Supramolecular Chemistry: From Concepts to Applications 9783110595604

Citation preview

Stefan Kubik Supramolecular Chemistry

Also of Interest Inorganic and Organometallic Polymers Narendra Pal Singh Chauhan, Narendra Singh Chundawat,  ISBN ----, e-ISBN ----, EPUB ----

Advanced Materials Theodorus van de Ven, Armand Soldera (Eds.),  ISBN ----, e-ISBN ----, EPUB ----

Reversible Deactivation Radical Polymerization Nikhil K. Singha and Jimmy Mays (Eds.),  ISBN ----, e-ISBN ----, EPUB ----

Silicon-based Polymers and Materials Jerzy J. Chruściel,  ISBN ----, e-ISBN ----, EPUB ----

Host-Guest Chemistry Brian D. Wagner,  ISBN ----, e-ISBN ----, EPUB ----

Stefan Kubik

Supramolecular Chemistry From Concepts to Applications

Author Prof. Dr. Stefan Kubik Technische Universität Kaiserslautern Fachbereich Chemie - Organische Chemie Erwin-Schrödinger-Str. 54 67663 Kaiserslautern Email: [email protected]

ISBN 978-3-11-059560-4 e-ISBN (PDF) 978-3-11-059561-1 e-ISBN (EPUB) 978-3-11-059357-0 Library of Congress Control Number: 2020943952 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2021 Walter de Gruyter GmbH, Berlin/Boston Cover image: Stefan Kubik Typesetting: Integra Software Services Pvt. Ltd. Printing and binding: CPI books GmbH, Leck www.degruyter.com

For my family Daniela, Miriam, and Jakob and my teachers Peter Klein, Günter Wulff, and Julius Rebek Jr.

Preface The year 2017 marked the 50th anniversary of supramolecular chemistry, whose origin is generally considered to date back to 1967, when Charles Pedersen’s first paper on crown ethers was published. One could argue that the field is, in reality, much older because supramolecular aspects were investigated even before 1967. Nonetheless, Pedersen’s publication undoubtedly served as a starting point for supramolecular chemistry to develop into the prominent research field it is today. With the almost explosive developments over the last five decades, supramolecular chemistry has played an important role in shaping the face of modern chemistry, partly because of its multidisciplinary character that allows bridging many scientific disciplines. A modern education in chemistry would therefore be incomplete without including some relevant aspects. I wrote this book with the intention to provide this basic knowledge. One of my objectives was to outline the differences between supramolecular chemistry and the chemistry taught in the early courses of chemistry studies. Supramolecular systems are, for example, typically held together by weak interactions, which causes entropy to have a profound influence on their thermodynamic stability. Moreover, the interactions are reversible and most systems are therefore highly dynamic, which requires a special view to understand their behavior. In addition to these fundamental aspects, I have devoted a large fraction of the book to the many fascinating topics that are associated with supramolecular chemistry today. Readers will thus obtain an overview of the field, but can also use the book as a reference since key concepts are easily identified by the questions distributed among the chapters. The first one is the following: How is this book organized?

The book starts with an introduction in which the term supramolecular chemistry is defined, and the historic development of the field and its current relevance explained. The next two chapters are then devoted to the fundamentals of molecular recognition processes. These chapters introduce the formalism of describing and characterizing the interactions of molecules and the different natures of these interactions. Many key concepts are explained in this context, and a good grasp of these aspects helps understanding the behavior of the systems described in later chapters. The fourth chapter focuses on the most important classes of receptors, their structures, syntheses, and binding properties. These receptors allow the complexation of a variety of different substrates and the chapter thus concludes with a classification indicating which receptor is used for which substrate type. The following chapters then deal with various specific topics of supramolecular chemistry, with the fifth chapter introducing strategies to assemble molecules and the sixth chapter concentrating on interlocked molecules. Chapters 7–10 address https://doi.org/10.1515/9783110595611-202

VIII

Preface

how to control molecular motion or develop supramolecular catalysts, carriers, or probes. These chapters also give first insight into applications of supramolecular systems. This aspect is examined in more detail in the final chapter to also outline potential directions into which the field might head in the future. Who could benefit from this book?

Although the book primarily addresses readers who are only starting to familiarize themselves with supramolecular chemistry and would like to obtain an accessible introduction, it may also be useful for those who already have a good understanding of the basic concepts and are interested in specific aspects. There is therefore no single way to read this book. Starting with the first three chapters is likely helpful to understand the concepts and terminology. All other chapters can be read independently, and additional information is hopefully located quickly with the available crossreferences. Are all aspects of supramolecular chemistry treated?

The short answer is no. The field is so large that I had to omit certain topics to retain the introductory character of the book. I decided to place the focus on supramolecular chemistry in solution, concentrating on discrete and structurally characterized complexes or assemblies, but to exclude or only briefly mention aspects relating to the solid state (clathrates, crystal engineering, molecular tectonics), materials (polymers, foldamers, gels, nanoparticles), or other larger assemblies (micelles, vesicles). I made this selection not because the latter aspects are less relevant but because I felt that they are outside the scope of an introductory textbook. I am aware that this view might not be shared by everyone, and I therefore apologize to all those who miss certain topics. I also apologize to all whose work or names I did not mention, although they made important contributions to the field. I had to make choices, which meant that I had to leave out many fine examples, even from friends and colleagues whose work I admire. I thank the people who supported me during the time of writing, most importantly my family. I thank Julia Bartl, Arne Lützen, and Konrad Tiefenbacher, who read parts of the book and made many helpful suggestions. I am also grateful to the team at DeGruyter for motivating me to embark on this venture and for their continuing support. I hope that this book will turn out to be a valued companion for many. Kaiserslautern, June 2020

Contents Preface

VII

Questions discussed

XV

1

Introduction and overview Bibliography 6

1

2 2.1 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.2.5 2.2.6

Analyzing complex formation 7 Thermodynamic and kinetic aspects 7 Analytical strategies and techniques 17 Strategies 17 NMR spectroscopy 23 UV–vis spectroscopy 28 Fluorescence spectroscopy 30 Potentiometry 30 Isothermal titration calorimetry 32 Bibliography 36

3 Understanding molecular recognition 37 3.1 Modes of binding 37 3.1.1 General considerations 37 3.1.2 Ion–ion interactions 40 3.1.3 Ion–dipole interactions 42 3.1.4 Dipole–dipole interactions 44 3.1.5 Hydrogen bonding 45 3.1.6 Halogen bonding 54 3.1.7 Cation–π interactions 56 3.1.8 Anion–π interactions 60 3.1.9 Aromatic–aromatic interactions 62 3.1.10 Dispersion interactions 64 3.2 Binding energies 66 3.2.1 General considerations 66 3.2.2 Trend analyses 67 3.2.3 Double-mutant cycles 69 3.2.4 Molecular balances 71 3.3 Solvent effects 72 3.4 Predicting binding strength in solution 80 3.5 Guidelines for receptor design 85 3.5.1 Complementarity, preorganization, and induced fit 3.5.2 Chelate effect and macrocyclic effect 87

85

X

3.5.3 3.5.4

Contents

Multivalency and cooperativity Allosterism and cooperativity Bibliography 98

89 93

4 Hosting ions and molecules 101 4.1 Receptors 101 4.1.1 Crown ethers 101 4.1.2 Cryptands 114 4.1.3 Spherands 122 4.1.4 Cyclodextrins 125 4.1.5 Cyclophanes 137 4.1.6 Cyclotriveratrylenes 144 4.1.7 Calixarenes 153 4.1.8 Calixpyrroles 165 4.1.9 Resorcinarenes 169 4.1.10 Pillararenes 186 4.1.11 Cucurbiturils 193 4.1.12 Clefts and tweezers 204 4.1.13 Foldamers 215 4.2 Substrates 219 4.2.1 Inorganic and organic cations 220 4.2.2 Inorganic and organic anions 223 4.2.3 Zwitterions and ion pairs 228 4.2.4 Neutral organic molecules 231 Bibliography 235 5 5.1 5.2 5.3 5.3.1 5.3.2 5.3.3 5.3.4 5.4 5.4.1 5.4.2 5.4.3 5.5 5.5.1 5.5.2

Assembling molecules 245 Self-assembly and template effects 245 Self-assembly mediated by the hydrophobic effect 262 Self-assembly mediated by hydrogen bonds 266 Introduction 266 Rosettes 271 Capsules 278 Tubes 290 Self-assembly mediated by halogen bonds 295 Introduction 295 Helices 296 Capsules 297 Self-assembly mediated by coordination bonds 297 Introduction 297 Helices 302

Contents

5.5.3 5.5.4 5.5.5 5.6 5.6.1 5.6.2 5.6.3 5.7 5.7.1 5.7.2 5.7.3 5.7.4 5.8

Grids 308 Rings 309 Cages 314 Self-assembly mediated by covalent bonds Introduction 326 Rings 330 Cages 335 Dynamic combinatorial chemistry 341 Introduction 341 Casting 348 Molding 351 Self-assembly 356 Systems chemistry 358 Bibliography 360

326

6 6.1 6.2 6.2.1 6.2.2 6.3 6.3.1 6.3.2 6.3.3 6.4 6.4.1 6.4.2 6.5 6.5.1 6.5.2 6.5.3 6.6 6.6.1 6.7 6.7.1 6.7.2

Threading molecules 369 Molecular topology 369 Synthetic strategies 374 Molecular strategies 374 Supramolecular strategies 380 Syntheses using metal coordination 383 Catenanes 383 Knots 390 Rotaxanes 394 Syntheses using charge-transfer interactions 398 Catenanes 398 Rotaxanes 403 Syntheses using hydrogen bonds 404 Catenanes 404 Knots 414 Rotaxanes 416 Syntheses using halogen bonds 418 Rotaxanes 418 Syntheses using the hydrophobic effect 420 Knots 420 Rotaxanes 423 Bibliography 424

7 7.1 7.2

Controlling molecular motion 429 Introduction 429 Rotaxane-derived machines 439

XI

XII

7.3 7.4

Contents

Catenane-derived machines 451 Machines without mechanical bonds Bibliography 459

454

8 8.1 8.2 8.2.1 8.2.2 8.3 8.3.1 8.3.2 8.4

Mediating molecular transformations 463 Introduction 463 Stoichiometric transformations 471 Transformation by functional group participation Transformation by confinement 475 Catalytic transformations 481 Transformation by functional group participation Transformation by confinement 493 Self-replication 498 Bibliography 508

9 9.1 9.2 9.2.1 9.2.2 9.3 9.3.1 9.3.2 9.4

Transporting molecules 513 Introduction 513 Cation transport 518 Channels 518 Carriers 522 Anion transport 523 Channels 523 Carriers 525 Water transport 528 Bibliography 530

10 Detecting molecules 533 10.1 Introduction 533 10.2 Single analyte sensing 536 10.2.1 Direct optical sensing 536 10.2.2 Indirect optical sensing 542 10.2.3 Direct electrochemical sensing 10.3 Multiple analyte sensing 547 Bibliography 553

544

11 Applying supramolecular systems 555 11.1 Introduction 555 11.2 Applications in medicine 556 11.2.1 Drugs 556 11.2.2 Drug formulations 558 11.2.3 Imaging 559 11.3 Applications in separation processes

562

471

481

Contents

11.3.1 11.3.2 11.3.3 11.4 11.4.1 11.4.2 11.4.3 11.5 11.6 11.7 11.7.1 11.7.2 11.7.3

12 12.1 12.2

Index

Chromatography 562 Extraction 563 Precipitation 565 Applications in materials chemistry Self-assembled polymers 567 Polymer networks 569 Self-assembled gels 573 Applications in catalysis 574 Applications in molecular electronics Applications in consumer products Textiles 578 Food 578 Household 579 Bibliography 579

567

577 578

Appendices 583 Concentrations in a 1:2 binding equilibrium 583 Concentrations in an indicator displacement assay Bibliography 586 587

584

XIII

Questions discussed Preface VII How is this book organized? VII Who could benefit from this book? VIII Are all aspects of supramolecular chemistry treated? 1

2

3

4

5

6

Introduction and Overview 1 What is supramolecular chemistry? 1 How did supramolecular chemistry emerge and develop?

VIII

4

Analyzing Complex Formation 7 How are 1:1 complexation equilibria formally described? 7 What happens when binding equilibria become more complex? 10 How does complex stability relate to binding enthalpy and entropy? 12 Is there a relationship between complex stability and the rate of complex formation? How can complex stoichiometry be determined? 17 How can complex stability be determined? 21 How can binding equilibria be analyzed? 23 Understanding Molecular Recognition 37 Which general parameters influence complex formation? 37 Which types of interactions cause molecules to stay together? How strong are intermolecular interactions? 66 How does the solvent influence complex stability? 72 How does water mediate molecular recognition? 76 Can binding strength be predicted? 80 Which strategies exist to achieve strong binding? 85

40

Hosting Ions and Molecules 101 Which strategies exist to favor macrocyclization reactions? 116 How does NMR spectroscopy help to characterize receptor-substrate complexes? How do water molecules in a receptor cavity contribute to complex formation? How does the counterion influence the binding of an ion to a receptor? 231 Assembling Molecules 245 What is the difference between preorganization and predisposition? How do templates work? 255 Can molecules be sorted? 259 How much space does a substrate usually occupy in a receptor cavity? Can a liquid be porous? 338 Does a template always amplify the best binder? 345 Are mixtures of molecules always messy? 358 Threading Molecules 369 What is the difference between topology and structure? 369 What is a mechanical bond? 373 What is a co-conformation? 386 When is a template passive and when is it active? 395

https://doi.org/10.1515/9783110595611-204

15

254

281

144 200

XVI

7

Questions discussed

Controlling Molecular Motion 429 What is a molecular machine? 429 What is the difference between a shuttle and a switch? Can a molecular motor power a car? 457

8

Mediating Molecular Transformations Do synthetic enzymes exist? 468 Is only DNA able to replicate? 498

9

Transporting Molecules 513 How do polar species cross cell membranes? How is membrane transport studied? 515

10 Detecting Molecules What is a sensor?

533 534

11 Applying Supramolecular Systems What is all of this good for? 555 Is this the end? 579 12 Appendices 583 How do I do the math?

463

583

555

513

437

1 Introduction and overview What is supramolecular chemistry?

A substantial part of the training in chemistry focuses on molecular chemistry. We learn in this context about the correlation between the structure of a molecule and its reactivity and about the synthetic methods available to form covalent bonds. The respective theoretical framework allows us to rather reliably predict how molecules react or how they are synthesized just by looking at their structural drawings. Let us take the reaction shown in Figure 1.1 as an example. We see that catechol and bis(2-chloroethyl) ether afford a macrocyclic product with six ether groups along the ring when treated with sodium hydroxide. Considering the intrinsic reactivity of phenols and alkyl halides under basic conditions, we can attribute the formation of each of the four new C–O bonds to the initial deprotonation of a catechol hydroxy group. Subsequently, the so-formed nucleophilic phenolate reacts with an electrophilic carbon atom in the reaction partner that releases a chloride ion as the leaving group. Each bond formation thus comprises a nucleophilic substitution and the formal generation of one equivalent of water and one equivalent of NaCl.

O OH 2

+

2 Cl

O

OH

NaOH

O

1-Butanol, reflux 44–48%

O

O

Cl O O

Figure 1.1: Reaction between catechol and bis(2-chloroethyl) ether under basic conditions that ultimately affords the depicted macrocyclic oligoether.

This mechanism adequately describes the actual formation of the C–O bonds, but does it also explain the formation of the macrocyclic product? Maybe if only two molecules of catechol and two molecules of bis(2-chloroethyl) ether would be present as Figure 1.1 suggests, but molecules are in reality rarely as lonely as in reaction schemes. Even reactions performed on a small scale involve the participation of an extremely large number of molecules when all reaction partners, reagents, and solvent molecules are considered. These molecules are moreover in constant motion and permanently bump into each other, rendering reaction mixtures very crowded and dynamic environments. As a consequence, the probability is high that a phenolate ion not only meets the correct partner in the above reaction but also the linear intermediates formed on the way to the product, potentially causing the composition of the solution to become very complex. Experimentally, however, the formation of https://doi.org/10.1515/9783110595611-001

2

1 Introduction and overview

the cyclic product proceeds surprisingly selectively and in good yields, even at a catechol concentration of 1.3 M, far from the high dilution conditions often used in cyclization reactions to favor ring formation over the unwanted chain elongation. The mechanism of ether formation, that is molecular chemistry, does not provide a straightforward explanation for this observation, indicating that effects that transcend the actual bond formation reaction might operate in this synthesis. As it turns out, these effects are related to the presence of the sodium ions that, although unimportant for C–O bond formation, are no innocent bystanders but play an active role in the reaction. At its onset, the sodium ions originating from the base preferentially interact with the solvent molecules. Once linear oligoethers start to appear in solution, however, they serve as additional and more potent binding partners. The respective interactions are mediated by the sequence of oxygen atoms along the chains of these intermediates, causing them to wrap around the cation. This situation is shown schematically in Figure 1.2 for the immediate precursor of the product. The resulting proximity of the phenolate group to the electrophilic carbon atom at the opposite end of the chain explains why the sodium ions facilitate the ring closure and improve the efficiency of the reaction.

O O

O Na

O

O O

Cl

Figure 1.2: Structure illustrating how a sodium ion preorganizes the linear precursor for the cyclization that affords the product in the reaction shown in Figure 1.1.

The principles underlying sodium ion complexation will be explained in later chapters, where many other systems in which similar interactions play a role will also be presented. In this introduction, the above example should just serve to illustrate that noncovalent interactions can exert characteristic effects on the outcome of a reaction. They are also responsible for the stereoselectivity of transformations mediated by certain catalysts, but their relevance extends far beyond reaction control. In biological systems, for example, such interactions induce protein folding, the substrate selectivity of enzymes, signal transduction, transport, the conservation and transmission of the genetic code, and many other fundamental biochemical processes. They are also important in various other areas of chemistry such as medicinal and materials chemistry, but to molecular chemistry, which is primarily concerned with creating covalent bonds, they are not central. The realm of intermolecular interactions, instead, lies at the heart of supramolecular chemistry, which is a field of chemistry whose name was coined by one of the pioneers, namely, Jean-Marie Lehn, who chose the Latin prefix supra to indicate that supramolecular chemistry transcends molecular chemistry. Lehn wrote in 1995, “Beyond molecular chemistry based

1 Introduction and overview

3

on the covalent bond there lies the field of supramolecular chemistry whose goal is to gain control over the intermolecular bond. It is concerned with the next step in increasing complexity beyond the molecule towards the supermolecule and organized molecular systems […]” [1]. This definition implies that supramolecular systems consist of structurally defined assemblies of interacting molecules whose formation is controlled by organizational principles encoded within the structures of the individual components. Figure 1.3 illustrates this idea.

Figure 1.3: Schematic representation of the formation of a supramolecular assembly from shape-complementary objects or molecules.

The shapes of some of the objects shown in this illustration allow them to arrange themselves such that the four curved segments surround the circle. The final arrangement is stabilized by attractive interactions between the individual components. These components thus recognize each other, ignoring the square-shaped constituents that cannot be incorporated into the product. An important prerequisite for this process to work is that errors occurring on the way to the final structure, which derive, for example, from incorrectly connected components, are constantly corrected. The interactions responsible for the assembly therefore need to be reversible, rendering the overall system dynamic and subject to thermodynamic control. Based on these considerations, we arrive at the following definition for supramolecular chemistry, which focuses on two basic principles, namely, recognition and reversibility, and covers most of the systems and processes discussed in this book. Supramolecular chemistry is a field in chemistry that deals with molecular recognition phenomena mostly under thermodynamic control.

Although this definition may not cover all aspects of supramolecular chemistry, since kinetically controlled processes also sometimes play a role, it is preferable to very broad definitions, which state, for example, that supramolecular chemistry comprises the development of functional molecules. While supramolecular systems are certainly functional, as we will see, it is questionable whether the reverse is always true. Acid–base indicators, for example, are functional because their optical properties depend on the pH of the solution, but this property and its use has no relation to supramolecular chemistry. In this book, a narrower view is therefore preferred.

4

1 Introduction and overview

How did supramolecular chemistry emerge and develop?

The advent of supramolecular chemistry is closely related to the reaction shown in Figure 1.1. This reaction occurred as an unwanted side reaction during the synthesis of bis[2-(2-hydroxyphenoxy)ethyl] ether, performed in 1962 by Charles J. Pedersen at DuPont in Wilmington, Delaware (Figure 1.4).

2 HO

+ Cl

O

O

O

O

NaOH

O

O

O

HO

OH

O

O

1-Butanol Cl

H+

O

O

O

O

Figure 1.4: Synthesis of bis[2-(2-hydroxyphenoxy)ethyl] ether performed by Charles J. Pedersen.

Pedersen was interested in this reaction because he wanted to use the VO+ complex of the product as a catalyst for olefin polymerization. He was aware that the available tetrahydropyranyl-protected starting material contained 10% of unprotected catechol, but he did not consider this impurity to be problematic because he expected it to give rise to easily separable oligomeric and polymeric by-products. Unexpectedly, the presence of catechol also afforded the macrocycle shown in Figure 1.1 in a yield of 1%, which could have easily gone unnoticed if the unusual properties of this product would not have sparked Pedersen’s curiosity. He observed, for example, that this compound is insoluble in methanol but readily dissolves after the addition of sodium hydroxide, which he correctly attributed to the binding of the sodium cation in the center of the ring. Pedersen subsequently investigated, in more detail, the syntheses and properties of such cyclic oligoethers, for which he coined the term “crown ethers,” and published an extensive account of this work in 1967 in the Journal of the American Chemical Society, which deals with almost 50 derivatives [2]. This paper is generally considered to mark the birth of supramolecular chemistry. Pedersen‘s work served as an inspiration for several other groups to invent novel low-molecular compounds, so-called receptors, that possess cavities available for the incorporation of suitable substrates. The next major step in this direction was made by the Lehn group with the development of cryptands, bi- or tricyclic analogs of crown ethers, whose affinities for alkali metal ions is typically significantly higher than those of crown ethers. The work on cryptands was only the first of numerous contributions from Lehn to the field of supramolecular chemistry. His work ranges from the development of various receptors and catalysts to polymetallic coordination compounds and organic materials. Accordingly, entire areas of supramolecular

1 Introduction and overview

5

chemistry and the associated concepts and terms, including the term supramolecular chemistry itself, can be traced back to him. The third pioneer of the field, Donald J. Cram, introduced the first chiral crown ethers and showed that these compounds are capable of enantioselective substrate recognition. Cram thus transferred a fundamental concept to synthetic systems that is a consequence of the homochirality of biomolecules in Nature. He then turned to the development of supramolecular catalysts that not only mimic the substrate affinity of enzymes but also their ability to transform the bound substrate. Cram furthermore developed various new classes of receptors, introduced many concepts of supramolecular chemistry such as that of preorganization, and he invented the term “host–guest chemistry” to describe the interaction of a receptor with its substrate. This term is still in use because it aptly illustrates the hosting of the substrate by a receptor (one could argue, however, that the receptor is a hostel rather than a host), but it refers to only part of the much broader field of supramolecular chemistry. In 1987, the Royal Swedish Academy of Sciences awarded the Nobel Prize in chemistry to Pedersen, Lehn, and Cram for their pioneering work, particularly for “their development and application of molecules with highly selective structurespecific interaction, i.e. molecules that can ‘recognize’ each other and choose with which other molecules they will form complexes” [3]. It was furthermore emphasized that the relevance of the molecules developed by Cram, Lehn, and Pedersen reaches far into life sciences, as they allow mimicking processes, which were previously the exclusive domain of biomolecules. It would be incorrect to say that there was no supramolecular chemistry prior to the work of the three Nobel Prize winners. Fundamental concepts underlying supramolecular chemistry were actually established well before Pedersen’s seminal paper, and groups were involved in what was later termed host-guest chemistry prior to 1967. Table 1.1 gives an overview of selected concepts and discoveries, which precede supramolecular chemistry and which are considered to have contributed to laying the groundwork of the field. In spite of this important early work, the design and development of synthetic supramolecular systems only really began after the publication of Pedersen’s paper. A further rapid increase in the worldwide research activities can be noted after the Nobel Prize was awarded in 1987 when many creative scientists joined the field. At the same time, the research became more and more diverse, touching organic, inorganic, physical, theoretical, materials, and analytical chemistry, often in combination with biochemistry. This interdisciplinary research not only led to a better understanding of thermodynamic and kinetic aspects of molecular recognition phenomena but also to the development of a wide variety of novel supramolecular systems. In many cases, the inspiration came from Nature, resulting in the transfer of biochemical concepts to synthetic systems such as catalysis, allosteric control of substrate binding, induced fit, cooperativity, multivalency, information storage, replication, transport, motion, and so on. The progressive improvement of

6

1 Introduction and overview

Table 1.1: Concepts and discoveries reported prior to 1967 that are relevant for supramolecular chemistry.  Alfred Werner describes the concepts of coordination chemistry.  Emil Fischer introduces the lock-and-key concept to rationalize the substrate selectivity of enzymes.  Hans Pringsheim reports on the complexation of organic compounds by cyclodextrins, natural macrocyclic receptors that were originally discovered by Antoine Villiers in .  Linus C. Pauling publishes his seminal paper in the Journal of the American Chemical Society entitled “The Nature of the Chemical Bond” in which he also alludes to the hydrogen bond.  James D. Watson and Francis H. C. Crick describe the structure of the DNA double helix.  Daniel E. Koshland Jr. introduces the “induced fit” concept, which refers to the conformational changes proteins undergo upon substrate binding.  Daryl H. Busch proposes a classification for template effects.

the understanding of these principles moreover facilitated moving further and further away from natural models and allowed making new discoveries. The corresponding extensive research activities caused supramolecular chemistry to develop into a prestigious and influential field of research. The current significance is also reflected in the fact that ca. 4,800 publications appeared in 2019 containing the term “supramolecular,” that is, more than 10 articles per day. The number of articles related in a broader sense to supramolecular chemistry is probably even higher because the term “supramolecular” often goes unmentioned today. Clearly, supramolecular chemistry has contributed substantially to shaping the face of modern chemistry and will continue to do so. It is therefore worth taking a closer look at the different facets of this fascinating field, but this is not possible without a firm understanding of the basics. The relevant concepts are explained in detail in the next chapters.

Bibliography [1] [2] [3]

Lehn JM. Supramolecular Chemistry – Concepts and Perspectives. Weinheim, VCH, 1995, p. 2. Pedersen CJ. Cyclic polyethers and their complexes with metal salts. J. Am. Chem. Soc. 1967, 89, 7017–36. The Nobel Prize in Chemistry 1987 (Accessed April 30, 2020, https://www.nobelprize.org/ nobel_prizes/chemistry/laureates/1987/press.html).

2 Analyzing complex formation CONSPECTUS: Before we come to the actual forces that hold supramolecular systems together, general principles are introduced in this chapter on how to mathematically describe and experimentally characterize complex formations. Thus, this chapter provides an understanding of the basic concepts and methods underlying the characterization of the supramolecular complexes presented later in the book.

2.1 Thermodynamic and kinetic aspects How are 1:1 complexation equilibria formally described?

Molecular recognition processes occur between structurally complementary molecules and lead to complexes that are characterized by their stability, and by the number and arrangement of the interacting subunits. These processes can involve any number of molecules but we restrict the discussion initially to only two, which we name receptor and substrate. This distinction is not necessary. Many supramolecular systems in fact comprise binding partners that cannot be easily classified into these terms, but the above distinction helps to understand the following considerations. Examples of receptors and substrates are the host–guest systems discussed in Chapter 4, in which the receptor is typically larger and has a defined cleft or cavity to host the substrate (Figure 2.1).

+

R

+

S

C

Figure 2.1: Schematic representation of the complexation of a substrate S by a structurally complementary receptor R and the respective reaction scheme.

The equilibrium arrow in Figure 2.1 indicates that the formation of complex C is reversible and that the receptor R, the substrate S, and the complex coexist in solution. The larger the extent to which receptor and substrate are bound in the complex and the smaller the amounts of uncomplexed receptor and substrate, the more efficient the interactions. This efficiency is described in quantitative terms by

https://doi.org/10.1515/9783110595611-002

8

2 Analyzing complex formation

using Ka , the stability or association constant, which results from the law of mass action according to the following equation. Ka =

cC cR cS

(2:1)

Note that equation (2.1) specifies the amounts of receptor, substrate, and complex in concentrations (cR , cS , cC ) instead of dimensionless activities. This is more practical since activity constants are usually not available for the species involved in binding equilibria. The approximation of using concentrations is even justified to some extent because binding equilibria are often investigated in dilute solutions, but it causes the resulting stability constants to have dimensions. Stability constants associated with 1:1 equilibria, for example, have units of M−1 (L/mol) because the denominator in equation (2.1) contains a product of two concentrations. Supramolecular chemists prefer the use of stability constants to characterize binding equilibria, maybe because there is a direct correlation between magnitude and stability: the larger the Ka the more stable the respective complex. In biochemistry, binding efficiency is usually denoted in terms of dissociation constants Kd , which are the reciprocal values of stability constants (Kd = Ka− 1 ). Thus, Kd is expressed in units of M (mol/L) and becomes smaller with increasing complex stability. No matter which value one prefers, Ka and Kd belong to the key thermodynamic parameters to describe the stability of supramolecular complexes. They are characteristic for every receptor–substrate combination, but depend strongly on external influences such as temperature or solvent. Thus, a comparison of the performance of different receptors is only feasible if the stabilities of their complexes were quantified under comparable conditions. Equation (2.1) does not directly give access to the concentration of the complex because the concentrations of receptor and substrate in the equilibrium are unknown. These concentrations are, however, linked to the initial concentrations of receptor (c0R ) and substrate (c0S ) through the complex concentration (cC ) by the following mass balances. cR = c0R − cC

(2:2a)

cS = c0S − cC

(2:2b)

Combining equation (2.1) with (2.2a) and (2.2b) leads to (2.3), which is a quadratic equation in cC that can be solved in a straightforward manner. Ka =


cC =

c0R + c0S + Ka− 1 2

cC 


 − cC c0S − cC sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 0 2 cR + cS + Ka− 1 − − c0R c0S 4

c0R

(2:3)

(2:4)

9

2.1 Thermodynamic and kinetic aspects

Of the two possible solutions, only equation (2.4) with the minus sign in front of the square root correctly describes how cC depends on c0R , c0S , and Ka . The following arguments explain the reason: if Ka < 1 M−1, Ka− 1 becomes large and the terms in front and behind the minus sign approximately amount to Ka− 1 =2. As cC is almost zero under these conditions, the equation is only fulfilled if the two terms are subtracted. Equation (2.4) allows calculating how much of a 1:1 complex of known stability will form from a receptor and a substrate when starting from a mixture of both binding partners at given concentrations. To illustrate this relationship, Figure 2.2a shows how complex concentration changes when increasing amounts of a substrate are added to a receptor solution with c0R = 10−3 M for complexes with stability constants of 102, 103, 104, and 106 M−1.

(a)

1.0

0.08

cC (mM)

0.8

cC (mM)

(b)

0.10

0.6 0.4 0.2

0.06 0.04 0.02

0.0

0.00 0

1

2

3

4

5 c S 0 / c R0

0

1

2

3

4

5 c S 0 / c R0

Figure 2.2: Graphs showing how the complex concentration changes when increasing amounts of a substrate are added to a receptor solution of given concentration. The initial receptor concentration amounts to cR0 = 10−3 M (=1 mM) in diagram (a) and to cR0 = 10−4 M (=0.1 mM) in diagram (b). The curves represent complexes of different stability with the Ka amounting to 102 M−1 (orange), 103 M−1 (red), 104 M−1 (blue), and 106 M−1 (black). The dotted lines mark the 1:1 substrate/ receptor ratios in both graphs.

In all cases, the complex concentration progressively increases with increasing substrate concentration. The black curve, representing the most stable complex, exhibits an almost linear rise until a 1:1 substrate/receptor ratio is reached, showing that every substrate molecule added to the solution is consumed in the complex. Once saturation is reached, further substrate molecules do not cause any change because no more free receptor molecules are available. The shape of this curve is consistent with equation (2.4) when Ka becomes large and Ka− 1 is therefore negligible. Assuming that c0S = xc0R with x denoting the substrate/receptor ratio and Ka− 1 ~ 0, the rearrangement of equation (2.4) yields cC = 0.5½1 + x − absð1 − xÞc0R . The complex concentration cC thus increases linearly between 0 ≤ x ≤ 1 and remains equal to c0R if x > 1.

10

2 Analyzing complex formation

The curves describing the less stable complexes in Figure 2.2a become progressively shallower as Ka decreases. As a consequence, substantial amounts of uncomplexed receptor and substrate are still present in solution even if an excess of the substrate is added. For comparison, 97% of the receptor is complexed at a 1:1 substrate/receptor ratio in the case of the complex with a Ka of 106 M−1, whereas the corresponding fractions amount to 73%, 38%, and 8% for the complexes with stability constants of 104, 103, and 102 M−1, respectively. Figure 2.2b shows that the decrease of the receptor concentration also causes the curves to become shallower. As a consequence, higher amounts of the substrate are required to reach saturation with respect to the situation in more concentrated solutions. We conclude from these considerations that in order to saturate the receptor with the substrate one can either: – add an excess of the substrate, whereby the exact amount is determined by complex stability, – or increase the concentrations of receptor and substrate.

What happens when binding equilibria become more complex?

For complexes that do not have a 1:1 stoichiometry, the law of mass action in equation (2.1) is no more valid. However, the general strategy to mathematically describe such equilibria is not very different from that explained earlier. Let us assume that the receptor is able to bind a second substrate molecule, leading to the formation of a 1:2 receptor–substrate complex. In this case, complex formation is a stepwise process, which involves the initial formation of the 1:1 complex that is subsequently converted into the 1:2 complex. The overall reaction thus comprises two equilibria of which each is associated with an individual equilibrium constant. The following reaction schemes illustrate the two steps. The corresponding laws of mass action are specified in the equations (2.5a,b). R +

S Ð C11 Ka11 =

C11

+

c11 C cR cS

S Ð C12 Ka12 =

c12 C c11 C cS

(2:5a,b)

In addition to these laws of mass action, we again need mass balances that establish a relationship between the initial concentrations of the substrate c0S and the receptor 12 c0R , and the concentrations of the 1:1 complex c11 C , the 1:2 complex cC , the substrate cS , and the receptor cR in the equilibrium. These mass balances are specified in equations (2.6a) and (2.6b). Note the factor 2 in equation (2.6b), which reflects the fact that the 1:2 complex contains two substrate molecules. 12 cR = c0R − c11 C − cC

(2:6a)

11

2.1 Thermodynamic and kinetic aspects

12 cS = c0S − c11 C − 2 cC

(2:6b)

Combining equations (2.5) and (2.6a) and (2.6b) ultimately affords a cubic equation 12 that can be solved in c11 C and cC . This solution is derived in Appendix 12.1. Here, we only qualitatively assess how the concentrations of the different complex species vary depending on the receptor–substrate ratio and the stepwise binding constants by using the two examples shown in Figure 2.3.

(a)

(b)

1.0

0.8

cC (1:1/1:2) (mM)

cC (1:1/1:2) (mM)

1.0

0.6 0.4 0.2

0.8 0.6 0.4 0.2

0.0

0.0 0

1

2

3

4

5 c S0 / c R 0

0

1

2

3

4

5 c S0 / c R 0

Figure 2.3: Graphs showing the concentrations of the 1:1 complex (orange) and the 1:2 receptor–substrate complex (red) when different amounts of a substrate are added to a 10−3 M (=1 mM) receptor solution. The black curves denote the sum of the concentrations of the 1:1 and the 1:2 complexes. In diagram (a) Ka11 = 100 M−1 and Ka12 = 10,000 M−1 and in diagram (b) Ka11 = 10,000 M−1 and Ka12 = 100 M−1.

The graphs in Figure 2.3a are associated with a stepwise equilibrium characterized by a Ka11 that is smaller than Ka12 . The 1:1 complex is therefore only formed to a small extent, but this amount is converted almost completely into the 1:2 complex. As a consequence, the overall equilibrium is dominated by the 1:2 complex and the curve describing its formation has a clear sigmoidal shape. Figure 2.3b shows graphs describing a two-step equilibrium in which Ka11 > Ka12 . The 1:1 complex is in this case almost the only species in solution until the substrate is present in excess. Only at higher substrate concentrations, significant amounts of the 1:2 complex are formed at the expense of the 1:1 complex. All curves in Figures 2.2 and 2.3 differ characteristically in shape. They therefore provide crucial information about the strength and stoichiometry of the underlying interactions. Accordingly, binding equilibria can be characterized by following how the concentrations of one or more species involved in complex formation change when the receptor–substrate ratio is varied. The fitting of the obtained binding isotherms to a suitable mathematical model then allows assessing whether the assumed stoichiometry is correct, ultimately also affording the stability constants of the investigated system. An obvious mismatch between the experimental and the theoretical

12

2 Analyzing complex formation

isotherms immediately indicates that assumptions made when selecting the binding model, for example the expected complex stoichiometry, were incorrect. Sigmoidal binding isotherms, for example, should not be evaluated on the basis of 1:1 equilibria. Before we look at the methods available for following binding equilibria in more detail in Section 2.2, other pertinent thermodynamic parameters associated with the stability of a receptor–substrate complex must be introduced. How does complex stability relate to binding enthalpy and entropy?

The thermodynamic driving force of complex formation is described in quantitative terms with the following equation. ΔG0 = − RTlnKa

(2:7)

This equation establishes a correlation between the association constant Ka of a complex and the Gibbs free energy of its formation ΔG0 . The other parameters are the gas constant R and the temperature T. Complexation processes are thus exergonic, that is, associated with a negative Gibbs free energy if Ka > 1. These reactions occur spontaneously, with the extent to which the complex forms depending on the difference in the Gibbs free energy of the system prior to complex formation and in the equilibrium. The more negative the ΔG0 , the more stable the complex. Note that ΔG0 describes the thermodynamic driving force of complex formation if all binding partners are in their standard states, which for dissolved species means that their concentrations amount to 1 M. Relating the thermodynamics of complex formation to this standard state has the advantage that different systems can be compared, thus rendering ΔG0 a useful alternative to Ka to quantify complex stability. In contrast to Ka , however, ΔG0 refers to only a single and in most cases relatively unrealistic situation. Experimentally, complex formation is rarely investigated in 1 M solutions and even if both binding partners are (or can be) mixed at this concentration, their interaction causes the system to immediately leave the standard state because the progressive formation of the complex causes the concentrations of the free binding partners to decrease. As a consequence, the absolute value of ΔG (note the absence of the index 0 , which indicates that this value refers to states of the system other than the standard state) progressively decreases until the equilibrium is reached where ΔG equals zero. The effect of the external conditions on the ΔG of complex formation can qualitatively be derived by comparing the graphs in Figure 2.2a and b. The blue curve in Figure 2.2a indicates, for example, that complex formation is 73% complete at equimolar concentrations of the binding partners, whereas the extent of complex formation goes down to 38% if the initial concentrations of receptor and substrate are reduced from 10−3 to 10−4 M (Figure 2.2b). The thermodynamic driving force of complex formation, that is, the exergonicity of the reaction, is thus much lower in the

2.1 Thermodynamic and kinetic aspects

13

dilute solution although the Ka of the complex is unchanged. There even exist ratios of receptor, substrate, and complex, were the reverse reaction, namely, complex dissociation becomes favorable. We therefore have to carefully distinguish between the ΔG0 that refers to the standard state and allows a comparison of different systems and the ΔG that refers to the thermodynamics of the system under the experimental conditions, which not only has a different absolute value than ΔG0 but can even have a different sign. In this context, it should be emphasized that the term system refers to more than just the receptor and the substrate. The thermodynamics of complex formation are strongly affected by all molecules present in the mixture, most importantly the solvent molecules whose concentration substantially exceeds that of the binding partners. These solvent molecules characteristically mediate the strength of receptor–substrate interactions and therefore have a profound effect on the thermodynamics of the reaction. If, for example, the Gibbs free energy required to desolvate the binding partners cannot be compensated by the Gibbs free energy gained during the complex formation, the corresponding complex will not form in the respective solvent although it might be very stable in an environment in which solvation is not so strong. It is therefore useful to separate the overall Gibbs free energy of binding ΔG0 into the intrinsic free energy of complexation in the gas phase ΔG0intr and the free energies of solvation of the receptor ΔG0solv ðRÞ, the substrate ΔG0solv ðSÞ, and the complex ΔG0solv ðCÞ. The respective treatment yields equation (2.8), which shows that the binding process only becomes exergonic if the intrinsic binding strength overcompensates the difference between the free energies of solvation of the complex and of its components (ΔG0solv ðCÞ − ΔG0solv ðRÞ − ΔG0solv ðSÞ). ΔG0 = ΔG0intr + ΔG0solv ðCÞ − ΔG0solv ðRÞ − ΔG0solv ðSÞ

(2:8)

We will come back to the influence of the solvent on the thermodynamics of complex formation in Section 3.3. Correlating binding strength with the Gibbs free energy ΔG0 allows further breaking down the energetics of complex formation into binding enthalpy and entropy. The corresponding underlying formalism is based on the following Gibbs–Helmholtz equation. ΔG0 = ΔH 0 − TΔS0

(2:9)

The change in enthalpy is defined as the heat change of the system during complex formation at a constant pressure. Again, ΔH 0 refers to the standard state in which receptor and substrate are present at 1 M concentrations. Heat changes during a molecular recognition process are to a first approximation related to the actual interactions between the molecules involved in the complexation process, with an attractive receptor–substrate interaction producing a favorable exothermic ΔH 0 contribution to ΔG0 . The direct receptor–substrate interactions are, however, not

14

2 Analyzing complex formation

the only factors influencing ΔH 0 . Further contributions come from solvent effects that could add adverse effects to ΔH 0 if the free binding partners are more strongly solvated than their complex. Moreover, unfavorable enthalpic contributions also result from strained receptor or substrate conformations in the complex. The overall binding enthalpy associated with complex formation therefore depends on a balance of a variety of factors and can end up to be exothermic (ΔH 0 < 0) or endothermic (ΔH 0 > 0). In the latter case, complex formation must be associated with a sufficiently large positive entropy to become overall exergonic. Entropy refers to the order of a system, with a positive ΔS0 denoting the increase of disorder, which promotes complex formation. There are a number of factors that contribute to the entropy. A fundamental one is that any binding process involving two or more molecules coming together to form a complex is necessarily entropically unfavorable because the individual components lose degrees of freedom, the most important ones being translational and rotational mobility. The question therefore arises whether it is at all possible for a complexation process to be entropically favored. The answer once again lies in the solvent contributions, namely, the reorganization of solvent molecules, which are released from the solvation shells of receptor and substrate when they interact. As a consequence, complexation processes, which intrinsically lead to ordered assemblies, increase the disorder of the overall systems by allowing solvent molecules to gain freedom. Global disorder can therefore cause local order, which is a very important concept in supramolecular chemistry in general and selfassembly in particular as we will see in Chapter 5. More detailed aspects of solvent effects are treated in Section 3.3. Note that entropy is temperature dependent, which has consequences when complex formation is investigated at different temperatures. Another important aspect is that the entropy of a solution increases upon dilution. This effect explains why complexation equilibria shift toward the dissociated species if the concentration is reduced as shown in Figure 2.2. It should also be noted that entropy more strongly influences the formation of supramolecular complexes than the formation of covalent bonds. The reason is that the interactions that stabilize supramolecular complexes are much weaker than covalent bonds and the corresponding ΔH 0 therefore smaller. The substantial contributions of both enthalpy and entropy to the formation of a complex is the cause for a peculiar phenomenon that can have very frustrating consequences when trying to optimize the performance of a receptor by increasing the binding strength, that is, by making the binding enthalpy more negative. The characterization of the newly designed and sometimes laboriously synthesized receptor not seldom reveals that it is actually not much better than the previous receptor because the improvement of ΔH 0 at the same time causes the binding entropy to become less favorable. The marginal change of ΔG0 is therefore caused

2.1 Thermodynamic and kinetic aspects

15

by the opposing directions into which ΔH 0 and ΔS0 develop, an effect called enthalpy–entropy compensation. This effect is qualitatively explained as follows: if binding becomes stronger, the complex also becomes “tighter.” As a consequence, conformational degrees of freedom of the receptor and the substrate are lost, which affects ΔS0 unfavorably. Conversely, weak binding is entropically beneficial since the binding partners retain flexibility in the complex. Whether enthalpy–entropy compensation is a real thermodynamic phenomenon is a controversial topic. Compensating effects are indeed often observed for supramolecular systems, but it is unclear, given the many factors that contribute to ΔH 0 and ΔS0 , why a gain in one parameter should almost exactly be canceled out by a loss in the other. Is there a relationship between complex stability and the rate of complex formation?

As we have seen, thermodynamics predicts that a negative ΔG0 causes complex formation to proceed spontaneously. This does not necessarily imply that complexation is also fast, although many binding equilibria in supramolecular chemistry are indeed associated with small activation barriers and are hence often (almost) diffusion controlled. There are, however, exceptions of which one is illustrated in Figure 2.4.

(a)

(b)

Free binding partners

Constrictive binding

G0

G0

G‡ G‡

G0

Complex Reaction coordinate

Free binding partners

G0 Complex Reaction coordinate

Figure 2.4: Energy profiles associated with complexation equilibria of which one has a small activation barrier (small ΔG‡) and leads to a stable complex (large ΔG0 ) (a) and the other has a large activation barrier (large ΔG‡) but leads to thermodynamically not very stable complex (small ΔG0 ) (b).

Figure 2.4a shows the energy profile of a complexation reaction where binding is associated with a substantial exergonic stabilization of the complex and a small activation barrier ΔG‡ that is easily overcome. In such a case, the equilibrium lies far

16

2 Analyzing complex formation

on the side of the complex. However, the system is dynamic, meaning that the complexes constantly form and dissociate. The situation shown in the energy profile in Figure 2.4b is different. In this case, the complex is thermodynamically not significantly favored over the free binding partners. However, complex formation and dissociation have to overcome substantial activation barriers. Such a situation arises, for example, if the receptor and the substrate have to adopt strained conformations to allow substrate exchange as in the hemicarcerands described in Section 4.1.9. Such complexes are therefore inert although their thermodynamic stability is low. To describe the behavior of these systems in quantitative terms, Cram introduced the term constrictive binding, which relates the Gibbs free energy of the transition state to the Gibbs free energy of the binding partners (Figure 2.4b) [1]. In other words, constrictive binding is the free energy that must be invested to reach the transition state from the uncomplexed state, rendering it also a measure for the rate of complex formation. The reaction profile in Figure 2.4a exhibits a small activation barrier and we therefore expect complex formation to be fast. But what does fast exactly mean in this context? To obtain information in this respect, it is useful to correlate the stability constant Ka of the complex with the rate constants associated with its formation and dissociation. This treatment is based on the rate equations (2.10a,b) in which kon represents the rate constant of complex formation and koff that of dissociation. R + S

kon

Ð koff

dcC = kon cR cS dt

C

−

dcC = koff cC dt

(2:10a,b)

Once the reaction reaches the thermodynamic equilibrium, the rates of complex formation and dissociation are the same (steady state), allowing us to write the following expression. kon cR cS = koff cC

(2:11)

The rearrangement of (2.11) yields equation (2.12), which shows that the stability constant Ka is given by a ratio of rate constants. kon cC = = Ka koff cR cS

(2:12)

Stable complexes thus form more rapidly than they dissociate. Note that this statement also applies to the steady state of an equilibrium, where the rates of complex formation (kon cR cS ) and dissociation (koff cC ) end up being the same because the rate constants are multiplied with concentrations. In the case of very stable complexes, for example, a large kon is multiplied with two very small concentrations, while a smaller koff is multiplied with a large cC .

2.2 Analytical strategies and techniques

17

The correlation in equation (2.12) allows us to estimate the lifetime of typical supramolecular complexes in solution. Assuming that complex formation is diffusion controlled (kon ~ 109 M−1 s−1) and that the complex has a Ka of 106 M−1, which represents a rather stable complex according to Figure 2.2a, we end up with a koff of 103 s−1. Thus, the complex has a lifetime in the order of milliseconds, which is rather short on the human timescale. The important lesson is that even highly stable complexes rapidly form and dissociate in solution and that the static picture suggested by reaction schemes such as the one in Figure 2.1 is misleading.

2.2 Analytical strategies and techniques 2.2.1 Strategies The only value required to derive the stability constant from the law of mass action in equation (2.1) is the complex concentration cC (in the case of higher complexes, the concentrations of all complexes present) because the concentrations of the receptor cR and the substrate cS are accessible from the mass balances (equations (2.2)). Unfortunately, most analytical techniques used for binding studies do not yield absolute concentrations but rather provide information to what extent the equilibrium shifts relative to the initial state if the concentrations of one or both binding partners are changed. Therefore, binding studies generally involve titrations during which the concentrations of the binding partners are varied while measuring a physical property that correlates with cC . These titrations afford binding isotherms, such as those shown in Figures 2.2 and 2.3, and fitting these isotherms to the mathematical model underlying complex formation then yields Ka . How can complex stoichiometry be determined?

The crucial aspect in this context is choosing the correct model because meaningful stability constants are only obtained if all aspects of complex formation, particularly complex stoichiometry, have been correctly taken into account in the mathematical treatment. It is therefore helpful to know the composition of the complex prior to recording binding isotherms, although information in this respect can sometimes be derived from the isotherms themselves. The sharp bend of the isotherms associated with the most stable complex in Figure 2.2a clearly indicates that this complex has a 1:1 stoichiometry, for example. Conversely, the black isotherm in Figure 2.2b reaches saturation approximately at a 1:2 receptor–substrate ratio, suggesting that two substrate molecules are bound to the receptor. Once the complexes become less stable, these correlations are, however, not reliable. Moreover, sigmoidal shapes of binding isotherms are important indications that higher complexes

18

2 Analyzing complex formation

are present, but they provide little information about the actual stoichiometry. Therefore, independent methods are usually needed to determine the composition of the complex. A classical approach relies on Job’s method of continuous variations, which Paul Job introduced in 1928 to determine the stability and stoichiometry of coordination complexes [2]. The underlying strategy involves the preparation of a series of solutions containing the receptor and the substrate at the same overall concentration but in different ratios, followed by measuring the amount of complex (or a property that correlates with complex concentration) in each solution. The idea underlying this method is illustrated for a 1:1 and a 1:2 complex in Figure 2.5. 0.50 c C (RS) c S0

(a)

c 0R 0

0.1

0.2

0.3

0.4 0.5 0.6 Mole fraction X

0.7

0.8

0.9

1.0

0.67 c C (RS2) c S0

(b)

c 0R 0

0.1

0.2

0.3

0.4 0.5 0.6 Mole fraction X

0.7

0.8

0.9

1.0

Figure 2.5: Schematic representation of the changes in complex concentration in solutions containing a receptor and a substrate at a constant overall concentration but in different ratios for a 1:1 complex (a) and a 1:2 receptor–substrate complex (b). The mole fraction X is defined as cS =ðcS + cR Þ.

Figure 2.5a shows that the concentration of a 1:1 complex should be largest in the solution containing the binding partners in equal amounts. The absolute amount of the complex at this mole fraction depends on its stability, but any reduction of either the substrate or the receptor limits the maximum amount of complex that can be formed, causing complex concentration to decrease. The situation is comparable for complexes of higher stoichiometry. If complex formation involves a 1:2 receptor–substrate complex, for example, as in Figure 2.5b, the maximum amount of the complex is formed at a mole fraction of 0.67. In general, the mole fraction X at which the amount of complex is maximum can be calculated by using equations (2.13a) or (2.13b), depending on how X is defined.

