and for each θ there is a unique set of t1 –t5 (14). In order to obtain Eq. [46a], the PASS condition, i.e., Eq. [29] in ref. (14), was used. When the five π pulses are equally spaced, we have f (θ) =
b=
1 2 2 3 γ G0 T 108
[46c]
A plot of the function f (θ) versus θ for both PASS-16 and PASS-32 is provided in Fig. 20.5. Figure 20.5 indicates that f(θ) is only a function of θ. This is because the same set of time delays (t1 –t5 ) is found for the same value of θ, which is independent of the total number of evolution increments along the θ dimension in a 2D PASS experiment (14). The amplitude part in Eq. [37] for the acquisition dimension t2 is described again by that of Eq. [38]. Putting it all together, Eq. [45] becomes FID(θ, t2 ) =exp(−(T /T2 )) ⎫ ⎧ (k) exp(−ikθ)exp(−D γ 2 G 2 T 3 f (θ))× ⎪ ⎪ a ⎪ 0 ∞ ⎨ ⎬ ⎪ 0 × exp i(ωiso + kωr )t2 exp − 13 D0 γ 2 G02 t23 ⎪ ⎪ ⎪ ⎭ ⎩ exp(−t /T ) k=−∞ ⎪ 2
2
[47] According to the theoretical prediction given in Eq. [47], the spin–spin relaxation time T2 causes a fixed attenuation of the total signal according to exp(−(T /T2 )), where T denotes the rotor period, but this term does not affect the quality of the 2D PASS in separating the spinning sidebands by order. Molecular diffusion also causes signal attenuation; however, the attenuation factor is θ dependent, i.e., according to the factor exp(−D0 γ 2 G02 T 3 f (θ)).
f(θ)
0.0 0.03 0.0 0.02 0.0 0.01 0.0 0.00 0
PASS-32 PASS-16 0
10
20
30
40
Pitch (θ) Fig. 20.5. The relationship between f(θ) and θ for PASS-32 and PASS-16.
Slow Magic Angle Sample Spinning
351
The detailed evaluation of the θ-dependent term in the presence of large magnetic susceptibility field using the sample of glass beads + H2 O will be given in Section 3.
3. Model Experiments and Typical Results
3.1. 1 H T1 and T2 Experiments
Unless otherwise specified, all 1 H-NMR experiments were performed on a Varian–Chemagnetics 300-MHz Infinity spectrometer, with a proton Larmor frequency of 299.982 MHz. A standard Chemagnetics CP/MAS probe with a 7.5-mm pencil-type spinner system was used. In order to spin at low frequencies, the rotor was equipped with a flat drive tip, and an airflow restriction was used in the driver channel. The spinning rate was controlled using a commercial Chemagnetics MAS speed controller under the automated control mode. By marking the rotor with three evenly spaced marks, the frequency stability was better than ±0.3 Hz at a spinning rate from 1 to 200 Hz. Spinning rates higher than 200 Hz were obtained after removing the airflow restriction in the driver channel and by replacing the flat drive tip with a standard tip. Glass beads with diameters of 210–250 µm were used as a model tissue sample. The glass beads were loaded into the rotor and then tap water was added, resulting in a homogenous mixture of beads and H2 O. Because of the large magnetic susceptibility field from the glass beads, this model system represents an upper limit that could be obtained in a biological sample, i.e., tissues close to an air cavity or a tissue–bone interface. Unless otherwise specified, the beads + H2 O system will be the model phantom system for experiments throughout this chapter. Further experimental details can be found in refs. (16, 22) for the 1 H 2D PASS experiments and the 1 H PHORMAT, respectively. The only difference between the published experiments and those described in this chapter is that no water suppression is applied in the current experiment because H2 O is the signal to be observed. The 1 H T1 of the H2 O in the mixture of glass beads + H2 O, as measured by the conventional inversion recovery method at spinning rates ranging from 30 to 500 Hz, is 1.59 ± 0.03 s. This observation confirms the prediction by Eq. [34b] that T1 is independent of molecular diffusion. 1 H T2 measured by a CPMG pulse sequence π − (τ − π − τ −)n − acq 2
352
Hu
Table 20.1 1 H T of H O in a mixture of glass beads + H O at different 2 2 2 spinning rate Spin Rate (Hz)
T2f (ms)
Percentage of T2f
T2s (ms)
Percentage of T2s
T2 (one exp fit)
100
–
–
27
100
27
200
14
76
70
24
31
400
11
69
76
31
33
500
13
65
86
35
39
where τ is the rotor period of the spinning, is generally multi exponential and can fit well with two components. The resultant values obtained at spinning rates from 100 to 500 Hz are summarized in Table 20.1. Recall that in our measurement, n started from 1 instead of 0, and the component with a very short T2 value is significantly reduced at a low spinning rate compared to that of the component with a longer T2 value. Despite this fact, Table 20.1 still indicates that the measured T2 value is decreased at a lower spinning rate due to molecular diffusion. This is because the π pulses in the CPMG pulse sequence used are less efficient in suppressing the effect of molecular diffusion at a low spinning rate than that at a higher spinning rate. 3.2. 1 H PHORMAT Experiments
Figure 20.6 shows the spectra along the isotropic dimension of the PHORMAT experiment at various sample spinning rates ranging from 1 to 50 Hz. The projection spectra along the acquisition dimension at these spinning rates were found essentially to be the same with a linewidth of 3745 Hz, defined as the full width at the half-height positions of the resonant line. It is known from Fig. 20.6 that line narrowing along the isotropic dimension of the PHORMAT experiment increases at increasing spinning rate. For example, a line narrowing factor of 22 was obtained at 50 Hz, while only a factor of 4 was achieved at 1 Hz. A plot of the isotropic linewidth subtracted by the natural linewidth obtained at a sample spinning rate of 1 kHz, ν1/2 , versus the spinning rate f is provided in Fig. 20.7, from which the following relationship is obtained with a correlation coefficient of 0.9991: ν1/2 = 911.31 × f −0.4718
[48]
The results given by Eq. [48] can be compared with the the−t1 ) oretical prediction using Eq. [44], since the term exp(−2 T3T 1 in Eq.[44] causes only a fixed attenuation at a low spinning rate because t1