2.2 Analytical strategies and techniques

19

X=

number of substrate molecules in the complex if total number of binding partners in the complex

X=

cS cS + cR

(2:13a)

X=

number of receptor molecules in the complex if total number of binding partners in the complex

X=

cR cS + cR

(2:13b)

Thus, a plot of the mole fraction vs. the amount of complex should yield a bellshaped curve with the position of the extremum providing information about complex composition. The actual equations describing these curves are based on the equations we derived previously [e.g., equation (2.4) for a 1:1 complex]. Examples of such Job plots are depicted in Figure 2.6. 1.0

1.0

(a)

0.8 c C / c C (max)

c C / c C (max)

0.8

(b)

0.6 0.4 0.2

0.6 0.4 0.2

0.0

0.0 0.0

0.2

0.4

0.6 X

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

X

Figure 2.6: Job plots for 1:1 binding equilibria. The overall concentration cS0 + cR0 amounts to 10−2 M in graph (a) with the curves representing complexes with stability constants Ka of 104 (black), 103 (red), and 102 M−1 (blue). The Ka is 104 M−1 in graph (b) and the total concentration cS0 + cR0 is varied between 10−2 (black), 10−3 (red), and 10−4 M (blue). The concentration cC ðmaxÞ equals ðcR0 + cS0 Þ=2.

Figure 2.6 shows that, like the binding isotherms in Figure 2.2, the Job plots become shallower as complex stability or the total initial concentrations of the binding partners decrease, reflecting the simultaneous decrease of the extent of complex formation. Because a reliable determination of the position of the extremum is not easy for shallow curves, recording meaningful Job plots requires carefully chosen experimental conditions. In addition, there are other factors that sometimes render Job plots problematic. One is illustrated in Figure 2.7, which shows examples of Job plots for 1:2 receptor–substrate binding equilibria, calculated by using the equations derived in Appendix 12.1. The black curves in Figure 2.7 show Job plots for a complex whose stepwise formation is associated with a Ka12 that is two orders of magnitude larger than Ka11 . Thus, substantial amounts of the 1:2 complex are present at each mole fraction so that the maximum of the Job plot ends up at the expected value of 0.67, independent of

20

2 Analyzing complex formation

1.0

1.0

(a)

0.8 cC (tot) / cC (max)

cC (1:2) / cC (max)

0.8

(b)

0.6 0.4 0.2

0.6 0.4 0.2

0.0

0.0 0.0

0.2

0.4

0.6 X

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

X

Figure 2.7: Job plots for 1:2 receptor–substrate binding equilibria. The overall concentration cS0 + cR0 amounts to 10−2 M in both graphs. The black curves denote a complex with Ka11 = 100 M−1 and Ka12 = 10,000 M−1 and the red curves a complex with Ka11 = 10,000 M−1 and Ka12 = 100 M−1. Graph (a) shows the concentrations of only the 1:2 complexes cC ð1:2Þ and graph (b) the total concentrations cC ðtotÞ = 2=3 cC ð1:1Þ + cC ð1:2Þ. The concentration cC ðmaxÞ equals ðcR0 + cS0 Þ=3.

whether the formation of the 1:2 complex is followed individually or in combination with the 1:1 complex. If, however, Ka12 is significantly smaller than Ka11 , as in the red curves, the 1:1 complex dominates in solution. As a consequence, the maximum of the Job plot is either located at a too high or too low mole fraction, depending on whether the plot shows only the concentration of the 1:2 complex as in graph (a) or the combined concentrations of the 1:1 and 1:2 complex as in graph (b). Neither Job plot thus reflects the true receptor stoichiometry, showing that situations exist where Job plots are unsuitable for assessing complex composition. This problem becomes more pronounced when complex formation goes beyond 1:2 equilibria. Moreover, even if the maximum of the respective Job plot is located at the correct mole fraction, it is impossible on the basis of a Job plot to differentiate complexes that have the same ratio of binding partners, for example, 1:1 and 2:2 complexes, and differentiating between maxima at 0.67 (1:2 complex) or 0.75 (1:3 complex) is experimentally challenging. These aspects have led some authors to proclaim “the death of the Job plot” [3]. Indeed, although there are situations where Job plots are useful, it is advisable to also consider additional and, ideally, independent means to determine the binding model. Such methods can include a careful statistical analysis of how well the fitted binding isotherm describes the experimental data to detect deviations from the chosen model, the results of crystallographic analyses, or mass spectrometric investigations. The latter, in particular, provide detailed information about the composition of the species present in a solution if the method of ionization is mild enough to transfer the complexes into the gas phase without decomposition.

2.2 Analytical strategies and techniques

21

How can complex stability be determined?

Once the composition of a complex is established, its stability is derived by fitting binding isotherms, calculated on the basis of the correct composition, to the experimental data [4]. Such nonlinear regressions involve varying Ka and any additional parameter associated with the analytical technique used for recording the binding isotherms until convergence, that is, until the calculated isotherm fits the experimental one best. At times when it was difficult to perform such nonlinear regressions computationally, linearized forms of the equations underlying the binding isotherms were popular because they allowed the determination of stability constants by linear regressions. These methods rely on approximations, however, and therefore only yield meaningful results if the approximations are valid. In the case of the Benesi–Hildebrand method, for example, one binding partner has to be present in excess so that its concentration can be assumed to be constant during the titration. Today, these methods are outdated and should not be used because a range of programs are available with which nonlinear regressions can easily be performed. There are even online tools that expediently allow the calculation of stability constants from binding isotherms without having to make approximations or assumptions [5]. A few guidelines still need to be followed to ensure that the results of these regression analyses are accurate. The quality of the data depends on the shape of the binding isotherm. For example, if the isotherm is too steep, like the black curve in Figure 2.2, the curved region that mainly encodes for complex stability is too small. Such curves sometimes allow deriving a lower limit of complex stability but not an exact value. Conversely, binding isotherms that end before a substantial amount of the complex is present, such as the orange curves in Figure 2.2, contain too little information to allow calculating Ka . As a rule of thumb, nonlinear regressions give the most reliable results if the isotherms cover complex formation between 20% and 80%. An aspect complicating the evaluation of higher equilibria is that the corresponding isotherms need to be fitted to a large number of independent parameters. Nonlinear regressions therefore sometimes yield different solutions depending on the parameters chosen as starting points. To avoid this problem, it is helpful to have estimates for one or more of these unknowns from independent measurements. Moreover, it is sometimes useful to reduce the overall concentrations of the interacting species in binding studies to suppress the formation of higher complexes, thus simplifying the binding model. Because stability constants depend so sensitively on external parameters, only constants determined under analogous conditions are comparable. That the respective measurements should have been performed in the same solvent or at the same temperature is obvious, but what about stability constants determined under

22

2 Analyzing complex formation

comparable conditions albeit with different methods? In principle, the value of a stability constant should be independent of how it was determined. However, different analytical techniques usually require different concentrations. NMR titrations are, for example, performed at millimolar concentrations while the concentrations for UV–vis titrations are one to three orders of magnitude lower. As a consequence, the binding isotherms resulting from different types of measurements vary, covering different regions of the complexation equilibrium. Moreover, different analytical techniques probe different properties of the complex. Thus, changing the analytical technique can also cause changes in the binding constants but the differences should not be large and they need to be discussed and interpreted on the basis of the conditions of the measurements. Finally, it is impossible to compare binding constants of complexes differing in composition because they have different units and are therefore incommensurable. Unfortunately, structural changes in a receptor sometimes cause changes in the binding model, which raises the questions about if and how the properties of such receptors can be compared. In such cases, it can be useful to compare the corresponding stepwise binding constants, for example those relating to the formation of the 1:1 complex. This is, however, neither a general solution nor is it necessarily correct because a crucial property of the receptor related to the formation of the higher complexes is neglected. To address this problem, the BC50 parameter has been introduced [6]. This so-called median binding concentration is defined as the total concentration of the receptor necessary for binding 50% of the substrate. BC50 is therefore independent of the binding model but specifies a receptor concentration at which a certain amount of the substrate is consumed. The lower this concentration, the higher is the affinity of the receptor. BC50 depends on the fraction of the receptor that is bound in the complex and if this fraction becomes negligible, BC50 becomes constant. This concentration is termed the intrinsic median binding concentration, BC050 . Thus, BC050 represents the maximum binding ability of the receptor toward the substrate. Since BC050 equals the dissociation constant Kd for 1:1 complexes, it can be viewed as a “global” affinity constant. The following example should illustrate the usefulness of the BC050 parameter. The two receptors shown in Figure 2.8 bind chloride anions in CDCl3/CD3CN, 80:20 (v/v). Under these conditions, other processes also occur, including the dimerization of the receptors, the ion-pairing of the chloride anion with its counterion, or the binding of the receptors to the whole ion-pair. Although all individual equilibrium constants are known, their comparison does not clearly indicate which of the receptors is the more efficient. A clear distinction is possible, however, on the basis of the corresponding BC050 parameters. These parameters amount to 4.6 × 10−3 M for 2.1 and 3.7 × 10−4 M for 2.2, demonstrating that the pyrrole-derived receptor 2.2 has a more than one order of magnitude higher chloride affinity than the urea-derived receptor 2.1.

2.2 Analytical strategies and techniques

RO O

HN

HN HN NH O

O H N

O

OR

R R NH

23

RO

NH HN H N R

NH O

O HN H N

O 2.1 (R = CMe2CH2CMe3)

O

O 2.2 (R = Benzyl)

Figure 2.8: Tripodal chloride receptors 2.1 and 2.2 whose chloride affinities derive from the cyclically arranged NH groups that form hydrogen bonds with the anion.

How can binding equilibria be analyzed?

2.2.2 NMR spectroscopy Complex formation usually produces characteristic changes in the NMR spectra of the receptor and/or the substrate, rendering NMR spectroscopy a useful tool to investigate molecular recognition processes. 1H NMR spectroscopy, in particular, is one of the most frequently used methods because it has several advantages over other techniques. One is that the extent to which a signal shifts or changes its integral upon complex formation correlates directly with the complex concentration without having to consider parameters that establish a relationship between the measured physical property and the concentration like the molar absorptivity coefficients in UV–vis spectroscopy (Section 2.2.3). 1H NMR spectroscopy is also rather versatile as it only requires the presence of protons in the investigated binding partners but no specific structural elements such as chromophores. Finally, since the spectroscopic changes produced by complex formation reflect the environment of receptor and substrate protons in the complex, NMR spectroscopy also provides structural information about the complex, unlike the other techniques discussed. An important aspect that has to be considered in the context of NMR spectroscopy is the timescale of the method in relation to the lifetime of the complex. In this respect, two cases must be distinguished: equilibria that are slow on the NMR timescale and those that are fast. What does this mean and what are the consequences? The NMR timescale defines whether protons, whose environment changes in a dynamic process during the measurement, appear as two individual signals in the spectrum or collapse into an averaged signal. Examples of dynamic processes are

24

2 Analyzing complex formation

conformational changes of a molecule, the exchange of acidic protons, or complexation equilibria. Whether signals of dynamic systems are resolved in the NMR spectrum is determined by Heisenberg’s uncertainty principle and depends on the excitation energy and the lifetime τ of the states that are in equilibrium. This lifetime, or the corresponding rate constant of interconversion, is estimated by using equation (2.14), which strictly applies only to dynamic processes that follow firstorder rate laws and involve exchanging nuclei that produce signals of equal intensity in the NMR spectrum. π 1 kc = pffiffiffi Δν = 2.22 Δν = τc 2

(2:14)

In this equation, Δν refers to the distance in Hertz of the signals belonging to the two exchanging protons when the exchange is slow. The parameter τc specifies the lifetime that just causes the signals to become indistinguishable, that is, to coalesce. If the actual lifetime of the system is larger than τc (or the rate constant of interchange is smaller than kc ), two signals are observed and if it is smaller, only an averaged signal results. Let us consider a specific example. If we assume that complex formation causes a signal of the receptor to shift by 1 ppm in the 1H NMR spectrum and that we work on a 300 MHz NMR spectrometer, Δν amounts to 300 Hz. Therefore, coalescence occurs if the exchange process has a lifetime of 1.5 ms (kc = 666 s−1) at the temperature at which the NMR spectrum is recorded. Processes that have longer lifetimes (slower processes) at the same temperature lead to two signals and those with shorter ones (faster processes) to only an averaged signal. We have seen previously that the slow step in complexation equilibria is complex dissociation (which is also the one following a first-order rate law) and that a complex with a Ka of 106 M−1 whose formation is diffusion controlled has a koff of 103 s−1 (Section 2.1). We would therefore expect averaged signals for the complexed and uncomplexed binding partners in this case. More stable complexes should produce separate signals and the appearance of separate signals for free and complexed binding partners in the NMR spectrum indeed often indicates that a very stable complex is formed. Note, however, that the rate of complex formation also depends on activation barriers. Consequently, equilibria are also slow if complex dissociation is associated with a large activation energy even if the thermodynamic stability is not very high (Section 2.1). It should be emphasized that the above estimation has been made under the assumption that Δν amounts to 300 Hz. If the extents to which signals shift in the spectrum upon complex formation are smaller or larger, other lifetimes result. It is even possible that complex formation causes some signals of the free and complexed binding partners to appear separately in the NMR spectrum and others to appear averaged if the extents to which complexation affects the chemical environment of different nuclei vary strongly. Another important aspect is that Δν depends

2.2 Analytical strategies and techniques

25

on the frequency of the NMR spectrometer. A 1 ppm difference of two signals in the spectrum correlates with 300 Hz on a 300 MHz machine but with 800 Hz on an 800 MHz spectrometer. Thus, NMR timescales depend on the spectrometer used for the measurement and processes that are fast on one spectrometer can become slow when using a more powerful machine. Deriving stability constants from NMR spectra of equilibria that are in slow exchange is relatively straightforward. A prototypic example of a series of NMR spectra is depicted in Figure 2.9. The signal in the bottom spectrum should correspond to a proton in the free receptor. This signal progressively disappears as the substrate concentration increases, while a new signal appears in the spectrum, which is produced by the same proton now residing in the complex. At a certain substrate concentration, no signal of the free receptor is observed anymore.

Complex signal

Increasing substrate concentration

Receptor signal

2.3

2.2

2.1

2.0

1.9 (ppm)

Figure 2.9: Exemplary 1H NMR spectra illustrating the effect of complex formation on a receptor signal if the free and complexed species are in slow exchange with respect to the NMR timescale. The amount of added substrate increases from the bottom to the top spectrum.

Since the initial concentrations of receptor and substrate c0R and c0S are known, and since the complex concentration cC is linked to these concentrations by mass balances, it is possible to quantitatively estimate cC , cR , and cS by integrating the respective signals. In the above example, cC is calculated by using equation (2.15), for

26

2 Analyzing complex formation

instance, which further allows accessing cR and cS on the basis of equations (2.2a,b) if a 1:1 complex is formed. Ð complex signal Ð (2:15) cC = Ð c0 complex signal + receptor signal R Once these concentrations are known, Ka is calculated by using the respective law of mass action [equation (2.1)]. Although a single spectrum is sufficient, in principle, to determine Ka in this case, it is better to evaluate a series of spectra differing in the substrate/receptor ratio to minimize the error. This method is applicable only if the complexes are not too stable. In a situation where all of the added substrate is consumed in the complex, cS and/or cR are almost zero, which prevents calculating a binding constant. If complex formation is fast on the NMR timescale, the estimation of Ka involves recording spectra of a series of solutions containing the receptor and the substrate in various ratios. The concentration of the component whose signal shift is followed is usually kept constant during such a titration and the concentration of the other component is increased stepwise. Under these conditions, the equilibrium progressively shifts toward the complex, which is reflected in the continuous shift of a diagnostic signal in the NMR spectrum. Figure 2.10 shows an example of such a titration during which increasing amounts of the substrate are added to a receptor solution with the concentration c0R .

1.0

/ ppm

0.8

0

0.6 0.4 0.2 0.0

4.0 3.9 3.8

3.7

3.6

3.5

3.4

3.3 3.2 δ (ppm)

0

1

2

3

4

5 c S0 / c R 0

Figure 2.10: Exemplary 1H NMR spectra illustrating the effect of complex formation on a receptor signal if the free and complexed species are in fast exchange with respect to the NMR timescale. The amount of substrate increases from the bottom to the top spectrum. The right graph shows a plot of the substrate/receptor ratio vs. the relative shift changes Δδ = jδ − δ0 j. The chemical shift δ0 corresponds to the signal of the free receptor and δ to the chemical shifts of the signals in the mixtures. The line in the right graph shows the binding isotherm, resulting from fitting the individual data points to a 1:1 binding equilibrium.

2.2 Analytical strategies and techniques

27

In this case, information about cC is derived according to equation (2.16) from the extent to which the receptor signal at δ0 shifts toward the corresponding signal of the complex denoted as δmax . cC =

δ − δ0 0 c δmax − δ0 R

(2:16)

If a signal of the substrate is followed in the NMR titration, c0R has to be replaced in equation (2.16) by c0S . Combining equation (2.16) with equation (2.4) affords an expression that correlates the experimentally determined chemical shift δ with the stability constant Ka . The estimation of Ka involves fitting the observed binding isotherm to equation (2.17). The only other parameter that is typically unknown is δmax , which therefore also has to be fitted during the nonlinear regression. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 0
0 0 2 cR + cS + Ka− 1 δmax − δ0 @c0R + c0S + Ka− 1 (2:17) δ = δ0 + − − c0R c0S A 2 4 c0R We saw in Section 2.1 that the reliability of such a nonlinear regression depends on the shape of the binding isotherm. Since NMR spectroscopy is not a very sensitive technique, binding studies are usually conducted at millimolar concentrations of the binding partners. As a consequence, the upper limit of binding constants that can reliably be determined with NMR titrations amounts to 104 M−1, in certain cases 105 M−1. This estimation is consistent with the binding isotherms in Figure 2.2, which show that the binding isotherm of a complex with a Ka of 104 M−1 is almost too steep to allow the exact estimation of complex stability by nonlinear regression if c0R = 1 mM. A way to access higher binding constants by NMR spectroscopy is to perform competitive titrations, which typically involve adding increasing amounts of the substrate, whose affinity for the receptor should be determined, to a complex of known stability. These titrations initially afford the ratio of the two stability constants Ka ð1Þ=Ka ð2Þ from which the unknown stability constant is determined by considering the known one. Information about binding enthalpy and entropy is obtained by performing NMR titrations at different temperatures and using van’t Hoff plots to determine ΔH 0 and ΔS0 . The underlying equation (2.18) results from a combination of equations (2.7) and (2.9). ln Ka = −

ΔH 0 − 1 ΔS0 T + R R

(2:18)

Thus, plotting T − 1 vs. ln Ka should yield a straight line with a slope of − ΔH 0 =R and an intercept of ΔS0 =R. This method is, however, associated with a number of drawbacks. One is that ΔH 0 and ΔS0 are not necessarily independent of temperature. Another is that, for practical reasons, temperature-dependent measurements can often be performed only over a limited range of temperatures. As we will see later, isothermal titration calorimetry (ITC) allows the direct determination of the enthalpy and entropy of a

28

2 Analyzing complex formation

binding process in a single titration (Section 2.2.6), which renders this method much better suited to access the individual thermodynamic parameters of complex formation.

2.2.3 UV–vis spectroscopy UV–vis spectroscopy can be used for binding studies if the receptor and/or the substrate contain suitable chromophores whose optical properties are affected by complex formation. This is usually the case if the environments of the chromophores differ in the complexed and the uncomplexed states. A typical effect is the increase or decrease of the intensity of an absorption band, which is followed while changing the receptor–substrate ratio. Because the electronic excitations that give rise to these absorption bands are fast, UV–vis spectroscopy in principle allows distinguishing the free and the complexed binding partners. In reality, however, the corresponding bands overlap and mainly differ in their intensity. In addition, the molar absorptivity coefficient of the complex is generally unknown. Therefore, UV–vis spectroscopic binding studies also have to rely on titrations and the fitting of the unknown parameters to the obtained isotherms. A typical strategy involves setting up a series of samples that contain a fixed concentration of one component, for example the receptor, and varying concentrations of the respective binding partner, followed by measuring the absorption intensity of the solutions Aobs at a given wavelength. For the mathematical treatment of the resulting binding isotherm, Aobs has to be correlated with the complex concentration cC . Provided that the Beer–Lambert law is valid under the conditions of the measurement, Aobs is expressed as the sum of the absorptions of all species in solution, which in this case are the receptor, the substrate, and the complex (equation (2.19)). Aobs = AR + AS + AC

(2:19)

The ensuing mathematical treatment is facilitated by choosing the conditions such that only two of the interacting species absorb at the wavelength at which Aobs is recorded. Assuming that the substrate is UV–vis silent under the conditions of the measurement, equation (2.19) leads to the simpler expression (2.20) in which the individual absorptions are correlated with the respective concentrations on the basis of the Beer–Lambert law. This equation is valid if the measurements are performed in the usual 1 cm cells. Aobs = AR + AC = εR cR + εC cC

(2:20)

By using the mass balance in equation (2.2a), equation (2.20) can be rearranged to afford (2.21). Combining equations (2.4) and (2.21) then yields the expression (2.22) that correlates the measured absorption Aobs with the Ka of a 1:1 complex. The term

2.2 Analytical strategies and techniques

29

εR c0R is the absorption of the free receptor in the absence of the substrate, whereas Δε = εC − εR is usually unknown. Therefore, Ka and Δε are fitted during the nonlinear regression to ultimately afford the stability constant of the complex.
 (2:21) Aobs = εR c0R − cC + εC cC = εR c0R + ðεC − εR ÞcC sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 0
0 0 2 cR + cS + Ka− 1 c0R + c0S + Ka− 1 0 @ (2:22) Aobs = εR cR + ðεC − εR Þ − − c0R c0S A 2 4 The major difference between NMR and UV–vis titrations is that the latter require significantly lower concentrations. As a consequence, binding constants between 106 and 107 M−1 can easily be determined, depending on the molar absorptivity coefficient of the participating binding partners. Another advantage of UV–vis titrations is that, in comparison to NMR titrations, lower amounts of receptor and substrate are necessary. A disadvantage is that the changes in the UV bands usually provide little structural information about the complexes formed. Another strategy used to obtain binding constants from UV–vis spectrometric measurements is the picrate extraction method developed by Cram to determine the stability of crown ether complexes. This method does not involve recording binding isotherms but measuring the extent to which a metal salt partitions between an aqueous solution and an organic solution containing a receptor. The concentrations of the metal ions in both phases are estimated indirectly by UV–vis spectroscopy from the intensities of the absorption band of the picrate counterion. The mathematical treatment of the measurements is based on the following equations. ðM · PicÞorg + Rorg Ð ðM · R · PicÞorg Maq + Picaq Ð ðM · PicÞorg

Ka =

Kextraction Kdistribution

(2:23a)

cðM · PicÞ cM cPic

(2:23b)

cðM · R · PicÞ cM cPic cR

(2:23c)

Kdistribution =

Maq + Picaq + Rorg Ð ðM + · R · PicÞorg Kextraction = Ka =

cðM · R · PicÞ cðM · PicÞ cR

(2:24)

Equation (2.23a) specifies the binding constant of the complex between the receptor and a metal picrate, both in the organic phase. The distribution coefficient of the metal salt between the aqueous and the organic phase is given by equation (2.23b) and the corresponding extraction coefficient in equation (2.23c). Combining all three equations yields expression (2.24), which shows that complex stability is calculated from the ratio Kextraction / Kdistribution . These measurements thus involve two extractions, one in which the metal salt is extracted from the aqueous to the organic phase in the absence of the receptor,

30

2 Analyzing complex formation

and an analogous one with the receptor present. The respective concentrations of the metal salt in the organic layers are quantified by measuring the absorption of the organic solutions and calculating the picrate concentrations with the help of the known molar absorptivity coefficient. The concentration in the aqueous phase is calculated by considering the mass balance and the initial metal salt concentration. The receptor is considered to be insoluble in the aqueous phase while the salt is considered to be ion-paired in the organic phase.

2.2.4 Fluorescence spectroscopy Because of the sensitivity of fluorescence spectroscopy and the corresponding low concentrations required, fluorescence titrations allow the determination of very high binding constants between 106 and 1010 M−1. This technique requires the presence of a fluorescent chromophore in at least one of the binding partners. In the absence of quenching, the mathematical treatment to derive stability constants from the binding isotherms of fluorescence titrations is very similar to that used for UV–vis titrations. If there is a linear correlation between the observed fluorescence Fobs and the concentrations of the emitting compounds and if one binding partner is nonfluorescent, we start from an expression that is closely related to (2.20). Fobs = FR + FC = kR cR + kC cC

(2:25)

Equation (2.25) can then be transformed in a similar manner as before to yield equation (2.26), which, together with the expression for cC in equation (2.4), forms the basis of the nonlinear regression analysis of a 1:1 binding isotherm. Fobs = kR c0R + ðkC − kR ÞcC

(2:26)

Note that if only the complex is fluorescent, that is, if fluorescence is turned on upon complex formation, expression (2.26) simplifies even further to give Fobs = kC cC . Modified versions of these equations must be used if Fobs is affected by quenching [4].

2.2.5 Potentiometry Potentiometric titrations provide information about the strength of receptor–substrate interactions in systems that engage in protonation equilibria. A typical example of such a receptor is the polyazamacrocycle 2.3, containing seven amino groups along the ring, whose protonation is characterized by the equilibria specified in Figure 2.11. The corresponding cumulative protonation constants log β are determined electrochemically by measuring the change of pH when titrating a solution of the fully protonated form of 2.3 with NaOH followed by fitting of the resulting titration curve.

2.2 Analytical strategies and techniques

31

Substrates that interact with the receptor influence these protonation constants. If, for example, the substrate prefers to interact with the cationic form of the receptor, complex formation stabilizes the protonated state. Conversely, if the free electron pair of a nitrogen atom forms a coordinate bond to a metal ion, protonation of this nitrogen atom becomes more difficult. Recording the protonation curves in the presence of the substrate, fitting them to the underlying equilibria, and comparing the obtained log β values with those of the free receptor thus provides information about the efficiency of substrate binding. If the substrate is also involved in protonation equilibria, the corresponding protonation constants need to be determined independently.

Equilibria: NH HN HN

NH NH

H N

HN

2.3 (= L) COO OOC

COO

Benzene-1,2,3-tricarboxylate (= A3–)

log βnH

L

log βnA

L

log Kn

HnLn+

+

HnLn+

n H+

+

n H+

+ +

A3–

HnLA(n–3)+

A3–

HnLA(n–3)+

n

log βnH

log βnA

log Kn

3

27.7

30.9

3. 2

4

34.1

39.2

5.1

5

37.8

46.1

8.3

6

40.0

51.0

11.0

7

41.9

54.6

12.7

Figure 2.11: Structures of the polyazamacrocycle 2.3 (L) and benzene-1,3,5-tricarboxylate (A3−) and definition of the cumulative protonation constants log βHn and log βAn and the stepwise binding constant log Kn on the basis of the corresponding equilibria. The data in the table summarizes selected cumulative and stepwise equilibrium constants of complexes of 2.3 and benzene-1,3,5tricarboxylate with different degrees of protonation. The log Kn are given by the difference log βAn − log βHn .

As an example, Figure 2.11 summarizes the cumulative protonation constants obtained when titrating 2.3 alone and together with benzene-1,3,5-tricarboxylate [7]. The table shows that each log βAn value is larger than the corresponding log βHn value of the free receptor, showing that the interaction of the tricarboxylate with the macrocycle stabilizes the respective protonated species. The stepwise binding constants log Kn , that is, the differences log βAn − log βHn , specify the extent of this stabilization. These constants are therefore equivalent to the association constants obtained with other methods. In the above example, the log Kn values show that the complex becomes more stable with increasing degree of protonation, which is consistent with

32

2 Analyzing complex formation

the strengthening of the electrostatic interactions that underlie substrate binding as the charge of the receptor increases (Section 3.1.2). How the distribution of the species formed in such complexation equilibria varies with pH is often depicted in a graphical fashion in speciation diagrams such as the one shown in Figure 2.12.

100 80 % relative to A3–

A3–

LAH52+ LAH63+ LAH4+

60 LAH74+

40

LAH3 20 0 3

4

5

6

7

8

9 pH

Figure 2.12: Speciation diagram of complexes of 2.3 (L) and benzene-1,3,5-tricarboxylate (A3−), differing in the degree of protonation as a function of pH in 0.15 M aqueous NaClO4 at 298 K. The concentration of the free anion is shown in red.

The stability constants determined by potentiometry depend on a number of factors, including the temperature, the concentrations of the binding partners, the concentration and type of background electrolyte, and the ionic strength of the medium. Therefore, results are only comparable if they have been measured under identical conditions.

2.2.6 Isothermal titration calorimetry Isothermal titration calorimetry (ITC) is a method that involves directly measuring the reaction enthalpy of complex formation. Calorimeters used for this purpose feature two thermostatted cells, one in which the actual titration is performed and a reference cell containing only the solvent. During the measurement, a solution of one binding partner is added in several injections from a syringe to the solution of the other complex component, usually the receptor, and the heat generated or consumed during complex formation is recorded. This is done either by measuring the resulting temperature difference between the two cells or by compensating the temperature difference with a corresponding energy input. Thus, very small heat flows can accurately be determined. Calorimetry is widely used in life sciences for the investigation of biological systems operating in water. With the availability of calorimeters that allow working in

33

2.2 Analytical strategies and techniques

organic solvents, this technique has also become popular in supramolecular chemistry. Its major advantage with respect to other analytical methods is that calorimetric measurements afford the full thermodynamic profile of complex formation in a single titration and often even allow deriving the complex stoichiometry from the shapes of the binding curves. ITC is also very versatile as it does not require the presence of special structural features in the binding partners. Only in relatively rare cases when complex formation is entirely entropy driven, lacking a measurable heat change, it is not possible to use ITC for the determination of binding strength. Figure 2.13a shows the primary result of an ITC titration, comprising a series of pulses that reflect the temperature changes during each titration step. The direction of the peaks indicate that complex formation is exothermic in this case. An endothermic reaction would consequently produce peaks in the opposite direction. The integration of these peaks yields the heats Q associated with each addition that correlate with the amounts of complex formed. Because ITC titrations do not involve a series of individual measurements but the stepwise addition of small aliquots of one binding partner to a solution of the other one, the resulting binding isotherms differ from the isotherms in Figure 2.2 and resemble curves associated with acid–base titrations. The first additions usually produce large heat changes because the binding partner in the cell is initially present in excess so that the other binding partner gets fully complexed. The heat changes then become progressively smaller as the amount of free binding partner in the cell decreases. At the end of the titration, only the heat of dilution of the added compound is measured. The thermal effects associated with this dilution are normally eliminated prior to the fitting of the binding isotherms by performing another titration during which the solution in the syringe is added to the pure solvent. The measured heats are then substracted from those of the actual titration. The final isotherm of an ITC titration thus has a sigmoidal shape as shown in Figure 2.13b.

5

(a) Q per addition

0

dQ/dt

–5 –10 –15 –20 –25 0.0

0.5

1.0

1.5

2.0

2.5

3.0 3.5 cS0 / cR0

2 0 –2 –4 –6 –8 –10 –12 –14 –16

(b)

0.0

0.5

1.0

1.5

2.0

2.5

3.0 3.5 cS0 / cR0

Figure 2.13: Thermogram resulting from an ITC titration (a) and binding isotherm reflecting the normalized heats per peak as a function of the molar ratio of receptor and substrate (b).

34

2 Analyzing complex formation

The mathematical formalism to derive the binding constant from such an isotherm is based on the following considerations. The heat produced or generated during each addition is proportional to the overall enthalpy ΔH 0 of the process, the amount of complex formed during the corresponding addition, and the cell volume V0 according to the equation (2.27). Q = V0 ΔH 0 cC

(2:27)

This equation establishes the correlation between the measured parameter Q and the unknown complex concentration cC . Unlike the techniques discussed before, i 0, i however, the total concentrations c0, R and cS of receptor and substrate differ for every injection. These concentrations are given by equations (2.28a,b) in which v is the injection volume and the expression ð1 − v=V0 Þ specifies the reduction of the concentration after i injections to a volume V0 , which remains constant throughout the titration. These equations are valid when the cell contains the receptor and the syringe the substrate.   v i i 0 = c 1 − c0, R R V0 "  # v i 0, i 0 cS = cS 1 − 1 − V0

(2:28a)

(2:28b)

Since every injection is associated with the consumption or release of heat proportional to the change in the concentration of complex, the heat change between injection i − 1 and i is given by the following equation.       1 v V0 ΔH 0 i v i−1 = (2:29) cC − cC 1 − Q = 0 Qi − Qi − 1 1 − V0 V0 vc0S vcS Combining this equation with equation (2.4) affords an expression that allows fitting the binding isotherms to obtain the two unknowns ΔH 0 and Ka . Based on these parameters, ΔG0 is calculated from Ka [equation (2.7)] and ΔS0 from ΔG0 and ΔH 0 [equation (2.9)]. Another parameter that is usually also determined during the nonlinear regression is the inflection point of the isotherm, which provides information about complex stoichiometry. The ratio of the binding partners at which this inflection point is observed is termed the stoichiometry number n. In the case of a 1:1 complex, n resides at an equimolar ratio of the binding partners ðn = 1Þ, whereas n values of 0.5 or 2 are consistent with 1:2 or 2:1 complexes, respectively. Values of n significantly deviating from a reasonable binding stoichiometry often indicate that the actual concentration of one or both binding partners is different from the assumed one because of impurities in the samples used.

35

2.2 Analytical strategies and techniques

A single ITC titration thus affords information about the binding model as well as the full thermodynamic profile of the binding process, rendering ITC a powerful method for binding studies. The accessible range of stability constants lies between 101 M−1 and 108 M−1, depending on the concentrations chosen. This range can be extended beyond 108 M−1 by performing competitive titrations. The reliability of Ka is mainly determined by the accuracy with which the binding isotherms can be fitted as we have seen before for other methods. If complex stability is too low, the curves are too flat, precluding reliable fitting. If, on the other hand, complex stability is too high, the step of the curve is too steep. Ideal curves are S-shaped with a well-defined curvature. To estimate the optimal concentration of the binding partner in the cell that should be present for the measurement to afford a useful binding isotherm (e.g. c0R ), the c parameter (Wiseman parameter) has been introduced. This parameter is defined in the following equation. c = c0R Ka

(2:30)

The Ka in this equation is the known or assumed stability constant of the complex. If the product of c0R and Ka , that is, the Wiseman parameter c, is between 5 and 500, the curvature of the binding isotherms is sufficiently well defined to allow fitting. If c < 5, no reliable fitting is possible. If c > 500, only the binding enthalpy and the complex stoichiometry can be accurately determined. Figure 2.14 shows a series of ITC binding isotherms to illustrate the relationship between their shape and the c value.

10.0

–ΔH 0 (kJ mol–1)

8.0 6.0 4.0 2.0 0.0 0.0

0.5

1.0

1.5

2.0 c S0 / c R 0

Figure 2.14: Examples of ITC binding isotherms corresponding to different c values. The concentration cR0 amounts to 10−2 M for all curves and Ka is varied between 102 M−1 (c = 1, red), 103 M−1 (c = 10, blue), 104 M−1 (c = 100, orange), and 106 M−1 (c = 10,000, black). A ΔH0 of –10 kJ mol−1 was chosen in all cases.

36

2 Analyzing complex formation

Based on the information provided in this chapter about how binding equilibria are treated mathematically, how thermodynamics influences the extent of complex formation and kinetics the rate, and how the parameters describing binding equilibria are determined experimentally, we now turn to the forces that control the interaction of molecules.

Bibliography [1] [2] [3]

[4] [5] [6] [7]

Cram DJ, Blanda MT, Paek K, Knobler CB. Constrictive and intrinsic binding in a hemicarcerand containing four portals. J. Am. Chem. Soc. 1992, 114, 7765–73. Job P. Formation and stability of inorganic complexes in solution. Ann. Chim. Appl. 1928, 9, 113–203. Hibbert DB, Thordarson P. The death of the Job plot, transparency, open science and online tools, uncertainty estimation methods and other developments in supramolecular data analysis. Chem. Commun. 2016, 52, 12792–805. Thordarson P. Determining association constants from titration experiments in supramolecular chemistry. Chem. Soc. Rev. 2011, 40, 1305–23. Online Tools for Supramolecular Chemistry Research and Analysis (Accessed April 30, 2020, http://supramolecular.org). Vacca A, Francesconi O, Roelens S. BC50: a generalized, unifying affinity descriptor. Chem. Rec. 2012, 12, 544–66. Bencini A, Bianchi A, Burguete MI, Dapporto P, Doménech A, García-España E, Luis SV, Paoli P, Ramírez JA. Selective recognition of carboxylate anions by polyammonium receptors in aqueous solution. Criteria for selectivity in molecular recognition. J. Chem. Soc., Perkin Trans. 1994, 2, 569–77.

3 Understanding molecular recognition CONSPECTUS: When molecules meet in solution and experience attractive interactions they stay together for a certain amount of time. This chapter deals with the binding forces that are responsible for these interactions. First, however, a number of general factors contributing to the efficiency of complex formation are outlined.

3.1 Modes of binding 3.1.1 General considerations

Which general parameters influence complex formation?

One key parameter that affects the stability of a complex is the size of the contact interface shared by the binding partners. The larger this area, the stronger the intrinsic interactions between the molecules can become (Section 3.5.3). This aspect is illustrated in Figure 3.1.

(a)

(b)

(c)

(d)

Figure 3.1: Dependence of the extent to which molecules come into contact during complex formation on their shape and size complementary. The case of an optimal shape and size complementarity is shown in (a). In (b), the binding partners are not shape complementary, while in (c) and (d) they are shape but not size complementary.

Assuming that the two binding partners are rigid, the spherical substrate in Figure 3.1a is clearly better suited to efficiently approach the inner surface of the concave binding site of the receptor than the square-shaped one in Figure 3.1b. Shape selectivity leads to strong binding, however, only if the binding partners are also complementary in size. Substrates that are too large cannot enter the receptor cavity (Figure 3.1c), while those that are too small fill only part of the available space, significantly diminishing the size of the contact interface (Figure 3.1d).

https://doi.org/10.1515/9783110595611-003

38

3 Understanding molecular recognition

To a first approximation, one would therefore expect the situation in Figure 3.1a to be most favorable for high complex stability, but this is only correct if further conditions are fulfilled. A decisive one derives from the fact that most molecules contain polar bonds, causing an anisotropic distribution of charges along their surfaces. As a consequence, contacts between molecules only favor complex formation if oppositely charged surface regions approach each other, that is, if the respective surfaces are not only shape and size complementary, but also electronically complementary. Most types of noncovalent interactions that contribute to the stabilization of supramolecular complexes can indeed be rationalized on the basis of an attraction between electron-deficient and electron-rich parts of the binding partners. These interactions are often classified according to the functional group or structural unit that causes the anisotropic charge distribution (hydrogen bonding, aromatic interactions, etc.), which is convenient but could make it difficult to identify common principles. To clearly illustrate that most types of noncovalent interactions can be understood on the basis of simple electrostatic considerations, electrostatic potential surfaces are shown for many compounds presented in this book. These surfaces are generated by mapping the potential energy of a virtual charged probe onto a molecular surface of constant electron density (usually 0.002 electrons/Å2), followed by color coding, thus providing information as to where in this molecule a binding partner with a known charge distribution experiences an attractive and where a repulsive interaction. An example is shown in Figure 3.2. (a)

(b)

(c)

O N H

Figure 3.2: Molecular structure (a), stick model (b), and electrostatic potential surface (c) of N-methyl acetamide. The color coding covers a potential range from −120 to +120 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

The red surface surrounding the carbonyl oxygen atom of N-methyl acetamide in Figure 3.2 indicates that this atom features a negative potential as expected for an electronegative atom. The blue color surrounding the proton at the nitrogen atom signifies a positive potential in this region. These areas are therefore available for electrostatic interactions with a positively and a negatively charged or polarized binding partner, respectively. Apparently, electrostatic potential surfaces are useful tools to estimate whether certain molecules are able to interact with each other and to predict the preferred binding mode (and we will see in Section 3.4 that they even allow the prediction of

3.1 Modes of binding

39

binding strength in certain cases). However, potential binding modes can only be reliably derived by visual inspection if the scales of the color coding are properly chosen. Very large ranges between the minimum and the maximum potential are unsuitable because they cause areas to appear apolar (green) that could actually feature a substantial positive or negative potential, whereas very small scales overestimate potentials. Whenever electrostatic potential surfaces are used in this book to illustrate binding properties, the range of the electrostatic potentials is therefore specified. Moreover, analogous scales are used whenever possible to allow comparison. It must be emphasized that the rationalization of noncovalent interactions solely on the basis of electrostatic forces is straightforward and usually sufficient for supramolecular complexes, but that other binding mechanisms cannot be neglected in some cases. An important one invokes orbital interactions, that is, the overlap of a filled orbital of a donor with an empty orbital of an acceptor. These interactions involve the transfer of electrons from one binding partner to the other, explaining the term charge-transfer interactions. The participating orbitals can be σ, π, or n orbitals. Of particular importance in this respect are charge-transfer interactions between the high-lying HOMO of electron-rich aromatic systems and the low-lying LUMO of electron-poor ones. The respective orbital mixing causes the appearance of a new band in the UV–vis spectrum, the so-called charge-transfer band. We will come back to this type of interactions in Section 3.1.9. Besides the actual binding forces, another aspect influencing complex formation and stability is the flexibility of the binding partners. Complex formation is usually accompanied by a conformational reorganization of the receptor and/or the substrate, which can result in changes of the arrangement of functional groups in the regions where the contact occurs, causing the optimization of attractive interactions and thus reinforcing binding. Conformational changes also allow the binding partners to correct a shape or size mismatch. By contracting or expanding its cavity, for example, a receptor can adapt to the size of the substrate. These effects typically have a favorable effect on binding enthalpy, but the restriction of the conformational mobility of the binding partners during complex formation also has entropic consequences. Entropy becomes unfavorable if the conformational changes associated with complex formation become too large or if binding becomes too tight, potentially causing the situation in Figure 3.1a to be less beneficial than one where the substrate retains a certain mobility inside the receptor cavity. This relationship between receptor conformation and complex stability is described in more detail in Section 3.5.1. Finally, solvent effects have a profound influence on complex stability as already mentioned in Section 2.1 (also see equation (2.8)). Solvent molecules promote complex formation even in the case of weak direct interactions if the solvation of the complex is more favorable than that of the individual binding partners. Conversely, solvent molecules prevent intermolecular interactions if the binding partners are too strongly solvated. These effects are discussed in Section 3.3.

40

3 Understanding molecular recognition

The overall stability of a complex thus depends on the interplay of various factors. In the following sections, we first learn about the direct forces responsible for the components of a supramolecular complex to stay together, and we concentrate in this context on noncovalent interactions. Subsequently, methods used to experimentally assess the strength of noncovalent interactions are presented, and we then look at how conformational flexibility and the solvent influence complex stability. Applications in supramolecular chemistry of coordinative interactions, which have a substantial covalent character, are explained in later chapters. Which types of interactions cause molecules to stay together?

3.1.2 Ion–ion interactions The prototype of an electrostatic interaction is the attraction between two oppositely charged ions (Figure 3.3a). The potential energy E associated with this interaction is estimated by using equation (3.1), which derives from Coulomb’s law. E=

q1 q2 4πεε0 r

(3:1)

In this equation, q1 and q2 denote the elementary charges of the ions (±1.602 × 10−19 C per positive or negative charge), ε0 the permittivity of the vacuum (8.854 × 10−12 C V−1 m−1), ε the relative permittivity of the medium, and r the distance of the ions. The interaction thus becomes attractive (E < 0) if the ions are oppositely charged, that is, if q1 and q2 have different signs. Moreover, the inverse correlation of E with ε and r shows that the interaction strength becomes weaker as the distance of the ions increases or the medium becomes more polar. This dependence is illustrated in Figure 3.3b. According to Figure 3.3b, the strength of an ionic interaction in the gas phase easily amounts to several hundred kJ mol−1, comparable to covalent interactions. Medium effects, however, cause a substantial attenuation, reducing E by ca. two orders of magnitude when going from the gas phase to water. It must be emphasized that the estimates in Figure 3.3b do not reflect realistic binding strengths because equation (3.1) does not properly take into account atomic or molecular properties. First, this equation only applies to point charges, but ions have finite radii and therefore approach each other only within a certain distance. Once the electron clouds of the ions come too close, their interactions become repulsive, which is not considered in equation (3.1). Second, effects of the medium on complex stability are only partly due to the relative permittivity of the solvent. The solvation of the binding sites, that is, the direct interaction between binding

3.1 Modes of binding

(a)

(b)

0

41

–6 kJ mol–1

q1

q2 r

E (kJ mol–1)

–103 kJ mol–1 –200

–400 –496 kJ mol–1 –600 2

3

4

5

6

7

8

9

10 r (Å)

Figure 3.3: Schematic representation of two oppositely charged ions interacting at the distance r (a) and dependence of the potential energy E on the distance r in the gas phase (ε = 1, orange), in chloroform (ε = 4.8, blue), and in water (ε = 78, red) (b) (relative permittivities refer to T = 20 °C). The dotted line denotes the distance of a Na+ and a Cl− ion in NaCl (2.8 Å).

partners and solvent molecules as well as the complexation-induced reorganization of solvent molecules (Section 3.3) also play decisive roles. Moreover, if the receptor cavity is shielded from the solvent as, for example, in deep cavitands (Section 4.1.9) or proteins, the permittivities experienced by the substrate inside and outside the binding site substantially differ. Finally, ions come in many different structures. Examples are inorganic anions, which can be linear (N3−), trigonal planar (NO3−), tetrahedral (SO42−), or octahedral (PF6−). Organic ions have an even larger structural variability, causing the strength of the interactions to depend on the direction from which the binding partner approaches. In the case of organic anions, inductive and mesomeric effects furthermore cause the charge to be distributed across a relatively large region of the molecule. This is shown in Figure 3.4a,b for the tetramethylammonium cation and the benzoate anion. The respective electrostatic potential surfaces illustrate that the positive charge of the tetramethylammonium cation is distributed across all four methyl groups, consistent with the higher electronegativity of nitrogen with respect to carbon. Similarly, the negative charge of the benzoate anion is distributed equally over both oxygen atoms because of conjugation and the phenyl ring also features a substantial negative potential, which is the basis of a type of noncovalent interactions termed “cation–π interactions” (Section 3.1.7). Electrostatic interactions between charged molecules are therefore usually weaker and depend on many more factors than equation (3.1) suggests. This equation correctly predicts, however, that the efficacy of the interactions directly correlates with the number of charges on the binding partners.

42

3 Understanding molecular recognition

(a)

(b)

–

(c)

O

N

N H2

O NH2

N H2 H2N

NH2 H H2N 2 N

CN NC

CN

3–

Co NC

CN CN

H7(3.1)7+

Figure 3.4: Molecular structures and electrostatic potential surfaces of the tetramethylammonium cation (a) and the benzoate anion (b). The color coding covers a potential range from −500 to +500 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively. The heptaprotonated form of the polyazamacrocycle 3.1 and [Co(CN)6]3− shown in (c) form a complex that is mainly stabilized by ion–ion interactions.

Macrocyclic polyammonium ions, such as the heptaprotonated form of the azacrown 3.1, and complex anions, for example [Co(CN)6]3− (Figure 3.4c), mainly interact with each other by electrostatic interactions (Section 4.1).

3.1.3 Ion–dipole interactions Ion–dipole interactions occur between permanently charged atomic or molecular species and polar uncharged molecules. In the electrostatic model, the strength of these interactions depends on the relative extents of the attractive and repulsive forces between the ion and the partial charges in the polarized molecule. This situation is illustrated in Figure 3.5, where a cation is oriented at a distance r to a diatomic polar molecule with the partial charges δ + and δ–. – r– d q

r+ + r

Figure 3.5: Schematic representation of a cation interacting with a polar diatomic molecule at the distance r and the angle θ. The distances between the ion and the individual partial charges δ+ and δ– are denoted with r+ and r–, and the distance between the partial charges in the diatomic molecule with d.

The cation experiences an attractive interaction with δ–, arranged at a distance of r–, whereas the interactions with δ+, located at a distance of r+, is repulsive. The resulting balance of attraction and repulsion is given by equation equation (3.2a), which derives from (3.1) and treats the electrostatic interactions of the ion with the

3.1 Modes of binding

43

individual partial charges separately. Considering the relevant geometric parameters and the dipole moment μ, which is defined by the magnitude and the distance of the charge separation in the polar molecule, equation (3.2a) is transformed into equation (3.2b) that differs in several aspects from (3.1).   q δ+ δ− (3:2a) + E= 4πεε0 r r + r− E=

qμ cosθ 4πεε0 r2

(3:2b)

First, the charge of one ion in equation (3.1) is replaced by the dipole moment of the polar molecule. As μ is a vector pointing into the direction of the negative partial charge, ion–dipole interactions are, unlike ion–ion interactions, directional. This aspect is reflected in the cosθ term. A cation (q > 0) oriented at an angle θ of 0° to the positive end of the dipole experiences a maximum repulsive interaction. If θ = 90°, attractive and repulsive interactions cancel out so that E becomes zero, and the interactions become most attractive at θ = 180°. Finally, E is proportional to the inverse squared distance as a consequence of the combination of attractive and repulsive interactions in the ion–dipole interactions. The strength of these interactions thus decreases more rapidly than that of ion–ion interactions. To estimate the extent to which the strengths of ion–ion and ion–dipole interactions differ, we must compare q and μ=r. Since dipole moments have an order of magnitude of 10−30 C, and since distances are in the Ångström range, μ=r amounts to ca. 10−20 C. The elementary charge, on the other hand, amounts to ca. 10−19 C and a single ion–dipole interaction can therefore be expected to be at least one order of magnitude weaker than a single ion–ion interaction if other parameters such as charges are comparable. Ion–dipole interactions nevertheless become substantial if multiple dipoles are involved. A good example is the interaction of ions with the water molecules in their solvation shells. Hydration enthalpies thus provide useful information about how the nature of the ion influences binding strength. Selected values are shown in Table 3.1 [1]. According to Table 3.1, the strength of ion–dipole interactions correlates with the charge density of the ion, with smaller cations or anions interacting more strongly with water molecules than respective larger ones. The reason is that distributing the same charge over a larger volume attenuates the electrostatic interactions as also indicated by the electrostatic potential surfaces of the halides depicted in Figure 3.6a. The larger the halide, the less negative the electrostatic potential and, hence, the weaker the electrostatic interactions with an oppositely charged binding partner.

44

3 Understanding molecular recognition

Table 3.1: Hydration enthalpies ΔH0hydr and ionic radii for a selection of inorganic cations and anions. Cation +

Li Na+ K+ Rb+ Cs+ Mg+ Ca+

0 ΔHhydr (kJ mol−1)

Ionic radius (Å)

− − − − − −, −,

. . . . . . .

Anion

0 ΔHhydr (kJ mol−1)

Ionic radius (Å)

− − − −

. . . .

F− Cl− Br− I−

(a)

(b)

F–

Cl –

Br –

I–

Na+

Cl –

Figure 3.6: Electrostatic potential surfaces of the four halides fluoride, chloride, bromide, and iodide (a) and preferred orientation of a water molecule with respect to a cation and an anion (b). The potentials amount to −1,072.5 kJ mol−1 for F−, −797.0 kJ mol−1 for Cl−, −687.4 kJ mol−1 for Br−, and −616.3 kJ mol−1 for I− according to the respective calculations.

When comparing ions of similar size but different charges, for example Li+ and Mg2+, the hydration of the doubly charged ion is much stronger, which is consistent with equation (3.2b). Moreover, anions are typically more strongly hydrated than their corresponding isoelectronic cations. This trend may be counterintuitive at first sight because F− is larger than Na+ and Cl− larger than K+. It is consistent, however, with the orientation of water molecules in the first hydration shell of ions. A water molecule approaches a cation with its oxygen atom but an anion with the positively polarized protons as shown schematically in Figure 3.6b. Since hydrogen atoms are smaller than oxygen atoms, water molecules approach anions at a smaller distance, causing the respective ion–dipole interactions to be stronger than those between water molecules and cations. Typical complexes in supramolecular chemistry stabilized by ion–dipole interactions are the cation complexes of crown ethers (Section 4.1).

3.1.4 Dipole–dipole interactions Two or more uncharged but polar molecules can engage in dipole–dipole interactions. These interactions occur, for example, in solvents with permanent dipole

45

3.1 Modes of binding

moments such as chloroform, acetone, or DMSO. They cause orientations of the individual dipoles in which their positive and negative ends preferentially approach each other. The strength of these interactions depends on the magnitude of the dipole moments and the relative arrangement of the dipoles in space. Equation (3.3) describes the underlying relationship mathematically. E= −

μ1 μ2 ð2cosθ1 cosθ2 − sinθ1 sinθ2 cos’Þ 4πεε0 r3

(3:3)

This equation is closely related to equation (3.2b) but expectedly contains a product of two dipole moments μ1 μ2 . Since dipole moments are positive, the negative sign renders the interaction attractive at an appropriate orientation of the dipoles. The last term in equation (3.3) accounts for the dependence of E on the arrangement of the dipoles in space, defined by the angles θ1 , θ2 , and φ as shown in Figure 3.7.

Relative twist angle 2 1

r

Figure 3.7: Schematic representation of the interaction of two dipoles at the distance r. The angles θ1 , θ2 , and φ appear in equation (3.3) and describe the relative arrangement of the dipoles in space.

As we have seen in the previous chapter, the formal replacement of one charge in equation (3.1) by μ=r causes the interaction energy of ion–dipole interactions to be weaker than ion–ion interactions. Replacing both charges leads to a further reduction, rendering dipole–dipole interactions the weakest in this series. In addition, the interaction energy of dipole–dipole interactions falls off most rapidly with distance because E correlates with r − 3 instead of r − 2 or r − 1 .

3.1.5 Hydrogen bonding At this point, we turn to one of the most important types of noncovalent interactions, namely, hydrogen bonding. Hydrogen bonds contribute to the stabilization of protein structures or to the formation of double-stranded DNA, and they are therefore crucial for the function of these and many other biomolecules. Because of this importance, hydrogen bonds are generally treated as an individual class of noncovalent interactions, but in terms of electrostatics, they are closely related to ion–dipole or dipole–dipole interactions, depending on whether one binding partner is an ion or a neutral polar molecule. The characteristic structural feature of hydrogen bonds is

46

3 Understanding molecular recognition

that the dipole is generated by a polar bond between an electronegative atom and a proton, with the presence of this proton having important implications for the strength of the interactions. The size of protons, in particular, allows two molecules that interact by hydrogen bonding to approach each other at a smaller distance than possible for binding partners engaging in conventional dipole–dipole interactions. This proximity is beneficial for complex stability, as already noted in Section 3.1.3, where we have seen that hydration is stronger if water molecules approach an ion with the protons. It must be stressed that reducing hydrogen bonds to electrostatic interactions alone is a simplification, which is, however, reasonable for the moderately strong hydrogen bonds found in the majority of supramolecular systems. Hydrogen bonds in the strictest sense are more complex types of interaction, comprising a combination of electrostatics, polarization, and dispersion (Section 3.1.10), and very strong hydrogen bonds even have a partial covalent character. This complexity is one of the reasons for the difficulty in providing a clear-cut definition for a hydrogen bond. In 2011, the IUPAC recommended the following one [2]: The hydrogen bond is an attractive interaction between a hydrogen atom from a molecule or a molecular fragment X–H in which X is more electronegative than H, and an atom or a group of atoms in the same or a different molecule, in which there is evidence of bond formation.

According to this definition, a hydrogen bond can be denoted in the following manner: X − H    Yð − ZÞ X–H is the hydrogen bond donor as it carries the proton that is essential for the interaction. X is typically an electronegative heteroatom such as nitrogen, oxygen, or a halogen atom, but any molecular fragment that causes H to have a more or less pronounced partial positive charge is allowed. The hydrogen bond acceptor, that is, the binding partner of the proton, is a neutral or charged atomic species Y or a molecular species Y–Z. Since the above definition does not limit the nature of Y to certain atoms or molecular fragments, all structural elements that interact with a positively polarized hydrogen atom are permitted as hydrogen bond acceptors – not only electronegative heteroatoms but, for example, also π-systems. The electrostatic attraction between the proton and Y causes the X–H bond to become longer and weaker upon hydrogen bonding. These effects can conveniently be followed by IR and NMR spectroscopy. The weakening of the X–H bond causes the X–H vibration band to move to lower wavenumbers and to become substantially

3.1 Modes of binding

47

more intensive and broader. Simultaneously, new vibrational modes belonging to the H. . .Y bond arise. In the 1H NMR spectrum, hydrogen bond formation is associated with a deshielding of the proton whose signal in the spectrum thus moves to higher ppm values. The extent of the transfer of the proton from the hydrogen bond donor to the hydrogen bond acceptor is influenced by the acidity of the donor and the basicity of the acceptor. Thus, donor/acceptor combinations exist where complete proton transfer is thermodynamically favored. Proton abstraction leads, however, to an ion-pair that forms a salt bridge stabilized, in part, by hydrogen bonding but also by a substantial contribution from ionic interactions. An example is the carboxylate–guanidinium ion pair shown in Figure 3.8a, which is formed from a carboxylic acid and a guanine derivative.

H

H

(a)

O R

H N +

R

NH2

O H

N

O ‒ O

F

H F

NH2 H

H (b)

H N + H N

‒

Bifluoride

O

H

O

Enol of acetylacetone

Figure 3.8: Structure and formation of the carboxylate-guanidinium ion-pair (a) and examples of systems with very strong hydrogen bonds (b).

In special cases, the proton ends up to be symmetrically bound by two short hydrogen bonds between X and Y, normally with X = Y. These hydrogen bonds are often very strong, potentially even stronger than weak covalent bonds (50–150 kJ mol−1 in the gas phase). Examples are the hydrogen bonds in the bifluoride anion or in the enol of acetylacetone whose stability arises from a substantial covalent character produced by the sharing of the electrons between the hydrogen atom and the two flanking heteroatoms (Figure 3.8b). Apart from these special cases, hydrogen bonds are often of intermediate strength (15–50 kJ mol−1) and can also be very weak, not substantially stronger than dispersion interactions ( bromide > chloride > fluoride. This electrostatic view helps rationalizing the influence of structural effects on the strength of halogen bonding. The effectivity of an organic iodide to act as halogen bond donor improves, for example, as the electronegativity of the group X to which the iodine atom is bound increases. The reason is a shift of electron density along the X–I bond into the direction of X. This shift results in an increase of the size and the positive potential of the σ-hole as the comparison of the electrostatic potentials of methyl iodide and trifluoromethyl iodide in Figure 3.15 illustrates. The electrostatic model also explains the pronounced directionality of halogen bonds (Figure 3.15c). Since even small deviations from the optimal angle of 180° cause the halogen bond acceptor to experience a significantly smaller positive electrostatic potential at the halogen atom, halogen bonds are preferentially collinear with the X–Hal bond. Again, the electrostatic model is insufficient to explain all aspects of halogen bonds. Weak interactions are mediated to a substantial extent by dispersion, for example, while charge-transfer becomes important in the case of strong interactions. The electrostatic model helps, however, to understand the situation occurring in most supramolecular systems. In this context, the following comparison with hydrogen bonds is useful. Focusing first on the hydrogen or halogen bond donor shows that any effect of X in X–H or X–Hal that increases the positive potential of the hydrogen atom or the σ-hole of the halogen atom is beneficial for binding strength. Thus, the more electron-withdrawing X, the stronger binding. While halogen bonding allows tuning binding strength by varying the nature of the halogen atom, this possibility is missing in hydrogen bonding where only protons are allowed in the donor. Another difference is that the anisotropic charge distribution of halogen atoms causes halogen bonds to feature a much larger degree of directionality than observed for hydrogen bonds. Finally, protons are small and hard whereas iodide, for example, is large and soft, causing halogen bond donors to favor softer accetors. As a consequence, binding is less susceptible to solvent effects (halogen atoms are less strongly solvated than protons). Halogen bonding has therefore been considered to be the “hydrophobic ‘sister interaction’ to hydrogen bonding” [8]. Halogen bonding has been extensively used for crystal engineering, that is, the programmed assembly of crystalline materials from building blocks that engage in

56

3 Understanding molecular recognition

defined and predictable interactions. One example is the cocrystal of 1,2diiodotetrafluoroethane and N,N,Nʹ,Nʹ-tetramethylethylenediamine in which linear chains are found, stabilized by halogen bonds between the iodine and the nitrogen atoms (Figure 3.16a). An example of a receptor whose binding in solution is based on halogen bonding is the tripodal benzene derivative 3.2 (Figure 3.16b) [9]. The three iodine atoms in the aromatic side chains of 3.2 engage in interactions with, for example, a chloride anion in acetone, leading to a 1:1 complex with a log Ka of 4.3. Harder oxoanions such as hydrogensulfate or nitrate interact with 3.2 much less efficiently (log Ka < 1). Other examples of receptors whose substrate recognition involves halogen bonding are presented in Section 6.6.

(a)

F

(b) F N

F F I

I

F

F

F

F

N

F I

F F n

F

F

F I

O

F

O I

O O

F

O O 3.2

Figure 3.16: Schematic representation of the polymeric assembly that results when 1,2diiodotetrafluoroethane and N,N,Nʹ,Nʹ-tetramethylethylenediamine are cocrystallized (a), and structure of the tripodal receptor 3.2 with three aromatic residues containing iodine atoms as halogen bond donors (b).

3.1.7 Cation–π interactions We have seen that functional groups or structural elements in a molecule that induce an anisotropic charge distribution allow this molecule to interact with a complementary binding partner. In this and the following chapters we look at how π-systems engage in such interactions. Typical π-systems like those in alkenes or aromatic compounds are normally considered to be apolar, and it is therefore not immediately evident how they promote binding. Again, looking at the respective electrostatic potential surfaces helps in understanding the principles of the underlying interactions. Figure 3.17 shows that benzene features a pronounced negative potential on both ring faces, whereas the potential is positive along the rim. The reason is the polarization of the six CH bonds that causes a shift of electron density from the peripheral protons to the center of the ring (recall that C is more electronegative than H and that the electronegativity of C increases with increasing s-character of the orbital that is used for making the CH bond). The six dipoles of the CH bonds of benzene thus add up to produce a permanent quadrupole moment, characterized by two

3.1 Modes of binding

(a)

57

(b)

Dipole moments Figure 3.17: Electrostatic potential surface of benzene viewed from the top (a) and from the side (b). The end-to-end aligned dipoles that make up the quadrupole moment are shown in the side view. The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

end-to-end aligned dipoles. Although these two dipoles compensate each other, causing benzene to be overall apolar, a cation approaching the center of one face of a benzene ring experiences an attraction. This electrostatic interaction is the basis of the cation–π interaction, which is not restricted to aromatic π-systems but is also observed for alkenes such as ethylene for similar reasons. Cation–π interactions have been extensively studied in the gas phase, in the solid state, and in solution. First evidence that they exist came from gas phase studies, which demonstrated that the ΔH 0 of the interaction between benzene and a potassium ion in the gas phase is worth −80 kJ mol−1, rendering it even slightly stronger than the interaction of K+ with water (−75 kJ mol−1) [10]. Binding strength expectedly increases as the charge density of the cation rises, amounting to −117 kJ mol−1 for Na+ and −160 kJ mol−1 for Li+, but these energies only reflect the upper limit of the strength of a cation–π interaction. As in the case of ion–ion interactions, the presence of solvent molecules causes the interactions to become weaker, so that their contributions to a binding event in water typically only ranges between 1 and 3 kJ mol−1, for example [11]. The structures of a series of complexes between alkali metal salts and so-called lariat ethers (Section 4.1), comprising a crown ether moiety with appended substituents, also provide strong evidence for the attractive nature of cation–π interactions [12]. An example is the lariat ether 3.3, which adopts an extended conformation in the solid state with the two phenol groups oriented away from the ring (Figure 3.18a,b). In the absence of cation–π interactions, a cation binding to the crown ether should have no pronounced effect on this conformation and remain ion paired in the complex with its counterion. If cation–π interactions contribute to the stabilization of the complex, however, cation binding should induce a conformational reorganization. The crystal structure of the potassium iodide complex of 3.3 (Figure 3.18c), in which the potassium ion is sandwiched between the two phenol rings, illustrates that the latter is the case. Thus, cation–π interactions are sufficiently strong in this system to outperform the strength of the electrostatic interactions between the oppositely charged ions.

58

3 Understanding molecular recognition

(a)

(b)

(c)

OH O N

O

O

N O HO 3.3

Figure 3.18: Molecular structure of lariat ether 3.3 (a) and crystal structures of the uncomplexed form of 3.3 (b), and of the corresponding potassium iodide complex (c). The potassium ion is shown in blue and the iodide ion in purple.

Similar close contacts between cationic and aromatic residues as seen in the KI complex of 3.3 are found in the crystal structures of protein complexes, for example in the structure of nicotinic acetylcholine receptor binding to acetylcholine. In this case, the cationic trimethylammonium head group of the substrate comes into close contact with the indole moiety of a tryptophan residue of the protein, thus providing evidence for the importance of cation–π interactions also in biological systems. The dependence of the strength of cation–π interactions on the geometry of the complex and on the nature of the cation can be explained in terms of the electrostatic model. The interactions are strongest, for example, if the cation resides directly above or below the site of the aromatic system that features the largest negative potential – in the case of benzene, the center of the ring. Deviations from this optimal arrangement cause the weakening of binding as a consequence of the reduction of the electrostatic attraction. Moreover, the binding strength in the gas phase correlates with the charge density of the ion as expected for electrostatic interactions, decreasing as the ion becomes larger in the order Li+ > Na+ > K+ > Rb+. Solvent effects cause deviations from this trend. In water, for example, the two smallest and most strongly hydrated ions Li+ and Na+ rank at the end of the affinity scale, forming complexes that are even less stable than those of Rb+. Complex stability in solution thus depends on the balance between the intrinsic strength of the cation–π interaction and the degree to which the binding partners are solvated, with strong solvation causing the complex to become less stable. We have already seen this influence of solvation on the overall binding strength when discussing equation (2.8). Interestingly, solvation effects cause the strength of cation–π interactions to fall off less strongly with increasing polarity of the solvent than other types of interactions. The reason is that if both binding partners are polar, both of them are better solvated in more polar solvents and therefore less prone to interact. In the case of cation–π interactions, however, desolvation of one partner, namely the π-system, is usually not strongly solvent-dependent and relatively easy even in water.

59

3.1 Modes of binding

Other factors affecting binding strength are the structure and substitution pattern of the aromatic system that interacts with the cation. Again, the best way to assess these factors involves analyzing the electrostatic potential surfaces of the respective aromatic compounds. Figure 3.19 shows these surfaces for a selection of substituted benzene derivatives and heterocyclic aromatic compounds.

NH2

Eint –133.1 (kJ mol–1)

–118.4

–112.5

O

OH

CF3

–111.1

–81.2

N H

CN

NO2

–66.9

–58.6

N H

Figure 3.19: Electrostatic potential surfaces of aniline, toluene, benzene, phenol, trifluoromethylbenzene, benzonitrile, nitrobenzene, furan, pyrrole, and indole. The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively. The calculated energies Eint for the monosubstituted benzene derivatives, taken from ref. [13], describe the interaction of a sodium ion with the respective aromatic compound in the gas phase.

At first sight, the influence of aromatic substituents on the strength of cation–π interactions roughly correlates with the reactivity of the respective aromatic systems in electrophilic aromatic substitution reactions according to Figure 3.19, with activating substituents strengthening the interaction and deactivating ones weakening it. However, a closer inspection reveals deviations. Phenol, for example, has an activated π-system but interacts less strongly with Na+ than benzene. Furan is also substantially more reactive in SEAr reactions than benzene but is nevertheless a weaker cation binder. Indeed, it was found that the ability of aromatic systems to participate in cation–π interactions does not correlate with the polarization of the π-system, but with Hammett’s σmeta parameter, which describes the inductive effects of substituents on the aromatic ring [14]. Calculations confirmed that the effects of aromatic substituents primarily arise from the influence of their dipole moments on the electrostatic potential of the π-system [13].

60

3 Understanding molecular recognition

Important information about the stabilizing effects of cation–π interactions came from the characterization of the binding properties of a family of cyclophanes, a representative of which is compound 3.4 (Figure 3.20) [15]. This compound was originally developed as a receptor for apolar substrates in water. With a log Ka of 8.4 in 10 mM borate buffer, the complex of 3.4 with the N-methylquinolinium cation is, however, more than two orders of magnitude more stable than the complex with the neutral but otherwise similarly shaped 4-methylquinoline (log Ka = 5.9). Since the carboxylate groups point away from the receptor cavity, their electrostatic contributions to binding is small and the higher affinity of 3.4 for the positively charged substrate is therefore due to the contribution of cation–π interactions. Many of the receptors with aromatic residues presented in Section 4.1 bind to cations through cation–π interactions.

‒

‒ OOC

COO

O

O

N N +

3.4 log Ka O

8.4

5.9

O

‒

OOC

COO

‒

Figure 3.20: Structure of cyclophane 3.4 and binding constants log Ka of the complexes of 3.4 with the N-methylquinolinium cation and 4-methylquinoline in 10 mM aqueous borate buffer.

3.1.8 Anion–π interactions If cation–π interactions exist, what about anion–π interactions? Aromatic systems are, of course, usually considered to be electron-rich because of the electrons delocalized in the π-orbitals, which renders anion–π interactions unlikely. We have seen, however, that cation–π interactions are not governed by π-effects. Moreover, nucleophiles are known to react with electron-poor aromatic systems. There must therefore be a driving force that allows a nucleophile to engage in interactions with a π-system and, indeed, electron-poor carbocyclic or heterocyclic aromatic compounds feature a pronounced positive potential, allowing them to attract an anion (Figure 3.21). Computational work indicates that anion–π interactions mainly rely on two effects. One is the electrostatic attraction between the anion and the positive potential

3.1 Modes of binding

F F

NO2

F

F

F F

61

N O2N

NO2

N N

Figure 3.21: Electrostatic potential surfaces of hexafluorobenzene, 1,3,5-trinitrobenzene, and 1,3,5-triazene. The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

of the electron-deficient aromatic system. This interaction is further reinforced by the ion-induced polarization of the π-system, which arises when the anion approaches the aromatic ring. Once close enough, the anion causes a shift of the electron density to the opposite side of the ring, thus inducing a dipole moment, the positive end of which faces the anion. In spite of the evidence from computations that anion–π interactions are attractive, the experimental proof of their stabilizing effects has not been straightforward. In crystal structures, for example, anions are found in various arrangements with respect to close-by π-systems, which are not always consistent with anion–π interactions. For example, anions approaching the substituted carbon atom of a benzene ring coincide with the trajectories of nucleophiles when forming Meisenheimer complexes during SNAr reactions, therefore not providing clear evidence for the involvement of anion–π interactions. Moreover, planar anions such as nitrate, carbonate, or azide can engage in π–π interactions with arenes. In certain systems, however, arrangements of anions with respect to electron-deficient aromatic rings exist that are consistent with even very strict views of anion–π interactions. Examples are the bromide salt of the quaternary ammonium ion 3.5 and the tetrabutylammonium bromide complex of benzamide 3.6a, which both feature the anion right above the pentafluorophenyl moieties (Figure 3.22) [16, 17]. These structures thus provide similar evidence for the attractive nature of anion–π interactions as the lariat ether 3.3 does for cation–π interactions (Figure 3.18 and Section 3.1.7). There is also evidence for the stabilizing effects of anion–π interactions in solution. An early example is the sulfonamide 3.7a with an appended pentafluorophenyl ring (Figure 3.22) [18]. While the corresponding benzene derivative 3.7b does not exhibit any measurable affinity for halides in chloroform, 3.7a binds weakly to halides under the same conditions (Ka between 20 and 34 M−1). According to the experimental

62

3 Understanding molecular recognition

N +N

R F F

F

SO2 NH R

NH F F

O

F F 3.5

F

F F

R R

R R

3.6a R = H

3.7a R = F

3.6b R = Ph

3.7b R = H

Figure 3.22: Molecular structures of the ammonium cation 3.5, the benzamides 3.6a,b, and sulfonamides 3.7a,b, and modes of interactions of 3.5 and 3.6a with a bromide ion in the solid state. The tetrabutylammonium counterion in the complex of 3.6a is omitted for reasons of clarity.

results, halide affinity is due to anion–π interactions and not due to differences in the acidity of the sulfonamide NH groups of both receptors. Unfortunately, the crystal structures of the halide complexes of 3.7a provide no evidence that anion–π interactions are also operative in the solid state. By contrast, the arrangement of the anion in the crystal structures of 3.6a demonstrates that anion–π interactions contribute to the halide affinity of 3.6a and 3.6b (Figure 3.22) [17]. Ascribing the binding event exclusively to the interaction between the anion and the electron-deficient arene is generally difficult, however, which is why anion–π interactions are still a somewhat less well-established interaction type than the related cation–π interactions. The importance of the field is nevertheless progressively increasing, especially with the work on the use of anion–π interactions for ion transport or catalysis (Section 9.3.1).

3.1.9 Aromatic–aromatic interactions The fact that aromatic systems feature regions with positive and negative electrostatic potentials along their surfaces enables them to engage in electrostatic interactions not only with charged binding partners, as in cation–π and anion–π interactions, but also with themselves. In order for these interactions to become attractive, regions with opposite electrostatic potentials must come close to each other, rendering only certain arrangements favorable. In the case of benzene, for example, the T-shaped or edge-toface arrangement of two benzene molecules is the most stable one as it involves positively polarized hydrogen atoms along the edge of one ring approaching the face of the second ring where the electrostatic potential is negative. This arrangement is, in fact, the basis for the herringbone structure of crystalline benzene (Figure 3.23a,b). Alternatively, the planes of two benzene rings can also be arranged in a parallel fashion if structural effects within a molecule or in a receptor–substrate complex prevent the perpendicular orientation. A parallel arrangement of aromatic rings is

63

3.1 Modes of binding

(a)

(b)

(c)

(d)

(e) F

F

F

F

F

Edge-to-face

Displaced

Stacked

F

Stacked

Figure 3.23: Possible arrangements of two aromatic rings. The T-shaped (a) and the displaced (c) arrangements are attractive. The stacked arrangement of two benzene rings (d) is repulsive but the alternating stacked arrangement of benzene and hexafluorobenzene is attractive (e). The herringbone structure of crystalline benzene (b) is characterized by multiple edge-to-face interactions.

sometimes also induced by solvent effects because it is associated with a more efficient shielding of hydrophobic surfaces than the T-shaped structure. The interaction is attractive when the rings are displaced, so that the positively polarized edge of one ring is located above or below the face of the second ring (Figure 3.23c), but it is repulsive when the rings stack, rendering this arrangement not possible (Figure 3.23d). The term π–π stacking, although frequently found in the literature, is therefore misleading as it indicates that attractive interactions between π-systems could cause them to stack, which is not the case [19]. Stacking becomes favorable, however, if different aromatic compounds interact whose electrostatic potentials have opposite signs in the center of the rings. An example is the mixed crystal of benzene and hexafluorobenzene in which the two components are stacked in an alternate fashion (Figure 3.23e). The stabilization caused by the electrostatic interactions between the rings is reflected in the melting point of this crystal, which amounts to 24 °C, ca. twenty degrees higher than the melting points of either benzene (6 °C) or hexafluorobenzene (4 °C). The interaction between electron-rich and electron-poor π-systems goes beyond electrostatic attraction if the HOMO of the donor and the LUMO of the acceptor are properly aligned and the energy gap between them is sufficiently small, allowing the orbitals to mix. In this case, charge-transfer interactions are possible, that is, the transfer of a fraction of electronic charge from the donor to the acceptor. We have seen that orbital mixing contributes to many kinds of noncovalent interactions, rendering its role in aromatic–aromatic interactions not unique. Charge-transfer interactions between π-systems often have the special feature, however, of causing a color change of the solution because the energy required to excite an electron from the HOMO into the LUMO of the complex lies in the visible region of the electromagnetic spectrum. Complex formation can thus be followed by UV–vis spectroscopy. An example of a system in which a charge-transfer interaction plays a role is the orange-red

64

3 Understanding molecular recognition

colored complex between the electron-rich 1,4-dimethoxybenzene and the receptor 3.8, which features two electron-poor paraquat moieties [20] (Figure 3.24). We come back to this system in Sections 4.1.5 and 6.4.

E + OCH3

hvvis

OCH3 + LUMO (acceptor)

HOMO (donor)

1,4-Dimethoxybenzene

N

N

N

N

+

+

3.8

Figure 3.24: Schematic representation of the orbital mixing that underlies charge-transfer interactions between an electron-rich and an electron-poor π-system and structures of 1,4-dimethoxybenzene and receptor 3.8 that form a complex stabilized by charge-transfer interactions.

3.1.10 Dispersion interactions We now come to the weakest type of noncovalent interactions, namely, dispersion or van der Waals interactions. These interactions explain why apolar atoms or molecules lacking anisotropic charge distributions such as noble gases or alkanes interact and can, for example, be liquefied. To mathematically describe how the potential energy E of this interaction depends on the distance r of the binding partners, the empirically derived Lennard–Jones potential in equation (3.4) is generally used.  12 6  σ σ (3:4) − E=ε r r According to equation (3.4), E is divided into a repulsive (positive sign) and an attractive (negative sign) term. The parameter σ in both terms corresponds to the distance at which attraction and repulsion cancel each other out, causing E to become zero (σ = r). The parameter ε describes the hardness of the interaction. This parameter relates to the polarizability of the binding partners, with soft–soft interactions having larger ε values than hard–hard ones. The dependence of E on r is depicted graphically in Figure 3.25a. The curves in Figure 3.25a illustrate that the Lennard–Jones potential is repulsive at small distances where the orbitals of the interacting molecules start to

65

3.1 Modes of binding

(a) 8

(b)

6

tBu tBu tBu

tBu

E

4

tBu

2 0

tBu tBu tBu tBu

–2 2

3

4

5

6

7

tBu tBu tBu

8 r (Å)

Figure 3.25: Lennard-Jones potential functions for σ = 3.2 Å and two different values of ε with the red curve describing a soft-soft interaction (larger ε) and the blue one a hard-hard interaction (smaller ε) (a). In (b), the structures of 1,1,1,2,2,2-hexaphenylethane and 1,1,1,2,2,2-hexakis(3,5-ditert-butylphenyl)ethane are shown, of which the former compound is unstable, while the latter is stabilized by the dispersion interactions mediated by the tert-butyl groups.

overlap and Pauli repulsion occurs (electrons have to move to higher orbitals to ensure that each occupies an individual quantum state). At distances larger than σ, there is a small region where the interactions are attractive. This attraction then quickly vanishes as r increases with a distance dependence of r − 6 . The attractive nature of the interactions within this small range of distances is often explained by using the model of oscillating dipoles: the movement of the electrons in one molecule causes a fluctuating dipole moment, which in turn induces a complementary dipole moment in the respective binding partner, thus resulting in a short-lived but attractive interaction. This view is also consistent with the increasing strength of induced dipole-induced dipole interactions with the increasing polarizability of the binding partners. Dispersion interactions are weak if only two small atoms or molecules are involved (explaining the low boiling points of noble gases). They become stronger, however, if the contact interfaces between two molecules increase in size. Thus, boiling points of alkanes increase as the chains become longer. Intramolecular dispersion interactions are also responsible for the stabilization of sterically highly crowded molecules. 1,1,1,2,2,2-Hexaphenylethane is, for example, unstable because of the steric congestion of the six phenyl groups. The more crowded 1,1,1,2,2,2hexakis(3,5-di-tert-butylphenyl)ethane (Figure 3.25b) should therefore be even less stable but the opposite is the case. The reason is that the interdigitation of the tertbutyl groups with the aromatic residues gives rise to intramolecular dispersion interactions, which stabilize this molecule to such an extent that isolation and recrystallization become possible. Since supramolecular complexes are often structurally rather complex, featuring a number of polarizable groups in close proximity, intermolecular dispersion interactions lead to a similar substantial stabilization. Even if these interactions are intrinsically weak, they must therefore not be neglected.

66

3 Understanding molecular recognition

3.2 Binding energies 3.2.1 General considerations

How strong are intermolecular interactions?

All types of interactions discussed so far are associated with characteristic binding strengths, and we have seen equations in Section 3.1 that allow quantifying how efficient the corresponding interactions are. These equations also permit ranking the different interactions according to their strength. An ion–dipole interaction should, for example, be ca. one order of magnitude weaker than an ion–ion interaction, as discussed in Section 3.1.3, because of the replacement of one elementary charge q in equation (3.1) with the term μ=r. Consequently, dipole–dipole interaction should be even weaker. We arrive at the same trend by looking at the distance dependence of the respective interactions. Two particles interacting at a distance r by means of ion–ion interactions should experience a much weaker interaction at the same distance if the interactions are due to ion–dipole or dipole–dipole interactions because the strengths of these interactions correlate with r − 2 or even r − 3 , respectively, rather than r − 1 . Accordingly, dispersive interactions are intrinsically the weakest ones. The sequence in which the different types of interactions are discussed in the previous chapter thus roughly reflects the order in which their strengths decrease. Gas phase studies or calculations allow quantifying intrinsic interaction energies as shown for a few selected examples in Section 3.1. These energies, however, typically overestimate how much the respective interaction is worth in solution and in a real system. The reason is that, among other factors, medium effects combined with specific effects of solvation strongly affect the interaction strength (Section 3.3). The high relative permittivity of water weakens ion–ion interactions as shown in Section 3.1.2, for example, and specific hydration further reduces these interactions to such an extent that they become negligible. On the other hand, the interactions between large hydrophobic surfaces are often reinforced in water, so that van der Waals interactions end up to be apparently stronger than ion–ion interactions under certain conditions. A statistical analysis of reported binding constants shows that typical receptor– substrate complexes in organic solvents exhibit binding constants Ka ranging between 1 and 107 M−1 with an average of 103.4±1.6 M−1, which corresponds to a ΔG0 value of −19.8 ± 9.1 kJ mol−1 [21]. In water, the range of binding constants is similar, but the average is smaller by ca. 7 kJ mol−1. Thus, interaction energies in solution are much lower than those calculated for the gas phase. Another literature survey confirms that the different types of interactions discussed earlier are associated in solution with ΔG0 values between −1 and −40 kJ mol−1 (reflecting the above mentioned Ka values

3.2 Binding energies

67

between 1 and 107 M−1), but no clear correlation between binding strength and type of interaction is visible [11]. To obtain information about the actual contributions of specific interaction types in solution to molecular recognition processes, the determination of binding constants is therefore often not sufficient and special strategies are required. These strategies have moreover to account for the fact that complex formation in solution is practically never due to a single type of interaction. Dispersion interactions, for example, very often contribute to the interaction between two molecules, even if the main driving force of complex formation is due to other effects. Thus, strategies that should provide information about the effect of a specific type of interaction need to allow separating one contribution from another. Three useful methods devised to achieve this differentiation are presented in the following sections.

3.2.2 Trend analyses The underlying notion of trend analyses is that each contribution to the stabilization of a complex is associated with a characteristic free energy value and that the overall stability of the complex equals the sum of all individual values according to equation (3.5). ΔG0total =

n X

ΔG0i

(3:5)

i=1

This concept is illustrated in Figure 3.26a by using the example of the interaction of a macrocyclic receptor, containing a varying number of positively charged groups along the ring, with an anionic substrate. Although the complex of the singly charged receptor is mainly stabilized by one salt bridge (Coulomb attraction reinforced by a hydrogen bond), the Gibbs free energy of its formation generally not only reflects the creation of this salt bridge alone, but also comprises other factors such as further attractive or repulsive interactions, the entropic penalty associated with bringing receptor and substrate together, and the loss of conformational flexibility. Thus, no information about the extent to which the salt bridge contributes to the stability of this complex can be derived from a single ΔG0 value. Each additional salt bridge that becomes possible in the higher charged receptors is expected, however, to add a constant and characteristic energy term to ΔG0 if there are no cooperative effects that complicate the situation (Section 3.5.3). The respective energetic contribution to complex stability can thus be determined individually by comparing the stabilities of a series of complexes. Indeed, the logarithmic binding constants log Ka of the anion complexes of macrocyclic polyammonium receptors, for example receptor 3.1 shown in Figure 3.4,

68

(b)

−ΔG 0 (kJ mol–1)

(a)

3 Understanding molecular recognition

80

60

− OOC

COO − − COO

40

COO − − COO

20 3

4

5

6

7

n

Figure 3.26: Schematic illustration of the complexes of an anion with macrocyclic receptors containing one to four positive charges that contribute to stability by the equivalent number of salt bridges (denoted as dotted lines) (a), and graph showing the dependence of the Gibbs free energy of formation of the complexes of terephthalate (blue) and benzene-1,2,3-tricarboxylate (red) on the degree of protonation of receptor 3.1 (Figure 3.4) (b) [7].

often increase linearly with the degree of protonation, obeying the expression 3.6, where zR and zA are the charges of the receptor and the anion, respectively. log Ka = a zA zR + b

(3:6)

As log Ka values are directly correlated with ΔG according to equation (2.7), plotting ΔG0 vs. the degree of protonation should yield a straight line with the slope a zA in which the term a reflects the energetic contribution of each salt bridge to the overall stability. That this treatment is indeed feasible is illustrated in the graphs in Figure 3.26b, which depict how the free energies ΔG0 that are associated with the formation of the complexes of 3.1 (Figure 2.11) with terephthalate and benzene-1,3,5-tricarboxylate vary with the number of protonated amino groups. For both complexes, the free energies of complex formation increase linearly when going from the triprotonated to the fully protonated form of the receptor. The slopes of the lines differ because of the different charges of the anions. By eliminating the effect of zA , an estimate of a is thus obtained, which indicates that, independent of the anion, each salt bridge stabilizes these complexes by −4.7 kJ mol−1. An analysis of a large number of complexes between charged binding partners indeed shows that salt bridges in water are associated with a surprisingly constant energetic contribution of –(5 ± 1) kJ mol−1, relatively independently of the type of receptor or substrate [11]. Estimates for other types of interactions such as hydrogen bonding or van der Waals interactions can be derived in a similar manner. 0

3.2 Binding energies

69

3.2.3 Double-mutant cycles The idea of obtaining information about the contribution of a specific type of noncovalent interactions to binding strength by comparing the stability of a complex in which this interaction occurs with the stability of an analogous systems that lacks the respective interaction is attractive, but often fails to produce meaningful results. The reason is shown schematically in Figure 3.27a. The complex ABXY is stabilized by pairwise interactions between the shape complementary subunits a and b, and x and y. However, secondary attractive or repulsive interactions between a and y and between b and x, like those discussed in Section 3.1.5, which are denoted as red dotted lines in the figure, cannot be completely ruled out. Thus, replacing the groups x and y in the binding partners with the noninteracting “mutated” groups xʹ and yʹ, yielding the double-mutant complex AB, not only causes the disappearance of the primary interaction, about which information should be obtained, but also of the secondary interactions. As a consequence, no conclusive information about how strongly x interacts with y can be obtained from the difference ΔG0ABXY − ΔG0AB . A workaround involves the additional integration of the single mutants ABX and ABY in the analysis. (a)

a

x

(b) ΔG0ABXY = ΔG0ab + ΔG0xy + ΔG0xb + ΔG0ay

y b ABXY a

x' ABY

b

ΔG0ABX = ΔG0ab + ΔG0xb a

x

b

y'

ABX

y AB a x'

ΔG0ABY = ΔG0ab + ΔG0ay ΔG0AB = ΔG0ab ΔG0ABXY – ΔG0ABY = ΔG0xy + ΔG0xb ΔG0ABX – ΔG0AB = ΔG0xy (ΔG0ABXY – ΔG0ABY) – (ΔG0ABX – ΔG0AB) = ΔG0xy

b

y'

Figure 3.27: General strategy of a double-mutant cycle that provides quantitative information about the contribution of the interaction between x and y to the overall stability of complex ABXY (a) and estimation of ΔG0xy from the Gibbs free energies ΔG0ABXY , ΔG0ABX , ΔG0ABY , and ΔG0AB , encompassing the sums of the individual energy contributions ΔG0ab , ΔG0xy , ΔG0ay , and ΔG0bx to the stability of the respective complexes (b).

The estimation of ΔG0xy is then based on the notion that all contributions to binding strength occurring in the different complexes are additive and not affected by the structural changes. Hence, the stabilities of the individual complexes reflect the sums of the corresponding individual contributions ΔG0ab , ΔG0xy , ΔG0ay and ΔG0bx , allowing

 the estimation of ΔG0xy from the difference ΔG0ABXY − ΔG0ABY − ΔG0ABX − ΔG0AB [or
0 
 equivalently ΔGABXY − ΔG0ABX − ΔG0ABY − ΔG0AB ] as derived in Figure 3.27b.

70

3 Understanding molecular recognition

The following conditions must be met for this treatment to work: (i) the structures of the different complexes must be very similar, (ii) the energetic contributions of the individual interactions must be additive, (iii) the interaction between a and b must be sufficiently strong to guarantee that the complex AB holds together even in the absence of the interacting groups x and y, and (iv) the mutated groups xʹ and yʹ must not interact. Double-mutant cycles provide information, for example, to what extent the interactions between pairs of amino acid subunits along a protein backbone contribute to the stability of the folded state. In this case, they involve replacing each of these amino acids or both of them with subunits that do not engage in interactions. In addition, several synthetic systems allow constructing double-mutant cycles [22]. An example is shown in Figure 3.28. In this system, the primary interaction between the binding partners 3.9a and 3.10a relies on hydrogen bonding. In addition, the respective complex is stabilized by edge-to-face aromatic interactions between the peripheral aromatic substituents.

O N

N H

H

O

O

H N

N H

O

3.9a.3.10a

O N

N H

H

O

O

O

H N

N H

N

N H

O

O

H N

N

O

H

H

O

3.9a.3.10b

3.9b.3.10a O N

N H

H

O

O

H N

N H

O

3.9b.3.10b

Figure 3.28: A double-mutant cycle that allows the quantification of the contribution of edge-toface aromatic interactions (red dotted lines) in the complex between 3.9a and 3.10a.

71

3.2 Binding energies

Replacing these substituents with aliphatic residues affords the analogous compounds 3.9b and 3.10b that allow establishing a double-mutant cycle by characterizing the stability of the four complexes 3.9a·3.10a, 3.9b·3.10a, 3.9a·3.10b, and 3.9b·3.10b. This treatment yields a ΔG0 of −1.3 kJ mol−1 for the magnitude of the terminal edge-toface aromatic interactions. The stabilization becomes stronger if one aromatic ring is electron deficient and the other electron rich, and repulsive if the interaction occurs between the edge of a phenyl ring and the face of a pentafluorophenyl ring, which is consistent with the electrostatic model. The results depend sensitively on the arrangement of the interacting subunits in the complex according to studies involving binding partners derived from different core structures. Structurally related complexes, however, provide reliable correlations about how structural parameters affect the strengths of certain types of interactions. Besides aromatic interactions, double-mutant cycles have also been used to characterize cation–π interactions, CH–π interactions, and others.

3.2.4 Molecular balances A third approach to characterize the strengths of individual types of noncovalent interactions relies on designed compounds that adopt two well-defined conformational states, one in which the interaction is possible and another, in which it is not [23]. This approach is illustrated schematically in Figure 3.29. It relies on a core structure whose conformational flexibility allows two potentially interacting groups x and y to be either closely arranged in space (folded state) or arranged at a distance too large to allow for interactions (unfolded state). The respective conformational equilibrium is described by an equilibrium constant K, specifying to what extent the folded state is favored or disfavored over the unfolded one. The ratio of both states in solution, which allows the calculation of K, is usually obtained by NMR spectroscopy since

(a)

(b) O

O O

x y

O

K

x

K

F3C

CF3

y N N 3.11 (folded)

N N 3.11 (unfolded)

Figure 3.29: Working principle of a molecular (torsion) balance to estimate the strength of interactions between the groups x and y (a), and molecular structure of the balance 3.11, derived from Tröger’s base that allows determining the strength of the edge-to-face interaction between the 4-trifluoromethylbenzene group and the benzene ring (b).

72

3 Understanding molecular recognition

typical systems used in such analyses feature conformational equilibria that are slow on the NMR timescale. The ratio of the two conformations is thus determined by integrating the corresponding NMR signals. The equilibrium constant K then allows the calculation of ΔG0xy by using equation (3.7). More reliable values for ΔG0xy are obtained by combining this concept with a double-mutant cycle, that is, by also considering the conformational equilibria of analogs in which either one or both substituents are replaced by noninteracting ones. ΔG0xy = − RT ln K = − RT ln

cfolded cunfolded

(3:7)

A compound that allows such analyses is the Tröger’s base derivative 3.11, which is known by the term “molecular torsion balance” or molecular balance for short. This compound exists in two conformations of which one is stabilized by an edge-to-face interaction between the 4-trifluoromethylbenzene unit and the benzene ring. The effect of this intramolecular interaction stabilizes the folded conformation in chloroform by −2.4 kJ mol−1 over the unfolded one, affording an estimate for the interaction strength in this system [24]. Molecular scales are attractive study objects, but several aspects must be considered when using them. One is the influence of the solvent on the ratio of the folded and unfolded state. This influence is particularly pronounced if the conformational rearrangement is associated with a substantial reorganization of solvent molecules surrounding the subunits whose interaction is probed. Another aspect is that the determined interaction energies depend on the exact arrangement of the substituents, which is defined by the core structure of the balance. Since this arrangement could differ from the one the same substituents prefer in the absence of the scaffold, the transfer of the results to noncovalently assembled systems could be problematic. An advantage of molecular balances is that the interactions occur intramolecularly, within a covalently assembled structure, so that no entropic penalty associated with bringing two molecules together has to be paid. Thus, even very weak or repulsive interactions can be characterized. Various interaction types were studied by using molecular balances, including CH–π, OH–π, NH–π, π–π, dispersion, and solvophobic interactions [23].

3.3 Solvent effects How does the solvent influence complex stability?

Imagine a complex that has been thoroughly characterized in the gas phase. How does dissolving this complex affect its structure and stability? To answer this

3.3 Solvent effects

73

question, one has to evaluate the effects of the large number of molecules that surround the complex in solution. These molecules can influence the mutual arrangement of the binding partners, but these effects are difficult to predict or generalize. More obvious is that the solvent defines the relative permittivity ε of the medium. We have seen in the previous chapter that the parameter ε influences the strength of electrostatic interactions between two particles, with higher relative permittivities causing the interactions to become weaker. Table 3.2 shows that the relative permittivities of typical solvents used in supramolecular chemistry vary over two orders of magnitudes, so that correspondingly large variations in binding energies are expected in different solvents as a consequence of this effect alone. Table 3.2: Selection of parameters describing the properties of solvents typically used in binding studies [26]. Solvent Water DMSO DMF Acetonitrile Methanol Acetone Dichloromethane Tetrahydrofuran Chloroform Benzene ,-Dioxane Tetrachloromethane a

ε

μ × 1030 a

π*

α

β

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

Dipole moment in C m.

Solvent effects do not only involve the relative unspecific influence on the medium, but also encompass direct interactions between solvent molecules and solutes. The dipole moments μ collected in Table 3.2 demonstrate, for example, that polar solvent molecules bind to solutes through various types of electrostatic interactions (ion–dipole, dipole–dipole, or hydrogen bonding). The Kamlet–Taft parameters π*, α, and β provide an even better insight into the propensity of solvents to engage in such specific interactions [25]. These parameters characterize solvents according to their polarizability (π*), and hydrogen bond donating (α) and accepting (β) ability. The greater the values of α and β, for example, the better the respective solvent molecule is capable of acting as a binding partner in a hydrogen bond. Accordingly, water is a better hydrogen bond donor than acceptor (consistent with the stronger

74

3 Understanding molecular recognition

hydration of anions in water with respect to cations of similar size and charge). DMSO, on the other hand, is an even stronger hydrogen bond acceptor than water but cannot serve as a donor. Tetrahydrofuran and 1,4-dioxane are also strong acceptors, while methanol is a good donor as a result of the acidic proton in the OH group. Hence, solvent molecules have properties similar to those of the binding partners in a complex, allowing them to participate in and thus affect the binding equilibrium. The equilibrium shown in Figure 3.30a, where a single solvent molecule Solv competes with the actual substrate S for the binding site at the receptor R, should serve as an example. At first sight, S should not have great difficulties in displacing Solv from R if the interactions between R and S are stronger than those between R and the solvent molecule. One has to consider the large excess of the solvent molecules, however, which causes them to have a pronounced influence on the equilibrium, even in the case of weak solvation.

(a)

R...Solv + S

(b)

R...S + Solv

O

O N H O O

N H

O +

N H

O +

O

O N H

Figure 3.30: Scheme showing an equilibrium in which a single solvent molecule Solv competes with the substrate S for the binding site of a receptor R (a), and specific example of the competition of 1,4-dioxane in the formation of the acetamide dimer (b).

This solvent effect is relevant, for instance, when two molecules of N-methylacetamide dimerize by the formation of a hydrogen bond between the NH group of one molecule and the C=O group of the other (Figure 3.30b). In tetrachloromethane, this dimer has a Ka of 4.7 M−1 [27]; other authors reported a Ka of 24 M−1 [28]. By contrast, practically no dimerization occurs in 1,4-dioxane. The reason cannot be a medium effect because the two solvents have the same relative permittivity (Table 3.2). Instead, the stability of the N-methylacetamide dimer depends on structural parameters of the solvent: 1,4-dioxane contains oxygen atoms that can engage in hydrogen bonding (β = 0.37) with the NH group of N-methylacetamide, thus interfering in the interactions of two N-methylacetamide molecules. Tetrachloromethane, on the other hand, does not have a large propensity to act as a hydrogen bond acceptor (β = 0.10). Even if the hydrogen bond between 1,4-dioxane and N-methylacetamide is expected to be not as strong as that between two amides (Section 3.1.5), the sheer number of solvent molecules causes the dimerization equilibrium of N-methylacetamide to shift almost completely to the dissociated side in 1,4-dioxane.

3.3 Solvent effects

75

These considerations can be generalized by including solvent molecules in the complexation equilibrium shown in Figure 2.1. The updated Figure 3.31 illustrates that solvent molecules, which occupy the receptor cavity and surround (parts of) the substrate prior to complex formation, are released when the binding partners come together. Despite the large number of solvent molecules involved in this equilibrium, complex formation has a good chance to outperform solvation because the complex is stabilized through multiple interactions distributed over large contact interfaces. If solvation becomes too strong, however, it impairs or even suppresses the formation of the complex.

+

+

Rsolv

+

Ssolv

Csolv

Released solvent molecules

Figure 3.31: Schematic representation of a complexation process that considers the solvation of the substrate S, the receptor R, and the respective complex C. The white circles represent the solvent molecules.

The competition between complex formation and solvation is quantitatively estimated by using equation (2.8) (Section 2.1). In this equation, the overall Gibbs free energy associated with complex formation ΔG0 equals the sum of the intrinsic affinity of the binding partners in the gas phase ΔG0intr and the term ΔG0solv ðCÞ − ΔG0solv ðRÞ − ΔG0solv ðSÞ. This term attributes the energetic effect of solvent reorganization to the difference between the Gibbs free energy gained by solvating the complex and the energies required to desolvate its components. Since solvent reorganization encompasses an enthalpic and an entropic contribution, it is helpful to separate both parameters, which leads to equation (3.8).  0 0 0 ðCÞ − ΔHsolv ðRÞ − ΔHsolv ðSÞ ΔG0 = ΔG0intr + ΔHsolv (3:8)  − T ΔS0solv ðCÞ − ΔS0solv ðRÞ − ΔS0solv ðSÞ The enthalpic terms in this equation describe how the strength of the direct interactions between solvent and solute molecules affect complex stability. If solvation is weak, the overall enthalpic effect of solvent reorganization on complex stability is small. In polar solvents, however, in which the solutes are strongly

76

3 Understanding molecular recognition

0 0 0 solvated, ΔHsolv ðCÞ − ΔHsolv ðRÞ − ΔHsolv ðSÞ can have a large positive value. The reason is that fewer molecules are generally needed to solvate the complex than its com0 ðCÞ cannot ponents, so that the enthalpy gained from solvating the complex ΔHsolv 0 0 compensate the enthalpies ΔHsolv ðRÞ and ΔHsolv ðSÞ associated with the desolvation of the binding partners. The resulting positive enthalpy term thus reduces the absolute value of ΔG0intr . This effect is the reason why, as a rule of thumb, complex stability decreases as the polarity of the solvent rises or, more precisely, as the ability of the solvent molecules to engage in strong and specific interactions with the solutes improves. While the reduction of the number of molecules involved in solvation when going from the individual binding partners to the complex could have an adverse enthalpic effect on stability, it is entropically advantageous. Figure 3.31 shows that complex formation is associated with the transfer of solvent molecules from solvation shells into the bulk. As a consequence, the unfavorable entropy associated with the solvation of the complex ΔS0solv ðCÞ benefits from the favorable desolvation of the receptor and the substrate. The term ΔS0solv ðCÞ − ΔS0solv ðRÞ − ΔS0solv ðSÞ therefore usually returns a positive result, which is beneficial for the overall ΔG0 . While this effect may not be large in unstructured solvents, it becomes decisive when solvation has a strong effect on solvent organization. This is again the case in polar solvents, and a favorable entropic effect of solvent reorganization in these solvents can thus compensate and sometimes even overcompensate an adverse enthalpic term. The situation becomes even more complex when considering that the solvent molecules released upon complex formation subsequently engage in interactions with other solvent molecules in the bulk, which adds further enthalpic and entropic contributions to the binding event (Section 3.4). This brings us to the special case of the solvent water in which the interactions between solvent molecules are sometimes more important for complex formation than the direct interactions between the binding partners.

How does water mediate molecular recognition?

Based on the above considerations, it is not immediately evident how molecular recognition should work in a solvent with the high relative permittivity of water whose molecules furthermore have the pronounced tendency to engage in hydrogen bonding. That water does not prevent the interaction of molecules is obvious because natural systems, for which molecular recognition is absolutely essential, are operative in water. It should be noted that the comparison with Nature is probably not fully appropriate in this context because many binding events in biological systems occur inside of cavities buried within folded proteins, where the relative

3.3 Solvent effects

77

permittivity is lower than in the surrounding aqueous environment. Molecular recognition occurs in the lipophilic environment of the bilayers of membranes or in the cytosol, where the concentration of water is lower than normal. Water molecules are still present under these conditions and can therefore potentially compete. Another aspect worth mentioning is that the ability of molecules to interact in water, even in the absence of a complex biochemical machinery, has likely been an important prerequisite for the development of life on this planet. Molecular recognition in water is therefore not only possible but also highly relevant in different contexts. We will see in Section 4.1 that there are indeed a number of structurally relatively simple receptors that bind to suitable substrates in water by using the rather weak interactions presented above, but to understand the reasons, we first have to look at the properties and structure of water in the liquid phase. In contrast to many other solvent molecules, water molecules are both hydrogen bond acceptors and donors, causing them to bind to each other relatively efficiently, which explains the unusually high boiling point of water with respect to the hydrogen compounds of other elements. The resulting hydrogen bond network is responsible for the ordered structure of ice in which each oxygen atom is surrounded in a tetrahedral fashion by four hydrogen atoms, two of which are bound covalently and two by hydrogen bonds. This order is not completely lost in liquid water in which each water molecule forms on average 3.6 hydrogen bonds to its neighbors at room temperature. Water molecules form hydrogen bonds also to polar and, especially, to ionic 0 < 0), solutes. These interactions typically lead to an enthalpic stabilization (ΔHhydr but the effect on entropy is unfavorable because the water molecules in the hydration shell of these solutes are more ordered than in the bulk. Since enthalpy usually outweighs entropy, a negative Gibbs free energy of hydration ΔG0hydr results, explaining why the strong solvation of polar or ionic substances in water disfavors molecular recognition. This situation changes, however, for solutes whose hydration exhibits a different thermodynamic signature. If, for example, the hydration of a solute features the usual entropic disadvantage but is enthalpically neutral, interactions between such solutes benefit from the entropically favorable release of water molecules. Conversely, if water molecules have to give up hydrogen bonds when hydrating a solute, the release of these water molecules when such solutes engage in interactions leads to a gain in enthalpy. The strong and defined interactions between water molecules in liquid water thus cause the effects of solutes on solvent structure to have much larger enthalpic and/or entropic consequences than in other solvents. Not all details are fully understood yet, but several models emerged in recent years, also driven by intensive research activities in supramolecular chemistry, that help understanding the underlying principles. These models are based on concepts that correlate the influence of the dissolution of a nonpolar organic compound in water, with which water molecules do not

78

3 Understanding molecular recognition

form direct hydrogen bonds, with changes in the water structure. These changes sensitively depend on the sizes and shapes of the dissolved species. As a consequence, the recovery of the original water structure when these molecules form a complex is associated with a characteristic thermodynamic signature [29]. In the case of small (almost) spherical compounds – examples are linear, branched, or cyclic alkanes or benzene derivatives – the creation of a cavity that allows these compounds to be incorporated into the water structure is enthalpically almost neutral or even slightly exothermic. This potentially counterintuitive observation is explained by the ability of the water molecules in the immediate vicinity of sufficiently small solutes to retain all their hydrogen bonds, and the fact that the strength of these bonds even slightly increases upon cavity formation. The increased strength and order of the water molecules lining the cavity causes entropy to decrease, however. Thus, merging two cavities into a slightly larger one, which accommodates two binding partners that were previously hydrated individually, does not cause a large change in enthalpy but a pronounced gain in entropy because it is associated with the release of ordered water molecules into the bulk. This gain in entropy is the main reason why nonpolar solvents and water do not mix. The underlying process is associated with the term “hydrophobic effect” (or solvophobic effect in general). One also finds the term “hydrophobic interactions” in the literature, which is misleading because it suggests that phase separation is caused by direct attractive interactions between nonpolar compounds, which is not the case. Once the nonpolar compounds come together, they engage in van der Waals interactions, but the driving force of aggregation is predominantly the entropic gain associated with water reorganization. This thermodynamic signature changes if the solutes increase in size because the creation of a cavity in water that hosts larger spherical or planar apolar solutes does not allow all surrounding water molecules to retain their hydrogen bonds. In this case, hydration becomes enthalpically unfavorable but has a positive entropic component because the water molecules that cannot form the maximum number of hydrogen bonds gain translational freedom. As a consequence, the solvent reorganization associated with the interaction of large apolar solutes in water is exothermic and entropically unfavorable, opposite to the hydrophobic effect discussed earlier. This combination of enthalpy and entropy is sometimes referred to as the “nonclassical” hydrophobic effect. It should be noted that binding processes involving large ions in water feature analogous thermodynamic signatures for similar reasons, although the orientation of the water molecules surrounding these charged solutes differs from that around nonpolar solutes. The thermodynamically favored desolvation of these ions that drives their complexation has been termed “chaotropic effect” [30]. The relationship of size of the solute and the thermodynamic signature of hydration is summarized in Figure 3.32, which indicates that there should be a crossover

3.3 Solvent effects

Small convex surface

79

Large convex or planar surface

or

Neutral or slightly favorable entropy Unfavorable

enthalpy

Unfavorable Favorable

Figure 3.32: Dependence of the thermodynamic signature of the solvation of apolar solutes in water on size and shape.

point where complexation changes from entropically favored to enthalpically favored as the sizes of the binding partners increase and/or their shapes change. A third case arises for solutes featuring concave surfaces that need to be hydrated in water, for example the inner surfaces of receptor cavities. Again, the ability of a water molecules to retain all or most of their hydrogen bonds or not when solvating a cavity decisively depends on cavity size and shape. Large and shallow cavities can be hydrated without a substantial loss of hydrogen bonds. The release of the included water molecules is therefore mostly entropy driven. There are, however, receptors with relatively small and rigid cavities that accommodate only a few water molecules featuring a lower than optimal number of hydrogen bonds. This cavity water has been termed “high-energy water” because the hydrogen bond deficit is enthalpically costly. The recovery of the hydrogen bonds when releasing these water molecules into the bulk is therefore strongly exothermic (but usually entropically unfavorable) [31]. In Section 4.1.11, we look more closely at receptors whose substrate recognition benefits from this effect. Concluding, apolar solutes produce distinct changes in the bulk structure of water that translate into characteristic effects on enthalpy and entropy. The thermodynamics of binding processes in water thus reflect to a much larger extent than in other solvents the contributions of solvent reorganization associated with complex formation. These contributions can be so pronounced that they cause molecules to aggregate in water that otherwise have no large propensity to interact. The corresponding binding processes range from the formation of structurally defined complexes between complementary binding partners to more complex self-assembled structures such as micelles or vesicles, of which the latter two are not observed outside the aqueous environment.

80

3 Understanding molecular recognition

3.4 Predicting binding strength in solution Can binding strength be predicted?

The above discussion shows that the stability of a complex in solution is determined by a complex interplay of various effects associated with direct interactions between the binding partners, solvation, and solvent reorganization. One could therefore argue that any model attempting to reliably predict binding strength in solution by using a few easily accessible parameters is likely to fail. Christopher A. Hunter showed, however, that this is not the case [32]. His assumptions, approximations, and main results are described in this chapter. Molecular recognition events are treated in Hunter’s model as shown schematically in Figure 3.33a. Accordingly, the formation of a complex R···S between the receptor R and the substrate S requires the solvent molecules to leave the binding sites in R and S, in turn allowing them to interact with themselves (Solv···Solv). The corresponding solvent-solvent interactions extend the reaction shown in Figure 3.30a by the influence of solvophobic interactions that cause the

(a)

(c)

(b) F3C F3C CF3 O H

+

O H9C4 P C4H9 C4H9

R...S + Solv...Solv

F3C CF3 F3C O H O H9C4 P C4H9 C4H9

α > αS

Solute–solvent interactions dominate

Solute–solute interactions dominate

Solvent–solvent interactions dominate

Solute–solvent interactions dominate

αS α

R...Solv + S...Solv

α < αS

0

βS

β < βS

β > βS

β Figure 3.33: Competition of the solvation of the receptor R and the substrate S in a binding equilibrium with the formation of the complex R···S, which considers that the released solvent molecules can also interact with themselves (a). The equilibrium describing the hydrogen bond formation between perfluoro-tert-butyl alcohol and tri-n-butylphosphine oxide is shown in (b), and (c) illustrates that the predominating interactions in solution depend on the ratio of the hydrogen bond donating and accepting ability of the solutes (α and β, respectively), in relation to the corresponding parameters αS and βS of the solvent. The blue quadrants in the diagram show the regions where complex formation is favorable, either because the interactions between R and S are stronger than those to the solvent or because the solvent molecules interact so strongly that they promote R and S to come together. Conversely, the red quadrants denote the regions where the solvation of R and S is so strong that complex formation does not occur.

3.4 Predicting binding strength in solution

81

position of the binding equilibrium to not only depend on the relative strengths of the interactions in R···S, R···Solv, and S···Solv, but also on how strongly the solvent molecules bind to each other. If the solvent molecules prefer to stay together rather than bind to R and S, for example, the association of R and S is promoted by the solvent, which is the hallmark of the hydrophobic effect. Hunter argues that complex formation can be estimated solely on the basis of electrostatic arguments. Dispersion interactions, for example, approximately cancel out in solution, rendering their contribution to the overall binding affinity small. Binding strength can thus be estimated by considering the electrostatic potentials of the binding partners in the regions of the molecules that come into contact in the complex. In the absence of steric restrictions, these molecules are assumed to adopt the optimal mutual arrangement, so that the maxima and minima of the electrostatic potentials determine the interaction strength. All noncovalent interactions are therefore essentially treated as electrostatic interactions between positively and negatively polarized regions of the interacting molecules. This approximation immediately shows a limitation of the method: if similar values of the electrostatic potential are distributed over larger areas of a molecule as in aromatic residues, a structurally less defined complex structure results, rendering the approach unreliable. In cases where it is valid, however, it should be possible to estimate the binding strength on the basis of a set of universal hydrogen bond parameters that describe the hydrogen bond donor and acceptor properties of the binding partners. These parameters must be chosen such that they allow estimating the stability of each species involved in the equilibrium in Figure 3.33a in terms of Gibbs free energies. Further assuming that these energies are additive, the overall ΔG0H − bond of the equilibrium is given by equation (3.9). 


 ΔG0H − bond = − αβ + αS βS + αβS + αS β = − ðα − αS Þ β − βS (3:9) The parameters α and β in this equation denote, respectively, the hydrogen bond donor and acceptor strengths of the molecules involved in complex formation, and αS and βS the corresponding parameters of the solvent molecules. Equation (3.9) thus relates the contribution of the electrostatic interaction to complex formation to the strength of the interactions between R and S (αβ) and between two solvent molecules (αS βS ). Since α and β and the corresponding parameters of the solvent are all positive, the negative sign in front of the first sum in equation (3.9) is required to yield a negative binding energy. The competition of the desolvation of R and S with the interactions in R···S and Solv···Solv is accounted for by the second term in equation (3.9), with the respective products αβS and αS β becoming larger as solvation becomes stronger. It must be stressed that while the meaning of α and β is the same as that of the Kamlet–Taft parameters, the purpose of the two parameters is here to provide information about binding energies and not solvent properties. The values of α and β therefore differ from those in Table 3.2.

82

3 Understanding molecular recognition

According to independent studies, quantitative thermodynamic information can indeed be derived from parameters describing the ability of a molecule to engage in hydrogen bonding. These studies show that the association constants of many hydrogen-bonded complexes follow the general expression (3.10). log Ka = c1 αH2 βH2 + c2

(3:10)

In this equation, αH2 and βH2 again describe the hydrogen bond donor and acceptor properties of the interacting species. The factor c1 reflects the solvent effect on the strength of the interactions. It increases as the polarity of the medium decreases. The constant c2 typically amounts to −1.0 ± 0.1, independent of the solvent and the binding partners. The decrease of log Ka by ca. one order of magnitude caused by c2 , which translates into a ΔG0 value of 6 kJ mol−1 at 298 K, thus seems to be a fundamental property of binding events and has been linked to the energetic cost of bringing two molecules together in solution to form a complex. Although equations (3.9) and (3.10) describe similar aspects, they are not directly comparable. Equation (3.9) only concentrates on the energetic contribution of the hydrogen bonding interactions ΔG0H − bond but does not allow calculating the overall binding strength because the effect of c2 is missing. Conversely, equation (3.10) considers the solvent effect only in terms of the global constant c1 and not by separating the hydrogen bonding donor and acceptor properties of the solvent. Hunter’s approach to combine both equations is based on the experimental values of αH2 and βH2 . These values are converted to the α and β scale such that equation (3.11) allows calculating ΔG0 . The solvent parameters αS and βS are obtained in a similar fashion, assuming that these values do not change substantially when going from dilute solutions in which they were determined to the bulk. For compounds for which αH2 and βH2 are not available, α and β can be estimated computationally from the values of the electrostatic potentials of the respective molecule, allowing also new compounds to be included in the model. ΔG0 = − ðα − αS Þðβ − βS Þ + 6 kJ mol − 1

(3:11)

To confirm that equation (3.11) accurately predicts binding energies, the experimental and calculated stabilities of the complex between perfluoro-tert-butyl alcohol and trin-butylphosphine oxide in 13 different solvents were compared (Figure 3.33b) [33]. The agreement was mostly excellent with only one exception (n-decanol), which was attributed to the fact that the hydrogen bond parameters in this case did not properly reflect the solvent properties. An illustrative example of the usefulness of the method is also the comparison of the experimental stability of the N-methylacetamide dimer in tetrachloromethane and 1,4-dioxane with the respective calculated values (Section 3.3). By using the hydrogen bonding parameters for an amide (α = 2.9, β = 8.3), tetrachloromethane

3.4 Predicting binding strength in solution

83

(αS = 1.4, βS = 0.6), and an alkyl ether (αS = 0.9, βS = 5.3), equation (3.10) affords a ΔG0 of −5.6 kJ mol−1 for the stability of the N-methylacetamide dimer in tetrachloromethane, which translates into a Ka of 9 M−1. In ether, a ΔG0 of zero results, which in very good agreement with the experiments. Equation (3.10) also allows deriving general information about solvent effects on complex stability. The diagram in Figure 3.33c shows how the formation of R···S depends on the hydrogen bond donor and acceptor strength of the solvent. The diagram is divided into four quadrants whose boundaries meet at the values corresponding to αS and βS where no binding occurs (ΔG0 = 6 kJ mol−1). In the region where solutes are better hydrogen bond donors (α > αS ) and better hydrogen bond acceptors (β > βS ) than the solvent, the interactions between R and S are stronger than solvation and complex formation dominates the equilibrium. Conversely, the complex is also formed in the region where the solvent-solvent interactions dominate (α < αS , β < βS ). However, if solutes are better hydrogen bond donors than the solvent (α > αS ) but poorer hydrogen bond acceptors (β < βS ) or vice versa (α < αS , β > βS ), they are too strongly solvated and complex formation is not possible. The exact shapes and sizes of the regions favoring (blue) and disfavoring (red) complex formation depend on the exact values of αS and βS as illustrated in Figure 3.34. Figure 3.34a shows the hypothetical situation in the absence of a solvent (αS = 0, βS = 0). Expectedly, all interactions are attractive under these conditions, but the actual binding strengths strongly depend on the hydrogen bonding properties of the compounds that form the complexes. Selected hydrogen bond donors and acceptors are depicted along the axes of the diagram to illustrate the correlation between hydrogen bonding strength and structure. Accordingly, weakly polarized compounds such as alkanes, alkenes, or aromatic compounds are located at the lower end of the scale, whereas compounds with pronounced dipole moments have large values of α and β and therefore interact stronger. In DMSO (αS = 0.8, βS = 8.9) the diagram is dominated by the top left quadrant (Figure 3.34b). The reason is that most compounds along the top axis are weaker hydrogen bond acceptors than DMSO itself (β < βS ). As a consequence, donors prefer to interact with the solvent. DMSO is thus a good solvent to dissolve compounds that form strongly hydrogen-bonded aggregates in the solid state but a bad one for promoting complex formation by hydrogen bond formation. Figure 3.34c shows that chloroform (αS = 2.2, βS = 0.8) is much better suited in this respect. Because of the low βS value, chloroform does not compete with most hydrogen bond acceptors. Thus, hydrogen bond donors whose α is larger than 2.2 interact with almost all acceptors in chloroform, even the weakest ones. Weak donors are well solvated in chloroform, however, giving rise to the red quadrant in the lower right corner of the diagram. Finally, Figure 3.34d shows the situation in water (αS = 2.8, βS = 4.5). It should be noted that the treatment of water is somewhat more complex and based on an adapted version of equation (3.10), which accounts for the fact that water does not

84

3 Understanding molecular recognition

(a)

(b) 5

5 –45 –35

4

+35 +25

4

–25

+15

3

3

α

α

–15

2

+5

2 –5

1

1

0 0 0 1

2

3

4

(c)

5 β

6

7

8

9 10

0 0

0

1

2

3

4

5 β

(d)

5

6

7

8

9 10

5 –15

–5

4

4

+5

–5

3 2

0

α

α

3 0

2 +5 +5

1

1 +15

+15

0 0

1

2

3

4

5 β

6

7

8

9 10

0 0

1

2

3

4

5 β

6

7

8

9 10

Figure 3.34: Functional group interaction profiles between solutes with different values of α and β in a hypothetical medium with αS = βS = 0 (a), in DMSO (αS = 0.8, βS = 8.9) (b), in chloroform (αS = 2.2, βS = 0.8) (c), and in water (αS = 2.8, βS = 4.5) (d). The contour lines denote the values of ΔG0 in kJ mol−1 at intervals of 1.25 kJ mol−1 with blue regions showing attractive and red regions repulsive solute-solute interactions. Representative structures of hydrogen bond acceptors and donors are shown along the x- and the y-axis, respectively.

interact directly with very apolar solutes but, rather, prefers to include them in a cavity that is lined by water molecules strongly interacting with themselves. The respective diagram is, however, closely related to the three other ones in that there are two red regions in the top left and the bottom right corner where solvation of the binding partners opposes complex formation. Due to the large values of αS and βS , the blue region in the top right corner is small, indicating that only the strongest

3.5 Guidelines for receptor design

85

hydrogen bond donors and acceptors are able to form complexes in water. In contrast to the other diagrams, there is also a substantial blue region in the bottom left corner, consistent with the observation that water mediates the interactions between apolar molecules as a consequence of the hydrophobic effect. This model thus provides deep insight into some of the most important factors that control complex formation in solution. The fact that even quantitative predictions about binding strength can be made by attributing complex formation to electrostatic interactions alone is the reason why the discussion of the different types of interactions in Section 3.1 concentrates on the contributions of these polar effects. With its recent enhancements and refinements that extend the treatment to charged binding partners, for example, Hunter’s model has developed into a valuable tool for understanding molecular recognition processes and how they are influenced by the solvent.

3.5 Guidelines for receptor design Which strategies exist to achieve strong binding?

At this point, we have seen that various types of intermolecular interactions cause molecules to stay together. We have also learned about effects that influence binding strength and about the impact of the solvent on complex formation. In these contexts, we have already seen a few examples of supramolecular receptors, cyclic and acyclic ones, some with binding sites along the ring and some within appended substituents. Obviously, receptors come in widely different shapes and structures, but their design is usually based on a set of general guidelines. The most important ones are outlined in the following chapters.

3.5.1 Complementarity, preorganization, and induced fit A good starting point to derive an important concept that often underlies receptor design is Figure 3.1, illustrating schematic structures of potential receptor–substrate combinations. Of the four scenarios shown in this figure, the first one is normally most efficient. The reason is that only in this case, the receptor cavity and the substrate are perfectly complementary in size and shape. If both binding partners are additionally complementary in their electronic properties, stronger binding is expected than in situations where the shape of the substrate does not match the shape of the cavity, or where the substrate is too large or too small. Size, shape, and electronic complementarity of the binding partners are thus decisive factors for a stable complex.

86

3 Understanding molecular recognition

The concept of complementarity finds its roots in the lock-and-key principle formulated by Emil Fischer in 1894. Fischer used this analogy to rationalize the substrate selectivity of enzymes by proposing that the substrate fits the binding pocket of the enzyme like a key its lock. Fischer wrote “Um ein Bild zu gebrauchen, will ich sagen, dass Enzym und Glucosid wie Schloss und Schlüssel zu einander passen müssen, um eine chemische Wirkung auf einander ausüben zu können” [To use a picture, I would like to say that enzyme and glucoside have to fit together like lock and key to have a chemical effect on each other.] [34]. The idea is not restricted to biological molecular recognition processes but is also applicable to receptors in supramolecular chemistry. It is illustrated in Figure 3.35, which shows that the shape complementarity of the binding partners in (a) should allow complex formation, whereas the lack of complementarity in (b) should prevent the complex from being formed.

(a)

(b)

Perfect shape complementarity

No shape complementarity

Lock-and-key fit

Induced fit

Figure 3.35: Schematic illustration of the lock-and-key and induced fit concepts. The binding partners in (a) are perfectly shape complementary and should thus be able to form a complex according to the lock-and-key principle, whereas no binding is expected for the binding partners in (b). Complex formation still occurs by an induced fit if the receptor is able to adapt to the structure of the substrate. Nevertheless, the receptor is better preorganized for the substrate in (a) than in (b).

Although intriguing, comparing molecules with locks and keys is not entirely appropriate because molecules are never as rigid as a key. Indeed, Daniel E. Koshland Jr. showed that the active sites of enzymes must not be shape complementary to the substrates (or more specifically to the transition state of the reaction that they catalyze) prior to binding. Instead, it needs to be able to adapt to the structure of the substrate by a conformational reorganization. This process was likened by Koshland to the fitting of a hand in a glove. An empty glove adopts all kinds of shapes but can still distinguish the right from the left hand if it is used [35]. The induced fit model introduced by Koshland therefore provides a more realistic view of a binding process, independently of whether proteins or small receptors are involved. Figure 3.35b shows that it allows complex formation even if the empty cavity does not match the shape of the substrate.

3.5 Guidelines for receptor design

87

The ability of receptors to adapt to the structural requirements of the substrate is actually an advantage. If the receptor is completely rigid, it needs to be perfectly shape-complementary for optimal binding. Even small deviations from the optimal structure lead to a drop in affinity. More flexible receptors, on the other hand, respond to these deviations by optimizing substrate binding and thus reinforcing the complex. If the corresponding conformational reorganization is not too extensive, the enthalpy gained by the structural adjustments easily overcompensates the adverse entropy term associated with the loss of conformational freedom. If, however, the conformations of the receptor in the absence and the presence of the substrate differ substantially, complex stability suffers. The reason is mainly entropic as complex formation is associated with a significant loss of conformational flexibility. An additional adverse enthalpic term arises if the conformation of the receptor in the complex is strained or otherwise disfavored. Donald J. Cram realized how strongly structural changes in the binding partners caused by complex formation affect complex stability and formulated the principle of preorganization, which states that “the more highly hosts and guests are organized for binding […] prior to their complexation, the more stable will be their complexes” [36]. Combining the three concepts of complementarity, induced fit, and preorganization thus leads to the first guideline for receptor design: A receptor should be complementary in size, shape, and electronic properties to the substrate and it should also have a certain flexibility to be able to optimize the interactions in the complex. A receptor should, however, not be too flexible, i.e., it should be sufficiently well preorganized to minimize the structural adjustments necessary for substrate binding.

3.5.2 Chelate effect and macrocyclic effect Receptor–substrate interactions generally do not rely on a single point of contact but on a combination of several interactions distributed over a larger contact interface. The overall efficiency of binding actually greatly benefits from these multiple contacts since a hypothetical complex in which the substrate interacts with an equivalent number of isolated binding partners is significantly less stable. An example from coordination chemistry should illustrate this point. The formation of the tetraamminecopper(II) complex [Cu(NH3)4]2+ in water is a stepwise process in which water molecules are progressively replaced from the copper center by ammonia molecules. The overall equilibrium is characterized by a log Ka of 12.6. The same reaction with the linear tetraamine 3.12 (Figure 3.36a) has an impressive eight orders of magnitude larger log Ka of 20.9 [37]. The reason cannot primarily relate to the intrinsic strength of the underlying metal–nitrogen interactions, which is likely

88

3 Understanding molecular recognition

(a) [Cu(H2O)6]2+ +

log Ka = 12.6 4 NH3 N H

HN [Cu(H2O)6]

2+

log Ka = 20.9

[Cu(NH3)4]2+

H N

HN

NH

HN Cu2+

+

NH

HN

3.12 NH

HN [Cu(H2O)6]2+ +

log Ka = 24.8

NH

HN Cu

NH

HN

2+

NH

HN

3.13 (b) O

O

O

O

log Ka = 2.1

K+ + O

O

O K+

O

O O

O

O

3.14 O

O O

O K+ +

O

O O

log Ka = 6.1

O

O K+

O

O O

3.15 Figure 3.36: Comparison of the coordination of ammonia, the linear tetraamine 3.12, and the cyclic tetraamine 3.13 to [Cu(H2O)6]2+ (a), and of the podand 3.14 and crown ether 3.15 to K+ (b). Coordinated water molecules in the products are not shown for reasons of clarity. The binding constants of the podand and crown ether complexes were determined in methanol at 25 °C [38].

comparable in both complexes. The difference lies in the entropy. In the case of [Cu (NH3)4]2+, each step of complex formation comprises the attachment of a new ammonia molecule. The associated entropy of restricting the translational degrees of freedom of the ligand therefore has to be paid four times. By contrast, the entropic term associated with bringing two molecules together has to be invested only once when Cu2+ (or more precisely [Cu(H2O)6]2+) interacts with the first nitrogen atom in ligand 3.12. Each subsequent coordination step is an intramolecular reaction, opposed in entropy only by the restriction of the conformational mobility of the ligand. This effect, termed chelate effect, generally causes the complexes between

3.5 Guidelines for receptor design

89

metal centers and multidentate ligands to be more stable than the complexes between the same metal and the corresponding number of monodentate ligands. Figure 3.36a shows that the coordination of ligand 3.12 still suffers from the fact that this ligand has to fold around the metal center. One could say in the terminology of supramolecular chemistry that 3.12 is not well preorganized for complex formation. A way to address this issue involves using tetraamine 3.13 as ligand, the cyclic analog of 3.12. In 3.13, the cost of the restriction of conformational mobility and the stabilization of a folded conformation required for metal coordination has already been paid for during the cyclization reaction, rendering the complex of this ligand with Cu2+ ca. four orders of magnitude more stable than that of 3.12 (log Ka = 24.8) [37]. The stabilization of the complex caused by the use of a cyclic ligand has been termed macrocyclic effect. This effect is usually even more pronounced for macrobicyclic ligands (Section 4.1.2), where the term macrobicyclic effect is used. These principles are not restricted to coordination chemistry but also apply to supramolecular chemistry. The complex of the podand 3.14 with K+ is, for example, significantly less stable than that of the crown ether 3.15 (Figure 3.36b) as a consequence of the macrocyclic effect. It should be stressed that chelate effects act in almost every supramolecular complex because of the distribution of different types of interactions such as ion–dipole interactions, hydrogen bonding, and dispersion interactions across the whole contact interfaces of receptor–substrate complexes. Dispersion interactions may, for instance, be too weak to stabilize a complex alone, but if stronger interactions bring the binding partners together, weaker ones contribute to the overall stability. In terms of thermodynamics, chelate and macrocyclic effects are also not always entropic in origin. Solvation effects or unfavorable conformations of the receptor in the absence of the substrate sometimes also cause enthalpy to dominate. We can anyway conclude: Multidentate interactions within supramolecular complexes have a positive effect on stability and combining this chelate effect with preorganization is particularly advantageous.

3.5.3 Multivalency and cooperativity Multivalency and cooperativity are concepts that refer to complexes stabilized by multiple receptor substrate interactions. In contrast to the complexes discussed in the previous chapter, which were stabilized by interactions at a single common interface, multivalent systems comprise at least two but usually more individual and typically clearly separated binding sites within each binding partner. Complex

90

3 Understanding molecular recognition

formation thus occurs between a receptor with several recognition units that can but not need be identical and a complementary polyfunctional substrate. This situation is illustrated in Figure 3.37.

(a)

(b)

(c)

Figure 3.37: Schematic illustration of a monovalent interaction (a) in comparison to divalent (b) and trivalent (c) interactions between receptors containing two or three binding sites and substrates with the equivalent number of complementary subunits.

Complexes stabilized by multivalent interactions are more stable than monovalent ones. This trend is closely related to the chelate effect, which also involves multiple points of attachment. There is, however, an important difference. The chelate effect describes the increase in binding strength caused by replacing a certain number of monodentate binding partners in a complex with a multidentate analog that engages in the equivalent number of interactions. To assess the magnitude of the chelate effect, we thus (somewhat inappropriately because of the different stoichiometries) compare the stability of a higher order complex such as [Cu(NH3)4]2+ with that of the corresponding 1:1 analog [Cu·3.12]2+. By contrast, in the case of multivalency, we compare the stability of a monovalent complex, depicted schematically in Figure 3.37a, with that of a multivalent counterpart (Figure 3.37b,c). Both complexes have the same stoichiometry and the binding constants are therefore directly comparable. The increase of the stability of a multivalent complex with respect to the monovalent analog simply reflects the fact that the multivalent system is stabilized by more than one point of attachment, with each attachment point contributing to the overall stability with an individual Gibbs free energy contribution. The overall stability of theX multivalent complex thus derives from the sum of the individualY binding N N 0 ΔGi , or the product of the respective association constants Ki . energies i=1 i=1 a As a consequence, multivalent interactions are always stronger than monovalent ones even if not every interaction contributes equally to the overall binding strength. To differentiate the stabilities of complexes stabilized by a monovalent interaction or by multivalent interactions, one uses the term affinity to describe the strength of the monovalent complex and avidity for the multivalent one. This brings us to the concept of cooperativity. In order to understand what cooperativity means, one has to consider that multivalent complexes are formed in a stepwise manner. The formation of the complex shown in Figure 3.37b, for example, involves the initial binding of one subunit of the receptor to one subunit of the

91

3.5 Guidelines for receptor design

substrate, affording the partially bound acyclic intermediate (RR·SS)o (Figure 3.38). This intermediate then cyclizes to yield the cyclic divalent complex (RR·SS)c and if this cyclization does not proceed efficiently, higher complexes such as RR·(SS)2 are also formed. Polymeric complexes with multiple receptors binding to multiple substrates are also possible under certain conditions, but the formation of these complexes can be neglected if SS is present in excess so that the equilibrium simplifies to the situation shown in Figure 3.38.

(RR.SS)o

SS

RR

SS

RR.(SS)2

4 Ka + 2

+

1/2 K a EM

+

(RR.SS)c

SS

Figure 3.38: Stepwise formation of the cyclic complex (RR·SS)c between a divalent receptor RR and a divalent substrate SS via the acyclic intermediate (RR·SS)o. In the case of high chelate cooperativity, the cyclic product dominates in the equilibrium. If cooperativity is low, however, also higher complexes such as RR·(SS)2 are formed to a substantial extent.

The ratio of (RR·SS)o and (RR·SS)c in this equilibrium depends on how strongly the formation of the cyclic complex is favored over the intermolecular interactions. If the cyclic complex forms very efficiently once (RR·SS)o is present, the cooperativity of the overall process is positive. It is negative if the cyclic complex does not form, for example because the tether linking the two binding sites in the substrate does not allow the formation of (RR·SS)c. In this case, RR·(SS)2 becomes the favored product. To assess cooperativity, one therefore has to evaluate the strength of the intramolecular binding event relative to the intermolecular one. This strategy is comparable to the treatment of the chelate effect, explaining why the term chelate cooperativity is used to describe the behavior of such systems [39, 40]. The effect of chelate cooperativity on the equilibria in Figure 3.38 is estimated on the basis of the following considerations. Assuming that the intrinsic interaction between a monovalent receptor and a monovalent substrate is associated with the

92

3 Understanding molecular recognition

stability constant Ka , the law of mass action describing the formation of the acyclic complex (RR·SS)o from RR and SS can be rearranged to afford equation (3.12). cðRR · SSÞo = 4 Ka cRR cSS

(3:12)

The statistical factor 4 in this equation arises from the fact that a single interaction between RR and SS can occur in four different ways as shown in Figure 3.39a.

(a)

S2

S1

S1

R1

S2

S2

R2

R1

R2

R1

S1

S1

S2

R2

R1

R2

(b) S1

S2

S2

S1

R1

R2

R1

R2

Figure 3.39: Possibilities of a ditopic receptor RR to interact with a ditopic substrate SS (b). The receptor subunits R1 and R2 are identical as are the substrate subunits S1 and S2. The differentiation is only made to illustrate that there are four ways in which these binding partners interact, leading eventually to indistinguishable complexes (RR·SS)o. The four complexes in (a) converge to give two possible arrangements of the closed complex (RR·SS)c as shown in (b).

The equilibrium between the open complex (RR·SS)o and its cyclized form is then given by equation (3.13) in which Ka is again the intrinsic intermolecular binding constant of the respective monovalent complex, and 1/2 the statistical factor for the cyclization during which the four possible combinations of the intermediate converge to afford two microspecies of the product (Figure 3.39b). 1 cðRR · SSÞc = Ka EM cðRR · SSÞo 2

(3:13)

Since the formation of the cyclic from the open complex is an intramolecular process, the actual equilibrium constant is dimensionless. Relating this reaction to Ka thus requires the introduction of a factor with the dimension of a concentration to render the equation correct. This factor is termed effective molarity EM. It is a measure of the extent to which the intramolecular process is favored over the

3.5 Guidelines for receptor design

93

intramolecular one, which becomes evident by looking at the rearranged equation (3.14). cðRR · SSÞc 1 = Ka EM cðRR · SSÞo 2

(3:14)

The ratio of the cyclic and the open form of the complex is thus given by the product of the intrinsic stability of the monovalent complex RS and EM. The larger EM, the larger the amount of the cyclic complex in the equilibrium. This effective molarity can be associated with the substrate concentration that has to be reached until the intermolecular complex outcompetes the intramolecular complex. The product Ka EM in equation (3.14) is a measure for cooperativity. If Ka EM ≪ 1, the partially bound complex is more stable than the cyclic one. In this case, the overall equilibrium is unaffected by the cyclic complex and the system behaves as if the substrate would be monovalent. Positive cooperativity is observed if Ka EM ≫ 1, causing (RR·SS)c to be the major species over a wide concentration range. The strong coupling of the individual binding events in the case of high chelate cooperativity thus leads to an all-or-nothing behavior, with the equilibrium either featuring the fully dissociated binding partners or the fully formed complex but no partially bound complexes. A typical example of such a system is the DNA double helix, which is stabilized by multiple interactions between the individual polynucleotide strands. Since every interaction strongly favors the subsequent one, only the fully formed double helix is present under normal conditions. Conversely, when DNA is heated in solution, strand dissociation occurs within a relatively narrow temperature range in which all interactions simultaneously break. The structural integrity of multivalent complexes featuring chelate cooperativity nevertheless depends on the relative concentrations of the binding partners. If the substrate concentration exceeds EM, the 1:1 complexes dissociate and higher complexes are formed. With regard to receptor design, one can therefore state that: Multivalent interactions are always stronger than monovalent ones. If complex formation furthermore features positive chelate cooperativity, partially bound species are usually underrepresented in the binding equilibrium, while the fully bound complex dominates.

3.5.4 Allosterism and cooperativity Allosteric behavior is another feature that is closely linked to receptors with more than one binding site. In this case, however, the receptors do not interact with multivalent substrates but with several identical or different monovalent ones. The underlying concept comes from biochemistry where it describes the control of enzymatic activity by compounds, so-called effectors, that bind to enzyme pockets different from the actual active sites. Effector complexation typically triggers a conformational change of

94

3 Understanding molecular recognition

the enzyme that either results in the activation of enzymatic activity or the shutdown. In the first case, the effector molecule acts as an activator and in the second case as an inhibitor. In supramolecular chemistry, multivalent receptors exhibit allosteric behavior if the interaction with the first substrate affects the affinity of the other binding site(s). Such processes therefore also comprise an element of cooperativity, which in this case describes the extent to which the different intermolecular binding steps influence each other. If the binding of the first substrates favors the subsequent binding steps, the system exhibits positive cooperativity, whereas negative cooperativity occurs if the first binding step weakens the subsequent interactions. To differentiate this behavior from the chelate cooperativity described in the previous chapter, which involves only one intermolecular process followed by intramolecular ones, the term allosteric cooperativity is used [39, 40]. A prototypic example of a natural system exhibiting positive allosteric cooperativity is hemoglobin, where the binding of each oxygen molecule promotes the binding of the next one until all four binding sites are occupied. Strong positive cooperativity therefore results in an all-or-nothing behavior also in the case of allosteric systems, with the equilibrium featuring a sharp transition from the fully dissociated to the fully bound complex. In order to assess allosteric cooperativity, one has to distinguish the microscopic association constants, characterizing the individual interactions at each receptor subunit, and the experimental association constants that additionally include the statistical degeneracy of the system in the case of identical binding sites. This aspect should be illustrated by using the example of a divalent receptor RR that interacts with two substrate molecules S shown in Figure 3.40.

RR .S

RR S

RR .S2 S

K a1 (exp)

+ 2

K a2 (exp)

+ 2 K a11

1/2 K a12

Figure 3.40: Stepwise complexation of two substrate molecules S by a divalent receptor RR. The microscopic association constants Ka11 and Ka12 describe the intrinsic strength of each binding step and the statistical factors 2 and 1/2 reflect the degeneracy of the partially bound intermediate.

The corresponding equilibrium comprises two steps, the initial formation of the partially bound intermediate RR·S, whose formation is associated with the microscopic association constant Ka11 , and the subsequent formation of the fully formed complex RR·S2, characterized by Ka12 . Since the interaction of a monovalent substrate S with the divalent receptor RR gives rise to two microspecies in which the substrate binds

3.5 Guidelines for receptor design

95

to either the first or the second binding site of the receptor, a statistical factor of 2 has to be introduced when correlating the experimental stability constant Ka1 ðexpÞ with Ka11 as shown in Figure 3.40. Both microspecies then yield the same product, so that Ka2 ðexpÞ of the second step amounts to 1/2 Ka12 . The cooperativity of the overall process depends on the direction and the extent to which the presence of S in the partially bound intermediate influences the strength of the interactions in the second step. The parameter used to describe this effect is the interaction parameter α, which is the ratio of the two microscopic association constants according to equation (3.15). α=

Ka12 Ka11

(3:15)

In the absence of cooperativity, the intrinsic interactions at both binding sites have the same strength, which means that Ka11 = Ka12 and α = 1. Because of the statistical factors that correlate the microscopic and experimental binding constants, this relationship translates into 0.5 Ka1 ðexpÞ = 2 Ka2 ðexpÞ or Ka1 ðexpÞ = 4 Ka2 ðexpÞ. The experimental stepwise binding constants of the equilibria shown in Figure 3.40 thus differ by a factor of 4 for a noncooperative system and any deviation from this ratio indicates cooperative behavior. If α < 1, the interactions in the intermediate RR·S are stronger than in the fully bound state RR·S2. RR·S2 is therefore only populated to a significant extent if S is present in excess. If complex formation is positively cooperative (α > 1), however, there is a substantial thermodynamic driving force for the binding of the second substrate molecule once the intermediate complex has been formed. The curves in Figure 2.3 illustrate how the substrate concentrations affect the extent of complex formation in the case of negative and positive cooperativity. If α is very large, the intermediate state is never populated and the system features all-or-nothing behavior. Whether a system exhibits allosteric cooperativity is estimated by using the socalled Hill plot. The respective treatment is based on the idea that complex formation involves a single equilibrium characterized by the binding of a multivalent receptor to n guests. The corresponding law of mass action is given by equation (3.16a). Rearranging this equation yields (3.16b). Ka = Ka cnS =

cC cR cnS c0R

cC − cC

(3:16a) (3:16b)

96

3 Understanding molecular recognition

Expressing the ratio cC =c0R as the degree of saturation Θa then leads to 3.16c after further rearrangement.   Θa = log Ka + n log cS (3:16c) log 1 − Θa Plotting log½Θa =ð1 − Θa Þ vs. log cS yields a graph, the Hill plot, whose slope n at 50% saturation, that is, at log½Θa =ð1 − Θa Þ = 0, provides information about the cooperativity of the reaction. If the Hill coefficient n amounts to 1, complex formation is noncooperative. Negative cooperativity (α < 1) is reflected by a Hill coefficient n < 1 and positive cooperativity (α > 1) by n > 1. To illustrate the shapes of such graphs, Hill plots are depicted in Figure 3.41 that describe the complex formation of a ditopic receptor such as the one shown in Figure 3.40 [39]. 3

log (θa/(1‒θa))

2 1 0

0

1

2 3 log (K'cS )

Figure 3.41: Hill plots describing the complex formation of a ditopic receptor if substrate binding proceeds with no cooperativity (α = 1) (black), with positive cooperativity (α = 100) (red), or with negative cooperativity (α = 0.01) (blue). The graphs show that positive cooperativity leads to a slope larger than unity (n > 1) at log½Θa =ð1 − Θa Þ = 0, while negative cooperativity leads to a slope smaller than unity (n < 1). The use of log ðK′cS Þ as the x-axis, where
0.5 causes the curves to pass through the K′ = Ka11 Ka12 origin for different values of α.

An allosteric behavior of supramolecular receptors can be realized in different ways. The one most closely resembling the allosteric behavior of natural systems relies on conformationally coupling two or more binding sites within a receptor [41]. Possible situations are illustrated in Figure 3.42. These examples illustrate the positive and negative allosteric behavior of receptors with different binding sites (heterotopic systems). If the binding of the first substrate causes a conformational reorganization of the receptor that improves the preorganization of the other binding site, the receptor exhibits positive cooperativity as in Figure 3.42a. Conversely, the stabilization by the first substrate of a receptor conformation that is unsuitable for the uptake of the second substrate results in negative cooperativity (Figure 3.42b). Cooperativity is further mediated by interactions between the bound substrate molecules. If the two substrates are oppositely charged, for example, attractive electrostatic interactions between the bound ions stabilize the final complex, typically resulting in strong positive cooperativity as in some ion-pair receptors (Section 4.2.3). Since this mechanism of cooperativity differs from allosteric

97

3.5 Guidelines for receptor design

(a)

Binding of

Favors binding of

(b) Binding of

Inhibits binding of

(c) N N

O

O O

O K+

O

K+

O N

N

O

O

O Crown ether conformation suitable for potassium complexation

3.16

O

W(CO)6

OC OC N OC W N OC

O

O O

O

O

Crown ether conformation less suitable for potassium complexation

Figure 3.42: Schematic representation of the modes of binding of allosteric receptors exhibiting positive heterotopic (a) and negative heterotopic (b) cooperativity. The allosteric receptor 3.16 depicted in (c) has a binding site for a transition metal ion and for an alkali metal ion, with the conformational coupling of the two sites causing the coordination of the bipyridine unit to a transition metal ion to weaken the binding of the potassium ion to the crown ether moiety.

cooperativity, which involves conformationally coupled binding sites, it is associated with the term intermolecular cooperativity [42]. Homotopic systems feature two or more binding sites for the same substrate but behave otherwise in a similar fashion. Receptor 3.16 is a classic example of an allosteric system exhibiting negative heterotopic cooperativity [43]. This receptor has two different binding sites, the 2,2ʹ-bipyridyl subunit for the complexation of a transition metal ion, and the crown ether moiety that can accommodate a potassium ion. In the absence of either substrate, 3.16 prefers a conformation with diverging nitrogen atoms at the 2,2ʹ-bipyridyl unit to minimize the repulsion of the lone pairs. As a consequence, the crown ether adopts a distorted conformation, unsuitable for K+ complexation. The addition of potassium ions induces a conformational change, allowing complex formation to occur. In the resulting receptor conformation, the two nitrogen atoms are oriented toward one another, but further rigidifying this arrangement by coordination to W(CO)6 induces a crown ether conformation, which is again

98

3 Understanding molecular recognition

unfavorable for potassium complexation. The transition metal ion thus acts as an allosteric inhibitor. One can conclude: The use of allosteric effects is an attractive means to control the substrate affinity of synthetic receptors with suitable effector molecules.

Bibliography [1] [2]

[3]

[4]

[5]

[6] [7]

[8] [9] [10] [11] [12] [13]

[14] [15]

[16]

Marcus Y. Ion Properties. Marcel Dekker: New York, 1997. Arunan E, Desiraju GR, Klein RA, Sadlej J, Scheiner S, Alkorta I, Clary DC, Crabtree RH, Dannenberg JJ, Hobza P, Kjaergaard HG, Legon AC, Mennucci B, Nesbitt DJ. Definition of the hydrogen bond (IUPAC Recommendations 2011). Pure Appl. Chem. 2011, 83, 1637–41. Schneider HJ, Hydrogen bonds with fluorine. Studies in solution, in gas phase and by computations, conflicting conclusions from crystallographic analyses. Chem. Sci. 2012, 3, 1381–94. Djurdjevic S, Leigh DA, McNab H, Parsons S, Teobaldi G, Zerbetto F. Extremely strong and readily accessible AAA−DDD triple hydrogen bond complexes. J. Am. Chem. Soc. 2007, 129, 476–7. Kelly TR, Bridger GJ, Zhao C. Bisubstrate reaction templates. Examination of the consequences of identical versus different binding sites. J. Am. Chem. Soc. 1990, 112, 8024–34. Hamilton AD, Van Engen D. Induced fit in synthetic receptors: nucleotide base recognition by a molecular hinge. J. Am. Chem. Soc. 1987, 109, 5035–6. Desiraju GR, Ho S, Kloo L, Legon AC, Marquardt R, Metrangolo P, Politzer P, Resnati G, Rissanen K. Definition of the halogen bond (IUPAC Recommendations 2013). Pure Appl. Chem. 2013, 85, 1711–3. Bulfield D, Huber SM. Halogen bonding in organic synthesis and organocatalysis. Chem. Eur. J. 2016, 22, 14434–50. Sarwar MG, Dragisic B, Sagoo S, Taylor MS. A tridentate halogen-bonding receptor for tight binding of halide anions. Angew. Chem. Int. Ed. 2010, 49, 1674–7. Ma JC, Dougherty DA. The cation–π interaction. Chem. Rev. 1997, 97, 1303–24. Biedermann F, Schneider HJ. Experimental binding energies in supramolecular complexes. Chem. Rev. 2016, 116, 5216–300. Gokel GW, Barbour LJ, Ferdani R, Hu J. Lariat ether receptor systems show experimental evidence for alkali metal cation–π interactions. Acc. Chem. Res. 2002, 35, 878–86. Wheeler SE, Houk KN. Substituent effects in cation/π interactions and electrostatic potentials above the centers of substituted benzenes are due primarily to through-space effects of the substituents. J. Am. Chem. Soc. 2009, 131, 3126–7. Mecozzi S, West Jr. AP, Dougherty DA. Cation–π interactions in simple aromatics: electrostatics provide a predictive tool. J. Am. Chem. Soc. 1996, 118, 2307–8. Kearney PC, Mizoue LS, Kumpf RA, Forman JE, McCurdy A, Dougherty DA. Molecular recognition in aqueous media. New binding studies provide further insights into the cation–π interaction and related phenomena. J. Am. Chem. Soc. 1993, 115, 9907–19. Albrecht M, Müller M, Mergel O, Rissanen K, Valkonen A. CH-directed anion–π interactions in the crystals of pentafluorobenzyl-substituted ammonium and pyridinium salts. Chem. Eur. J. 2010, 16, 5062–9.

Bibliography

99

[17] Giese M, Albrecht M, Krappitz T, Peters M, Gossen V, Raabe G, Valkonen A, Rissanen K. Cooperativity of H-bonding and anion–π interaction in the binding of anions with neutral π-acceptors. Chem. Commun. 2012, 48, 9983–5. [18] Berryman OB, Hof F, Hynes MJ, Johnson DW. Anion–π interactions augments halide binding in solution. Chem. Commun. 2006, 506–8. [19] Martinez CR, Iverson BL. Rethinking the term “pi-stacking”. Chem. Sci. 2012, 3, 2191–201. [20] Odell B, Reddington MV, Slawin AMZ, Spencer N, Stoddart JF, Williams DJ. Cyclobis(paraquatp-phenylene). A tetracationic multipurpose receptor. Angew. Chem. Int. Ed. Engl. 1988, 27, 1547–50. [21] Houk KN, Leach AG, Kim SP, Zhang X. Binding affinities of host-guest, protein-ligand, and protein-transition-state complexes. Angew. Chem. Int. Ed. 2003, 42, 4872–97. [22] Cockroft SL, Hunter CA. Chemical double-mutant cycles: dissecting noncovalent interactions. Chem. Soc. Rev. 2007, 36, 172–88. [23] Mati IK, Cockroft SL. Molecular balances for quantifying noncovalent interactions. Chem. Soc. Rev. 2010, 39, 4195–205. [24] Fischer FR, Schweizer WB, Diederich F. Substituent effects on the aromatic edge-to-face interaction. Chem. Commun. 2008, 4031–3. [25] Marcus Y. The properties of organic liquids that are relevant to their use as solvating solvents. Chem. Soc. Rev. 1993, 22, 409–16. [26] Marcus Y. The properties of solvents. John Wiley & Sons: Chichester, 1998. [27] Klotz IM, Franzen JS. Hydrogen bonds between model peptide groups in solution. J. Am. Chem. Soc. 1962, 84, 3461–6. [28] Krikorian SE. Determination of dimerization constants of cis- and trans-configured secondary amides using near-infrared spectrometry. J. Phys. Chem. 1982, 86, 1875–81. [29] Hillyer MB, Gibb BC. Molecular shape and the hydrophobic effect. Annu. Rev. Phys. Chem. 2016, 67, 307–29. [30] Assaf KI, Nau WM. The chaotropic effect as an assembly motif in chemistry. Angew. Chem. Int. Ed. 2018, 57, 13968–13981. [31] Biedermann F, Nau WM, Schneider, HJ. The hydrophobic effect revisited – studies with supramolecular complexes imply high-energy water as a noncovalent driving force. Angew. Chem. Int. Ed. 2014, 53, 11158–71. [32] Hunter CA. Quantifying intermolecular interactions: guidelines for the molecular recognition toolbox. Angew. Chem. Int. Ed. 2004, 43, 5310–24. [33] Cook JL, Hunter CA, Low CMR, Perez-Velasco A, Vinter JG. Solvent effects on hydrogen bonding. Angew. Chem. Int. Ed. 2007, 46, 3706–9. [34] Fischer E. Einfluss der Configuration auf die Wirkung der Enzyme. Ber. Dtsch. Chem. Ges. 1894, 27, 2985–93. [35] Koshland Jr. DE. The key-lock theory and the induced fit theory. Angew. Chem. Int. Ed. Engl. 1994, 33, 2375–8. [36] Cram DJ. Preorganization – from solvents to spherands. Angew. Chem. Int. Ed. Engl. 1986, 25, 1039–57. [37] Clay RM, Corr S, Micheloni M, Paoletti P. Non-cyclic reference ligands for tetraaza macrocycles. Synthesis and thermodynamic properties of a series of α,ω-di-N-methylated tetraaza ligands and their copper(II) complexes. Inorg. Chem. 1985, 24, 3330–6. [38] Haymore BL, Lamb JD, Izatt RM, Christensen JJ. Thermodynamic origin of the macrocyclic effect in crown ether complexes of sodium(1+), potassium(1+), and barium(2+). Inorg. Chem. 1982, 21, 1598–602. [39] Hunter CA, Anderson HL. What is cooperativity? Angew. Chem. Int. Ed. 2009, 48, 7488–99.

100

3 Understanding molecular recognition

[40] Ercolani G, Schiaffino L. Allosteric, chelate, and interannular cooperativity: a mise au point. Angew. Chem. Int. Ed. 2011, 50, 1762–8. [41] Kremer C, Lützen A. Artificial allosteric receptors. Chem. Eur. J. 2013, 19, 6162–296. [42] von Krbek LKS, Schalley CA, Thordarson P. Chem. Soc. Rev. 2017, 46, 2622–37. [43] Rebek Jr. J, Wattley RV. Allosteric effects. Remote control of ion transport selectivity. J. Am. Chem. Soc. 1980, 102, 4853–4.

4 Hosting ions and molecules CONSPECTUS: Now that we understand the principles behind noncovalent interactions and the methods to characterize complexes formed between complementary binding partners, it is time to look at the actual types of receptors commonly found in supramolecular chemistry. So many receptors have been described in the past 50 years that a comprehensive overview is clearly impossible. Instead, we focus on the most important receptor families and learn about their basic structures and properties. In the last part of this chapter, we reverse the viewpoint and look at complex formation from the perspective of the substrate, which will allow us to derive guidelines how to design a receptor for a given target.

4.1 Receptors 4.1.1 Crown ethers Introduction: Crown ethers are macrocyclic polyethers, typically but not necessarily derived from polyethylene glycol. They were discovered by Charles J. Pedersen whose first paper about crown ethers dealt with the syntheses and properties of almost 50 derivatives [1]. We saw in Chapter 1 that the first member of this rapidly growing family of receptors, dibenzo-18-crown-6 4.1 (Figure 4.1), was a chance discovery. Having isolated a small amount of this side product in the synthesis of bis [2-(2-hydroxyphenoxy)ethyl] ether, Pedersen set out to elucidate its structure. In one experiment, he checked whether the isolated compound contained free phenolic OH groups by measuring the UV–vis spectra of a solution in methanol before and after the addition of NaOH. To his surprise, both spectra were identical, showing that the product did not contain an acidic proton. The addition of NaOH did, however, cause a substantial improvement of the solubility of 4.1 in methanol, which Pedersen correctly attributed to the complexation of the sodium ion by the crown ether. He said in his Nobel Prize lecture: “It seemed clear to me now that the sodium ion had fallen into the hole in the center of the molecule and was held there by the electrostatic attraction between its positive charge and the negative dipolar charge on the six oxygen atoms symmetrically arranged around it in the polyether ring” [2]. Pedersen mentions at the end of his landmark publication that 4.1 also possesses “unusual physiological properties,” including oral toxicity, which shows that crown ethers interfere in biological processes. A similar observation had been made around the same time for certain macrocyclic natural substances and was also associated with

https://doi.org/10.1515/9783110595611-004

102

4 Hosting ions and molecules

O

O O

O

O

O HN

O

O O

O

N H

O NH

O O

O

O

HN

O

O

O

NH

O

4.1 O

O

O

O

H N O

4.2 Figure 4.1: Structures of dibenzo-18-crown-6 4.1 and valinomycin 4.2.

their binding to alkali metal ions. In 1964 it was reported, for example, that the antibiotic valinomycin 4.2 (Figure 4.1) stimulates mitochondrial potassium transport. In 1967, the same year in which Pedersen’s article on crown ethers came out, it was shown that valinomycin is able to complex potassium ions and transport them along cell membranes. The relationship in terms of structure and cation binding properties between valinomycin and crown ethers was quickly recognized and several subsequent publications indeed showed that valinomycin and crown ethers both mediate the membrane transport of metal ions, potentially explaining their toxicity (Section 9.2.2). It can be speculated whether the more or less simultaneous discovery of the cation-binding properties of crown ethers and natural ionophores such as valinomycin contributed to the immediate attention that Pedersen’s publication found. In any case, this publication initiated intensive research in a field that would shortly thereafter be called supramolecular chemistry. Nomenclature: Pedersen suggested the name “crown” for macrocyclic polyethers because it adequately describes the shape of many of these compounds in their cation complexes. Later, “crown” was replaced by “crown ether,” likely because the addition of the word “ether” helps to clarify that the term relates to a chemical compound. The term “coronand” is used as a synonym for crown ethers, which allows referring to their complexes as “coronates.” Another way to denote a complex in supramolecular chemistry is based on the use of the mathematical symbol of inclusion ⊂. Accordingly, the abbreviation [K+⊂dibenzo-18-crown-6] stands for the potassium complex of crown ether 4.1.

103

4.1 Receptors

Since the naming of crown ethers by using the systematic nomenclature would be too cumbersome, Pedersen invented a simpler method, which is still in use. Accordingly, the term “crown” is used as the basis for the name. A preceding number specifies the overall number of atoms in the ring, and the number of the oxygen atoms is appended to the name. Thus, an 18-membered macrocycle with six oxygen atoms is named 18-crown-6. The presence of other groups or heteroatoms is denoted in a prefix (e.g., “benzo” for a benzene ring, “aza” for nitrogen, or “thia” for sulfur). Crown ether 4.1 is thus called dibenzo-18-crown-6. To illustrate these rules, a selection of crown ethers is shown together with their respective names in Figure 4.2. This figure also illustrates that the nomenclature introduced by Pedersen is also readily applicable to macrocyclic polyethers that do not derive from ethylene glycol but have more than two carbon atoms between the heteroatoms.

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

O

12-Crown-4

O

O

O

15-Crown-5

18-Crown-6

O

O

O

21-Crown-7

O

O

O

O

NH

HN

O

S

O

O

O

NH

HN

S

O

O

O

O

Dicyclohexano-18-crown-6

4,7,13,16-Tetraaza18-crown-6

O O O

1,7-Dithia12-crown-4

22-Crown-6

Figure 4.2: Examples of crown ethers together with their names according to the nomenclature introduced by Pedersen.

Synthesis: Crown ethers are usually prepared through Williamson ether syntheses by treating a diol and a ditosylate in the presence of a base. Dichlorides, dibromides, or diiodides can be used instead of ditosylates but the latter usually afford the products in higher yields. An example is the synthesis of 18-crown-6 4.3 shown in Figure 4.3a. The corresponding reaction between tetraethylene glycol and the ditosylate of tetraethylene glycol affords the desired [1 + 1] macrocyclization product 18-crown-6, but also the [2 + 2] product 36-crown-12 or even larger rings, in addition to polymeric side products. Since the Williamson ether synthesis is irreversible, the ratio with which these products are formed is kinetically controlled, in other words the product whose formation is associated with the lowest Gibbs free energy of activation dominates in the final reaction mixture. If the macrocyclic product is not

104

4 Hosting ions and molecules

(a)

O

O

O OH O

O +

O

O OH

O

O

TsO Base

O

O

O

O

O

O

+ O

O O

TsO

O

O O

(b)

O O

+

K

18-Crown-6 (4.3) 36-Crown-12 [1+1] macrocyclization [2+2] macrocyclization + larger rings and polymeric side products

O –

O

O O

O

OTs

Figure 4.3: Synthesis of 18-crown-6 from tetraethylene glycol and the ditosylate of tetraethylene glycol in the presence of a base (a) and structure illustrating how the binding of a potassium ion to the linear precursor of 4.3 mediates the ring closure (b).

strained, the ring closure has no enthalpic disadvantage and product formation is therefore mainly determined by entropy. To assess this influence, two opposing effects of the intramolecular and intermolecular reaction pathways have to be considered. Intramolecular reactions are generally entropically more favorable than intermolecular ones because they do not cause a reduction of the number of molecules in solution. Therefore, intramolecular reactions yielding small or medium-sized rings usually proceed rather efficiently. The larger the distance between two reacting groups in the linear precursor, the less favorable the entropic term of the intramolecular reaction becomes, however, because the linear precursor loses more and more degrees of conformational freedom when adopting the conformation required for the cyclization. Macrocyclizations are therefore typically associated with small yields. Consequently, a complex reaction mixture featuring substantial amounts of polymeric materials is normally expected for reactions such as that shown in Figure 4.3a, which is indeed the case when tetra-n-butylammonium hydroxide is used as base. In the presence of potassium tert-butoxide, however, the yield of 18-crown-6 is high, exceeding 90% under optimal conditions [3]. This dramatic improvement in comparison to the reaction mediated by tetra-n-butylammonium hydroxide is due to the influence of the potassium ions on the course of the reaction. These ions likely do not affect the formation of the first ether linkage between the starting materials. Once the direct precursor of the product is present, however, they interact with it, inducing the folding of the chain and thus bringing the two end groups into close proximity (Figure 4.3b). Potassium binding thus reduces the unfavorable activation entropy of the macrocyclization because the adverse term associated with restricting the

4.1 Receptors

105

conformational mobility is paid for by complex formation. The Gibbs free energy of activation of the macrocyclization is therefore lower in the presence of potassium ions than in their absence, causing macrocyclization to be faster than unwanted side reactions. In the terminology of supramolecular chemistry, the potassium ions act in this reaction as templates, which control the outcome of the reaction by favoring the [1 + 1] macrocycle over other products. The term template was introduced in 1964 by Daryl H. Busch who proposed the following definition [4]:

A chemical template organizes an assembly of atoms, with respect to one or more geometric loci, in order to achieve a particular linking of atoms.

Busch realized that there are two types of templates, those that influence the structure of the transition state of an irreversible reaction and those that thermodynamically stabilize a certain product of a reversible reaction. We will come back to this classification in Section 5.1, where the differences between thermodynamic and kinetic templates are explained in more detail. At this point it shall suffice to note that potassium ions (or other metal ions) act as kinetic templates in crown ether syntheses. They promote the rate with which the desired macrocycle is formed and their role in the reaction is therefore related to that of a catalyst. The major difference compared to a catalyst is that stoichiometric amounts of the potassium ions are usually present in the reaction because they are introduced together with the base. Since quaternary ammonium ions cannot interact with the oxygen atoms of the reaction intermediates, they cannot exert a template effect, explaining why the synthesis of 4.3 shown in Figure 4.3a does not proceed efficiently in the presence of tetra-n-butylammonium hydroxide. Binding Properties: Crown ethers that contain only oxygen atoms along the ring prefer to bind electropositive cations such as alkali metal or alkaline earth metal ions. Larger crown ethers also interact with organic substrates containing hydrogen bond donors such as protonated amino acids. Before we come to these systems, we first look at the different binding modes available for crown ethers to interact with metal ions. These binding modes mainly depend on the ratio of the diameter of the ring and the radius of the bound ion as the crystal structures shown in Figure 4.4 illustrate. If the radius of the cation matches the radius of the available cavity as in the case of [K+⊂18-crown-6] (Figure 4.4a), complexes are formed with the metal ion located exactly in the center of the ring. If the ring is too small to allow incorporation, the cation remains perched above or below the ring, potentially allowing it to recruit a second crown ether, which leads to the formation of sandwich complexes as in the K+ complex of 15-crown-5 (Figure 4.4b). Conversely, very large rings either fold around the ion as in [K+⊂dibenzo-30-crown-10] (Figure 4.4c) or bind two metal ions

106

4 Hosting ions and molecules

(a)

(b)

(d)

(c)

(e)

Figure 4.4: Crystal structures of the complexes of K+ with 18-crown-6 4.3 (a), 15-crown-5 (b), dibenzo-30-crown-10 (c), and valinomycin 4.2 (d), and of Na+ with dibenzo-30-crown-10 (e). Only the acidic protons of valinomycin are shown in (d) for reasons of clarity.

simultaneously as in the respective Na+ complex (Figure 4.4e). In all cases, the oxygen atoms of the crown ethers orient themselves such that they approach the bound ion, indicating that complex formation is due to electrostatic interactions between the positively charged substrate and the negatively polarized oxygen atoms as proposed by Pedersen. It is also instructive to inspect how valinomycin manages to bind a potassium ion. Figure 4.4d shows the crystal structure of the respective complex. Accordingly, valinomycin adopts a folded bracelet-like conformation that is stabilized by six hydrogen bonds between each NH group and the carbonyl group of the following valine moiety. Due to this intramolecular conformational stabilization, the six remaining carbonyl groups of the α-hydroxy carboxylic acids point into the center of the cavity, allowing them to coordinate to the metal ion in an almost perfect octahedral fashion. The cavity dimension is optimal for K+, but also allows the inclusion of larger metal ions such as Rb+. The valinomycin backbone is too rigid, however, to contract and thus bind to Na+ or even Li+. The optimal size fit of the K+ ion in the valinomycin cavity, the coordination of the cation by the six carbonyl oxygen atoms, and the simultaneous stabilization of the complex by six intramolecular hydrogen bonds results in a log Ka of 4.5 in methanol [5]. In comparison, the log Ka of the Na+ complex

4.1 Receptors

107

of valinomycin amounts to only 0.8 under the same conditions, due to the smaller size of the Na+ ion and its more difficult desolvation. The structures in Figure 4.4 suggest that complex stability should correlate with the number of attractive contacts between the receptor and the substrate. This trend is indeed observed in the gas phase, where the affinity for a given cation increases from 12-crown-4 over 15-crown-5 to 18-crown-6 [6]. With respect to the cation, complex stabilities in the gas phase progressively decrease in the order Na+ > K+ > Rb+ > Cs+ as a consequence of the decreasing charge densities in the same direction. Thus, Na+ and 18-crown-6 form the most stable complexes under these conditions. In solution, effects of solvation cause deviations from this trend. To illustrate this effect, the stabilities of the alkali metal ion complexes of 12-crown-4, 15-crown-5, and 18-crown-6 in methanol are summarized in Figure 4.5 [7].

7

log Ka

12-Crown-4

15-Crown-5

18-Crown-6

Li+

–

1.2

–

Na+

1.5

3.3

4.4

K

+

1.6

3.5

6.1

+

1.7

2.8

5.4

Cs+

1.6

2.7

4.6

Rb

6 5 log Ka

Cation

4 3 2 1

Li+

Na+

K+

Rb+

Cs+

Figure 4.5: Dependence of the stability of crown ether complexes on ring size and on the size of the cation. All binding constants refer to methanol as the solvent.

Figure 4.5 shows that the correlation between complex stability and the number of oxygen atoms is retained in solution. However, complex stability does not progressively become smaller with increasing size of the cation but shows a plateau for the Na+ and K+ complexes of 15-crown-5 complexes and a pronounced peak for [K+⊂18-crown-6]. The complexes of the smaller ions are obviously less stable than expected in methanol, which can be attributed to ion solvation. This relationship is best understood on the basis of equation (2.8) derived in Section 2.1 (ΔG0 = ΔG0intr + ΔG0solv ðCÞ − ΔG0solv ðRÞ − ΔG0solv ðSÞ). Assuming that ΔG0solv ðCÞ and ΔG0solv ðRÞ do not change strongly for a series of crown ethers and their complexes (which is certainly an approximation because these compounds differ in structure), the overall stability of a crown ether complex ΔG0 depends on the balance between ΔG0intr and ΔG0solv ðSÞ. Since cation-crown ether interactions and cation solvation are both mediated by ion–dipole interactions, cations that strongly bind

108

4 Hosting ions and molecules

to crown ethers are also strongly solvated. As a consequence, the desolvation of small ions with a high charge density causes a more pronounced reduction of ΔG0intr than that of larger more weakly solvated ones. The opposing effect of cation solvation on cation binding therefore helps to rationalize the cation selectivity of crown ethers in polar protic solvents. Additional effects contribute to the observed selectivity trends. A particularly strong interaction is expected, for example, for binding partners that match in size. This effect likely explains the particularly high K+ affinity of 18-crown-6 or the larger than expected Na+ affinity of 15-crown-5. However, if size match alone would be responsible for the cation selectivity of crown ethers, the Li+ or Na+ affinity of 15-crown-5 should be much higher than experimentally observed according to the geometric parameters collected in Table 4.1, showing that other factors are equally or even more important.

Table 4.1: Geometric parameters of selected alkali metal ions and crown ethers. Alkali metal ion +

Li Na+ K+ Rb+ Cs+

Ionic radius (Å)

Crown ether

Inner cavity radius (Å)

. . . . .

-Crown- -Crown- -Crown-

.–. .–. .–.

-Crown-

.–.

Another factor influencing complex stability is the conformational flexibility of crown ethers that is markedly reduced upon complex formation. Whereas small crown ethers such as 12-crown-4 can be assumed to be relatively well preorganized for complex formation, larger systems preferentially adopt a conformation in the absence of cations with diverging oxygen atoms to avoid the repulsion of free electron pairs as shown in Figure 4.6a. Complex formation thus induces a considerable conformational reorganization of the receptor, which is disadvantageous for binding. With regard to the conformations in the complexes, crown ethers derived from ethylene glycol have the advantage that the synclinal conformation at the O–C–C–O bonds (torsion angles around ±60°) that is required for a converging arrangement of the oxygen atoms to form an unstrained five-membered chelate ring with the cation, is easily accessible and even favored by the stereoelectronic gauche effect. Longer alkyl chains between the oxygen atoms preferentially adopt antiperiplanar conformations, which are often less suited for complex formation, explaining why crown ethers that do not derive from ethylene glycol typically exhibit a strongly reduced cation affinity.

109

4.1 Receptors

(a)

(b) O

O

O O

O O O

O

O O

O

O O

O

O H R N H + H O O O

Figure 4.6: Schematic representation of the conformational reorganization of a crown ether upon metal ion complexation (a) and mode of ammonium ion binding to 18-crown-6 (b).

Consistent with the ion–dipole interactions underlying complex formation, crown ether complexes of the doubly charged alkaline earth metal ions should be more stable than those of alkali metal ions, but this trend is again only seen for the weakly solvated Ba2+ ions. Ca2+ binding is even weaker than the complexation of singly charged alkali metal ions because the desolvation of this small and doubly charged ion is difficult. Complex stability is moreover strongly dependent on the solvent. In water, for example, the cation complexes of crown ethers are typically at least two orders of magnitude less stable than in methanol. Cation selectivity is, however, similar in both solvents, with [K+⊂18-crown-6] also exhibiting the highest stability in water among the complexes shown in Figure 4.5 (log Ka = 2.1) [7]. As mentioned before, ammonium ions are also suitable substrates for crown ethers. Because NH4+ and K+ have similar ionic radii and because of a good symmetry match, 18-crown-6 is particularly suited for such guests. The recognition involves the formation of three hydrogen bonds from ammonium NH groups to the crown ether oxygen atoms (Figure 4.6b). These interactions are, however, less efficient than those with potassium ions. The association constant log Ka of the NH4+ complex of 18-crown-6, for example, only amounts to 4.1 in methanol, two orders of magnitude below the association constant of the K+ complex [8]. Organic ammonium ions bind even weaker because of steric effects between their substituents and the surrounding 18-crown-6 framework. Cram nevertheless used the affinity of crown ethers for ammonium ions to show for the first time that a fundamental concept of Nature, namely that of enantioselective substrate recognition, can be transferred to a synthetic system. The structure of this classic receptor, the chiral bis (BINOL)-derived crown ether (R,R)-4.4, is shown in Figure 4.7. Receptor (R,R)-4.4 allows the separation of the enantiomers of amino acid ammonium salts such as the perchlorate salt of D/L-phenylglycine methyl ester. These substrates bind to the crown ether in a similar fashion as shown in Figure 4.6b. Binding occurs on both sides of (R,R)-4.4, but since the receptor is C2 symmetric, identical arrangements result. In these complexes, the orientation of the guest is determined by the steric effects of the flanking BINOL groups as illustrated in

110

4 Hosting ions and molecules

Me O

Me O O

O

O

D -Phenylglycine

Me

O Me

CO2Me

O O

Me O

H3N

O H

H

O

O O H H O Me O O More stable complex

methyl ester O Me O O O O H H

O O O O (R,R )-4.4

H3N

CO 2Me

L -Phenylglycine

methyl ester

O H H Me O

O

Less stable complex because of steric hindrance of the phenyl group and the adjacent naphthyl residue

Figure 4.7: Structure of the chiral crown ether (R,R)-4.4 (a) and preferred arrangements of D- and an L-phenylglycine methyl ester in the complexes with (R,R)-4.4 (b). Because of a larger steric hindrance in the complex with the L-amino acid, the complex of the D-enantiomer is more stable.

Figure 4.7. The methylated BINOL moiety exerts the largest steric hindrance, only allowing the proton at the α-carbon atom of the bound amino acid to be located in close proximity. The other substituents then either occupy the cleft between the BINOL units or are arranged along the face of the other naphthyl moiety. Since the side chains of the substrates are typically larger than the ester groups, they prefer occupying the larger cleft. As a consequence, (R,R)-4.4 binds D-amino acids stronger than the respective L-enantiomers. Although the overall stability of these complexes is relatively low, (R,R)-4.4 even allows the separation of enantiomers on a preparative scale. To this end, an ingenious method was devised by Cram that makes use of the enantioselective transport of an amino acid from an aqueous solution across an organic phase into a receiving aqueous solution. The principle is shown in Figure 4.8. The experimental setup comprises two U-tubes, one containing a chloroform solution of (R,R)-4.4 and the other a chloroform solution of the (S,S)-enantiomer of the receptor. Both tubes are in contact in the center with an aqueous solution containing the racemic amino acid of which the D-enantiomer is transported into one of the receiving phases by (R,R)-4.4. Conversely, the enantiomeric receptor mediates the transport of the other amino acid enantiomer. As a consequence, each receiving phase predominantly contains a single enantiomer at the end of the experiment (enantiomeric excess (e.e.) 86–90%). This experiment illustrates the use of crown ethers to mediate the transport of substrates across liquid membranes. For other applications of crown ethers, see Chapter 11.

111

4.1 Receptors

Aqueous solution of racemic amino acid

Aqueous solution enriched in L-amino acid

Aqueous solution enriched in D-amino acid (R,R)-4.4 in CHCl3 selectively mediating the transport of the D-amino acid via the respective complex

(S,S)-4.4 in CHCl3 selectively mediating the transport of the L-amino acid via the respective complex

Figure 4.8: Method of separating the racemate of amino acid ammonium ions. Each receptor enantiomer in the chloroform solutions mediates the transport of one enantiomer of the substrate so that each aqueous receiving phase contains mainly enantiomerically pure compounds at the end of the experiment.

Derivatives: The number of currently known crown ethers or cation receptors that derive from crown ethers is huge. At this point, we therefore only concentrate on a few general strategies that have been used to modulate the binding properties of these systems. Probably the most obvious one involves replacing the oxygen atoms along the ring with other heteroatoms. The most widely used elements in this context are nitrogen and sulfur but others (e.g. Se and Te) have been used as well. Heteroatoms that are less electronegative than oxygen generally interact weaker with electropositive cations, causing cation affinity to drop as the number of such heteroatoms along the ring increases. This effect is clearly evident along the series of receptors shown in Table 4.2 [9]. Conversely, the affinity for softer transition metal ions increases in the same direction as also illustrated in the table.

Table 4.2: Effect of the replacement of oxygen atoms in 18-crown-6 with nitrogen atoms on the log Ka values of the respective K+ complexes in methanol and Ag+ complexes in water at 298 K. Cation

O O O

O

O

O

O

O K+ Ag+

. .

N H

O

O

O

O

O . .

. .

N H

H N

O O

112

4 Hosting ions and molecules

Introducing trivalent nitrogen atoms into the ring has the advantage that it allows the subsequent attachment of further substituents. Bridging two opposing nitrogen atoms in a crown ether with a suitable chain, for example, leads to the bicyclic cryptands, which are discussed in the next chapter. Appending substituents to crown ethers gives rise to lariat ethers (one substituent) or bibracchial lariat ethers (two substituents). The term lariat, which normally refers to a rope in the form of a lasso, illustrates very well that these receptors consist of a linear chain attached to a loop. Lariat ethers combine the properties of cyclic crown ethers with those of their acyclic analogs, the so-called podands. We will look more closely in the next chapter at how binding properties are affected when going from a linear podand to a macrocyclic coronand and further to a lariat ether and a bicyclic cryptand. Here, just recall that lariat ethers played an important role in the characterization of cation–π interactions as we have seen in Section 3.1.7. The replacement of most or even all oxygen atoms of a crown ether by nitrogen atoms leads to polyazamacrocycles, another large family of receptors that serves either as ligands for transition metals or, after protonation, as anion receptors. Two important macrocyclic tetraamines that have found widespread use for the coordination of transition metal ions are 1,4,7,10-tetraazacyclododecane 4.5 (cyclen) and 1,4,8,11tetraazacyclotetradecane 4.6 (cyclam) shown in Figure 4.9a. Both compounds form highly stable complexes with, for example, Cu2+ and Zn2+ in water ([Cu⊂4.5]2+, log Ka = 24.8; [Zn⊂4.5]2+, log Ka = 16.2; [Cu⊂4.6]2+, log Ka = 27.2; [Zn⊂4.6]2+, log Ka = 15.5) [10]. The metal ions have square pyramidal coordination geometries with the nitrogen atoms at the base and the counterion in the apical position (Figure 4.9b). The lability of the bond to the counterion renders such metal complexes useful building blocks for the construction of anion receptors. Their chemistry otherwise falls into the domain of coordination chemistry, which is why these systems are not further discussed here. Instead we focus on the properties of metal-free polyamines, an example of which is compound 4.7 in Figure 4.9c.

(a)

(c)

(b)

NH HN HN HN

4.5

NH

HN

NH

NH

HN

NH

4.6

HN M NH HN

NH

(Distorted) square pyramidal

NH

HN NH HN 4.7

Figure 4.9: Structures of cyclen 4.5 and cyclam 4.6 (a), schematic structure of the metal complexes of these ligands (b), and structure of the polyaza analog 4.7 of 18-crown-6 (c).

4.1 Receptors

113

These polyamines are the nitrogen-containing counterparts of crown ethers. They are protonated to a substantial degree in aqueous solution, rendering them cationic and therefore well suited to interact with anions by electrostatic interactions. In this context, several guidelines apply. First, the degree of protonation of macrocyclic polyamines is pH dependent. Thus, the number of positively charged groups along the ring increases with decreasing pH, causing complex stability to increase in the same direction. With regard to the protonation sequence, ammonium groups in a partially protonated polyamine tend to be arranged at the largest possible distance to reduce charge repulsion. Thus, nonadjacent amino groups are protonated first and the least basic ones last if at all when the pH of the solution is decreased. It is important in this context that secondary amino groups have a larger propensity to accept a proton than tertiary amino groups. The introduction of tertiary amines into strategic position of such polyamines therefore offers a means to control the protonation pattern. The distance of the amino groups along the receptor scaffold also strongly affects the conditions required for protonation, with smaller distances making it more difficult to achieve a high degree of protonation. Protonation moreover affects the receptor conformation because ammonium groups tend to move away from each other, causing the molecular framework of the receptor to expand upon protonation. Finally, ammonium groups along macrocyclic or polymacrocyclic systems preferentially adopt orientations with the protons located outside of the cavity. As a consequence, the receptor has to undergo a conformational reorganization to allow the formation of hydrogen bonds to an included anion. How these factors affect anion affinity has been systematically investigated by using complex anions as substrates such as the octahedral complexes [Fe (CN)6]4− and [Co(CN)6]3−. The comparison of the stabilities of the corresponding complexes provides information, for example, about the influence of the charge of the anion on complex stability without strong interferences by other factors. These studies showed that (i) the complexes of the higher charged [Fe(CN)6]4− anion are more stable than those of [Co(CN)6]3−; (ii) complex stability increases with increasing protonation degree of the receptor; (iii) for a given protonation degree, complex stability increases with increasing charge density, causing smaller macrocycles to form more stable complexes than larger ones. These straightforward correlations can break down, however, for anionic substrates such as oxoanions that additional engage in hydrogen bonding interactions with protonated amino groups.

114

4 Hosting ions and molecules

4.1.2 Cryptands Introduction: In 1969, two years after Pedersen’s first publication about crown ethers, two consecutive short communications by Bernard Dietrich, Jean-Marie Lehn, and Jean-Pierre Sauvage appeared in Tetrahedron Letters, both written in French, dealing with the synthesis, structural characterization, and cation affinity of macrobicyclic analogs of crown ethers [11, 12]. It may be worth noting here that not only Lehn received the Nobel Prize for his contributions to supramolecular chemistry but also Sauvage, who was Lehn’s PhD student in 1969 and was awarded the Nobel Prize in chemistry in 2016 for his work on molecular machines (Chapter 7). Representative structures of the family of macrobicyclic receptors introduced by Lehn in these publications are shown in Figure 4.10. All of them feature nitrogen atoms as bridgeheads, connected via three ethylene glycol chains of varying length. They thus feature three-dimensional cavities that allow fully engulfing a cation and surrounding it from all sides. This binding mode explains why the cation complexes of these receptors were named cryptates, a term that first appeared in the title of one of the 1969 publications (“Les Cryptates”) and that derives from Latin crypta = vault, cave. The free receptors are, accordingly, called cryptands.

O

O N

O

O N

N N

O O

O

[2.1.1]Cryptand

O

O

N O

O

O

O

[2.2.2]Cryptand (4.8)

O

O

O

N O O

O O

O

N

N O

O

[3.3.2]Cryptand

O

O O

O O

[3.3.3]Cryptand

O

O

O O

O O

O O

Orthoester-derived cryptand

Figure 4.10: Representative structures of bicyclic crown ether derivatives, so-called cryptands, along with the corresponding names.

The family of cryptands for a long time mainly contained compounds with nitrogen atoms as bridgeheads. It was not until 2015 that cryptands were described by Max von Delius in which carbon atoms connect the three straps [13]. An example is depicted in Figure 4.10. These cryptands are synthesized by treating orthoesters with diols under appropriate conditions (Section 5.7) and their binding properties typically range between those of crown ethers and classic cryptands of similar size. Nomenclature: The archetypical cryptands containing only ethylene glycol chains between the nitrogen atoms are distinguished by the number of oxygen atoms in the straps. Accordingly, the smallest cryptand shown in Figure 4.10 is termed [2.1.1]cryptand and the largest one [3.3.3]cryptand. [2.2.2]Cryptand (4.8) is commercially available under the name Kryptofix® 222. A variety of other cryptands widely differing in the

115

4.1 Receptors

number and type of heteroatoms as well as the structures of the bridging straps have been described. Although a systematic way of naming such cryptands has been proposed [14], this nomenclature has not found widespread use. Synthesis: The classical approach introduced by Lehn to prepare cryptands is illustrated in Figure 4.11a by using the original synthesis of 4.8 as an example. The first step is the reaction of a diamine with a diacid chloride to afford a macrocyclic dilactam. This product is then reduced with lithium aluminum hydride to yield 4,13-diaza-18crown-6 in an overall yield of 75%. The treatment of this crown ether with the same acid dichloride as used in the first step produces the macrobicyclic product in 45% yield whose reduction with diborane proceeds almost quantitatively to afford 4.8 [11]. (a)

O

O

+

O

O

O

O

O

O Cl

Cl

O

O O

O N

45%

1 equiv.

N O

O

O

O

H2N

NH2

K2CO3, CH3CN

+ 2 equiv.

O

TsO O

36%

O

O

O

O

O

O

O

O

B2H6 N 98%

O

O

OTs

O

O

HN

NH

75% over 2 steps

O

O

O

LiAlH4

HN

NH

O

O

O

Cl

Cl

(b)

O

O

H2N

NH2

O

O

O

O

O

O

N

N

N

4.8

4.8

Figure 4.11: Syntheses of cryptand 4.8 via dilactams (a) or by using nucleophilic substitution reactions under the influence of a suitable template (b).

It has to be considered that the transformations affording the macrocycle in the first step and the macrobicycle in the third step each involve two separate reactions. The initial intermolecular one leads to an intermediate that still contains an unreacted acid chloride and an amino group. The subsequent reaction either proceeds intramolecularly to afford the desired product or intermolecularly, yielding unwanted oligomeric side products. The question therefore arises how the intramolecular reaction pathway can be favored.

116

4 Hosting ions and molecules

Which strategies exist to favor macrocyclization reactions?

One possibility is to use template effects as already mentioned in Section 4.1 and discussed in more detail in Section 5.1. If template effects cannot be used or prove to be inefficient, the reaction is usually performed at very low concentrations. Under these conditions, the probability that two molecules meet is reduced. Reactive groups thus preferentially react intramolecularly as opposed to intermolecularly. Such high dilution conditions have the disadvantage, however, that they require large amounts of solvents and large reaction flasks if the syntheses are to be performed on a large scale. The better option is therefore often to very slowly add one reaction partner to the solution of the second starting material by using, for example, a motorized syringe pump (pseudo-high dilution conditions). In this way, the reaction mixture always contains one component in a very low concentration and if the addition is slow enough, the conversion of this compound into the desired product takes place before the next drop of the reactant reaches the solution. The synthetic approach outlined in Figure 4.11a allows accessing a variety of cryptands, even ones differing in the structures of all three bridges. Such multistep syntheses are, however, relatively complex and time-consuming, rendering simpler strategies more attractive. One possibility is shown in Figure 4.11b. In this reaction, the treatment of one equivalent of a diamine and two equivalents of a ditosylate in the presence of potassium carbonate in acetonitrile affords 4.8 in a single step with a good yield of 36% [15]. Characteristic effects of the cation introduced with the base on the outcome and yield indicate that this synthetic strategy is mediated by template effects in a similar fashion as the crown ether syntheses discussed in the previous chapter. Binding properties: Figure 4.12 illustrates the binding modes of a potassium and an ammonium ion inside the cavity of 4.8 in the crystal structures of the respective complexes. Both cations are fully incorporated into the three-dimensional cavity and interact with the heteroatoms by ion–dipole interactions and, in the case of the NH4+ ion, also hydrogen bonds.

(a)

(b)

Figure 4.12: Crystal structures of the complexes of K+ (a) and NH4+ (b) with 4.8.

4.1 Receptors

117

Because bicyclic systems are conformationally less flexible than macrocycles, the number of possible binding modes in cryptand complexes is lower than those available for crown ethers (Figure 4.4). Besides incorporation of the cation into the cavity, the only other available binding mode involves external binding between two bridges, which usually happens if the cation is too large to fit inside the cavity. The first binding motif is, however, associated with larger binding constants. To illustrate the effect of the receptor structure on cation affinity, the stabilities of the potassium complexes of six different polyethers are compared in Table 4.3 [9]. While all binding constants refer to methanol as solvent, the methods with which they were obtained partly differ so that a comparison may not be fully appropriate. General trends can, however, be derived. Table 4.3: Comparison of the potassium affinity in methanol at 298 K of six receptor types differing in structure and the number of oxygen atoms along the cavity. Receptor

logKa

O

Receptor

.

O

O

O

O

O O

O

O

O

.

O O

N H

H N

logKa

N H

O

O

.

O

O

O

O

N

.

O

O O

O

Receptor

O

O

O

logKa

O

O O O

.

O

N

N O

O

O

O

.

Table 4.3 shows that the potassium affinity of 18-crown-6 is four orders of magnitude higher than that of the corresponding acyclic podand with the same number of oxygen atoms. We have seen this comparison already in Section 3.5.2, where the increase of binding strength caused by cyclization was associated with the macrocyclic effect. The podand has to fold around the cation to maximize its interactions, and entropic and/or enthalpic penalties associated with this conformational reorganization are detrimental for the overall binding strength. In the case of the crown ether, these unfavorable energetic contributions have already been paid for during macrocyclization, rendering cation binding stronger than that of the podand. According to the results of calorimetric studies, the lower binding affinity of the linear receptor is, in this particular case, practically exclusively due to the less favorable

118

4 Hosting ions and molecules

binding enthalpy with respect to the crown ether (podand: ΔH 0 = −8.7 kJ mol−1, TΔS0 = −5.8 kJ mol−1; crown ether: ΔH 0 = −13.4 kJ mol−1, TΔS0 = −5.1 kJ mol−1) [16]. Thus, the macrocyclic effect observed for the cation binding of these polyethers in methanol is not due to a more unfavorable binding entropy of the less preorganized system, as one might expect, but to the fact that the acyclic system has to adopt enthalpically unfavorable conformations during complex formation. Table 4.3 shows that going from the all-oxygen crown ether to the derivative with one nitrogen atom leads to a pronounced decrease of affinity as already explained in Section 4.1. Appending an ethylene glycol chain with two oxygen atoms to this nitrogen atom compensates this effect, but only to such an extent that the respective lariat ether ends up having the same potassium affinity as the crown ether without the substituent. A further improvement of binding strength only results if the substituent in the lariat ether is covalently connected to a second nitrogen atom in the ring. Importantly, the third bridge in the resulting cryptand overcompensates the detrimental effect of the second nitrogen atom on cation affinity, rendering the bicyclic receptor by far the best potassium binder in this series of compounds. Similar trends are observed for other receptor–substrate combinations. The generally highest affinity of the cryptands, which is due to their good preorganization and their ability to surround the substrate from all sides, has been termed macrobicyclic effect. Because cryptands are unable to adapt easily to the size of the bound substrate, these receptors exhibit a pronounced selectivity for the cation whose size best matches the available space inside the cavity. If the cation is too small, the rigidity of cryptands makes it difficult for them to contract and thus optimize the interactions with the substrate. Conversely, large cations do not fit into a too small cavity. Cryptands thus exhibit pronounced peak selectivities, with the [2.1.1]cryptand exhibiting the highest affinity for Li+ (log Ka = 8.0) in spite of the strong solvation of this cation. The alkali metal ion affinity of the [2.2.1]cryptand peaks at Na+ (log Ka = 9.7) and that of 4.8 at K+ (log Ka = 10.4) (Figure 4.13) [9]. Note also the much larger differences between the cation affinities of these cryptands (Figure 4.13) with respect to the differences observed for the more flexible crown ethers (Figure 4.5). The cryptand 4.8, for example, binds K+ eight orders of magnitude more strongly than Li+ and six orders of magnitude more strongly than Cs+. In the case of the corresponding 18-crown-6 complexes, the binding constants vary only by less than two orders of magnitude, illustrating the effect of an improved structural complementarity between the receptor and the substrate on binding selectivity. The formation of all of these complexes is fast on the human timescale. In the case of alkali metal ion binding by 4.8, for example, the kon rates are in the order of 108 M−1 s−1 [9]. According to equation (2.12), the peak selectivity observed for these complexes therefore results from the different rates of complex dissociation, with the dissociation of the K+ complex of 4.8 being much slower (koff = 1.8 × 10−2 s−1) than that of the other alkali metal ion complexes of 4.8.

4.1 Receptors

119

11

log Ka [2.1.1]Cryptand

[2.2.1]Cryptand

[2.2.2]Cryptand

Li+

8.0

5.4

2.6

Na +

6.1

9.7

8.0

K+

2.3

8.5

10.4

Rb +

1.9

6.7

9.0

Cs +

4 in D2O [98], for example. Interestingly, dodecyltrimethylammonium salts bind to 4.49 not with the cationic head group included into the cavity but with the alkyl chain. This alkyl chain adopts an unfavorable helical conformation to efficiently fill the available space [97]. The energy necessary to adopt this conformation is paid for by the dehydration of the dodecyl chain upon complex formation and the dispersion interactions between the folded chain and the cavity walls. (a) NaOOC COONa

NaOOC

N HN

N

N

O

O

O

R

R R 4.49 (R = C2H 5 )

(b) O

–

O

O

O

O R

NH

N

HN O

O

COONa

NH

O

O O

O

– O

–

O

O

O

O

O

–

O

O

O O

O

O

OO

–

– O

O

O

O

O

O

–

O

O

O

– O O

O

4.50

Figure 4.57: Molecular structure of 4.49 and calculated structure of its complex with a dodecyltrimethylammonium ion, showing the bridging water molecules and illustrating the helical arrangement of the dodecyl chain inside the receptor cavity (the ethyl groups in the resorcinarene ring are replaced by methyl groups) (a), and structure of octa acid 4.50 (b).

180

4 Hosting ions and molecules

The final example of a deep cavitand is the “octa acid” receptor 4.50 developed by Bruce C. Gibb (Figure 4.57b). This receptor comprises a resorcinarene core whose aromatic subunits are bridged via benzal groups so that it structurally resembles 4.44 rather than the deep cavitands 4.45 or 4.46. The aromatic side walls of 4.50 are made up by the aromatic units bound to the benzal methine groups and since these side walls are further linked in their respective 3- and 5-positions with 3,5dihydroxybenzoic acid moieties, 4.50 is rigid with a ~ 8 Å wide and ~8 Å deep hydrophobic cavity. The eight carboxylate groups along the “feet” of the resorcinarene unit and the rim of the cavity render 4.50 soluble in water at pH > 7. Under these conditions, 4.50 hosts guest molecules such as alkanes or aliphatic carboxylates [99]. In the case of alkanes, the filling of the cavity produces a large hydrophobic surface at the cavity opening that derives from the guest molecules themselves but also from the aromatic receptor subunits. As a consequence, such complexes are prone to dimerize, driven by the hydrophobic effect, affording capsules in which the guest molecules reside. This mode of binding is explained in Section 5.2. Substrates with polar functional groups such as alkyl carboxylates are bound in a 1:1 fashion because their polar head group resides at the cavity entrance where it mediates the hydration of this part of the receptor, thus weakining the hydrophobic effect and preventing self-assembly. The cavitand 4.50 binds hydrophobic guests mostly stronger than the similarsized β-CD. 1-Adamantanecarboxylate is bound, for example, with a log Ka of 6.7 in 10 mM phosphate buffer at pH 11.3 [100], whereas the corresponding β-CD complex is more than two orders of magnitude less stable (Table 4.5). Complex formation is strongly exothermic with a small adverse entropy term as observed for cyclodextrins. Interestingly, 4.50 also binds large weakly coordinating anions such as perchlorate, which is counterintuitive at first sight: why should a negatively charged receptor bind to an anionic substrate? However, the negatively charged groups in 4.50 are too far away to be able to repel an anion bound inside the cavity. Moreover, the inner cavity surface of 4.50 features a positive potential in the region of the methine protons that are oriented toward the interior, inducing affinity for anionic guests. Anion binding causes the interactions with hydrophobic guests in water to be influenced in a characteristic way by salts. Perchlorate, for example, directly competes with 1-adamantanecarboxylate for the receptor cavity and thus lowers the stability of the respective complex [100]. Instead of building a wall around the rims of resorcinarenes to obtain receptors with deep cavities that are open at one end, it is also possible to prepare resorcinarene derivatives with completely closed cavities. The respective synthetic strategy involves covalently connecting two resorcinarene moieties via four linking units, which gives rise to similar molecular cages as the CTV-derived ones we saw in Section 4.1.6. Such cages were first prepared in the Cram group by using the conformationally rigid cavitand 4.44 (Figure 4.54) as the basis [101]. Since 4.44 lacks functional groups that would allow bridging two subunits, the respective pyrogallol derivative

4.1 Receptors

(a)

R

HO O

O OH

O

OH

O

O

O R

O HOO

CH2BrCl K2CO3 DMF

R

R R

4.51 (R = CH2CH2Ph )

O

R

O

O O O O O O O ON O O O O R

(b)

R R

O

181

O

OO O

O O O OO

O R

R R

4.52 DMF

Figure 4.58: Synthesis of 4.52 from the corresponding pyrogallolarene-derived cavitand 4.51 (a), and calculated structure of an analog of 4.52 with R = CH3 containing an entrapped DMF molecule (b). The DMF molecule is shown in blue.

4.51 (Figure 4.58a), whose hydroxy groups are available for reactions with dihalogenalkanes under basic conditions, serves as the actual starting material. The treatment of 4.51 with CH2BrCl in dimethylacetamide (DMA), dimethylformamide (DMF), or dimethylsulfoxide (DMSO) affords 4.52 as the product, for example, which represents one of the first resorcinarene-derived molecular cages [102]. This cage usually contains a molecule of the solvent in its cavity, which was present during the synthesis and cannot escape because none of the portals along the cage surface is large enough to allow it to squeeze through. Different analytical techniques provide clear evidence for the presence of this guest. The 1H NMR spectrum of the product, for example, not only contains the guest signals, but these signals also appear at significantly higher field than in the spectrum of the uncomplexed guest. Thus, the guest resides inside the cavity where it experiences the shielding effect of the π-systems lining the cavity walls. In the mass spectrum, the product peaks have m/z ratios consistent with the mass of the cage plus the mass of the included guest molecule, also showing that the guest and the cavitand are intimately connected. Finally, crystal structures unambiguously demonstrate that the cavity is occupied. To illustrate the binding mode, the structure of the DMF complex of an analog of 4.52 with shortened side chains is shown in Figure 4.58b. Because of the permanent incarceration of the guests, Cram suggested the name carcerands for these molecular cages. The respective complexes are termed carceplexes. Carcerand syntheses often proceed surprisingly effective considering that seven molecules are brought together and eight covalent bonds are formed. The yields of the DMA, DMF, and DMSO complexes of 4.52 in the reaction shown in Figure 4.58a respectively amount to, 49%, 54%, and 61%, for example [102]. If the reaction is performed in the bulky solvent N-formylpipyridine (NMP), no product is obtained, indicating that suitable templates have to be present to mediate product formation. As shown in Figure 4.59, the influence of these templates materializes once two pyrogallolarene

182

2

4 Hosting ions and molecules

HO OH HO OH

HO OH O OH HO OH O OH

CH2BrCl K2CO3

DMF HO HO

DMF

CH2BrCl K2CO3 DMF

OH O O

HO O O OH HO O O OH

OH

OH OH

O O O

O

O O

O

O

Figure 4.59: Schematic course of a carcerand synthesis, illustrating the template effect of a solvent molecule.

units are connected via the first linking unit. Because of the chelate effect, the corresponding singly-linked intermediate has a substantially higher affinity for a suitable substrate than two individual receptor halves that are not yet linked. We have seen this effect when discussing cyclodextrin dimers in Section 4.1.4. During carcerand synthesis, the binding of a solvent molecule to the singly-linked intermediate preorganizes the two receptor subunits such that the second linkage can only be formed between correct pairs of OH groups. Once this second linkage is present, all further linkages must lead to the desired product. The solvent molecule remains inside the cavity during this reaction, and it thus induces its own incarceration. Performing carcerand syntheses in the presence of suitable templates in NMP as solvent, which is too large to act as a template, provides information about which compounds are particularly suited for inducing product formation. A correlation between template structure and efficiency can furthermore be established by using competing templates and analyzing the ratio with which the respective products are formed. According to these investigations, pyrazine gives the highest yields because it induces an optimal arrangement of the two pyrogallol units during the synthesis. The authors concluded that templation is driven “by an optimum of van der Waals interactions and a minimum of steric interactions between the guest/template molecule and the interior of the forming shell” [103]. While carcerands and their complexes are conceptually interesting, representing “a new phase of matter” according to Cram [102, 104], their use is limited because guest exchange is impossible. This deficit was addressed by developing hemicarcerands, which have larger portals along their surface, allowing guests to enter or leave the cavity. Note that the prefix “hemi” has a different meaning in the word hemicarcerand than in hemispherand or hemicryptophane. The latter two receptor types only feature half or a spherand or cryptophane in their structure. Hemicarcerands, however, contain two pyrogallol units but do not, like carcerands, permanently entrap their

4.1 Receptors

183

guests. Hemicarcerands are, by definition, closed-surface molecules with an enforced internal cavity whose complexes are kinetically inert at ambient conditions but permit guest exchange at elevated temperatures. Two strategies allow preparing such receptors. The first involves reducing the number of linkers between the subunits by, for example, using pyrogallolarenes in the synthesis lacking one or two OH groups along the rim. An example is hemicarcerand 4.53 (Figure 4.60). Alternatively, the length of the linkers between the two receptor halves can also be increased as in hemicarcerand 4.54. The advantage of the latter strategy is that hemicarcerands with larger distances between the two subunits also allow the complexation of more sizeable guests.

R

O

O

R

R

O

O

O

O

O

O

O

O

O

O

O OO

O

O R

OO O

O

O

R

4.53 (R = CH2 CH2 Ph)

R

O

O

O

O O O

O

O

O

O

O

O O

O R

OO O

O OO

O

O

R

R

R

R

R

O

O

R

R

R R

4.54 (R = CH2CH2 Ph)

Figure 4.60: Molecular structures of hemicarcerands 4.53 and 4.54.

These hemicarcerands, like carcerands, are produced during the synthesis as hemicarceplexes, containing entrapped solvent molecules. In contrast to carcerands, these solvents molecules are, however, expelled from the cavity by heating for a prolonged time in solvents such as xylene, 1,3,5-trinitrobenzene, or chlorobenzene that are too large to enter the cavity. Removing DMF or DMA from the cavity of 4.53 at 165 °C takes 24 h and 12 h, respectively, which illustrates how difficult it is for guests to exit a hemicarcerand even if a suitable portal is available [105]. Filling hemicarcerands is achieved in a similar manner by heating in the presence of an excess of potential guest molecules. The estimation of the time it takes for guests to leave or enter hemicarcerand cavities under defined conditions provides information about the activation barriers associated with these processes. For instance, the free activation energies for a DMA molecule to enter and leave 4.54 in o-xylene-d10 at 100 °C amounts to 98.4 kJ mol−1 and 113.9 kJ mol−1, respectively [106]. The difference of these two energy values reflects the thermodynamic stability of the complex in terms of ΔG0 .

184

4 Hosting ions and molecules

The corresponding −15.5 kJ mol−1 correlates with an association constant Ka of 150 M−1, which shows that the stability of the complex is low. The fact that it is not dynamic at room temperature is therefore entirely due to the high activation barriers associated with complex formation and dissociation. Accordingly, such complexes are kinetically inert at room temperature but thermodynamically not very stable. Breaking down the ΔG0 of complex formation into the enthalpic and entropic contribution shows that both parameters contribute roughly equally to the overall complex stability (ΔH 0 = −6.3 kJ mol−1, TΔS0 = 9.2 kJ mol−1) [106]. The small ΔH 0 indicates that the direct interactions between the guest and groups lining the inner surface of the hemicarcerand are weak. Small differences in the strength of these interactions explain the effects of guest structure on hemicarceplex stability and they have also been invoked to explain differences in the templating ability of structurally related guests, but the reluctance of a hemicarcerand to release its substrate is mainly due to kinetics. The gain in entropy upon complex formation likely has various reasons such as the desolvation of the guest, the structural relaxation of the host, or the release of gas molecules from the cavity. To describe these kinetic effects quantitatively, Cram introduced the concept of constrictive binding [107], mentioned already in Section 2.1. In brief, this term describes the free energy of the transition state relative to the free energy of the separate binding partners (Figure 2.4). Constrictive binding therefore essentially describes how difficult it is in terms of activation energy for a substrate to enter the cavity of a hemicarcerand. This energetic barrier is mainly due to the strain induced in the hemicarcerand or in the guest when the latter squeezes through the portals while entering or leaving the cavity. Calculations provided information about possible pathways of these processes. For 4.53, they showed that guest egress involves the folding away of the acetal CH2 groups connecting the aromatic subunits in the pyrogallol moieties [108]. This gating process is comparable to the opening of a “french door.” Alternatively, the enlargement of a hemicarcerand side portal also proceeds via a “sliding door” mechanism. In this case, the conformational reorganization involves the whole molecular skeleton of the hemicarcerand without a pronounced outward motion of CH2 groups (Figure 4.61).

(b)

H

(a) O O O

O

O O O

O O O

O

O

O O O

O

O

Sliding door

OO O

H

French door

H

O OH H O O O OH H

H

O

O

Figure 4.61: Schematic illustrations of the “french door” (a) and “sliding door” (b) gating mechanisms that allow the egress of a substrate from the cavity of a hemicarcerand.

O O O

O

4.1 Receptors

185

One of the most remarkable applications of hemicarceplexes is the stabilization of highly reactive substrates. The first example was reported by the Cram group, comprising the complex between 4.53 and cyclobutadiene. Cyclobutadiene can be generated in solution but is unstable and rapidly undergoes a series of transformations, starting with the dimerization through a Diels-Alder reaction. Cram showed that it is possible to “tame” cyclobutadiene inside the cavity of hemicarcerand 4.53 [109]. The synthetic strategy involves the initial preparation of the hemicarceplex of 4.53 with α-pyrone as the guest. The irradiation of this hemicarceplex in chloroform initiates a two-step reaction sequence inside the cavity that starts with the electrocyclic reaction of the 1,3butadiene part of α-pyrone to produce a bicyclic intermediate, which then eliminates CO2 in a cycloreversion process to yield the product (Figure 4.62). CO2 leaves the cavity, but cyclobutadiene stays inside. Under these conditions, cyclobutadiene, which had prior to Cram’s work only been characterized at 8 K in an argon matrix, is stable even at room temperature. Since the hemicarcerand moreover prevents it from reacting with another molecule, cyclobutadiene can be spectroscopically characterized. 1H NMR spectroscopy shows, for example, that cyclobutadiene has a singlet ground state and gives rise to a signal at 2.27 ppm when bound to the hemicarcerand, ca. three ppm upfield from the proton signal of an isolable cyclobutadiene derivative with three bulky alkyl groups along the ring. O O

O

h·ν

h·ν

+

CO 2

O

Figure 4.62: Reaction sequence performed inside the hemicarcerand 4.53 for the preparation of cyclobutadiene from α-pyrone.

This strategy also allows stabilizing and characterizing other reactive intermediates. The group of Ralf Warmuth described the generation of o-benzyne from benzocyclobutanedione inside the cavity of a hemicarcerand, for example [110]. When irradiating the corresponding hemicarceplex at 77 K, CO extrusion initially leads to the formation of benzocyclopropenone inside the cavity (Figure 4.63). This compound slowly converts into benzoic acid when the hemicarceplex is stored at room temperature in the presence of water. Continuing the irradiation at 77 K affords o-benzyne under the extrusion of another CO molecule. Again, 1H NMR spectroscopy provides clear evidence for the presence of the product inside the cavity. This guest is so reactive, however, that it has to be characterized at 77 K. At higher temperatures, the encapsulated o-benzyne undergoes a Diels-Alder reaction with an aromatic ring lining the inner surface of the cavity. Further reactive intermediates stabilized within hemicarcerands include carbenes [111]. The limitation of the concept is that the reactions inside the cavity have to be initiated by triggers that reach the substrate, the most important of which is

186

4 Hosting ions and molecules

COOH O O

h·ν

O

H2O

77 K –CO h·ν

77 K –CO R

R

>175 K O

O

O

Reaction with inner surface of hemicarcerand

Figure 4.63: Reaction sequence performed inside a hemicarcerand for the generation of o-benzyne from benzocyclobutanedione. The benzocyclopropenone intermediate reacts with water to afford benzoic acid (top right) or loses another molecule of CO. The thus formed o-benzyne adds to an aromatic subunit along the inner surface of the hemicarcerand as shown in the bottom right corner of the reaction scheme.

light, but small molecules such as O2, CO2, or H2O are also able to pass the portals. Larger reagents are, however, unable to reach the bound substrate.

4.1.10 Pillararenes Introduction: Considering the long history of CTVs, calixarenes, and resorcinarenes in supramolecular chemistry, it is surprising that a fourth receptor family, also formed by treating a phenol derivative and an aldehyde, only entered the field in 2008 [112]. These cyclophanes were first described by Tomoki Ogoshi who showed that the reaction between 1,4-dimethoxybenzene and paraformaldehyde in the presence of a Lewis acid affords a five-membered paracyclophane in which the aromatic subunits are linked by methylene groups in their 2- and 5-positions. In the meantime, reaction conditions are known to also access larger rings (up to the 10-membered one), but the yields are often low. Therefore, only two members of this cyclophane family are more widely used, namely those with five or six aromatic subunits, which combine good synthetic accessibility and interesting binding properties [113]. Representative crystal structures of both compounds are depicted in Figure 4.64. These structures differ profoundly from those of most cyclophanes discussed earlier. While the aromatic subunits in CTVs, calixarenes, and resorcinarenes are titled, producing bowl-shaped structures with a wider and a narrower cavity opening, those in cyclophanes with 1,4-substituted benzene subunits are arranged almost parallel to the main axis of the ring. A cylindrical shape thus results, reminiscent of the symmetric pillars that constitute the Parthenon in Athens, explaining why this family of receptors is known as pillararenes.

4.1 Receptors

(a)

(b) R RO O

R R O O

O

187

R

O OR R R = C3H 7

R O

O O R R

R O R

R RO

O

O

O

O RR O

R

R OR O

O R

OR R

O

R = C3H 7

Figure 4.64: Molecular structures and crystal structures of the pillar[5]arene (a) and the pillar[6]arene (b) derived from 1,4-dipropoxybenzene. The protons in both structures are not shown for reasons of clarity.

The reason for the tubular structure of pillararenes is the substitution pattern of the aromatic subunits that causes a linear arrangement of the CH2–Ar–CH2 bonds. By contrast, the CH2–Ar–CH2 subunits are bent in ortho- or metacyclophanes. The aryl groups linking the CH2 group can moreover be regarded as extensions of the C–C bonds in cyclopentane and since the angle with which the aryl groups are arranged is almost the same as the interior angles in a pentagon (109° vs. 108°), the five-membered pillararene forms easily without producing strain. The six-membered analog is a bit distorted as illustrated by the slightly tilted arrangement of the aromatic subunits in the pillar[6]arene shown in Figure 4.64b, but the corresponding derivative with free OH groups has the conformation of an almost ideal hexagon. Pillar[5]arene has a cavity diameter of ca. 5 Å, which is similar to α-cyclodextrin, while the cavity diameter of pillar[6]arene amounts to 7 Å, intermediate between β- and γ-cyclodextrin. The substituents in the aromatic subunits of pillararenes are oriented at an angle with respect to the main axis of the ring. As a consequence, each substituent can adopt two distinct orientations with respect to the main axis, leading to altogether eight conformations for a pillar[5]arene in which all substituents are identical. All of these conformations are chiral, with the planar aromatic subunits representing the stereogenic elements. In terms of stereochemistry, the conformations are grouped into four diastereomers, each consisting of a pair of enantiomers. In the case of pillar[6]arene, three of the eight possible diastereomeric conformations are meso forms so that thirteen conformational isomers result. Of these conformations, only a few are relevant, however. In the case of pillararenes with OR groups, the most stable conformations are those with all substituents aligned, which also have the highest symmetry (C5 for pillar[5]arene and C6 for pillar[5]arene). In all other conformations, OR groups from adjacent subunits must be arranged in close proximity, which is sterically unfavorable. The flipping of all aromatic rings causes the interconversion of the two enantiomeric Cn symmetric conformations of pillararenes (Figure 4.65a). The rate of this conformational equilibrium depends on the size of the aromatic substituents but these substituents have to be

188

4 Hosting ions and molecules

(a)

(b)

Not involved in hydrogen bonding

Figure 4.65: Selected calculated conformations of pillar[5]arene and pillar[6]arene. In (a), the enantiomeric C5 symmetric conformations of a pillar[5]arene with identical substituents are shown. Stereochemically, the left structure represents the pR,pR,pR,pR,pR enantiomer and the right one the respective pS,pS,pS,pS,pS enantiomer. In (b), the preferred conformations of a pillar[5]arene (left) and pillar[6]arene (right) with free OH groups are depicted. The OH group that is not involved in a hydrogen bond in the pillar[5]arene is marked.

much larger than in calixarenes to prevent the conformational interconversion. Dodecyl chains, for example, are still not large enough since they can wriggle through the cavity of a pillar[5]arene. Not until the pillar[5]arene carries bulky cyclohexylmethyl substituents is it conformationally stable, allowing the chromatographic separation of the possible isomers [114]. This situation is different for pillararenes with free OH groups. In the case of pillar[6]arene, the even number of subunits allows a conformation in which all OH groups engage in intramolecular hydrogen bonds. The respective conformation is a meso form in which the orientations of the aromatic rings with respect to the ring axis alternate (Figure 4.65b). By contrast, an uninterrupted seam of hydrogen bonds is impossible in the unsubstituted pillar[5]arene because of the odd number of subunits. The most stable conformation is therefore that with the maximum number of hydrogen bonds, comprising four subunits arranged in an alternating fashion. At room temperature, pillar[5]arene is flexible but the conformational equilibrium can be sufficiently slowed down by lowering the temperature to observe the unsymmetrical conformer by 1H NMR spectroscopy. Nomenclature: The ring size of pillararenes is specified like that of calixarenes by including the number of subunits in square brackets between the two components of the

189

4.1 Receptors

name. The five- and six-membered unsubstituted pillararenes are therefore called pillar[5]arene and pillar[6]arene, respectively. Substituents are denoted by using the same prefixes as in the names of the individual aromatic building blocks. Thus, the pillar[5] arene derived from 1,4-dimethoxybenzene is termed 1,4-dimethoxypillar[5]arene, although this name is technically not correct since the respective pillararene contains ten and not two methoxy groups. Pillararenes with different substituents along the ring are sometimes called co-pillararenes. To specify the position of the substituents in these compounds, each aromatic subunit is assigned a letter and the cavity openings are distinguished by using the numbers 1 and 2. An A1/B1 co-pillararene accordingly has two substituents on the same side of the ring in neighboring aromatic residues. Synthesis: Most pillararene syntheses are based on the treatment of 1,4-dialkoxybenzenes with paraformaldehyde in the presence of a Brønsted or Lewis acid [113]. Mechanistically, the reaction proceeds similar to the formation of resorcinarenes. The first step is an electrophilic aromatic substitution, leading to a benzyl alcohol derivative (Figure 4.66a). Since all positions in symmetrical 1,4-dialkoxybenzenes are equivalent, the product invariably contains the new substituent in the ortho position of an alkoxy group. The introduction of the next substituent, which can occur in any step of the synthesis, is then controlled by a combination of steric and electronic effects. All remaining positions in the benzyl alcohol are adjacent to an alkoxy group

OMe

(a)

O MeO H+

O

OH

H

H

‒H

Most strongly OMe activated

+

H Acid

H + H OMe Sterically hindered

OMe OMe OH HO

+ +H ‒ H2O

OMe

OMe

MeO

‒H HO

OR

Lewis or Brønsted acid paraformaldehyde

RO

OR

etc.

HO

+

OMe

OMe

Br

OH

(b)

OMe

+

OMe

OMe

RO

H OH

RO

OR

OEt RO

OR

EtO Lewis or Brønsted acid

Lewis acid

Brønsted acid

Figure 4.66: Mechanism of the acid-catalyzed reaction between 1,4-dimethoxybenzene and formaldehyde, explaining the preferred substitution pattern of the aromatic subunits in pillararenes (a), and overview of the reaction conditions of pillararene syntheses starting from different aromatic precursors (b).

190

4 Hosting ions and molecules

and therefore activated, but the position adjacent to the benzyl alcohol group is sterically hindered and the next reaction therefore does not occur there. The para position of the CH2OH group is more strongly activated than the corresponding meta position so that the final product preferentially contains the two new substituents in positions 2 and 5 of the ring. Note that the sequence in which the various electrophilic substitutions occur is not important. All steps are reversible and the structure of the preferred product is therefore subject to thermodynamic control. In his first pillar[5]arene synthesis, Ogoshi used BF3·OEt2 as Lewis acid and 1,2dichloroethane as solvent and isolated the product with a yield of 22% [112]. The influence of the acid, the solvent, and other parameters on product formation was subsequently systematically studied, leading to optimized conditions that allow the synthesis of pillar[5]arene in yields above 80%. The correct choice of the solvent is crucial. 1,2-Dichloromethane typically affords the five-membered ring in high yields, whereas more bulky solvents such as chloroform are unsuitable. This influence of the solvent is likely due to a template effect. Being a linear molecule, 1,2dichloroethane binds well to pillar[5]arene and thus templates its formation. By contrast, the bulkier chloroform is too large to be included into the pillar[5]arene cavity and therefore impairs the cyclization of the respective precursor. Pillar[6]arenes are usually more difficult to access but with the right choice of template, they can also be synthesized on a preparative scale. In chloroform, for example, the reaction of 1,4-diethoxybenzene with paraformaldehyde and FeCl3 as catalysts affords ca. 30% of each, the corresponding pillar[5]arene and pillar[6]arene. When using 1,4-bis(cyclohexylmethoxy)benzene, paraformaldehyde, BF3·OEt2, and working in chlorocyclohexane as a large bulky template, the respective pillar[6]arene is obtained in a high yield of 87% with only 3% of pillar[5]arene as side product. Besides 1,4-dialkoxybenzenes, other aromatic precursors can also be employed for pillararene syntheses. When using 2,5-dialkoxybenzyl alcohols or 2,5-dialkoxybenzyl bromides, no paraformaldehyde is required because the methylene groups are already present in the starting materials (Figure 4.66b). When using 1,4-alkoxy-2,5-bis(ethoxymethyl)benzenes and a Brønsted acid, the reaction proceeds via an ipso-substitution accompanied by the release of diethoxymethane [115]. Binding Properties: The intrinsic receptor properties of pillararenes are characterized by the substantial negative electrostatic potential of their inner cavity surfaces, which results from the electron-rich aromatic subunits (Figure 4.67a). Substrate binding therefore mainly relies on charge-transfer interactions with electron-poor πsystems, CH–π interactions, and cation–π interactions. The substrate scope of pillararenes thus resembles that of the other cyclophanes discussed before. The major difference to calixarenes and resorcinarenes is that pillararenes have cavities open at both ends. Substrate molecules can therefore thread through the annulus, which renders the binding mode of pillararenes more similar to that of cyclodextrins or cucurbiturils than to that of many other cyclophanes.

191

4.1 Receptors

(a)

(b) R RO O

R R O O

4.55a R O

4.55a + N – H 2 PF 6

CN

NC

log Ka = 4.4 (in CDCl3) O

R

O OR R

4.55a R = CH3 4.55b R = H

O O R R

4.55b PF6 + N C8H 17

log Ka = 3.0 (in CDCl3) +

–

4.55b

+ H 17 C 8 N

log Ka = 3.1 (in CH3OH)

+ 2 PF 6 N C 8H 17

–

log Ka = 4.1 (in CH3OH)

Figure 4.67: Electrostatic potential surface of the pillar[5]arene 4.55a derived from 1,4dimethoxybenzene (a), and representatives substrates that interact with the pillararenes 4.55a and 4.55b together with the binding constants of the corresponding complexes (b). The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

Pillar[5]arene prefers to bind to linear aliphatic and monocyclic aromatic substrates, showing that the comparable diameter of the cavity to that of α-CD leads to an analogous substrate selectivity. Examples of substrates that bind to pillar[5] arene 4.55a and the analog with free OH groups 4.55b are collected in Figure 4.67b, together with information about the stabilities of the corresponding complexes. The respective binding constants are substantial, illustrating that pillararenes represent efficient receptors even for substrate whose binding involves relatively weak types of interactions. Because of the larger cavity diameter, pillar[6]arene prefers to bind bulkier substrates containing, for example, bicyclic aliphatic residues or quaternary ammonium ions. Derivatives: The structural variation of the side chains of pillararenes serves to influence binding properties or to link pillararenes to surfaces or other receptors. Many pillararenes are accessible directly from the respective 1,4-alkoxybenzene derivatives, but the introduction of functional groups along the cavity openings by the post-functionalization of suitable precursors is also possible. These conversions often start with the deprotection of the OH groups in a permethylated or otherwise alkylated pillararenes by treatment with boron tribromide. The modification of the free OH groups is then achieved in a number of ways, some of which are shown in Figure 4.68. The alkylation with propargyl bromide, for example, affords pillararenes with terminal alkyne groups that can be further converted into 1,2,3-triazoles under the conditions of the copper(I)-catalyzed azide–alkyne cycloaddition. The sulfonylation of the OH groups with triflic anhydride leads to valuable precursors for Pd(0) catalyzed cross-coupling reactions such as the Sonogashira–Hagihara reaction. The carboxymethylation with chloroacetates gives esters that can either be hydrolyzed to afford water-soluble pillararene carboxylates or converted into positively charged pillararenes containing quaternary ammonium ions by using the

192

4 Hosting ions and molecules

N N

O

O Br R

OH BBr3

Cl

RO

N

CO2Na

O

O NaOH

O

OTf2

CO2CH3 N N

CO2CH3

HO

NaO2C

OH O

TfO

O R

O

LiAlH4

H3CO2C OTf

R PdCl2(PPh3)2 Et3N

R

O

R-N3 Cu(I)

O

OR

N

PBr 3

HO

+ Br

N

O

O N(CH3 ) 3

O R

Br

O N +

Figure 4.68: Examples of synthetic methods available to vary the substituents in pillararenes.

reaction sequence shown in Figure 4.68. In addition, a number of strategies are known to prepare co-pillararenes with different substituents in defined positions along the ring [113, 116]. Possible structural variations are therefore manifold, which explains why pillararenes have found widespread use in supramolecular chemistry within only a few years. Applications in supramolecular chemistry are in the fields of self-assembly, sensing, or the development of interlocked molecules, including molecular machines. Of the large number of known pillararene derivatives, we focus on only two, namely the polyanionic and polycationic water-soluble ones containing carboxylate groups and quaternary ammonium ions in the periphery, respectively, to show the influence of the substituents on the binding properties. The ionic groups in these compounds either induce an affinity for cations, which are anyway the preferred substrates of pillararenes, or anions, for which pillararenes normally have no affinity. In the case of 4.56 (Figure 4.69), the negatively charged groups along the cavity enhance cation affinity. The complex between 4.56 and paraquat (N,Nʹ-dimethyl-4,4ʹ-

193

4.1 Receptors

R RO O

R R O O

R O

N

ONa

4.56 R =

O

R

O O R R

4.57 R =

N Paraquat

O O OR R

+

+

N +

Br

–

2 Cl

–

O S ONa O Sodium 1-octylsulfonate

Figure 4.69: Molecular structures of water soluble pillar[5]arenes 4.56 and 4.57 and structures of the corresponding preferred guests.

bipyridinium) has a log Ka of 4.9 in water, for example, which is larger than the log Ka of the complex between the neutral pillararene 4.55b and N,N’-dioctyl-4,4ʹbipyridinium in methanol (Figure 4.67b) [117]. If the pillar[6]arene with 12 carboxylate groups along the cavity is used, the paraquat complex has a log Ka of 8.0 [118], and the log Ka of the complex between paraquat and a pillar[7]arene with 14 carboxylate groups amounts to 9.5 [119] due to the better fit of the substrate into the cavities of the larger pillararenes in combination with the reinforcement of complex stability by the larger number of negatively charged substituents. Quaternary ammonium ions along the cavity opening induce affinity for anionic substrates. These groups thus overcompensate the intrinsic substrate affinity of pillararenes. Compound 4.57 (Figure 4.69) binds to 1-octane sulfonate with a log Ka of 4.1, for example [120]. Complex formation produces pronounced upfield shifts in the NMR spectrum of the signals of the protons in the octyl group of the guest. These protons give rise to eight distinct signals, with those belonging to the middle region of the octyl group exhibiting the largest upfield shift. Accordingly, the linear guest is threaded through the cavity of 4.57 such that its negatively charged head is oriented close to the cationic groups in the substituents.

4.1.11 Cucurbiturils Introduction: We have seen that the history of phenol formaldehyde-derived cyclophanes goes back to Adolf von Baeyer whose work laid the foundation for one of the first commercial plastic materials. Interestingly, the last family of macrocyclic receptors to be discussed has similar roots although its members differ structurally and in terms of binding properties from cyclophanes. Their discovery is linked to another synthetic polymer developed in the early days of polymer chemistry, namely the ureaformaldehyde or UF resin. In 1905, the group of Rolf Behrend in Hannover tested how glycoluril 4.58, a bicyclic bis(urea) derivative, behaves when treated with formaldehyde under the typical polymerization conditions (Figure 4.70a) [121]. They isolated a white crystalline material from concentrated sulfuric acid that proved to be very stable

194

4 Hosting ions and molecules

(a)

(b) O

HN H HN

NH H NH O

4.58

OO O N N N NN N

OO

O

N N N N NN

N N NN N N N N N N NN O O O OO O 4.59

Figure 4.70: Molecular structures of glycoluril 4.58 and cucurbit[6]uril 4.59 (a), and side and top view of the crystal structure of 4.59 (b).

and insoluble in water but formed complexes with a number of salts. Since the structure of this product could not be determined at that time, it initially entered the literature under the name Behrend’s polymer. In 1981, William L. Mock reproduced Behrend’s original procedure and showed, by using the modern analytical techniques at his disposal, that the product has the structure of 4.59 (Figure 4.70a) [122]. Mock thus demonstrated that Behrend’s polymer is actually a macrocycle with six glycoluril subunits linked at the nitrogen atoms via methylene units. According to the crystal structure, the glycoluril subunits are oriented with their protons pointing outward and with their carbonyl groups converging at both cavity openings (Figure 4.70b). The ring thus has a convex shape around the equator and a cavity that can be accessed from both sides. Mock suggested naming this compound cucurbituril because of its resemblance “to a gourd or pumpkin (family Cucurbitaceae), and by devolution from the similarly named (and shaped) component of the early chemists’ alembic” (i.e., the round bottom flask in modern chemistry) [122]. The comparison of the cucurbituril structure with the shape of a container suggests that Mock saw potential in cucurbiturils to act as receptors and he (and a few other groups) indeed continued to systematically study the binding properties of these compounds. Unfortunately, the studies had to be carried out in rather acidic solvent mixtures such as aqueous formic acid because of the low solubility of 4.59 in other media. This situation changed when the groups of Kimoon Kim and Anthony Day achieved the synthesis and isolation of other cucurbituril homologs, some of which are considerably better soluble than 4.59. Today, the cucurbituril family ranges from the five-membered to the ten-membered ring and an even larger 14-membered cucurbituril has also been described. With the availability of these compounds and the realization that some of them exhibit exceptional binding strength in aqueous media, cucurbiturils gradually developed into a popular family of receptors with the potential to substitute cyclodextrins in certain applications [123]. The structure of cucurbiturils does not change substantially with ring size. All homologs up to the octamer have a cylindrical shape, closely related to the structure of

4.1 Receptors

195

4.59. The height of the ring remains constant as it is determined by the distance between the carbonyl groups of the glycoluril units. The cavity diameter increases with increasing number of subunits and is larger in the equatorial region than at the cavity openings where the carbonyl groups reside. Figure 4.71 summarizes the structural parameters of the most important cucurbituril derivatives. This figure also shows the electrostatic potential surface of 4.59. Accordingly, the cavity openings of cucurbiturils have a pronounced negative potential, indicating that they can serve as binding sites for appropriately sized cations. The inner surface of the cavity is slightly negative and its concave shape makes it difficult for a guest molecule to make efficient contact. Note also the pronounced positive potential of the convex outer surface in the region of the methine protons, which explains the tendency of certain anions to associate with the outside of cucurbiturils, and the anion affinity of macrocyclic receptors that feature inverted glycoluril units such as the later discussed bambusuril.

(a)

h

(b)

dopening OO O N N N NN N N N NN N N O O O

OO

x O

N N N N NN NN N N N N OO

douter

CB[n]

O x

douter dopening h (Å) (Å) (Å)

V (Å3 )

0

CB[5]

4.4

2.4

9.1

68

1

CB[6]

5.8

3.9

9.1

142

2

CB[7]

7.3

5.4

9.1

242

3

CB[8]

8.8

6.9

9.1

367

5

CB[10]

11.7

10.0

9.1

691

Figure 4.71: Molecular dimensions of glycolurils (a) and electrostatic potential surface of cucurbit[6]uril 4.59 (b). The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively. The values of douter, dopening, and h take into account the van der Waals radii of the relevant atoms.

Nomenclature: The family name cucurbituril refers to the compounds derived from unsubstituted glycoluril 4.58. Ring size is denoted as we have seen for cyclophanes by specifying the number of glycoluril subunits in square brackets between the two parts of the name. Accordingly, compound 4.59 is a cucurbit[6]uril or CB[6] for short. For cucurbiturils with substituents along the equator, the number, type, and positions of the substituents must be specified and there are various ways of doing this. Especially straightforward is the naming of fully substituted cucurbiturils with only one type of substituent. A CB[5] with methyl groups in all glycoluril subunits is a decamethylcucurbit[5]uril or permethylcucurbit[5]uril, for example. Cucurbiturils with cyclohexane rings in the periphery are cyclohexanocucurbiturils. Synthesis: The synthesis of cucurbiturils involves the treatment of glycoluril or glycoluril derivatives with formaldehyde under acidic conditions. The required glycolurils are condensation products of 1,2-dicarbonyl compounds and urea. Unsubstituted 4.58

196

4 Hosting ions and molecules

is obtained from glyoxal, for example, dimethylglycoluril from butane-2,3-dione and cyclohexanoglycoluril from cyclohexane-1,2-dione. In the original CB[6] synthesis used by Behrend and Mock, the reaction between glycoluril and formaldehyde was carried out in hot concentrated H2SO4, affording only the six-membered ring. Product formation involves the initial hydroxymethylation of one or more nitrogen atoms of glycoluril (Figure 4.72). Under the acidic conditions, the hydroxymethyl groups eliminate water, affording cationic intermediates that react with a nitrogen atom from another glycoluril molecule, leading to the bridging of two subunits. Cyclic ethers such as compound A could also represent

O HN H HN

A

N H O N O

H H H N N O

H+

O H

N N H H H + –H

OH

H

O

O

O HN H HN

H + H

N

OH

Acid – H2O O

H H Acid

HN H HN

H NH

N N

H H H N N O

HN H HN

N

N

NH H NH

H H N O

O

O

N B exo

Acid – H 2O

HN H HN

O

O N

N

N

H H HN

N

N

NH H NH

H H N O

B endo

N

O + N

HN H HN

H N

OH

O

O Acid – H2 O O

O

+ +H H – 2O

O N N H H H + –H

OH

O

HN H HN

NH H NH

OH

O

O

O

OH H

etc.

O

Figure 4.72: Mechanism of the acid-catalyzed condensation reaction between glycoluril and formaldehyde. Possible intermediates on the way to the macrocyclic product such as the diether A or the stereoisomers of the glycoluril dimers Bexo and Bendo are marked.

4.1 Receptors

197

possible intermediates. All of these intermediates are rapidly consumed in the course of the reaction, initially affording ribbon-like oligomers that subsequently cyclize once a certain chain length is achieved. Since the underlying reactions are reversible, possible errors in the synthesis can be corrected. The bridging of two glycolurils leads to a C-shaped product Bendo, for example, with an endo arrangement of the two subunits, but also to the S-shaped isomer Bexo. The latter does not have the required curvature to allow cyclization but it can be transformed in the course of the reaction into the correct stereoisomer and further into the product. Under the harsh conditions employed, CB[6] is likely thermodynamically preferred although its formation is possibly also mediated by template effects of cationic species present in the reaction mixture. Kim and Day later showed that performing the reaction under milder conditions gives rise to a mixture of different cucurbiturils in which CB[6] typically dominates, but which also contains CB[5], CB[7], and CB[8] in useful quantities. Interestingly, also macrocycles are isolated from these mixtures containing one inverted glycoluril residue whose methine protons are oriented toward the interior of the cavity. These side products are believed to represent kinetic intermediates because they eliminate the incorrectly incorporated subunit in a ring contraction process when treated with acid at higher temperature, thus affording the next smaller ring. The product ratios in these syntheses depend on the solvent, the temperature, the type of acid, and also on whether templates such as metal ions are present or not [124]. In combination with suitable separation techniques, which take advantage of the different solubilities of the cucurbituril homologs, their receptor properties, and their elution behavior on suitable stationary phases, the available procedures permit the synthesis of many members of this receptor family on a preparative scale. Binding Properties: Cucurbiturils have a nonpolar cavity that is lined at both openings by cyclically arranged oxygen atoms. These atoms converge, suggesting that they engage in similar ion–dipole interactions with metal ions as the oxygen atoms in crown ethers. Indeed, 4.59 binds to alkali and alkaline earth metal ions with a selectivity that depends on ring size as shown by the stability constants collected in Table 4.8 [125]. While CB[5] exhibits peak selectivity for K+ similar to

Table 4.8: Comparison of the affinity of cucurbiturils and 18-crown-6 to various metal ions in terms of log Ka values. The binding constants were measured in water at 298 K. receptor

Li+

Na+

K+

Rb+

Cs+

Ca+

CB[] CB[] (.) CB[] CB[] -crown-

. . . . –

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

198

4 Hosting ions and molecules

log Ka

18-crown-6, the cation selectivity of the larger cucurbiturils is less pronounced. In general, however, all cucurbiturils bind to cations in water with a pronounced affinity. The relatively shallow selectivity of the larger congeners is attributed to the diameters of their cavity openings that are too large to allow the oxygen atoms to make efficient contact with the metal ions (the largest ion has a radius of 1.69 Å [cf. Table 3.1)], but the radius of the cavity opening of 4.59 is 1.95 Å (dopening = 3.9 Å). In contrast, the diameter of 18-crown-6 is significantly smaller (Table 4.1). Cation affinity explains why cucurbiturils are generally much better soluble in aqueous saline solutions such as aqueous Na2SO4, LiCl, KCl, CsCl, or CaCl2 than in water. The cations of these salts bind to the oxygen atoms at both cavity openings and thus mediate cucurbituril hydration. These metal ions furthermore seal off the cavity, causing apolar guest molecules such as THF to remain entrapped. The corresponding complexes are destroyed by the addition of acids that protonate the carbonyl oxygen atoms and thus prevent them from coordinating to the cation. Early investigations by Mock also revealed a high affinity of 4.59 for ammonium ions (in H2O/formic acid, 1:1 (v/v)) whose cationic head groups bind to the cyclically arranged carbonyl groups by ion–dipole interactions [126]. In this context, a characteristic dependence of complex stability on the length of the alkyl chain in primary alkyl ammonium salts or in α,ω-diammonium alkanes was found, which is shown in Figure 4.73.

7 6 5 4 3 2 1 0

2

4

6

8 10 Chain length n

Figure 4.73: Dependence of the stability of the complexes of 4.59 in H2O/formic acid, 1:1 (v/v) with alkyl ammonium ions H–(CH2)n–NH3+ (red circles) and α,ω-diammonium alkanes + H3N–(CH2)n–NH3+ (blue circles) on the chain length n.

Accordingly, the stability of the alkyl ammonium complexes increases strongly by ca. three orders of magnitude with increasing chain length until the butyl derivative. A further chain length increase then causes a pronounced drop in affinity with the heptyl derivative binding as weakly to 4.59 as the ethyl ammonium ion. The trend for α, ω-diammonium alkanes is similar but the maximum affinity is in this case

4.1 Receptors

199

observed for the substrates with five or six carbon atoms between the cationic head groups. The maximum log Ka value is moreover almost two orders of magnitude larger than that of the most stable alkyl ammonium complex. These trends reflect structural aspects of the complexes. Since guest protons oriented close to the carbonyl groups are deshielded and those included deep into the apolar cavity are shielded, the signal shifts observed in the 1H NMR spectra upon complex formation indicate that both types of substrates bind with their ammonium groups to the carbonyl groups of 4.59 while the alkyl groups occupy the cavity. The higher stability of the complexes of dialkylammonium ions is due to the fact that they are stabilized by ion–dipole interactions at both cavity openings. Among the monoammonium ions, the guest with a four-carbon chain binds strongest because it has the right length to fill the cavity. Ammonium ions with shorter alkyl chains are too small to fully replace the water molecules inside the cavity whereas longer alkyl groups protrude from the cavity opening opposite to where the ammonium group is bound. Both situations are energetically unfavorable because they require the hydration of apolar surfaces inside the cavity or of the substrate outside the cavity. In the case of diammonium ions, a chain length is best that allows both ammonium groups to bind at both cavity openings without producing strain, which is the case for the substrates with five and six methylene units. According to calorimetric measurements, complex formation is exothermic with a favorable or unfavorable entropy term, depending on the actual substrate [127]. The formation of the butylammonium complex is, for example, associated with a ΔH 0 of −26.8 kJ mol−1 and a TΔS0 of −2.9 kJ mol−1, whereas these parameters amount to −29.6 kJ mol−1 and 8.9 kJ mol−1, respectively, in the case of the 1,6-hexyldiammonium complex. We have seen similar thermodynamic signatures for the formation of cyclodextrin, cyclophane, and cavitand complexes and we will see below that they are also for cucurbiturils consistent with the notion of “high-energy” cavity water that is released upon complex formation. Note also that the complexes of 4.59 with α,ω-diammonium alkanes approach stability constants log Ka of up to 7, which is unusually high for synthetic receptors. Such high affinities are in fact characteristic for cucurbiturils, especially for CB[7], which binds to ferrocene derivatives, for example, with log Ka values up to 15.5 in water [128]. The record in terms of binding strength is held by the CB[7] complex of a diamantane diammonium ion, which with a log Ka of 17.8 has an almost attomolar stability [129]. Figure 4.74a shows a selection of guests that form such high-affinity complexes with CB[7] together with the thermodynamic parameters associated with complex formation.

200

4 Hosting ions and molecules

(a)

(b) N + N

N

+

+

Fe

+

NH3

Fe +

N +

N log Ka = 1 2.6 log Ka = 1 5.5 log Ka = 1 4.3 ΔH° = –90 kJ mol–1 ΔH° = –90 kJ mol–1 ΔH° = –81 kJ mol–1 –T Δ S° = 1 8 kJ mol–1 –T Δ S° = 2 kJ mol–1 –T Δ S° = 0 kJ mol–1 log Ka = 1 7.8

Figure 4.74: Structures of selected substrates that form highly stable complexes with CB[7] together with the thermodynamic parameters associated with complex formation (a), and crystal structure of the diamantane complex of CB[7] (b).

How do water molecules in a receptor cavity contribute to complex formation?

The extremely large affinities of these CB[7] complexes potentially arise from the tight fit of the guest molecules into the rigid cucurbituril cavity, which is illustrated by the crystal structure of the diamantane complex in Figure 4.74b. CB[7] is moreover very well preorganized for complex formation, rendering the loss of flexibility associated with guest binding small. A further contribution to complex formation could derive from the entropic gain caused by receptor dehydration. However, complex formation has a relatively small entropic component while the main thermodynamic driving force comes from the substantial negative enthalpy term. This strong exothermicity of complex formation is consistent with the tight interactions between receptor and substrate but it is striking that the binding enthalpy is relatively independent of the substrate structure. Therefore, another explanation for the exceptional binding properties of CB[7] was proposed by Werner M. Nau and Frank Biedermann, which is applicable also to other receptors exhibiting exothermic complex formation in water [31, 130]. The underlying ideas were briefly presented in Section 3.3, where the concept of “high-energy water” was introduced. This concept compares the situation of water molecules occupying concave regions of a molecule such as the cavity of a receptor with the situation in bulk water. In the bulk, every water molecule is engaged on average in 3.62 hydrogen bonds with surrounding water molecules, but this structure cannot be retained inside the confined cavities of receptors that are lined with apolar subunits. Depending on the size of these cavities, only a certain number of water molecules can be included, which mainly bind to themselves, causing the number of hydrogen bonds per water molecule to be lower than in bulk water. Based on these considerations, Nau and Biedermann estimated the average number of water molecules N occupying the cavities of typical receptors by molecular dynamics calculations. In addition, the respective water clusters were also structurally characterized to assess how many hydrogen bonds m per water molecule

4.1 Receptors

201

they contain (Figure 4.75) [131]. Results of these calculations are summarized for selected receptors in Table 4.9. 3

5

4

6

8

Figure 4.75: Examples of water cluster structures containing between three and eight water molecules.

Table 4.9: Parameters describing the cavity hydration of different receptors. N is the average number of water molecules in the respective cavity, m refers to the average number of hydrogen bonds per bound water molecules, and Z is the combined hydrogen bond deficit of these water molecules in comparison to bulk water. Z derives from equation (4.1). Receptor

N

m

Z

α-CD β-CD Calix[]arene Pillar[]arene CB[] CB[] CB[]

. . . . . . .

. . . . . . .

. . . . . . .

The parameters N and m now allow calculating by using equation (4.1), how many hydrogen bonds are regained overall if the corresponding water cluster is released from the receptor cavity into the bulk. In this equation, the number 3.62 denotes the average number of hydrogen bonds per water molecule in bulk water. Z = N ð3.62 − mÞ

(4:1)

According to Table 4.9, complex formation of all considered receptors should benefit more or less strongly from the release of cavity water and the associated recovery of hydrogen bonds. Because the formation of these hydrogen bonds represents an enthalpic gain, the release of cavity water contributes with a negative enthalpy term to

202

4 Hosting ions and molecules

complex formation. The overall binding enthalpy might still end up to be positive if other factors overcompensate this effect but if not, complex formation should be exothermic as indeed observed for many receptors in water. Table 4.9 also shows that the enthalpy gain of complex formation should be largest for cucurbiturils because their cavities contain more water molecules than those of other receptors. CB[7], in particular, has a most unfavorable combination of N and m values. There are other receptors where the water molecules occupying the cavities engage in even fewer hydrogen bonds, but these receptors also contain less cavity water so that Z, the number of hydrogen bonds regained when the cavity water is released, ends up to be smaller than for CB[7]. Receptors with larger cavities than CB[7] bind more water molecules whose overall hydrogen bond deficit is, however, not so pronounced. Thus, CB[7] should benefit the most from the release of cavity water among the considered receptors (Table 4.9), which is indeed observed. This concept thus provides a rationale why cucurbiturils are so potent receptors in aqueous solution, furthermore stressing the importance of considering solvent effects when interpreting binding properties. The substrate scope of cucurbiturils is immense [132], ranging from noble gases and simple aliphatic hydrocarbons to biologically relevant molecules such as amino acids, peptides, neurotransmitters, hormones, drugs, and toxins. Complex formation generally benefits from the release of cavity water but if ion–dipole interactions additionally contribute to stability as in the case of cations, the complexes are typically more stable by a factor of 10 to 100 than those of neutral substrates. A further interesting aspect of cucurbituril host-guest chemistry is that the large CB[8] even allows the incorporation of two guests, either two individual substrate molecules or two substituents originating from the same molecule. These complexes are particularly stable if they are reinforced by interactions between the two guests, which is the case if an electron-rich and an electron-poor aromatic π-system is bound. Examples are the complexes of CB[8] containing a paraquat moiety and an additional electron-rich aromatic guest such as 2,6-dihydroxynaphthalene, 1,4-dihydroxybenzene, tyrosine, dopamine, or tryptophan. To illustrate the binding mode in such complexes, the crystal structure of CB[8] incorporating 2,6-dihydroxynaphthalene and a paraquat derivative is depicted in Figure 4.76 [133]. The formation of these complexes is accompanied by the development of a charge-transfer band in the UV-vis spectrum. Accordingly, CB[8] represents a “molecular glue” that holds together appropriate binding partners in solution, for example polymers or peptides with suitable side chain substituents. Complex formation can moreover be conveniently monitored by the associated color change. Derivatives: One drawback of cucurbiturils is that they do not easily allow postfunctionalization because of the lack of functional groups. Structural variations therefore usually involve changes at an early stage of the synthesis, for example by using glycoluril precursors with different substituents. One method of postfunctionalization is the oxidation of the outward oriented methine protons of 4.59 with K2S2O8 to obtain the corresponding perhydoxylated CB[6] whose OH groups are available for further

4.1 Receptors

203

O OH OO O N N N NN N N N NN N N O O O

OO

+

O

N

HO

N N N N NN NN N N N N OO

O 3

+ N

OH

Figure 4.76: Molecular structures of CB[8], 2,6-dihydroxynaphthalene, and a paraquat derivative that form a ternary complex whose crystal structure is also shown.

modification. This oxidative method also allows producing monohydroxylated cucurbiturils, which represent valuable precursors for linking cucurbiturils to other molecules or supports. It is also possible to use a mixture of different glycolurils in cucurbituril syntheses, but these syntheses afford mixtures of products differing in the ratio and arrangement of the subunits along the ring. On the whole, the methods to structurally vary cucurbiturils are more limited than those available for other macrocycles and we therefore concentrate in this section on cyclic analogs of cucurbiturils that have gained importance as receptors. Acyclic cucurbiturils combining structural motifs of cucurbiturils and cyclophanes also exist and are discussed in Section 4.1.12. Hemicucurbiturils are derivatives of cucurbiturils containing ethyleneurea instead of glycoluril subunits. Each hemicucurbituril subunit can be regarded as half the subunit of a cucurbituril and since hemicucurbiturils therefore lack the second stabilizing row of methylene units, they are flexible. They prefer conformations with an alternating up-down arrangement of the carbonyl groups as shown for the hexamer in Figure 4.77a. The ethyleneurea subunits are moreover tilted inward, causing protons in the ethylene units to point into the cavity. These receptors thus feature significantly different binding properties than cucurbiturils, including the affinity for anions that is not observed for cucurbiturils. This difference is even more pronounced for another class of cucurbituril derivatives in which one ring of each glycoluril subunit contains alkylated nitrogen atoms so that only the other two nitrogen atoms are available for the linkages. Accordingly, the corresponding hexamer 4.60 (Figure 4.77b) has, like hemicucurbiturils, only one row of methylene groups and features alternately arranged glycoluril subunits that orient their convex sides toward the cavity interior. Regions with a positive potential, which reside on the outside of cucurbiturils (Figure 4.71b), thus line the inner cavity surface, inducing an affinity for anions (Figure 4.77c). Vladimir Sindelar, who first described receptor 4.60, proposed the

204

4 Hosting ions and molecules

(a)

O N

NO N N

N

O

N

N

O N

O N

N

N

N

O (b)

(c)

O N

N

O N O

NO N N N O

N

N

N

O N

NN NO N

O

O

N

N N

O N

N

N N

4.60

O

N N

(d)

O

Figure 4.77: Molecular structure and calculated structure of a hexameric hemicucurbituril (a), molecular structure of bambus[6]uril 4.60 (b), electrostatic potential surface of 4.60 (c), and crystal structure of the chloride complex of 4.60 (d). The protons in both structures are not shown for reasons of clarity. The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

name bambus[6]uril because of the resemblance of this compound to the stem of a bamboo plant [134]. Complex formation of 4.60 involves the incorporation of the anion into the center of the cavity. To illustrate this binding mode, the crystal structure of the chloride complex is shown in Figure 4.77d. Complex stability roughly correlates with the size of the anion, becoming larger as the ionic radius increases. As a consequence, a derivative of 4.60 with benzyl instead of methyl groups attached to the nitrogen atoms binds iodide (log Ka = 10.2) and perchlorate (log Ka = 10.3) with exceptional affinities in chloroform [135]. Importantly, water-soluble analogs of 4.60 have a slightly lower but still extraordinarily high anion affinity in water, placing them among the most potent receptors in this solvent for large, weakly coordinating anions. The thermodynamics of anion complexation indicates that complex formation also benefits from the release of cavity water.

4.1.12 Clefts and tweezers Introduction: So far, we have mostly discussed cyclic receptors and, indeed, most receptors in supramolecular chemistry are macrocyclic. The reason is that such

4.1 Receptors

205

receptors are often better preorganized for substrate binding and therefore exhibit higher substrate affinity than acyclic counterparts. Recall in this context the concept of the macrocyclic effect that was introduced in Section 3.5.2 and illustrated in Section 4.1 by comparing the binding properties of crown ethers and podands. Are acyclic receptors therefore unimportant? The clear answer is no. Nature itself relies on acyclic systems for molecular recognition, the most important of which are proteins and polynucleotides, and synthetic acyclic receptors can also be rather efficient as we will see in this and the following chapters. First, we concentrate on lowmolecular scaffolds in which suitably arranged functional groups produce well preorganized binding sites. In Section 4.1.13 we then discuss oligomeric receptors whose folding, determined in a predictable way by the type and sequence of monomer units, leads to binding properties comparable to those of proteins. The receptors discussed in this chapter have structures with concave regions where substrate binding takes place. Functional groups in the periphery of this binding site participate in substrate binding or mediate substrate selectivity. This concept should be illustrated by using the example of tripodal receptors derived from 1,3,5trisubstituted triethylbenzene [136]. In these systems, the triethylbenzene core serves as a rigid platform to which binding sites are attached in the 1-, 3-, and 5-position. Steric effects of the ethyl groups, which are associated with the term steric gearing, cause the six substituents along the ring to be alternately oriented in an up-down fashion as shown in Figure 4.78a. As a consequence, the groups responsible for substrate binding are arranged on one face of the benzene ring, surrounding a shallow cavity. They are therefore well preorganized to simultaneously engage in substrate binding. This receptor design is very versatile because it allows the recognition of substrates of widely different nature by just adapting the functional groups along the benzene ring. The receptors 4.61 and 4.62 in Figure 4.78b were designed, for example, for the recognition of cations (NH4+) and anions (citrate), respectively. We have seen other receptors 0 parameter (Figure 2.8). In Section 4.2.4 we of this type in Section 2.1 about the BC50 will see that they are also used to recognize monosaccharides.

(a)

+ HN

(b)

R

N N

N N

R

N N

R 4.61

H N +

HN NH NH

H H N + N

N H

N H 4.62

Figure 4.78: Preferred conformation of a 1,3,5-trisubstituted triethylbenzene derivative, illustrating the effect of the steric gearing (a), and molecular structures of the tripodal receptors 4.61 and 4.62 designed for the recognition of, respectively, an ammonium ion and citrate (b).

206

4 Hosting ions and molecules

Various terms are in use for acyclic receptors. Examples are molecular clefts, tweezers, pincers, or clips, not all of which are exactly defined and which are therefore sometimes used synonymously. In the following paragraphs, a selection of receptor families is presented in more detail. Rebek’s clefts: Julius Rebek Jr. described a number of preorganized acyclic receptors whose binding properties benefit from the characteristic structure of cis,cis1,3,5-trimethylcyclohexane-1,3,5-tricarboxylic acid or Kemp’s triacid [137, 138]. This cyclic tricarboxylic acid 4.63 (Figure 4.79) is locked into one ring conformation by the three methyl groups, which are bulkier than the planar carboxy groups and therefore occupy the equatorial positions. As a consequence, the three carboxy groups are forced into the axial positions of one ring face, rendering 4.63 an ideal basis for the development of receptors with converging binding sites. The strategy used by Rebek involved connecting two subunits of 4.63 to rigid aromatic scaffolds in such a way that their carboxy groups face each other, producing a binding site into which the substrate is included. Examples of such receptors

CO2H O N O

HO2C O N O

4.64a

CO2H CO2H CO2H

HO2C O N O

CO2H O N O

4.63

4.64b

CO2H O N O

N

HO2C O N O

4.64c

Figure 4.79: Molecular structure of Kemp’s triacid 4.63 and molecular and calculated structures of the clefts 4.64a, 4.64b, and 4.64c.

4.1 Receptors

207

are the bis(imides) 4.64a, 4.64b, and 4.64c (Figure 4.79), which are obtained by treating 4.63 with aromatic diamines. The methyl groups in the aromatic linkers prevent the rotation around the adjacent Ar–N bonds. They thus play a decisive role in receptor preorganization by stabilizing U-shaped structures. Because of the concave binding sites thus produced, these receptors are termed molecular clefts. In receptors 4.64a and 4.64b, the carboxy groups are sufficiently close to interact intramolecularly by hydrogen bonding. After deprotonation, they can simultaneously coordinate to divalent metal ions. Receptor 4.64c has a richer host-guest chemistry because of the larger distance of the two carboxy groups. In its diprotonated form, 4.64c binds to diamines, either by hydrogen bond formation as in the pyrazine complex or by salt bridge formation if the diamines are more basic, causing receptor deprotonation. Both binding modes are shown in Figure 4.80a,b. If the aromatic diamine engages in additional aromatic interactions with the acridine linker, complex stability increases. The log Ka of the pyrazine complex of 4.64c amounts to 3.1 in chloroform, for example, whereas the quinoxaline (1,4-diazanaphthalene) complex has a log Ka of 4.4. Since salt bridges are stronger than hydrogen bonds, the

(a)

(b) N

(c)

N

(d)

NH O

+ HN

HO

NH +

O

O HN

O

OH

Figure 4.80: Binding modes in the complexes of 4.64c with pyrazine (a), 1,4-diazabicyclo[2.2.2] octane (DABCO) (b), diketopiperazine (c), and oxalic acid (d). The binding of DABCO involves the transfer of two protons of the receptor to the substrate.

208

4 Hosting ions and molecules

1,4-diazabicyclo[2.2.2]octane complex is even more stable (log Ka = 5.2). These results demonstrate that the excellent preorganization of 4.64c induces high substrate affinity even in the absence of the macrocyclic effect. Cleft 4.64c also forms complexes with diketopiperazines (Figure 4.80c) and with dicarboxylic acids such as oxalic or malonic acid (Figure 4.80d). Zwitterionic α-amino acids are bound by a combination of hydrogen bonding, ionic, and aromatic interactions (if the amino acid contains an aromatic side chain), but complex formation involves the binding of two receptor molecules to a single substrate in this case. Another receptor type derived from Kemp’s triacid is obtained by treating 4.63 with ammonia to yield 4.65a (Figure 4.81a). This receptor features an A–D–A hydrogen bonding pattern at the imide group, inducing affinity for complementary substrates such as 9-ethyladenine. Its remaining carboxy group can be used for further structural variation. Receptors with different aromatic residues in this position, such as 4.65c, 4.65d, and 4.65e, allow estimating the extent to which aromatic interactions contribute to substrate binding, for example. When using the ΔG0 associated with the formation of the 9-ethyladenine complex of methylamide 4.65b as a reference, the affinities of the other receptors increase by a surprisingly constant energy contribution of ca. 1.8 kJ mol−1 per benzene ring. As a consequence, the anthryl moiety of 4.65e is responsible for ca. one third of the overall binding strength [139]. This series of receptors thus provides quantitative information about possible contributions of aromatic interactions to the stabilization of the DNA double helix. 9-Ethyladenine binds to these receptors either by Watson–Crick and Hoogsteentype base pairing (Figure 4.81b). Since both arrangements are possible, the U-shaped receptor 4.66 in which two imide subunits of 4.65 are arranged in a convergent fashion uses both binding modes for substrate recognition. The 9-ethyladenine complex of 4.66 has a log Ka of 4.0 in chloroform [140], which again demonstrates how efficient well-designed acyclic receptors can be. We will come back to such Kemp’s triacid-derived receptors in Chapters 8 about supramolecular catalysis. Zimmerman’s tweezers: The term “molecular tweezers” was introduced by Howard W. Whitlock for acyclic receptors with two aromatic substituents arranged in a parallel fashion at a distance that allows the intercalation of a suitable substrate [141]. The following examples of such receptors were developed in the group of Steven C. Zimmerman [142]. Receptor 4.67 comprises a 5,6,8,9-tetrahydrodibenz[c,h]acridine spacer to which two 9-anthryl groups are attached (Figure 4.82). The spacer arranges the substituents in a parallel fashion at a distance of ca. 7.2 Å as estimated from the distance of the respective protons in dihydrodibenz[c,h]acridine. While this distance is already well suited for the intercalation of planar aromatic systems (the inter-ring distance of π-systems in donor-acceptor complexes typically ranges between 3.2 and 3.5 Å), it can even be optimized if necessary by small adjustments of the angles between the anthryl units and the spacer 4.67. This receptor binds electron-poor aromatic guests, which are sandwiched between the anthryl groups. The log Ka of the

4.1 Receptors

(a)

Et

Et

N N

H N O OH O

4.65a ΔG°/ kJ

H N N H H CH3 O N O NH O

N N

H N H O

N N H H O N O NH O

N H N O NH O

4.65d

4.65e

–9.7

–11.4

–13.4

–15.1

0.0

–1.7

–3.7

–5.4

(c)

(b)

O

O NH

Et N

N

Et N

N N

H N N H H O N O NH O

Watson-Crick

N N H H O N O NH O

4.65c

mol–1

N

N H

4.65b mol–1

ΔG°Ar/ kJ

N

N N

H

Et

N

N N

O

Et

N

HN H O

209

HN

OO

OO

HN

NH

N H N O NH O

4.66

Hoogsteen

Figure 4.81: Molecular structures of imides 4.65a-e together with the ΔG0 values associated with their binding to 9-ethyladenine in chloroform and the ΔG0Ar values, calculated by subtracting the ΔG0 of 4.65b from the ΔG0 values of the other receptors (a). Structures of the Watson-Crick and Hoogsteen hydrogen bonding pattern between the ADA binding site of 4.64c and 9-ethyladenine are shown in (b), and (c) depicts the molecular structure of receptor 4.66 together with the calculated structure of its 9-ethyladenine complex.

2,4,5,7-tetranitrofluorene (4.68) complex amounts to 4.3 in chloroform, for example. An analog of 4.67 in which one anthryl units is replaced by an acridine unit has a ca. one order of magnitude lower affinity and the replacement of the second anthryl unit with an acridine unit causes a further drop of affinity. This trend shows that complex stability correlates with the electronic properties of the aromatic substituents, becoming lower if electron-rich residues are replaced with electron-poorer ones. By arranging a carboxy group between the two aromatic panels, aromatic interactions can be combined with hydrogen bonding interactions for substrate

210

4 Hosting ions and molecules

recognition. The corresponding receptor 4.69 binds 9-propyladenine with a log Ka of 4.4 in chloroform (Figure 4.82). The complex of the dimethylated N,N-dimethyl -9-ethyladenine has a considerably lower binding constant (log Ka = 2.6), clearly showing that the 9-propyladenine complex benefits from the reinforcement of the aromatic interactions by hydrogen bonding. Whether the binding of the substrate to the carboxy group involves a Watson–Crick or a Hoogsteen-like arrangement (Figure 4.83) can be deduced from the stability of the complex between 4.69 and the monomethylated analog of N,N-dimethyl-9-ethyladenine, N-methyl-9-ethyladenine.

Ar

N

O 4.67 (Ar = C6H5 )

O2N

NO2 O2 N

N Ar

NO2

4.68

COOH N

4.69 (Ar = p-C6 H4 NMe2) Figure 4.82: Molecular structures and calculated structures of molecular tweezers 4.67, 4.69, and molecular structure of 2,4,5,7-tetranitrofluorene 4.68. In the calculated structures, the aryl groups in the spacers are not shown for reasons of clarity.

For steric reasons, this adenine derivative almost exclusively adopts the conformation with the N-methyl group oriented away from the imidazolyl ring (Figure 4.83). It is thus well preorganized for binding to 4.69 in the Hoogsteen arrangement, while it has to adopt an energetically unfavorable conformation for Watson-Crick hydrogen bonding. Since the complexes of 4.69 with N-methyl-9-ethyladenine and 9-propyladenine have practically the same stability, one can conclude that no energetically costly conformational reorganization of N-methyl-9-ethyladenine has to occur prior to complex formation and that the Hoogsteen arrangement is therefore preferred in the complex. A possible explanation for this observation is that this binding mode allows a larger degree of overlap between the π-systems of adenine and the flanking anthryl groups. A derivative of 4.69 with a fully oxidized spacer has a ca. one order of magnitude larger

4.1 Receptors

N O

N

H H

N O

H

N

Pr

N Pr N

N O

N

N

H

N

N Pr

9-Propyladenine

H O

N H N

N N H

Me 2N Me

N

N

N

N

Me 2N H

211

N

Me

H N

N

N

Me

Me N

N

N Et N,N-Dimethyl-

N Et N-Methyl-

9-ethyladenine

9-ethyladenine

N

N

N

H N

N N

N Et

Figure 4.83: Watson-Crick- (left) and Hoogsteen-like (right) interaction of 9-propyladenine with the carboxy group of 4.69 and structures of the adenine derivatives used to assign which arrangement is more favorable.

9-propyladenine affinity because it is more rigid and the entropic cost of complex formation is therefore smaller. Klärner’s tweezers and clips: Another class of acyclic receptors was developed in the group of Frank-Gerrit Klärner [143]. Examples are compounds 4.70, 4.71, and 4.72 (Figure 4.84), which are characterized by an alternating arrangement of aromatic and norbornadiene subunits. The latter ensure a curved, ribbon-like arrangement of the aromatic subunits, each of which has one face oriented toward the interior of the cavity. The number of the bicyclic subunits moreover determines whether an approximately cyclic structure results as in 4.70 and 4.71, or a U-shaped one as in 4.72. These receptors are synthesized by a sequence of Diels-Alder reactions. All of them represent molecular tweezers in the sense of the Whitlock definition. Only the cyclic receptors are referred to as tweezers, however, while the U-shaped ones are usually termed clips. This differentiation should illustrate that complex formation of the tweezers involves the incorporation of the substrate into a cavity, while the clips sandwich the substrate between their two aromatic sidewalls. Figure 4.84 shows that the inner surface of the cavities of these receptors has a pronounced negative electrostatic potential. As a consequence, neutral or cationic electron-poor aromatic compounds are the preferred substrates. The cavity of 4.70 is too small to incorporate larger organic molecules but the extended version 4.71a forms stable complexes in chloroform with electron-deficient aromatic compounds such as 1,2,4,5-tetracyanobenzene (log Ka > 5) or N-methylpyridinium iodide (log Ka = 4.5). According to the crystal structure of the 1,2,4,5-tetracyanobenzene complex, this

212

4 Hosting ions and molecules

R

R

R R

R

R

a R=H b R = OH c R = OP(Me)O22‒

4.71

4.70

NC

4.72

O

CN

NH2 CN

NC

+ N

I

–

+ N

I

–

+

+

H2 N

NH3

N-MethylN-Methyl1,2,4,5-Tetracyanopyridinium iodide nicotinamide iodide benzene NH2

O H2 N

N O O N O P O P O + O O O – – HO OH NAD+

O HO

N

N N

NH2

HN O N H

O O

AcLysOMe

O

O S

N H

O O

TsArgOEt

OH

Figure 4.84: Molecular structures of the tweezers 4.70 and 4.71 and the clip 4.72. Below each receptor, the corresponding electrostatic potential surface is shown. The other structures depict suitable substrates. The color coding covers a potential range from −100 to +100 kJ mol−1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively.

substrate is fully incorporated into the cavity, with the nitrile groups protruding from the cavity openings and the positively polarized protons oriented toward the receptor π-systems, indicating that electrostatic interactions are partly responsible for complex stability. In solution, the binding mode is similar as demonstrated by the complexation-induced upfield shift of 3.6 ppm of the 1,2,4,5-tetracyanobenzene signal in the 1H NMR spectrum. 1,2,4,5-Tetracyanobenzene also binds to the clip 4.72b, but with a log Ka of 3.3 the corresponding complex is less stable than that of 4.71a. The reason is that 4.72b is unable to fully encapsulate the guest. Moreover, complex formation is associated with a structural adjustment of the receptor, involving the decrease of the distance of the aromatic sidewalls. This process is energetically costly, explaining the lower affinity of 4.72b in comparison to 4.71a. The induced fit observed for

4.1 Receptors

213

4.72b demonstrates that these receptors are able to adapt to the structure of the substrate if necessary because of the residual flexibility of the norbornadiene units. The properties of water-soluble derivatives of these tweezers and clips were studied in a collaboration of the Klärner group with that of Thomas Schrader. Complex formation often benefits from the hydrophobic effect as maybe expected for receptors featuring an apolar binding site. The affinity of 4.72c for N-methylnicotinamide iodide increases, for example, when going from methanol (log Ka = 4.2) to water (log Ka = 4.9) [144]. This clip also binds NAD+. In the respective complex, the nicotinamide moiety of NAD+ is sandwiched between the sidewalls of 4.72c, while the adenine moiety interacts with one sidewall from the outside. The spherical receptor 4.70c efficiently recognizes the cationic groups in the side chains of lysine and arginine in water [145]. Lysine, in the form of the protected derivative AcLysOMe, is bound in phosphate buffer (pH 7.6) with a log Ka of 3.6, whereas the log Ka of the complex of the protected arginine derivative TsArgOEt amounts to 3.2 (Figure 4.84). 4.70c and related molecular tweezers also recognize short peptides and even protein surfaces with exposed lysine and arginine side chains. As a consequence, these receptors are able to prevent the aggregation of pathological proteins involved in Alzheimer’s or Parkinson’s and even to dissolve already formed aggregates [146]. For details, see Section 11.2.1. Nolte’s clips: The group of Roeland J. M. Nolte used a glycoluril unit as the core structure for the preparation of molecular clips [147]. An example is 4.73, containing two 1,4-dimethoxybenzene-derived residues (Figure 4.85a). Such clips preferentially interact with planar aromatic guests, which are sandwiched between the two side walls and held tight by aromatic interactions. Guests that contain suitably arranged hydrogen bond donors additionally interact with the carbonyl units of the central glycoluril moiety by hydrogen bond formation. To illustrate this binding mode, the calculated structure of the resorcinol complex of 4.73, whose stability log Ka amounts to 3.4 in chloroform, is shown in Figure 4.85a [148]. Isaacs’s acyclic cucurbiturils: A drawback of Nolte’s clips and also the tweezers developed by Zimmerman is that they are self-complementary. The filling of the space between the aromatic subunits can thus be achieved by dimerization of two receptor molecules, which have to dissociate before complex formation with the substrate can take place. This tendency can be much reduced by giving acyclic receptors a curved shape with the end groups approaching each other at such a small distance that the cavity is practically closed. Lyle Isaacs demonstrated that this concept can be realized on the basis of glycoluril-derived receptors by replacing the single glycoluril unit in 4.73 by a C-shaped chain of four methylene-linked glycoluril units [149]. To illustrate the resulting structure, the crystal structure of one member of this receptor family is depicted in Figure 4.85b. The respective compound 4.74 contains anionic substituents in the terminating hydroquinone moieties to impart water solubility.

214

4 Hosting ions and molecules

(a)

O

MeO

OMe

N

N

N

N

Ph

Ph

MeO

(b) HO2C

OMe

O 4.73 O

O

O

N

N

N

N

O

N

N

N

N

H HO2C

O

O

O

N

N

N

N

H H

O

N

N

N

N

O

CO2H

O

CO2H

H

O

O

4.74

Figure 4.85: Molecular structure of the clip 4.73 and calculated structure of the complex of 4.73 with resorcinol to illustrate the hydrogen bonding interactions between the binding partners. The phenyl substituents in 4.73 are not shown for reasons of clarity. In (b), the molecular structure and crystal structure of the acyclic cucurbituril 4.74 is shown.

The binding properties of these so-called acyclic cucurbiturils are consistent with their structural relationship to cyclic cucurbiturils and cyclophanes. Accordingly, the combination of converging carbonyl groups around the cavity combined with aromatic residues induces a pronounced affinity for cations, which is further enhanced by the presence of the negatively charged substituents. Receptor 4.74, for example, binds to the dihydrochloride of 1,6-diaminohexane with a log Ka of 8.2 in 20 mM NaH2PO4 buffer at pH 7.4. For comparison, the log Ka values of the corresponding complexes of CB[6] and CB[7] amount to 8.5 in 50 mM NaCl and 8.0 in 50 mM NaOAc buffer (pH 4.74), showing that the cation affinity of acyclic cucurbiturils is comparable to that of the cyclic counterparts [150]. Because of their high affinity for suitable substrates and wide structural variability, acyclic cucurbiturils developed into a promising family of synthetic receptors for bio-medicinal applications as discussed in more detail in Section 11.2.

4.1 Receptors

215

Conclusion: The acyclic receptors presented in this chapter clearly demonstrate that a cyclic structure is not always necessary to achieve efficient or selective binding. Acyclic receptors can be as effective as cyclic ones, provided that they are equally well preorganized. With respect to synthesis, acyclic receptors additionally have the advantage that no macrocyclization step is required for the preparation. Such compounds should therefore not be overlooked in receptor development.

4.1.13 Foldamers Introduction: Although some natural receptors are macrocyclic, examples are ionophores such as valinomycin (Section 4.1), Nature more often makes use of acyclic molecules with a specific sequence of subunits for molecular recognition purposes. These compounds fold in a characteristic manner, thus producing cavities or clefts where substrate binding occurs. The shapes of these binding sites together with the precise distribution of substituents along their inner surfaces control substrate affinity and selectivity. These receptors are folded in the absence of the substrate and are therefore more or less preorganized for substrate binding although some conformational adjustments may occur during the binding event to optimize the interactions through an induced fit (Section 3.5.1). Proteins and polynucleotides are examples of such systems. Chemists not only elucidated many of the principles behind the folding and molecular recognition properties of such biosystems but have begun some time ago to also mimic these properties by using designed synthetic oligomers. These compounds are called foldamers, a term derived from the combination of the words “folding” and “oligomer.” By definition, a foldamer is “any oligomer that folds into a conformationally ordered state in solution, the structures of which are stabilized by a collection of noncovalent interactions between nonadjacent monomer units” [151]. Designing foldamers is often inspired by Nature. Amino acids are, for example, popular building blocks, but many systems also contain structural elements for which no natural counterpart exists. These synthetic oligomers adopt conformations found in biosystems such as sheets or helices, but they also produce structures and functions that differ from those of peptides and nucleotides. Since the folding pattern is determined by noncovalent interactions between the monomer units, aspects of supramolecular chemistry play a role in foldamer design. These aspects are particularly relevant when the noncovalent interactions are not only used intramolecularly for conformational stabilization but also intermolecularly to mediate the binding of a suitable substrate into a cavity created by the folded chain. In this context, the literature sometimes does not distinguish between systems that only assume the folded conformation once the substrate is present and those that are already folded in the absence of the substrate [152]. The outcome of complex formation is indeed similar in both cases, involving a substrate surrounded in the complex by a chain of receptor subunits and held tight by

216

4 Hosting ions and molecules

noncovalent interactions. Oligomers that only fold in the presence of the substrate can moreover be rather effective receptors, despite the lack of preorganization, because in contrast to podands, their complexes are often not only stabilized by the direct receptor–substrate interactions but also by intramolecular interactions within the folded chain. However, these receptors are strictly not foldamers according to the above definition. Here, we therefore only focus on receptors consisting of a chain of monomer units that is folded independently of whether the substrate is present or not and that have a binding site in the folded state available for substrate recognition. A few examples should illustrate the concept. Helices with cylindrical cavities: A straightforward approach for using a foldamer to create a binding site involves arranging its chain in a helical fashion around a sufficiently large cavity. Since foldamers are often constructed from sequences of identical subunits, such cavities are usually cylindrical in shape, having a constant diameter along the helix axis and two openings. The binding properties depend on how well the substrate fits into the cavity and how effectively it interacts with functional groups lining the inner surface. These parameters are controlled by the structures of the foldamer subunits and the length of the chain. A crucial aspect of foldamer design is the appropriate choice of intramolecular interactions that stabilize the required helical conformation. In the absence of direct interactions between the monomer units, solvophobic interactions are useful for this purpose. Oligomer 4.75 (Figure 4.86), for example, adopts an extended linear conformation in apolar media (CDCl3) but helical conformations in polar protic solvents (CD3OD) [153]. In most foldamers, the conformational stabilization relies on direct intramolecular interactions within the chain. Hydrogen bonds, in particular, can be arranged along the backbone in a predictable way by the proper choice of subunits as the examples of the oligohydrazide 4.76 [154] demonstrates. In the case of oligomer 4.77, comprising an alternating sequence of naphthyridine and pyrimidine units, the helical conformation is induced by the repulsion of the lone pairs of the nitrogen atoms that destabilize cisoid conformations of the bonds between the aromatic subunits, rendering the alternative transoid arrangements the preferred ones [155]. All three foldamers contain negatively polarized heteroatoms along the inner surface of their cavities, which induce affinity for positively charged guests or guests with hydrogen bond donors. The foldamer 4.77 allows the complexation of electropositive metal ions such as K+. Foldamers 4.75 and 4.76 have larger cavities, enabling them to host monosaccharides. Since 4.76 is only soluble in apolar media such as chloroform, it solubilizes alkyl glycosides under these conditions. The respective complexes have stabilities in the mM range. The foldamer 4.75 with solubilizing polyethylene glycol residues binds monosaccharides even in competitive protic media such as methanol or methanol/water, 10:1 (v/v) but not very efficiently (log Ka ca. 1). A property all these foldamers have in common is that their folding is dynamic, allowing the interconversion of enantiomeric helices of opposite sense. Both helices

4.1 Receptors

217

OR Ar N

N

RO

N

OR n

N

N

N

N N

N N RO

N

N H

4.75 (n ~ 80, R = (C2H4O)8CH3)

OR

NN I R O

RO OR O N H

O

O

O O

HN O NH

O

O

R

O HN O

Ar

N H Me

O

N

4.77 (Ar = 4-C6H4nBu) H N

O

N NN NN

N

R O

H N

O Me NH O HN O R

Ar

R

O

R R NH O NH O O H N R O 4.76 (R = C8H17)

Figure 4.86: Molecular structures of foldamers 4.75, 4.76, and 4.77. The calculated structures of 4.76 and 4.77 illustrate the helical arrangements of the chains that surround cavities whose inner surfaces are lined by polar functional groups. The residues R in 4.76 and the aryl groups in 4.77 are replaced by ethyl and methyl groups, respectively, for reasons of clarity. The hydrogen bonds stabilizing 4.76 are shown as dashed lines.

bind with equal strength to an achiral substrate, but if the substrate is chiral, the complexes of the two foldamer enantiomers are diastereomeric and therefore have different stabilities. As a consequence, one helix sense usually predominates if one enantiomer of a chiral substrate is present in excess. The chiral induction of a specific helix sense is conveniently monitored by recording the circular dichroism (c.d.) spectra of foldamer complexes. No c.d. bands are visible in the spectra of rapidly interconverting helices or foldamer complexes of achiral substrates (or the racemate of a chiral substrate) in which the P and M configured helices are present in equal amounts. The c.d. spectra of foldamer complexes of enantiomerically enriched substrates, on the other hand, feature distinct bands, reflecting the predominant chiral foldamer conformation. Since the two enantiomers of the substrate induce helices of opposite sense, the c.d. spectra of the corresponding complexes look like mirror images. Because of their conformational adaptability, receptors based on foldamers can thus be used to probe the e.e. of a chiral substrate. Moreover, if the correlation between the absolute configuration of the

218

4 Hosting ions and molecules

substrate and the sign of the band in the c.d. spectrum is known, it is also possible to determine which substrate enantiomer dominates in solution. Helices with bulged cavities: The group of Ivan Huc developed a toolbox of heterocyclic building blocks to construct helical oligoamide-based foldamers, conformationally stabilized along the chain by appropriately arranged hydrogen bond donors (e.g., amide NH groups) and acceptors (e.g., ring nitrogen atoms) [156]. Each building block has a characteristic influence on the folding pattern and the dimension of the resulting helix, so that it is possible to precisely control the preferred foldamer conformation by varying the sequence of the subunits. This approach allows the design of foldamers with noncylindrical cavities. An example is 4.78 (Figure 4.87a), which contains a central sequence of three pyridine-derived units, flanked at both sides by dimers of 8-substituted 2-quinolinecarboxylic acid. The preferred structure of this foldamer resembles the skin of a peeled apple, which is why 4.78 entered the literature under the name “molecular apple peel.” The cavity surrounded by the chain is wider in the central region and narrower at the two extremities, where the terminal quinoline units cap the helix openings. In contrast to the foldamers discussed above, the cavity created by 4.78 is therefore completely closed and spherical rather than cylindrical. It is only large enough to host a water molecule but variations in the lengths of such oligoamides, the structures of the monomers, and their sequence allows accessing foldamers with cavities sufficiently large to host organic molecules such as aliphatic alcohols or amines, dicarboxylic acids, or even monosaccharides. To illustrate the binding modes, crystal structures of the water complex of 4.78 and of foldamers with larger cavities containing 1,4butanediol or tartaric acid are depicted in Figure 4.87b-d. The lack of openings in the folded structures requires these oligoamides to undergo a substantial conformational reorganization during complex formation or dissociation. As a consequence, guest exchange is slow on the NMR timescale. Once bound, the substrates are moreover completely shielded from the environment, like in hemicarcerands. In contrast to hemicarcerands, however, the inner surface of these foldamers features hydrogen bond donors and acceptors with which the included substrates interact. Complex formation is thus associated with a considerable gain in ΔG0 as reflected in the appreciable binding constants determined for some of these complexes. The 1,4-butanediol complex of the foldamer shown in Figure 4.87c has a log Ka of 3.5 in chloroform, for example [157]. Binding is moreover very selective since substrates too bulky to be included into the cavity such as 1,3-butanediol, branched analogs of 1,4-butanediol, or 1,5-pentanediol are not bound. The selection of one helix conformation over the corresponding mirror image in the presence of chiral guests has also been observed for these oligoamides. D-Tartaric acid is, for example, preferentially bound by a right-handed helix [158]. Because of the high stability of this complex (log Ka > 6 in CDCl3/DMSO-d6, 99:1 (v/v)), analogous foldamers with suitable solubilizing groups also bind tartaric acid in more

4.2 Substrates

(a)

O BnO

O

O N H

N

N

N H

OiBu N

NH HN

NO2 O

N NH O

219

N

H N

4.78

NN OiBu OiBu

O2N O i BuO

(b)

(c)

(d)

Figure 4.87: Molecular structure of foldamer 4.78 that folds into a helix with a cavity large enough to host a water molecule (a), crystal structure of the corresponding complex (b), and crystal structures of related foldamers binding to 1,4-butanediol (c), and tartaric acid (d). Protons and side chains in the foldamers are not shown for reason of clarity.

competitive media such as methanol or even water. The largest substrates that can so far be bound are monosaccharides, again with good selectivity [159]. Such foldamers combine a large structural variability with the possibility to rather reliably predict the folding pattern. In addition, they have chiral threedimensional cavities in which substrates are shielded from the environment and engage in directed interactions with functional groups distributed along the inner surface. These features are not found in many other systems.

4.2 Substrates Let us now look at the receptors from the perspective of the substrate. In this context, imagine that you are expected to develop a receptor for a certain substrate. Usually, certain boundary conditions apply. For example, if your research group specializes in calixarenes, choosing cucurbiturils as receptors would probably not be opportune. It may also be that the receptor is expected to work in water. One then has to base the receptor design on scaffolds that are water-soluble and known to work in water.

220

4 Hosting ions and molecules

Maybe applications are targeted in a biochemical context or even in vivo. In this case, toxicological aspects need to be considered and using crown ethers as receptors, for example, would not be very wise. To simplify the situation, let us consider that no such boundary conditions exist and that the receptor can be freely chosen. In this case, the decision of which receptor to select can be based solely on its intrinsic binding properties. A good overview of which receptors are available for which type of substrate is therefore helpful. This chapter should give such an overview and should help identifying a suitable receptor for a given target.

4.2.1 Inorganic and organic cations We start with cationic substrates because after the discovery of crown ethers, cations were initially the main targets in supramolecular chemistry. Many receptors bind to positively charged guests, but their binding modes differ depending on the family to which they belong. To better understand the correlation, it is helpful to classify cations into transition metal ions, electropositive inorganic cations, and organic cations. Transition metals feature a rich coordination chemistry as a result of their ability to interact with Lewis-basic ligands. Some aspects relevant in coordination chemistry are equally important in supramolecular chemistry. The chelate effect, for example, explains the high thermodynamic stability of metal complexes with multidentate ligands, and the formation of kinetically labile complexes is subject to thermodynamic control as is the formation of supramolecular complexes that are held together by reversible interactions. Despite these relationships, the actual recognition of transition metal ions is best described by using the concepts of coordination chemistry and is therefore not treated here. Electropositive metal ions originate from elements in the first two groups of the periodic table of the elements. Some of these metal ions have pharmacological properties (lithium) or are responsible for maintaining the membrane potential of cells (sodium and potassium). Alkaline earth metal ions are components of the bone structure (calcium) and are responsible for hard water (calcium and magnesium). Salts of alkali or alkaline earth metal ions are moreover frequently used in synthetic procedures and the availability of receptors for these cations therefore has also practical implications (Chapter 11). Important types of interactions on which to base the recognition of these cations are ion–dipole interactions with electronegative heteroatoms, especially the oxygen atoms of ethers or carbonyl compounds. Important receptors for hard metal ions are crown ethers (Section 4.1). Their cation affinity and selectivity are controlled by ring size and also by the presence of additional substituents as in lariat ethers. Going from monocyclic crown ethers to bicyclic cryptands usually leads to an improvement of affinity and selectivity (Section 4.1.2). Other receptors containing ethylene glycol-derived subunits are also expected to interact with hard metal ions, even if not immediately qualifying as crown ether or cryptand.

4.2 Substrates

221

Spherands typically have an even higher cation affinity than cryptands (Section 4.1.3). They suffer, however, from a rather elaborate synthesis and they only bind well to the smallest cations (Li+, Na+). If other cations are targeted, hemispherands could represent alternatives but their cation affinity is lower than that of spherands because of a poorer preorganization. Calixarenes prefer to bind hard metal ions with their cyclically arranged OH groups along the narrower cavity opening (Section 4.1.7). A way to modulate the binding properties involves varying ring size and, hence, the number of available OH groups. A more versatile approach involves the modification of the OH groups by introducing additional binding sites as in the calixarene derivative 4.37 (Figure 4.44). Cucurbiturils also feature a cyclic arrangement of carbonyl groups along the cavity openings (Section 4.1.11). They therefore interact with metal ions but usually not with a pronounced selectivity. The complexation of cations at the two cavity openings can be used, however, to entrap suitable neutral substrates within the cucurbituril cavity. Table 4.10 summarizes the classes of receptors available for the binding of electropositive metal ions. Table 4.10: Classes of receptors available for the recognition of electropositive metal ions and organic cations together with their preferred binding modes. Substrate

Binding modes

Receptors

Alkali and alkaline earth metal ions

Ion–dipole interactions

Crown ethers, cryptands, spherands, calixarenes, cucurbiturils

Protonated amines with additional organic residues

Salt bridges potentially combined with hydrogen bonding

Negatively charged receptors such as appropriately substituted cyclodextrins, cryptophanes, calixarenes, deep cavitands, pillararenes, acyclic cucurbiturils, and molecular tweezers Or

Hydrogen bonding without additional salt bridges

Crown ethers, cryptands, hemispherands, cucurbiturils

Protonated organic amines Cation–π interactions and quaternary ammonium ions

Cryptophanes, calixarenes, deep cavitands, pillararenes

Quaternary ammonium ions

Cucurbiturils

Ion–dipole interactions

The third category of cationic guests are protonated amines or quaternary ammonium ions. These cations structurally differ in a wide range but they have in common that they are softer than electropositive metal ions, rendering them more prone to interact with a receptor by cation–π interactions. Important organic cations are the

222

4 Hosting ions and molecules

neurotransmitters acetylcholine, dopamine, noradrenaline, and adrenaline, or amino acids including side-chain methylated lysine and arginine derivatives that play key roles in controlling gene activity and expression. These substrates are characterized by one or more positively charged head groups and additional organic residues. Suitable receptors therefore often but not necessarily contain two binding sites, one for the cationic group and one for the other parts of the substrate. The complexation of the cationic group can be mediated by oppositely charged residues in the receptor. This mode of binding is found, for example, in cyclodextrins with anionic substituents along the cavity openings, water-soluble cryptophanes (e.g., 4.32, Figure 4.35), sulfonatocalixarenes (e.g., 4.39a or 4.39b, Figure 4.44), deep cavitands (e.g., 4.49 or 4.50, Figure 4.57), pillararenes (e.g., 4.56, Figure 4.69), molecular tweezers (e.g., 4.70c, Figure 4.84), and acyclic cucurbiturils (e.g., 4.74, Figure 4.85). The efficiency of the underlying electrostatic interactions varies with receptor structure. In the octa acid 4.50, for example, the negatively charged groups are more important to mediate water solubility than substrate affinity. All of these receptors have additional cavities that allow the incorporation of the uncharged part of the substrate, causing the associated release of solvent molecules to favorably contribute to complex stability. In some cases, the orientation of the substrate inside the cavity is controlled by the arrangement of the negatively charged groups. Organic cations are preferentially included into the deep cavitand 4.49 with their cationic head group oriented toward the carboxy groups lining the cavity opening, for example. The formation of these complexes is independent of whether the positively charged group of the substrate is a protonated amine or a quaternary ammonium group unless the negatively charged binding sites of the receptor are appropriately placed to allow hydrogen bond formation. In this case, the binding modes differ, involving a combination of electrostatic interactions and hydrogen bonding for protonated amines and only Coulomb attraction for quaternary ammonium ions. A similar distinction can be made if the receptors are neutral. Protonated amines form hydrogen bonds to the carbonyl groups of cucurbiturils, for example, while the binding of the same hosts to quaternary ammonium ions is mediated by ion–dipole interactions. Hydrogen bonds between oxygen atoms and protonated amines also stabilize the complexes of crown ethers, cryptands, or hemispherands (e.g., 4.18, Figure 4.20), while ion–dipole or even C–H···O interactions of the same oxygen atoms with quaternary ammonium ions are usually insufficient to induce binding (Section 3.1.5). Receptors with cavities lined by electron-rich aromatic subunits such as cryptophanes, calixarenes, deep cavitands, or pillararenes interact with cationic substrates by cation–π interactions. An overview of the receptors and the binding modes used to recognize organic cations is given in Table 4.10.

223

4.2 Substrates

4.2.2 Inorganic and organic anions It is of historic interest that the first receptors for anions were described by Chung Ho Park and Howard E. Simmons only one year after Pedersen reported the discovery of the cation-binding crown ethers [160]. These anion receptors were termed katapinands, from Greek καταπίνω (to swallow, to engulf) to illustrate that complex formation involves the complete incorporation of the guests into the receptor cavity. An example of a katapinand is 4.79 (Figure 4.88a) in which two nitrogen atoms are connected via three nonyl chains. In its diprotonated form, 4.79 incorporates anions such as halides between two bridgehead ammonium groups where they are held tight by a combination of electrostatic interactions and hydrogen bond formation. Anion affinity is, however, low. The chloride complex has a Ka of only 4 M−1 in CF3COOD/D2O, 1:1 (v/v), for example. This weak binding is partly due to the poor preorganization of this receptor, which prefers a conformation in the absence of anionic substrates with the protonated nitrogen atoms arranged at the maximum distance, because of charge repulsion, and the NH protons located outside of the cavity. Anion binding thus requires the receptor to undergo a considerable conformational reorganization from the preferred out–out to the less favorable in–in conformation as shown in Figure 4.88b. (a) N

N

4.79

(b) (CH2) 9

(CH2) 9 + NH

+ HN

(CH2)9 (CH2) 9 Out−out

+ HN

HN (CH2)9 (CH2) 9 Out−in

(CH2) 9 +

+

+ NH HN (CH2)9 (CH2) 9

–

(CH2) 9 +

NH – HN (CH2)9 (CH2) 9

+

In−in

Figure 4.88: Molecular structure of katapinand 4.79 (a), and schematic illustration of the conformational preorganization the diprotonated form of this receptor has to undergo during anion complexation (b).

Although the development of cation and anion receptors started almost simultaneously, the two fields progressed at considerably different paces. While the development of cation receptors flourished immediately after Pedersen’s first paper with the development of the cryptands by Lehn, the chiral crown ethers and spherands by Cram, and many other receptors, the field of anion coordination chemistry took longer to develop, mainly because anions are more demanding substrates than cations for the following reasons: – Inorganic anions are relatively large, therefore requiring receptors with larger cavities than cations.

224

4 Hosting ions and molecules

– Electrostatic interactions are weaker to an anion than to an isoelectronic cation because the anion is larger and therefore has a smaller electron density. – We have also seen in Section 3.1.3 that anions have higher free energies of hydration than cations with the same absolute charge and comparable size. Strong interactions are therefore required in water for a receptor to efficiently compete with the water molecules in the hydration shells of anions. – Many oxoanions are involved in protonation equilibria in water. Phosphate buffer, for example, contains almost equal amounts of HPO42− and H2PO4− ions at physiological pH, rendering anion recognition pH dependent. – Finally, inorganic anions come in different geometries such as spherical, linear, trigonal planar, tetrahedral, or octahedral, and (organic) polyanions have even more complex structures, which is an aspect that also has to be considered in receptor design. While these factors could have made anion receptors more difficult to design than cation receptors in the early days of supramolecular chemistry, many of these difficulties have been overcome. Anion coordination chemistry has thus developed into a mature and influential research field, due in part to the relevance of many anions. Potential anionic substrates range from inorganic anions such as nitrate and phosphate, which are responsible for the eutrophication of water bodies, over toxic anions such as cyanide and arsenate, to biologically relevant anions such as chloride, sulfate, and, most importantly, phosphate. In addition, more than 70% of the substrates and co-factors involved in biological processes are negatively charged. The most important ones are nucleotides, polynucleotides, phosphorylated carbohydrates or proteins, and biomolecules with carboxylate groups or sulfate esters. There is thus considerable interest in anion receptors for practical applications such as environmental monitoring or medicinal chemistry. The most important types of noncovalent interactions that allow anion binding are electrostatic interactions with a positively charged receptor or hydrogen bonding with a hydrogen bond donor. More recently, the use of anion–π interactions or halogen bonding has also become popular. Many anion receptors furthermore contain metal centers integrated within a suitable ligand framework to which an anion binds through coordinative interactions. Table 4.11 gives an overview of receptors that make use of these binding modes. Receptors that only use electrostatic interactions to interact with anions are relatively rare but we have encountered such receptors in Section 4.1.2 in the form of the macrotricyclic systems 4.10 and 4.11 (Figure 4.15), which contain quaternary ammonium ions in the bridgehead positions. These receptors adopt conformations with wide open cavities as a consequence of the electrostatic repulsion between the positively charged groups. These cavities are available for anion binding in water but the affinities are generally low.

4.2 Substrates

225

Table 4.11: Classes of receptors available for the recognition of anionic substrates and their preferred binding modes. Binding modes

Receptors

Electrostatic interactions

Macrocyclic or macropolycyclic receptors with quaternary ammonium ions

Electrostatic interactions and hydrogen bonding

Polyammonium-based receptors or receptors containing guanidinium or amidinium groups

Hydrogen bonding

Receptors with hydrogen bond donors along a chain or within a macrocyclic or macropolycyclic framework

Anion–π interactions

Receptors with electron-deficient aromatic subunits

Halogen bonding

Receptors with halogen atoms bound to electron-deficient aromatic subunits

Metal coordination

Ligand frameworks incorporating coordinatively unsaturated metal centers or metal centers with weakly bound ligands

Considerably higher affinities are typically observed for polyammonium-based receptors with protonated nitrogen atoms, which combine electrostatic interactions with hydrogen bonding. Examples are the macrocyclic receptor 4.7 (Figure 4.9) or the polyazacryptands 4.12 and 4.13 (Figure 4.16). In contrast to receptors with quaternary ammonium groups, the charge state of these systems depends on the pH so that anion affinity changes in a pH-dependent manner. Receptors that bind to anions through a combination of hydrogen bonds and electrostatic interactions cannot derive only from polyamines but can also contain guanidinium or amidinium groups. Both groups are strongly basic with pKa values typically ranging between 11 and 13 so that they remain protonated over a wide pH range. In the corresponding protonated states, they form two parallel hydrogen bonds reinforced by electrostatic interactions to oxoanions such as carboxylates, carbonates, phosphates, or sulfates as shown in Figure 4.89a. Amidinium and guanidinium groups are therefore effective binding motifs to induce affinity in a receptor for oxoanions. They are found, for example, in the tripodal receptor 4.62 shown in Figure 4.78b, or the receptors depicted in Figure 4.89b. Receptors that only bind to anions by hydrogen bond formation contain hydrogen bond donors along the molecular framework, often in the form of NH groups from amide, urea, thiourea, or pyrrole groups. Calix[4]pyrrole 4.40 (Figure 4.46) and the receptors shown in Figure 4.90a are examples. Hydroxy groups also serve as hydrogen bond donors but are less frequently used. Since these receptors are uncharged and relatively nonpolar, they are usually only soluble in organic media so that binding studies have to be performed in solvents such as chloroform, DMSO, acetone, or acetonitrile.

226

4 Hosting ions and molecules

(a)

(b)

H N H + N H

R1

H2N

O –

alkyl/aryl O

NH + NH

HN NH2 + HN

N R

N + N H H R = O SiPh2tBu

H H H N + N

Amidinium: R1 = alkyl, aryl Guanidinium: R1 = NH-alkyl, NH-aryl

R

NH2 Figure 4.89: Schematic representation of the interaction of a guanidinium group or an amidinium group with a carboxylate ion (a), and examples of receptors containing guanidinium groups as binding sites (b). Phosphate or sulfate ions are bound in a similar fashion as the carboxylate in (a).

(a) O

O

O N

NH

NH

HN

N

N

O

NH O NH

HN

N

O

HN NH

O N

NH

O N

O

O

N N

HN

HN

NH

N

O

O HN

O

HN H N

O

H N

O

O

(b)

N N

N N

N

N

N H

H

H

H

N

CN

NC

N

H

H

H H N

N

NC

CN

H

N CN

Figure 4.90: Examples of receptors containing NH (a) or CH (b) hydrogen bond donors for anion recognition.

The propensity of NH groups to interact with anions can be fine-tuned by varying the substituents at the nitrogen atoms. In general, electron-withdrawing groups increase the acidity of the NH groups and, consequently, their ability to act as hydrogen bond donors. A potential caveat is that certain anions deprotonate NH groups when they become too acidic, leading to ion pairs. These ion pairs are held together

4.2 Substrates

227

by electrostatic interactions but the structural integrity mediated by hydrogen bonds is potentially lost. Polarized CH bonds are also used for anion recognition. They are found, for example, in the two macrocyclic receptors depicted in Figure 4.90b, both developed in the group of Amar Flood, and the bambus[6]uril 4.60 (Figure 4.77) introduced in Section 4.1.11. Anion–π interactions or halogen bonding require the receptor to contain electron-deficient aromatic units, or halogen atoms bound to electron-withdrawing groups, respectively. We came across such systems in Sections 3.1.8 and 3.1.6 where the corresponding types of interactions were introduced. Although the use of these interactions in anion coordination chemistry is still limited, halogen bonding, in particular, is becoming increasingly popular because of the high directionality of halogen bonds and their lower susceptibility to solvent effects in comparison to hydrogen bonds. We will see additional uses of halogen bonds in Sections 5.4, 6.6, and 8.3.1. Metal-containing receptors are very common in anion coordination chemistry. They are often constructed by using multidentate ligands as molecular scaffolds that form chelate-type complexes with one or more transition metal ions, leaving at least one coordination site available for anion recognition. The macrocyclic and macrobicyclic polyamines 4.5, 4.6 (Figure 4.9), and 4.14 (Figure 4.17) described in Sections 4.1 and 4.1.2 are examples of such ligands. Metal ions most frequently used in such receptors are zinc(II), which is diamagnetic and thus allows the use of NMR spectroscopy to characterize the structure and the binding properties of such receptors, or copper(II). In the latter case, complex formation often causes a distinct color change of the solution, thus facilitating the binding studies. These receptors are rather efficient even in water since they benefit from the strength and directionality of coordinative interactions that are hardly affected by solvation effects. So far, we mainly considered binding motifs that target inorganic anions or the anionic residues in organic anions without distinguishing between these substrate types. Indeed, all receptors suitable for the first substrate type are also used for the second. The only difference is that the binding of organic anions such as carboxylates, nucleotides, or other phosphate esters is usually improved by introducing additional binding sites into the receptor that interact with the organic part of the substrate. This general principle, introduced in Section 3.5.3 about multivalent interactions, also applies to the receptors for organic cations discussed in the previous chapter. An application in anion coordination chemistry is the attachment of aromatic subunits to the framework of polyammonium receptors, which can engage in aromatic interactions with nucleobases, thus improving nucleotide affinity. An example is the receptor for adenosine triphosphate depicted in Figure 4.91.

228

4 Hosting ions and molecules

O

NH

HN

H N

N

N N H

N NH O

N

HN

Figure 4.91: Example of a receptor that combines a polyazamacrocycle with appended aromatic residues to mediate the simultaneous recognition of the triphosphate and the nucleobase group of ATP.

4.2.3 Zwitterions and ion pairs Zwitterion receptors should bind simultaneously to the cationic and the anionic group of the substrate. They are thus constructed by covalently connecting two complementary binding sites derived from receptor motifs described in the previous sections. One has to take care in the design of such receptors that the two binding sites do not interact with themselves, which would either lead to the collapse of the receptor cavity if the interactions occur intramolecularly or to the noncovalent polymerization of the receptor. Strategies to prevent these detrimental effects include the use of complementary recognition elements that exhibit no or only a small affinity for one another, connecting them through rigid spacers, or incorporating them in a rigid macro(bi)cyclic framework. Figure 4.92 shows a selection of zwitterion receptors to illustrate the concept. The crown ether-derived receptors recognize unprotected amino acids or short peptides. Their cation-binding parts interact with the ammonium groups of the substrates, whereas carboxylate binding is mediated by electrostatic interactions to either a quaternary ammonium ion, the protonated forms of the macrocyclic polyamine 4.7 (Figure 4.9), the macrotricyclic cage 4.11 (Figure 4.15), or a guanidinium group. The other receptors in Figure 4.92 are hemicryptophanes in which the CTV moieties mediate cation binding, whereas the amide NH groups at the opposite end of the cavity interact through hydrogen bonds with the anionic part of the zwitterionic substrate. These receptors bind to ω-aminocarboxylic acids, ω-aminosulfonic acids, and ω-aminophosphonic acids as we have seen in Section 4.1.6. Ion-pair receptors are conceptually very similar to zwitterion receptors in that they also combine binding sites for a cation and an anion. In contrast to zwitterions, the components of an ion pair are, however, not part of the same molecule. They nevertheless prefer to stay together due to Coulomb attraction, especially in nonpolar media where ions are not well solvated. Three binding modes are distinguished for such ion-pair receptors.

229

4.2 Substrates

O

N

N O

O O

O

O O

O O

O

O

O

N

O

O O

+ N

NH + NH

N HN

NH N

+

HN

N

N

+

+

N

O

NH HN O N

MeO

O

O

OMe OMe O O

O

O

MeO O

+

O

O

O

OMe OMe O NH

NH O

NH O O

NH N

HN

HN O

O

O

N

O2S N

O

O

HN

O

NH NH

O

O

Figure 4.92: Examples of ditopic receptors for the recognition of zwitterions.

The first involves the binding of both ions as a contact ion pair. In this case, complex formation does not require the energetically unfavorable separation of the cation and the anion. Cascade complexes such as that formed between the dicopper(II) complex of the polyazacryptand 4.14 and an azide anion in Figure 4.17 are examples of this binding mode. Another example is based on receptor 4.80 (Figure 4.93a), developed by Bradley D. Smith, that contains two amide groups for chloride binding and the crown ether moiety as a binding site for the cation [161]. The cavity is small enough to host KCl as a contact ion pair as illustrated by the crystal structure in Figure 4.93a. That binding of the first ion infuences the binding of the respective counterion with a positive cooperativity is demonstrated by the fact that the potassium complex of 4.80 binds chloride thirteen times more strongly in DMSO than 4.80 alone. Conversely, the chloride complex of 4.80 has a ca. 40-fold higher K+ affinity than 4.80.

N

O

230

4 Hosting ions and molecules

(a)

(b)

(c) O

O

O NH

HN

NH

O

N

N O

O 4.80

N H

HN

O

O N

O

O

O O O

O O 4.81

N

NH

HN H N

O

4.40

Figure 4.93: Molecular structures of the ion-pair receptors 4.80 (a), 4.81 (b), and 4.40 (c) and crystal structures of their KCl (a), NaCl (b), and CsF (c) complexes.

The second case involves the binding of the ions as a solvent separated ion pair. This binding mode is observed for 4.81, which is structurally related to 4.80 but has a larger cavity [162]. In the crystal structure of the NaCl complex of 4.81, a chloroform molecule resides between the two ions (Figure 4.93b). The resulting separation of the cation and the anion is energetically unfavorable, explaining why the potassium complex of 4.81 only has a ca. 9 times higher affinity for Cl− in DMSO-d6/CD3CN, 3:1 (v/v) than 4.81 alone. Finally, ion pairs can also be bound separately as host-separated ion pairs. Since the complete separation of the ion pair is energetically unfavorable, the communication between the cation and the anion is rarely fully prevented in such complexes. Moreover, interactions between the individual ions and the receptor partly compensate the energy required for ion-pair separation, causing the sequential binding of the two ions to often proceed with positive cooperativity. An example is the CsF complex of calix[4]pyrrole 4.40 (Figure 4.93c) [73] in which the fluoride anion is bound by hydrogen bonding to the NH groups. The cesium cation occupies the shallow bowl-shaped cavity opposite the anion binding site, where it is held tight by electrostatic interactions with the already present anion and by cation–π interactions with the pyrrole units as illustrated by the crystal structure in

4.2 Substrates

231

Figure 4.93c. As a consequence of this binding mode, 4.40 is able to extract CsCl and CsBr but not CsNO3 from an aqueous phase into nitrobenzene. How does the counterion influence the binding of an ion to a receptor?

The above examples of ion-pair receptors could suggest that the main driving force for the oppositely charged ions to come together in the complex stems from the complementary binding sites in the receptors. This assumption is, however, not correct since cations and anions are to a substantial degree paired in organic solvents. As a consequence, it is difficult if not impossible to study the binding of an ion in organic media without the interference from the counterion. Several studies indeed demonstrated that the complexation of quaternary ammonium ions in chloroform by cyclophanes [163] or calixarenes [164] is strongly affected by the counterion, with cation affinity typically decreasing in the order picrate > iodide > bromide > chloride > tosylate. This order is independent of the receptor and therefore anion specific. An analysis of the underlying equilibria by using double-mutant cycles showed that the intrinsic strength of cation–π interactions between the receptor and the cation is not affected by the anion, leading the authors to conclude that the observed anion effect is likely due to the “multiple equilibria that are present and [the] competition for interaction sites between the anion and cation” [165]. Describing the binding of a cation to a receptor in organic solvents with a single equilibrium that only considers the formation of the cation complex is thus an oversimplification. At least ion pairing has to be additionally taken into account, which directly competes with cation binding since the ion pair likely has to dissociate to some degree during the formation of the cation complex. As a consequence, the stronger the interactions in the ion pair, the weaker cation binding should be. This notion is consistent with the above dependence of cation affinity on the nature of the halide, which indicates that cation binding is particularly weak if the halide is small and has a high charge density. Conversely, anions with a lower charge density that should interact weaker with the counterion have smaller adverse effects on cation binding. In the case of organic anions, the ability of the host to arrange the anion in the vicinity of the cation without having to separate the ion pair plays an additional role. Ion-pairing phenomena thus strongly influence ion recognition in organic solvents and cannot be neglected. In water, on the other hand, ions of opposite charge are usually fully dissociated and strongly hydrated, rendering the binding of an ion to a receptor often independent of the counterion.

4.2.4 Neutral organic molecules The binding of polar neutral molecules to synthetic receptors benefits from dipole– dipole interactions or hydrogen bond formation and becomes particularly effective

232

4 Hosting ions and molecules

when several of these interactions are simultaneously employed. We have seen a number of receptors in the previous chapters that use hydrogen bonding for substrate recognition such as the Hamilton receptor 4.28 (Figure 4.30), the Kemp’s triacid derivatives 4.65b-e and 4.66 (Figure 4.81), and the foldamers with hydrogen bond acceptors and donors along their cavity. Other types of interactions that are used to mediate the binding of neutral substrates are aromatic interactions such as the edge-toface interactions that stabilize the durene complex of cyclophane 4.23 (Figure 4.28), the p-benzoquinone complex of cyclophane 4.27 (Figure 4.30), and the 1,2,4,5tetracyanobenzene complex of the molecular tweezers 4.71 (Figure 4.84). In addition, charge-transfer interactions between electron-rich and electron-deficient π-systems as in the 1,4-dimethoxybenzene complex of the blue box 4.26 (Figure 4.29) and the complexes of the molecular tweezers 4.67 (Figure 4.82) are also possible. Substrates with regions along their surface featuring pronounced positive or negative electrostatic potentials thus have anchor points with which a receptor can interact. Apolar substrates are, on the other hand, much more difficult to bind in apolar solvents. The reason is that the contributions of electrostatic interactions to binding are small. Moreover, apolar substrates are well solvated in organic media so that entering a receptor cavity where the environment is not very different from that in the bulk solvent represents no improvement unless large contact areas come in contact, giving rise to dispersion interactions. The release of solvent molecules from the receptor cavity could help but usually does not lead to a pronounced thermodynamic stability. A useful trick to force apolar substrates into a cavity is to perform the binding studies in solvents too large to be bound, but this works well only for receptors that have rather shielded cavities such as molecular cages. Examples are cryptophane-E 4.30, cryptophane-A 4.31, (Figure 4.35), or deep cavitands. Under these conditions, the cavity is empty in the absence of the substrate, which is thermodynamically unfavorable, and its filling therefore leads to a more stable state. Polar neutral substrates are thus relatively easy to bind in organic media, whereas the binding of apolar substrates is more difficult. In water, the situation is exactly opposite. Under these conditions, a thermodynamic advantage of complex formation is expected if the substrate has large apolar surfaces that cannot be hydrated well and/or if the receptor cavity contains water molecules, the release of which is enthalpically or entropically favorable. We came across a number of receptors that benefit from these effects. The most important ones are cyclodextrins, which represent a very versatile class of receptors for neutral substrates in water, but also sulfonatocalixarenes such as 4.39a (Figure 4.44), water-soluble pillararenes, or cucurbiturils. If, on the other hand, a polar substrate is targeted that is strongly solvated in water, binding is more difficult. In this context, carbohydrates probably represent the most challenging substrate class because they mainly differ from a cluster of water molecules by the exact arrangement of their OH groups and the presence of the aliphatic CH groups.

4.2 Substrates

233

Let us first look at how the binding of carbohydrates is achieved in apolar solvents. Under these conditions, carbohydrate recognition should not be very difficult since the receptor can target the OH groups of the substrate. Indeed, many carbohydrate receptors contain an appropriate arrangement of hydrogen bond donors and acceptors along the binding site. The helical foldamers mentioned in Section 4.1.13 are examples. Additional interactions useful for carbohydrate recognition are CH–π interactions between one or more aromatic substituents in the receptor and the ring faces of the substrates where the aliphatic CH groups reside. Many carbohydrate receptors thus combine suitable hydrogen bond donors and acceptors with one or more aromatic residues to allow CH–π interactions. The acyclic tripodal receptors containing a triethylbenzene core or the molecular cage shown in Figure 4.94a are examples. They efficiently bind in organic media such as chloroform to carbohydrates that contain long chain alcohols at their anomeric centers to make them soluble [166]. The same binding motifs are used to achieve carbohydrate binding in water. In this environment, it is necessary to also effectively shield the binding site. The receptors 4.82 and 4.83 shown in Figure 4.94b illustrate that this can be achieved by using cage-type receptor architectures, featuring aromatic residues as “roofs” to mediate CH–π interactions, and linkers as “bars” with multiple hydrogen bonding sites [167]. Both receptors moreover contain substituents in the periphery with multiple negative charges that mediate water solubility, help to shield the cavities, and possibly contribute to substrate recognition. These receptors were developed in the group of Anthony P. Davis, who introduced the term “temple receptors” to describe their structure. Receptors 4.82 preferentially binds to all-equatorial monosaccharides such as β-D-glucose 4.84a or β-D-glucosides such as the methyl glucoside 4.84b (Figure 4.94c). With a Ka of 9 M−1, the affinity for 4.84a is low in water but 4.82 was the first receptor demonstrating that monosaccharide binding is possible with designed receptors in this solvent. N-Acetylglucosamine 4.84c and its β-methyl glycoside 4.84d form significantly more stable complexes. The Ka of the complex between 4.82 and 4.84d amounts to 630 M−1, for example, which is very similar to that of the complex of the same monosaccharide with the natural carbohydrate binder Wheat Germ Agglutinin. With a log Ka of 4.3, 4.83 is the current record holder of glucose affinity in water [168]. Glucose is bound stronger by a factor of ca. 100 than galactose and by a factor of ca. 1000 than a range of other potential substrates, thus rendering this receptor even useful for medicinal applications. It should be mentioned in this context that carbohydrate recognition can be achieved in water also by using boronic acids, which form cyclic boronates with 1,2- and 1,3-diols [169]. Although substrate binding through covalent bonds strictly does not fall into the domain of supramolecular chemistry, the formation of boronates from carbohydrates has many hallmarks of supramolecular systems. First, boronate formation is reversible and relatively rapid even at room temperature and

234

4 Hosting ions and molecules

(a) HN

H2N NH2

NH

N H 2N NH N

N HN

HN HN

HN HN

NH NH

HN HN

O

NH

H N

H N

O

NH

NH

HN

O

(b)

HN

NH O

R HN

R

O

O

O

O O

O

HN O

O R

NH

R=

O –

O O

N H

O

O

O –

O

HN

O

NH 4.82

–

O

–

O

O

R

NH NH

O

O

R=

HN

O

R2

O O O – O O

–

HO HO

–

O O

OH O

– O

N H NH

N H

4.83

(c)

OH O

O

R

HN O HN

NH

O NH

O

HN HN

NH

–

O O

R

HN HN

O

O

HN O O

–

O

NH R

O O–

4.84a R 1 = H, R2 = OH OR 1

4.84b R 1 = CH3, R2 = OH 4.84c R 1 = H, R2 = NHAc 4.84d R 1 = CH 3, R2 = NHAc

Figure 4.94: Examples of carbohydrate receptors active in apolar organic solvents (a) and in water (b), and structures of monosaccharides that were investigated as substrates (c).

physiological pH. Second, it is possible to control selectivity by the structure of the boronic acid. A major advantage of using boronic acids for carbohydrate recognition in water is the strength of the covalent bonds that mediate binding. Figure 4.95a shows the equilibria underlying the reaction between a boronic acid and a diol. Accordingly, ester formation encompasses both trigonal and tetrahedral species. In general, the equilibrium involving the tetrahedral species is favored because the angle of the O–B–O bond in the tetrahedral structure is smaller and the cyclic boronate therefore less strained than in the ring with the trigonal

235

Bibliography

boron center. The equilibrium constants describing boronate formation thus depend on the conditions, especially the pH. Independent of these conditions, simple boronic acids such as the phenyl boronic acid 4.85 (Figure 4.95b) generally exhibit selectivity for D-fructose. A challenge in this field is therefore to change this intrinsic selectivity by the proper structural design of the receptor away from D-fructose to other sugars, preferentially D-glucose. First success in this respect came with the bis(boronic acid) 4.86, developed by Tony D. James and Seiji Shinkai, which binds monosaccharides through the formation of two boronate units, provided the geometry is favorable [170]. D-Glucose fits well inside the binding site and is bound with a log Ka of 3.6 in methanol/water, 1:2 (v/v) buffered to pH 7.8, while D-fructose and D-galactose are bound weaker under the same conditions with log Ka values of only 2.5 and 2.2, respectively. (a)

(b)

Ktri HO

OH +

R B OH

–H2O

O R B O

HO

HO

B

OH

OH

HO

OH

HO

B

B

N +OH– Ktet – OH R B OH OH

HO + HO

N

+OH–

–H2O

OH

4.85

4.86

R B O O

Figure 4.95: Equilibria describing the formation of boronates from a boronic acid and 1,2-ethanediol (a), and molecular structures of phenylboronic acid 4.85 and the bis(boronic acid)-derived glucose receptor 4.86.

Bibliography [1] [2] [3] [4] [5] [6]

Pedersen CJ. Cyclic polyethers and their complexes with metal salts. J. Am. Chem. Soc. 1967, 89, 7017–36. Pedersen CJ. The discovery of crown ethers (Nobel Lecture). Angew. Chem. Int. Ed. Engl. 1988, 27, 1021–7. Greene RN. 18-Crown-6: a strong complexing agent for alkali metal cations. Tetrahedron Lett. 1972, 13, 1793–6. Busch DH. Structural definition of chemical templates and the prediction of new and unusual materials. J. Inclusion Phenom. Mol. Recognit. Chem. 1992, 12, 389–95. Rose MC, Henkens RW. Stability of sodium and potassium complexes of valinomycin. Biochim. Biophys. Acta 1974, 372, 426–35. Schalley CA. Molecular recognition and supramolecular chemistry in the gas phase. Mass Spectrom. Rev. 2001, 20, 253–309.

236

[7]

[8]

[9] [10] [11] [12] [13] [14] [15] [16]

[17]

[18] [19]

[20]

[21] [22] [23] [24] [25] [26] [27]

4 Hosting ions and molecules

Arnaud-Neu F, Delgado R, Chaves S. Critical evaluation of stability constants and thermodynamic functions of metal complexes of crown ethers (IUPAC Technical Report). Pure Appl. Chem. 2003, 75, 71–102. Gokel GW, Goli DM, Minganti C, Echegoyen L. Clarification of the hole-size cation-diameter relationship in crown ethers and a new method for determining calcium cation homogeneous equilibrium binding constants. J. Am. Chem. Soc. 1983, 105, 6786–8. Izatt RM, Bradshaw JS, Nielsen SA, Lamb JD, Christensen JJ, Sen D. Thermodynamic and kinetic data for cation-macrocycle interaction. Chem. Rev. 1985, 85, 271–339. Bianchi A, Micheloni M, Paoletti P. Thermodynamic aspects of the polyazacycloalkane complexes with cations and anions. Coord. Chem. Rev. 1991, 110, 17–113. Dietrich B, Lehn JM, Sauvage JP. Diaza-polyoxa-macrocycles et macrobicycles. Tetrahedron Lett. 1969, 10, 2885–8. Dietrich B, Lehn JM, Sauvage JP. Les cryptates. Tetrahedron Lett. 1969, 10, 2889–92. Brachvogel RC, Hampel H, von Delius M. Self-assembly of dynamic orthoester cryptates. Nat. Commun. 2015, 6, 7129. Weber E, Vögtle F. Classification and nomenclature of coronands, cryptands, podands, and of their complexes. Inorg. Chim. Acta 1980, 45, L65–L67. Krakowiak KE, Bradshaw JS, Kou X, Dalley NK. One- and two-step metal ion templated syntheses of the cryptands. Tetrahedron 1995, 51, 1599–606. Haymore BL, Lamb JD, Izatt RM, Christensen JJ. Thermodynamic origin of the macrocyclic effect in crown ether complexes of sodium(1+), potassium(1+), and barium(2+). Inorg. Chem. 1982, 21, 1598–602. Graf E, Kintzinger JP, Lehn JM, LeMoigne J. Molecular recognition. selective ammonium cryptates of synthetic receptor molecules possessing a tetrahedral recognition site. J. Am. Chem. Soc. 1982, 104, 1672–8. Schmidtchen FP. Inclusion of anions in macrotricyclic quaternary ammonium salts. Angew. Chem. Int. Ed. Engl. 1977, 16, 720–1. Dietrich B, Dilworth B, Lehn JM, Souchez JP, Cesario M, Guillhem J, Pascard C. Anion cryptates: synthesis, crystal structures, and complexation constants of fluoride and chloride inclusion complexes of polyammonium macrobicyclic ligands. Helv. Chim. Acta 1996, 79, 569–87. Dietrich B, Guillhem J, Lehn JM, Pascard C, Sonveaux E. Molecular recognition in anion coordination chemistry – Structure, binding constants and receptor-substrate complementarity of a series of anion cryptates of a macrobicyclic receptor molecule. Helv. Chim. Acta 1984, 67, 91–104. Amendola V, Fabbrizzi L, Mangano C, Pallavicini P, Poggi A, Taglietti A. Anion recognition by dimetallic cryptates. Coord. Chem. Rev. 2001, 219, 821–37. Lehn JM. Dinuclear cryptates: dimetallic macropolycyclic inclusion complexes: concepts – design – prospects. Pure Appl. Chem. 1980, 52, 2441–59. Cram DJ. Preorganization – from solvents to spherands. Angew. Chem. Int. Ed. 1986, 25, 1039–57. Cram DJ, Lein GM. Host-guest complexation. 36. Spherand and lithium and sodium ion complexation rates and equilibria. J. Am. Chem. Soc. 1985, 107, 3657–68. Cram DJ, Dicker IB, Lauer M, Knobler CB, Trueblood KN. Host-guest complexation. 32. Spherands composed of cyclic urea and anisyl units. J. Am. Chem. Soc. 1984, 106, 7150–67. Crini G. Review: a history of cyclodextrins. Chem. Rev. 2014, 114, 10940–75. Saenger W. Cyclodextrin inclusion compounds in research and industry. Angew. Chem. Int. Ed. Engl. 1980, 19, 344–62.

Bibliography

237

[28] Manor PC, Saenger W. Topography of cyclodextrin inclusion complexes. III. Crystal and molecular structure of cyclohexaamylose hexahydrate, the water dimer inclusion complex. J. Am. Chem. Soc. 1974, 96, 3630–9. [29] Rekharsky MV, Inoue Y. Complexation thermodynamics of cyclodextrins. Chem. Rev. 1998, 98, 1875–1918. [30] VanEtten RL, Sebastian JF, Clowes GA, Bender ML. Acceleration of phenyl ester cleavage by cycloamyloses. A model for enzymic specificity. J. Am. Chem. Soc. 1967, 89, 3242–53. [31] Biedermann F, Nau WM, Schneider HJ. The hydrophobic effect revisited – studies with supramolecular complexes imply high-energy water as a noncovalent driving force. Angew. Chem. Int. Ed. 2014, 53, 11158–71. [32] Assaf KI, Ural MS, Pan F, Georgiev T, Simova S, Rissanen K, Gabel D, Nau WM. Water structure recovery in chaotropic anion recognition: high-affinity binding of dodecaborate clusters to γ-cyclodextrin. Angew. Chem. Int. Ed. 2015, 54, 6852–6. [33] Rekharsky MV, Inoue Y. Chiral recognition thermodynamics of β-cyclodextrin: the thermodynamic origin of enantioselectivity and the enthalpy-entropy compensation effect. J. Am. Chem. Soc. 2000, 122, 4418–35. [34] Wenz G. Cyclodextrins as building blocks for supramolecular structures and functional units. Angew. Chem. Int. Ed. Engl. 1994, 33, 803–22. [35] Wenz G. Influence of intramolecular hydrogen bonds on the binding potential of methylated β-cyclodextrin derivatives. Beilstein J. Org. Chem. 2012, 8, 1890–5. [36] Breslow R, Greenspoon N, Guo T, Zarzycki R. Very strong binding of appropriate substrates by cyclodextrin dimers. J. Am. Chem. Soc. 1989, 111, 8296–7. [37] Diederich F. Complexation of neutral molecules by cyclophane hosts. Angew. Chem. Int. Ed. Engl. 1988, 27, 362–86. [38] Odashima K, Itai A, Iitaka Y, Koga K. Biomimetic studies using artificial systems. 3. Design, synthesis, and inclusion complex forming ability of a novel water-soluble paracyclophane possessing diphenylmethane skeletons. J. Org. Chem. 1985, 50, 4478–84. [39] Smithrud DB, Sanford EM, Chao I, Ferguson SB, Carcanague DR, Evansek JD, Houk KN, Diederich F. Solvent effects in molecular recognition. Pure Appl. Chem. 1990, 62, 2227–36. [40] Kearney PC, Mizoue LS, Kumpf RA, Forman JE, McCurdy A, Dougherty DA. Molecular recognition in aqueous media. New binding studies provide further insights into the cation–π interaction and related phenomena. J. Am. Chem. Soc. 1993, 115, 9907–19. [41] Odell B, Reddington MV, Slawin AMZ, Spencer N, Stoddart JF, Williams DJ. Cyclobis(paraquatp-phenylene). A tetracationic multipurpose receptor. Angew. Chem. Int. Ed. Engl. 1988, 27, 1547–50. [42] Bühner M, Geuder W, Gries WK, Hünig S, Koch M. A novel type of cationic host molecules with π-acceptor properties. Angew. Chem. Int. Ed. Engl. 1988, 27, 1553–6. [43] Hunter CA. Molecular recognition of p-benzoquinone by a macrocyclic host. J. Chem. Soc. Chem. Commun. 1991, 749–51. [44] Allott C, Adams H, Bernad Jr. PL, Hunter CA, Rotger C, Thomas JA. Hydrogen-bond recognition of cyclic dipeptides in water. Chem. Commun. 1998, 2449–0. [45] Chang SK, Hamilton AD. Molecular recognition of biologically interesting substrates: synthesis of an artificial receptor for barbiturates employing six hydrogen bonds. J. Am. Chem. Soc. 1988, 110, 1318–9. [46] Collet A. Cyclotriveratrylenes and cryptophanes. Tetrahedron 1987, 43, 5725–59. [47] Canceill J, Collet A, Gottarelli G. Optical activity due to isotopic substitution. Synthesis, stereochemistry, and circular dichroism of (+)- and (–)-[2,7,12-2H3]cyclotribenzylene. J. Am. Chem. Soc. 1984, 106, 5997–6003.

238

4 Hosting ions and molecules

[48] Zimmermann H, Tolstoy P, Limbach HH, Poupko R, Luz Z. The saddle form of cyclotriveratrylene. J. Phys. Chem. B 2004, 108, 18772–8. [49] Canceill C, Lacombe L, Collet A. A new cryptophane forming unusually stable inclusion complexes with neutral guests in a lipophilic solvent. J. Am. Chem. Soc. 1986, 108, 4230–2. [50] Haberhauer G, Woitschetzki S, Bandmann H. Strongly underestimated dispersion energy in cryptophanes and their complexes. Nat. Commun. 2014, 3542. [51] Canceill J, Cesario M, Collet A, Guilhem J, Lacombe L, Lozach B, Pascard C. Structure and properties of the cryptophane-E/CHCl3 complex, a stable van der Waals molecule. Angew. Chem. Int. Ed. Engl. 1989, 28, 1246–8. [52] Collet A, Dutasta JP, Lozach B, Canceill J. Cyclotriveratrylenes and cryptophanes: Their synthesis and applications to host-guest chemistry and to the design of new materials. Top. Curr. Chem. 1993, 165, 103–29. [53] Canceill J, Lacombe L, Collet A. Analytical optical resolution of bromochlorofluoromethane by enantioselective inclusion into a tailor-made cryptophane and determination of its maximum rotation. J. Am. Chem. Soc. 1985, 107, 6993–6. [54] Garel L, Lozach B, Dutasta JP, Collet A. Remarkable effect of receptor size in the binding of acetylcholine and related ammonium ions to water-soluble cryptophanes. J. Am. Chem. Soc. 1993, 115, 11652–3. [55] Bartik K, Luhmer M, Dutasta JP, Collet A, Reisse J. 129Xe and 1H NMR Study of the reversible trapping of xenon by cryptophane-A in organic solution. J. Am. Chem. Soc. 1998, 120, 784–91. [56] Jacobson DR, Khan NS, Collé R, Fitzgerald R, Laureano-Pérez L, Bai Y, Dmochowski IJ. Measurement of radon and xenon binding to a cryptophane molecular host. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 10969–73. [57] Zhang D, Martinez A, Dutasta JP. Emergence of hemicryptophanes: from synthesis to applications for recognition, molecular machines, and supramolecular catalysis. Chem. Rev. 2017, 117, 4900–42. [58] Perraud O, Lefevre S, Robert V, Martinez A, Dutasta JP. Hemicryptophane host as efficient primary alkylammonium ion receptor. Org. Biomol. Chem. 2012, 10, 1056–9. [59] Perraud O, Robert V, Martinez A, Dutasta JP. The cooperative effect in ion-pair recognition by a ditopic hemicryptophane host. Chem. Eur. J. 2011, 17, 4177–82. [60] Perraud O, Robert V, Martinez A, Dutasta JP. A designed cavity for zwitterionic species: selective recognition of taurine in aqueous media. Chem. Eur. J. 2011, 17, 13405–8. [61] Zinke A, Ziegler E. Zur Kenntnis des Härtungsprozesses von Phenol-Formaldehyd-Harzen, X. Mitteilung. Ber. Chem. Chem. Ges. 1944, 77, 264–72. [62] Cornforth JW, D’Arcy Hart P, Nicholls GA, Rees RJW, Stock JA. Antituberculous effects of certain surface – active polyoxyethylene ethers. Brit. J. Pharmacol. 1955, 10, 73–86. [63] Gutsche CD, Muthukrishnan R. Calixarenes. 1. Analysis of the product mixtures produced by the base-catalyzed condensation of formaldehyde with para-substituted phenols. J. Org. Chem. 1978, 43, 4905–6. [64] Original image taken by an anonymous photographer and published under the CC0 1.0 Universal (CC0 1.0) Public Domain Dedication license. [65] Gutsche CD, Bauer LJ. Calixarenes. 13. The conformational properties of calix[4]arenes, calix[6]arenes, calix[8]arenes, and oxacalixarenes, J. Am. Chem. Soc. 1985, 107, 6052–9. [66] Gutsche CD, Gutsche AE, Karaulov AI. Calixarenes 11. Crystal and molecular structure of p-tert -butylcalix[8]arene. J. Inclusion Phenom. 1985, 3, 447–51. [67] Iwamoto K, Ikeda A, Araki K, Harada T, Shinkai S. “Cone”-“partial-cone” isomerism in tetramethoxycalix[4]arenes. influence of solvent polarity and metal ions. Tetrahedron 1993, 49, 9937–46.

Bibliography

239

[68] Arduini A, Ghidini E, Pochini A, Ungaro R, Andreetti GD, Calestani G, Ugozzoli F. p-t-Butylcalix[4]arene tetra-acetamide: a new strong receptor for alkali cations. J. Inclusion Phenom. 1988, 6, 119–34. [69] Pappalardo S, Parisi MF. Selective endo-calix complexation of linear alkylammonium cations by functionalized (1,3)-p-tert-butylcalix[5]crown ethers. J. Org. Chem. 1996, 61, 8724–5. [70] Cui J, Uzunova VD, Guo DS, Wang K, Nau WM, Liu Y. Effect of lower-rim alkylation of p‐sulfonatocalix[4]arene on the thermodynamics of host–guest complexation. Eur. J. Org. Chem. 2010, 1704–10. [71] Gale PA, Sessler JL, Král V. Calixpyrroles. Chem. Commun. 1998, 1–8. [72] Baeyer, A. Ueber ein Condensationsproduct von Pyrrol mit Aceton. Ber. Dtsch. Chem. Ges. 1886, 19, 2184–5. [73] Custelcean R, Delmau LH, Moyer BA, Sessler JL, Cho WS, Gross D, Bates GW, Brooks SJ, Light ME, Gale PA. Calix[4]pyrroles: old yet new ion-pair receptor. Angew. Chem. Int. Ed. 2005, 44, 2537−42. [74] Gross DE, Schmidtchen FP, Antonius W, Gale PA, Lynch VM, Sessler JL. Cooperative Binding of calix[4]pyrrole-anion complexes and alkylammonium cations in halogenated solvents. Chem. Eur. J. 2008, 14, 7822–7. [75] Gale PA, Anzenbacher Jr. P, Sessler JL. Calixpyrroles II. Coord. Chem. Rev. 2001, 222, 57–102. [76] Gale PA, Sessler JL, Král V, Lynch V. Calix[4]pyrroles: old yet new anion-binding agents. J. Am. Chem. Soc. 1996, 118, 5140–1. [77] Sessler JL, Gross DE, Cho WS, Lynch VM, Schmidtchen FP, Bates GW, Light ME, Gale PA. Calix[4]pyrrole as a chloride anion receptor: solvent and countercation effects. J. Am. Chem. Soc. 2006, 128, 37, 12281–8. [78] Peng S, He Q, Vargas-Zúñiga GI, Qin L, Hwang I, Kim SK, Heo NJ, Lee CH, Dutta R, Sessler JL. Strapped calix[4]pyrroles: from syntheses to applications. Chem. Soc. Rev. 2020, 49, 865–907. [79] Yoon DW, Hwang H, Lee CH. Synthesis of a strapped calix[4]pyrrole: structure and anion binding properties. Angew. Chem. Int. Ed. 2002, 41, 1757–9. [80] Gil-Ramírez G, Escudero-Adán EC, Benet-Buchholz J, Ballester P. Quantitative evaluation of anion–π interactions in solution. Angew. Chem. Int. Ed. 2008, 47, 4114–8. [81] Adriaenssens L, Ballester P. Hydrogen bonded supramolecular capsules with functionalized interiors: the controlled orientation of included guests. Chem. Soc. Rev. 2013, 42, 3261–77. [82] Peñuelas-Haro G, Ballester P. Efficient hydrogen bonding recognition in water using arylextended calix[4]pyrrole receptors. Chem. Sci. 2019, 10, 2413–23. [83] Niederl JB, Vogel HJ. Aldehyde-resorcinol condensations. J. Am. Chem. Soc. 1940, 62, 2512–4. [84] Högberg AGS. Cyclooligomeric phenol-aldehyde condensation products. 2. Stereoselective synthesis and DNMR study of two 1,8,15,22-tetraphenyl[14]metacyclophan3,5,10,12,17,19,24,26-octols. J. Am. Chem. Soc. 1980, 102, 6046–50. [85] Timmerman P, Verboom W, Reinhoudt DN. Resorcinarenes. Tetrahedron 52, 1996, 2663–704. [86] Abis L, Dalcanale E, Du Vosel A, Spera S. Structurally new macrocycles from the resorcinolaldehyde condensation. Configurational and conformational analyses by means of dynamic NMR, NOE, and T1 experiments. J. Org. Chem. 1988, 53, 5475–9. [87] Schneider U, Schneider HJ. Synthese und Eigenschaften von Makrocyclen aus Resorcinen sowie von entsprechenden Derivaten und Wirt-Gast-Komplexen. Chem. Ber. 1994, 127, 2455–69. [88] Tunstad LM, Tucker JA, Dalcanale E, Weiser J, Bryant JA, Sherman JC, Helgeson RC, Knobler CB, Cram DJ. Host-guest complexation. 48. Octol building blocks for cavitands and carcerands. J. Org. Chem. 1989, 54, 1305–12.

240

4 Hosting ions and molecules

[89] Tanaka Y, Kato Y, Aoyama Y. Molecular recognition. 8. Two-point hydrogen-bonding interaction: a remarkable chain-length selectivity in the binding of dicarboxylic acids with resorcinol-aldehyde cyclotetramer as a multidentate host. J. Am. Chem. Soc. 1990, 112, 2807–8. [90] Moran JR, Karbach S, Cram DJ. Cavitands: synthetic molecular vessels. J. Am. Chem. Soc. 1982, 104, 5826–8. [91] Tucker JA, Knobler CB, Trueblood KN, Cram DJ. Host-guest complexation. 49. Cavitands containing two binding cavities. J. Am. Chem. Soc. 1989, 111, 3688–99. [92] Moran JR, Ericson JL, Dalcanale E, Bryant JA, Knobler CB, Cram DJ. Vases and kites as cavitands. J. Am. Chem. Soc. 1991, 113, 5707–14. [93] Cram DJ, Choi HJ, Bryant JA, Knobler CB. Host-guest complexation. 62. Solvophobic and entropic driving forces for forming velcraplexes, which are 4-fold, lock-key dimers in organic media. J. Am. Chem. Soc. 1992, 114, 7748–65. [94] Soncini P, Bonsignore S, Dalcanale E, Ugozzoli F. Cavitands as versatile molecular receptors. J. Org. Chem. 1992, 57, 4608–12. [95] Rudkevich DM, Hilmersson G, Rebek Jr. J. Self-folding cavitands. J. Am. Chem. Soc. 1998, 120, 12216–25. [96] Biedermann F, Schneider HJ. Experimental binding energies in supramolecular complexes. Chem. Rev. 2016, 116, 5216–300. [97] Biros SM, Rebek Jr. J. Structure and binding properties of water-soluble cavitands and capsules. Chem. Soc. Rev. 2007, 36, 93–104. [98] Biros SM, Ullrich EC, Hof F, Trembleau L, Rebek Jr. J. Kinetically stable complexes in water: the role of hydration and hydrophobicity. J. Am. Chem. Soc. 2004, 126, 2870–6. [99] Laughrey Z, Gibb BC. Water-soluble, self-assembling container molecules: an update. Chem. Soc. Rev. 2011, 40, 363–86. [100] Gibb CLD, Gibb BC. Anion binding to hydrophobic concavity is central to the salting-in effects of Hofmeister chaotropes. J. Am. Chem. Soc. 2011, 133, 7344–7. [101] Jasat A, Sherman JC. Carceplexes and hemicarceplexes. Chem. Rev. 1999, 99, 931–68. [102] Sherman JC, Cram DJ. Carcerand interiors provide a new phase pf matter. J. Am. Chem. Soc. 1989, 111, 4527–8. [103] Chapman RG, Chopra N, Cochien ED, Sherman JC. Carceplex formation: scope of a remarkably efficient encapsulation reaction. J. Am. Chem. Soc. 1994, 116, 369–70. [104] Cram DJ. Molecular container compounds. Nature 1992, 356, 29–36. [105] Tanner ME, Knobler CB, Cram DJ. Hemicarcerands permit entrance to and egress from their inside phases with high structural recognition and activation free energies. J. Am. Chem. Soc. 1990, 112, 1659–60. [106] Cram DJ, Blanda MT, Paek K, Knobler CB. Constrictive and intrinsic binding in a hemicarcerand containing four portals. J. Am. Chem. Soc. 1992, 114, 7765–73. [107] Cram DJ, Tanner ME, Knobler CB. Host-guest complexation. 58. Guest release and capture by hemicarcerands introduces the phenomenon of constrictive binding. J. Am. Chem. Soc. 1991, 113, 7717–27. [108] Houk KN, Nakamura K, Sheu C, Keating AE. Gating as a control element in constrictive binding and guest release by hemicarcerands. Science 1996, 273, 627–9. [109] Cram DJ, Tanner ME, Thomas R. The taming of cyclobutadiene. Angew. Chem. Int. Ed. Engl. 1991, 30, 1024–7. [110] Warmuth R. o-Benzyne: strained alkyne or cumulene? NMR characterization in a molecular container. Angew. Chem. Int. Ed. Engl. 1997, 36, 1347–50. [111] Warmuth R. Inner-phase stabilization of reactive intermediates. Eur. J. Org. Chem. 2001, 423–37.

Bibliography

241

[112] Ogoshi T, Kanai S, Fujinami S, Yamagishi TA, Nakamoto Y. para-Bridged symmetrical pillar[5]arenes: their Lewis acid catalyzed synthesis and host-guest property. J. Am. Chem. Soc. 2008, 130, 5022–3. [113] Ogoshi T, Yamagishi TA, Nakamoto Y. Pillar-shaped macrocyclic hosts pillar[n]arenes: new key players for supramolecular chemistry. Chem. Rev. 2016, 116, 7937–8002. [114] Ogoshi T, Masaki K, Shiga R, Kitajima K, Yamagishi TA. Planar-chiral macrocyclic host pillar[5]arene: no rotation of units and isolation of enantiomers by introducing bulky substituents. Org. Lett. 2011, 13, 1264–6. [115] Cao D, Kou Y, Liang J, Chen Z, Wang L, Meier H. A facile and efficient preparation of pillararenes and a pillarquinone. Angew. Chem. Int. Ed. 2009, 48, 9721–3. [116] Strutt NL, Zhang H, Schneebeli ST, Stoddart JF. Functionalizing pillar[n]arenes. Acc. Chem. Res. 2014, 47, 2631–42. [117] Ogoshi T, Hashizume M, Yamagishi TA, Nakamoto Y. Synthesis, conformational and host–guest properties of water-soluble pillar[5]arene. Chem. Commun. 2010, 46, 3708–10. [118] Yu G, Zhou X, Zhang Z, Han C, Mao Z, Gao C, Huang F. Pillar[6]arene/paraquat molecular recognition in water: high binding strength, pH-responsiveness, and application in controllable self-assembly, controlled release, and treatment of paraquat poisoning. J. Am. Chem. Soc. 2012, 134, 19489–97. [119] Li Z, Yang J, Yu G, He J, Abliz Z, Huang F. Water-soluble pillar[7]arene: synthesis, pHcontrolled complexation with paraquat, and application in constructing supramolecular vesicles. Org. Lett. 2014, 16, 2066–9. [120] Ma Y, Ji X, Xiang F, Chi X, Han C, He J, Abliz Z, Chen W, Huang F. A cationic water-soluble pillar[5]arene: synthesis and host–guest complexation with sodium 1-octanesulfonate. Chem. Commun. 2011, 47, 12340–2. [121] Behrend R, Meyer E, Rusche F. I. Ueber Condensationsproducte aus Glycoluril und Formaldehyd. Liebigs Ann. Chem. 1905, 339, 1–37. [122] Freeman WA, Mock WL, Shih N.-Y. Cucurbituril. J. Am. Chem. Soc. 1981, 103, 7367–768. [123] Lagona J, Mukhopadhyay P, Chakrabarti S, Isaacs L. The cucurbit[n]uril family. Angew. Chem. Int. Ed. 2005, 44, 4844–70. [124] Cong H, Ni XL, Xiao X, Huang Y, Zhu QJ, Xue SF, Tao Z, Lindoy LF, Wei G. Synthesis and separation of cucurbit[n]urils and their derivatives. Org. Biomol. Chem. 2016, 14, 4335–64. [125] Zhang S, Grimm L, Miskolczy Z, Biczók L, Biedermann F, Nau WM. Binding affinities of cucurbit[n]urils with cations. Chem. Commun. 2019, 55, 14131–34. [126] Mock WL, Shih NY. Structure and selectivity in host-guest complexes of cucurbituril. J. Org. Chem. 1986, 51, 4440–6. [127] Meschke C, Buschmann HJ, Schollmeyer E. Complexes of cucurbituril with alkyl mono- and diammonium ions in aqueous formic acid studied by calorimetric titrations. Thermochim. Acta 1997, 297, 43–8. [128] Rekharsky MV, Mori T, Yang C, Ko YH, Selvapalam N, Kim H, Sobransingh D, Kaifer AE, Liu S, Isaacs L, Chen W, Moghaddam S, Gilson MK, Kim K, Inoue Y. A synthetic host-guest system achieves avidin-biotin affinity by overcoming enthalpy-entropy compensation. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 20737–42. [129] Cao L, Šekutor M, Zavalij PY, Mlinarić-Majerski K, Glaser G, Isaacs L. Cucurbit[7]uril⋅guest pair with an attomolar dissociation constant. Angew. Chem. Int. Ed. 2014, 53, 988–93. [130] Biedermann F, Uzunova V. D, Scherman O. A, Nau W. M, De Simone A. Release of high-energy water as an essential driving force for the high-affinity binding of cucurbit[n]urils. J. Am. Chem. Soc. 2012, 134, 15318–23. [131] Vaitheeswaran S, Yin H, Rasaiah JC, Hummer G. Water clusters in nonpolar cavities. Proc. Natl. Acad. Sci. U. S. A. 2004, 101, 17002–5.

242

4 Hosting ions and molecules

[132] Barrow SJ, Kasera S, Rowland MJ, del Barrio J, Scherman OA. Cucurbituril-based molecular recognition. Chem. Rev. 2015, 115, 12320–406. [133] Kim HJ, Heo J, Jeon WS, Lee E, Kim J, Sakamoto S, Yamaguchi K, Kim K. Selective Inclusion of a hetero-guest pair in a molecular host: formation of stable charge-transfer complexes in cucurbit[8]uril. Angew. Chem. Int. Ed. 2001, 40, 1526–9. [134] Svec J, Necas M, Sindelar V. Bambus[6]uril. Angew. Chem. Int. Ed. 2010, 49, 2378–81. [135] Lizal T, Sindelar V. Bambusuril anion receptors. Isr. J. Chem. 2018, 58, 326–33. [136] Hennrich G, Anslyn EV. 1,3,5-2,4,6-Functionalized, facially segregated benzenes – Exploitation of sterically predisposed systems in supramolecular chemistry. Chem. Eur. J. 2002, 8, 2219–24. [137] Rebek Jr. J. Recent progress in molecular recognition. Top. Curr. Chem. 1988, 149, 189–210. [138] Rebek Jr. J. Molecular recognition with model systems. Angew. Chem. Int. Ed. Engl. 1990, 29, 245–55. [139] Williams K, Askew B, Ballester P, Buhr C, Jeong KS, Jones S, Rebek Jr. J. Molecular recognition with convergent functional groups. VII. Energetics of adenine binding with model receptors. J. Am. Chem. Soc. 1989, 111, 1090–4. [140] Benzing T, Tjivikua T, Wolfe J, Rebek Jr. J. Recognition and transport of adenine derivatives with synthetic receptors. Science 1988, 242, 266–8. [141] Chen CW, Whitlock HW. Molecular tweezers: a simple model of bifunctional intercalation. J. Am. Chem. Soc. 1978, 100, 4921–2. [142] Zimmerman SC. Rigid molecular tweezers as hosts for the complexation of neutral guests. Top. Curr. Chem. 1993, 165, 71–102. [143] Klärner FG, Kahlert B. Molecular tweezers and clips as synthetic receptors. Molecular recognition and dynamics in receptor-substrate complexes. Acc. Chem. Res. 2003, 36, 919–32. [144] Fokkens M, Jasper C, Schrader T, Koziol F, Ochsenfeld C, Polkowska J, Lobert M, Kahlert B, Klärner FG. Selective complexation of N-alkylpyridinium salts: binding of NAD+ in water. Chem. Eur. J. 2005, 11, 477–94. [145] Fokkens M, Schrader T, Klärner FG. A molecular tweezer for lysine and arginine. J. Am. Chem. Soc. 2005, 127, 14415–21. [146] Schrader T, Bitan G, Klärner FG. Molecular tweezers for lysine and arginine – powerful inhibitors of pathologic protein aggregation. Chem. Commun. 2016, 52, 11318–34. [147] Sijbesma RP, Kentgens APM, Lutz ETG, van der Maas JH, Nolte RJM. Binding features of molecular clips derived from diphenylglycoluril. J. Am. Chem. Soc. 1993, 115, 8999–9005. [148] Reek JNH, Priem AH, Engelkamp H, Rowan AE, Elemans JAAW, Nolte RJM. Binding Features of Molecular Clips. Separation of the Effects of Hydrogen Bonding and π−π Interactions. J. Am. Chem. Soc. 1997, 119, 9956–64. [149] Ganapati S, Isaacs L. Acyclic cucurbit[n]uril-type receptors: preparation, molecular recognition properties and biological applications. Isr. J. Chem. 2018, 58, 250–63. [150] Ma D, Zavalij PY, Isaacs L. Acyclic cucurbit[n]uril congeners are high affinity hosts. J. Org. Chem. 2010, 75, 4786–95. [151] Hill DJ, Mio MJ, Prince RB, Hughes TS, Moore JS. A field guide to foldamers. Chem. Rev. 2001, 101, 3893–4012. [152] Juwarker H, Suka JM, Jeong KS. Foldamers with helical cavities for binding complementary guests. Chem. Soc. Rev. 2009, 38, 3316–25. [153] Waki M, Abe H, Inouye M. Helix formation in synthetic polymers by hydrogen bonding with native saccharides in protic media. Chem. Eur. J. 2006, 12, 7839–47. [154] Hou JL, Shao XB, Chen GJ, Zhou YX, Jiang XK, Li ZT. Hydrogen bonded oligohydrazide foldamers and their recognition for saccharides. J. Am. Chem. Soc. 2004, 126, 12386–94.

Bibliography

243

[155] Petitjean A, Cuccia LA, Lehn JM, Nierengarten H, Schmutz M. Cation-promoted hierarchical formation of supramolecular assemblies of self-organized helical molecular components. Angew. Chem. Int. Ed. 2002, 41, 1195–8. [156] Ferrand Y, Huc I. Designing helical molecular capsules based on folded aromatic amide oligomers. Acc. Chem. Res. 2018, 51, 970–7. [157] Bao C, Kauffmann B, Gan Q, Srinivas K, Jiang H, Huc I. Converting sequences of aromatic amino acid monomers into functional three-dimensional structures: second-generation helical capsules. Angew. Chem. Int. Ed. 2008, 47, 4153–6. [158] Ferrand Y, Kendhale AM, Kauffmann B, Grélard A, Marie C, Blot V, Pipelier M, Dubreuil D, Huc. Diastereoselective encapsulation of tartaric acid by a helical aromatic oligoamide. J. Am. Chem. Soc. 2010, 132, 7858–9. [159] Chandramouli N, Ferrand Y, Lautrette G, Kauffmann B, Mackereth CD, Laguerre M, Dubreuil D, Huc I. Iterative design of a helically folded aromatic oligoamide sequence for the selective encapsulation of fructose. Nat. Chem. 2015, 7, 334–1. [160] Park CH, Simmons HE. Macrobicyclic amines. III. Encapsulation of halide ions by in,in-1,(k + 2)diazabicyclo[k.l.m.]alkane ammonium ions. J. Am. Chem. Soc. 1968, 90, 2431–2. [161] Mahoney JM, Beatty AM, Smith BD. Selective recognition of an alkali halide contact ion-pair. J. Am. Chem. Soc. 2001, 123, 5847–8. [162] Deetz MJ, Shang M, Smith BD. A macrobicyclic receptor with versatile recognition properties: simultaneous binding of an ion pair and selective complexation of dimethylsulfoxide. J. Am. Chem. Soc. 2000, 122, 6201–7. [163] Bartoli S, Roelens S. Electrostatic attraction of counterion dominates the cation–π interaction of acetylcholine and tetramethylammonium with aromatics in chloroform. J. Am. Chem. Soc. 1999, 121, 11908–9. [164] Böhmer V, Dalla Cort A, Mandolini L. Counteranion effect on complexation of quats by a neutral calix[5]arene receptor. J. Org. Chem. 2001, 66, 1900–2. [165] Hunter CA, Low CMR, Rotger C, Vinter JG, Zonta C. The role of the counteranion in the cation– π interaction. Chem. Commun. 2003, 834–5. [166] Mazik M. Molecular recognition of carbohydrates by acyclic receptors employing noncovalent interactions. Chem. Soc. Rev. 2009, 38, 935–56. [167] Davis AP. Synthetic lectins. Org. Biomol. Chem. 2009, 7, 3629–38. [168] Tromans RA, Carter TS, Chabanne L, Crump MP, Li H, Matlock JV, Orchard MG, Davis AP. A biomimetic receptor for glucose. Nat. Chem. 2019, 11, 52–6. [169] Wu X, Li Z, Chen XX, Fossey JS, James TD, Jiang YB. Selective sensing of saccharides using simple boronic acids and their aggregates. Chem. Soc. Rev. 2013, 42, 8032–48. [170] James TD, Sandanayake KRAS, Shinkai S. A glucose-selective molecular fluorescence sensor. Angew. Chem. Int. Ed. Engl. 1994, 33, 2207–9.

5 Assembling molecules CONSPECTUS: The receptors introduced in the previous chapter have cavities surrounded by inwardly directed functional groups to mediate the recognition of suitable substrates. In this chapter, we look at recognition processes occurring between molecules with outwardly directed binding sites. Two or more of these molecules interact to yield larger assemblies, the structure and stability of which are controlled by the number and properties of their components. This chapter starts by introducing the principles underlying such self-assembly processes. We then look at the various molecular architectures that can be accessed, the interactions on which self-assembly is based, and the ways to control product formation by template effects. The chapter ends with a look at how mixtures of interacting molecules behave if they become increasingly more complex.

5.1 Self-assembly and template effects When a synthetic chemist assembles a molecule, he or she performs a series of reactions to establish the correct connectivity of the atoms in the product. Such syntheses allow the precise control over how the atoms are linked by carefully designing the nature and the order of the individual synthetic steps, taking into account selectivity issues and/or the compatibility of functional groups with reaction conditions that may require the use of protecting groups. This strategy ultimately allows accessing virtually any molecule of reasonable size, even strained or unstable ones, although the synthetic effort is high if the product is structurally complex. The overall approach is reminiscent of the manufacturing of a macroscopic object from its components on an assembly line in that the starting material is sequentially changed by adding, removing, or modifying groups in certain positions until the synthesis is complete. When a supramolecular chemist sets out to assemble molecules, she or he attempts something else. The synthetic concept in this case involves the use of carefully designed molecular building blocks (synthesized by using the covalent strategy described above) that are able to interact in a predictable way with themselves or with complementary binding partners by virtue of their in-built recognition motifs. The aim is to obtain larger aggregates with structures that are intimately linked to the shape and arrangement of the subunits from which they are composed. This noncovalent approach of assembling molecules therefore extends synthetic chemistry into the supramolecular realm. It allows accessing levels of structural complexity that are very difficult if not impossible to realize by using covalent chemistry. The underlying principles moreover differ in many aspects from those of (most) covalent syntheses. One difference is that noncovalent syntheses of even complex products proceed continuously from the starting materials to the final product without external interference and not in a stepwise fashion as in the case of covalent syntheses. Mechanistically, product formation involves continuous association and dissociation https://doi.org/10.1515/9783110595611-005

246

5 Assembling molecules

processes along an interconnected network of possible reaction pathways. The final product has to be thermodynamically favored under the chosen conditions or it does not form. Kinetically favored products are not accessible because the reversibility of noncovalent interactions ensures that incorrectly formed side products are ultimately converted into the product until the thermodynamic equilibrium is reached. This continuous error correction during the assembly is advantageous because it allows producing even structurally very complex products, ideally in quantitative yields, by only mixing the required components in the correct ratio as long as there is a thermodynamic driving force of product formation. Another aspect of noncovalent syntheses is that the products remain dynamic after they have been assembled. They may be stable under the reaction conditions, but nevertheless retain their ability to exchange building blocks if they were not kinetically trapped by postfunctionalization. Changing the external conditions could thus cause dissociation or rearrangement, which is a disadvantage because it compromises structural integrity, but may be also viewed as an advantage if responsiveness is targeted, that is, the ability of the product to change its structure or composition in response to external stimuli. It is worth mentioning in this context that the differences between covalent and noncovalent syntheses are most pronounced if the covalent approach involves irreversible and therefore kinetically controlled reactions. Covalent syntheses based on reversible bond formation share many of the characteristics of noncovalent syntheses as we will see in Sections 5.6 and 5.7. An analogy from the macroscopic world often used to illustrate the spontaneous assembly of the building blocks during a noncovalent synthesis is the construction kit whose components miraculously assemble in the correct fashion to afford the finished model when mixed or otherwise agitated. Indeed, molecules are permanently in motion in solution and potentially in search for a binding partner. However, they also autonomously stick together when properly arranged, which is usually not true for the components of a construction kit. Another aspect where the above analogy fails relates to thermodynamics because the finished model does not represent a more stable state in the macroscopic world than the disassembled components. The driving force is therefore missing for the components to interact, while in the case of molecules, the energetic gain associated with their interactions drives the assembly and its outcome. Molecules thus spontaneously assemble not only because they move around all the time, but also because the overall system reaches a more stable state in the end. This state ensues from the conditions as well as the structure and ratio of the building blocks but does not necessarily require additional external influences. The molecules thus find the way from the disassembled state to the product by themselves, which is why the overall process is termed “self-assembly.” Based on these considerations, the following definition for self-assembly processes can be derived.

5.1 Self-assembly and template effects

247

Self-assembly is the spontaneous, thermodynamically controlled association of molecules to yield larger aggregates, the structures of which are direct consequences of the shapes, arrangement, and ratio of the building blocks.

Although this definition does not exclude the receptor-substrate complexes discussed in the previous chapter, self-assembly is a much broader concept as it does not restrict the number of interacting molecules to two or only a few, and it makes no distinction between receptors and substrates. Moreover, to allow molecular building blocks to assemble into aggregates larger than themselves, they need to have divergent binding sites as opposed to typical receptors in which the binding sites are arranged in a convergent fashion around a cavity or cleft. The ultimate potential of self-assembly processes can best be appreciated by looking at examples from Nature. In this context, virus assembly is a particularly instructive example. Viruses consist of protein shells, so-called capsids, which surround and thus protect polynucleotide strands, either DNA or RNA, which are needed by the virus to reprogram infected cells and induce them to produce new virus particles. Although capsids have rather complex shapes, they are usually made up of multiple copies of a few or even only one protein to minimize the amount of genetic information encoding for their biosynthesis. The protein components of capsids must therefore be able to spontaneously assemble into the threedimensional shape characteristic for the final virus, an ability that derives from their shape, their folding pattern, and the distribution of amino acid side chains along their surfaces. This process is especially well understood for the tobacco mosaic virus (TMV) (Figure 5.1a).

(a)

(b) 100 nm

Figure 5.1: Transmission electron microscopic image of a tobacco mosaic virus (a) and top and side view of the crystal structure of a four-layer disk comprising 68 subunits of the TMV coat protein [2]. A protein subunit in this structure is highlighted in red. The image in (a) was taken from the International Committee on Taxonomy of Viruses (ICTV) database.

248

5 Assembling molecules

TMV infects various plants, especially tobacco, causing characteristic discoloration patterns on the leaves. It has a rod-like shape with a length of 300 nm and a diameter of 18 nm. Its shell comprises 2,130 identical protein subunits that surround, in a helical fashion, a single RNA strand with 6,395 nucleobase residues. The virus can be disassembled by treatment with a detergent (sodium dodecyl sulfate), and the proteins can subsequently be separated from the polynucleotides by precipitation with ammonium sulfate. When mixing the individual components again in an aqueous buffer, reassembly occurs, eventually affording fully reconstituted and infectious virus particles [1]. The mechanism of TMV formation involves a number of steps, starting with the self-assembly of 34 protein subunits into a disk that consists of two stacked 17membered rings. To illustrate this structure, the crystal structure of a four-layer disk comprising 68 individual protein subunits is shown in Figure 5.1b [2]. This disk has a central hole into which a short segment of the single-stranded RNA is subsequently included. RNA recognition is highly specific as it occurs at a specific nucleation site in the single-stranded viral RNA, which is located ca. 1,000 nucleobases from the 3′ end, while foreign RNAs that do not have this site are rejected. RNA binding then leads to a rearrangement of the disk into a helical structure that grows by the attachment of further protein disks and the simultaneous threading of the RNA through the central hole. Self-assembly stops once the RNA strand is fully buried [3]. Complex as it may seem, this process proceeds fully autonomously and extremely efficient. Every step is reversible so that errors can be corrected, rendering the final outcome being entirely driven by thermodynamics. It goes without saying that a structure of similar structural complexity would be impossible to obtain by covalent synthesis, demonstrating the huge potential of self-assembly processes to generate such large and ordered structures. The construction of virus capsids is only one of the many impressive thermodynamically controlled self-assembly processes found in Nature. Other important systems are DNA, which rarely exists in natural systems in the monomeric form, but at the first level of organization as a dimer of two complementary DNA strands. Moreover, in the nuclei of eukaryotic cells, DNA interacts with proteins, leading to compact aggregates that help protect the DNA from degradation. Proteins assemble for several reasons. The assembly of enzymes gives rise to multienzyme complexes, for example, in which the product from one reaction step is efficiently passed on to the enzyme that mediates the next step. Protein aggregation leads, however, also to the plaques found in the brains of patients suffering from Parkinson’s or Alzheimer’s disease. Cell membranes form spontaneously from the self-assembly of lipids, a process that is mainly driven by the hydrophobic effect. Although the hydrophobic groups buried within the bilayer are mainly held together by dispersion interactions that lack a directional component, the curvature of the resulting membranes is determined by

5.1 Self-assembly and template effects

249

the shapes of the individual building blocks (Section 9.1). Large vesicular structures thus result, ensuring the confinement of molecules or larger entities involved in biochemical processes within the cell. The latter example shows that structural organization occurs in biological systems at various levels, from the assembly of molecules such as lipids, proteins, or polynucleotides to the generation of single cells or multicellular living organisms (whose assembly further leads to societies). There are, however, fundamental differences in the outcome of these processes. The self-assembly of molecules is thermodynamically controlled, proceeding autonomously also in the absence of the complete biochemical machinery and leading to an equilibrium state. Living cells, on the other hand, are out-of-equilibrium systems. Hence, the term selfassembly only applies to the first case, while the interaction of self-assembled entities into larger functional systems that function out-of-equilibrium is termed self-organization. The concept of self-assembly is, of course, not restricted to biochemical systems, but can also be applied in a straightforward fashion to abiotic molecules. All that is needed are appropriate molecular building blocks that engage in intermolecular interactions to produce larger aggregates of defined structure and size. The exact mode of self-assembly depends on a number of factors, including the nature and number of groups interacting in the building blocks, their shape and mutual arrangement, the strength of the underlying interactions, the solvent, and the presence of templates. The corresponding principles should be derived by using the reaction schematically shown in Figure 5.2. These rules are applicable also to more complex systems. (a)

(b) (c)

Figure 5.2: Schematic representation of a self-assembly process involving a curved building block with two complementary recognition units that assembles into a macrocyclic tetramer (a). Alternative products such as an acyclic oligomer and rings of varying sizes are shown in (b) and (c), respectively.

250

5 Assembling molecules

Figure 5.2a shows the for this self-assembly process arbitrarily chosen ideal situation, which involves the formation of a macrocyclic tetramer from a single ditopic building block with complementary recognition units. This tetramer is only one of many possible products because self-assembly could alternatively lead to acyclic oligomers of varying lengths (Figure 5.2b) or smaller or larger rings (Figure 5.2c) so that the product mixture could potentially become very complex. To identify the factors controlling product distribution, it is helpful to differentiate self-assembly into oligomerization and cyclization steps as shown in Figure 5.3. K2intra

Kainter K3intra

Kainter K4intra

Kainter K5intra

K ainter K 6intra

K ainter etc. Figure 5.3: Stepwise self-assembly of a ditopic building block that leads to acyclic oligomers (shown in the center), which subsequently cyclize in intramolecular processes to afford the respective macrocyclic products (shown on the right). Each oligomerization step is associated with an intermolecular equilibrium constant Kainter , while the ring-chain equilibria have intramolecular constants Knintra that vary with the ring size n.

According to Figure 5.3, the oligomerization of the building blocks to afford linear products is a stepwise process, involving in each step the attachment of a monomeric building block to the chain end of a self-assembled oligomer. These oligomers have

5.1 Self-assembly and template effects

251

different lengths, but since the actual interactions underlying chain elongation are all identical, all equilibria are characterized by the same intermolecular equilibrium constant Kainter . The same functional groups responsible for chain elongation also mediate cyclization, but the extent of cyclization not only depends on the interaction strength but also on the ease with which the corresponding ring is formed. The intramolecular binding constants Knintra therefore vary with ring size n. When focusing only on the first reaction in Figure 5.3, the cyclization of the linear dimer and the competing chain elongation yielding the trimer, we notice that we came across similar equilibria in Section 3.5.3, where we discussed chelate cooperativity. In this context, we derived equation (3.14), which correlates the ratio of cyclized over linear species cc /co with the intrinsic strength of the intermolecular interactions Ka and the effective molarity EM. By adapting equation (3.14) to the formation of the cyclic dimer in Figure 5.3, we obtain equation (5.1) in which the strength of the noncovalent interaction underlying chain elongation and cyclization is denoted by Kainter , the ratio cc /co is expressed as the corresponding equilibrium constant K2intra , and in which statistical factors are absent because Kainter and Kaintra refer to experimental stability constants. K2intra = Kainter EM2

(5:1)

Rearrangement yields equation (5.2), which shows that the effective molarity quantitatively describes to what extent the intramolecular cyclization is favored over the intermolecular chain elongation. EM2 = K2intra = Kainter

(5:2)

More precisely, EM2 denotes the concentration at which the monomeric building block involved in the above self-assembly process needs to be present for chain elongation to proceed as efficiently as the cyclization of the dimer. Thus, the higher the EM2 , the stronger the cyclization is favored over chain elongation. Considering now that the self-assembly shown in Figure 5.3 allows the formation of rings of different sizes, we can generalize equation (5.1) by writing (5.3). Knintra = Kainter EMn

(5:3)

Accordingly, the formation of each ring is associated with a characteristic effective molarity, and the extents to which these EMn values differ determines the product ratio. If various rings have similar EMn values, for example, they will all be present in the final mixture. If, however, the EMn of one macrocycle is much larger than those of the others, this ring will dominate the equilibrium. Should the macrocyclic tetramer be the dominant product, for example, EM4 must be larger than the other EMn values. A specific example of this situation is the macrocyclization of 5.1 (Figure 5.4).

252

5 Assembling molecules

H O R2

O R2

N

N

H N

H

N

R1 N

R2 N

N

N

O

N

H

H N

O N H

N N H

H N

H

H

H

N H N N

O N

O

N R1

O O

R1 = 4 O R1

N

N

N

N

O N

N

O N

H N

H N

H

H

H N

N

O

N

N

N

R1

O N

N

O R2

5.1

H N

O

N H

H H

H

O O

R2 =

R1 N

O

H N H

N

O

R2

O H

5.2

Figure 5.4: Self-assembly of the ditopic building block 5.1 containing cytidine and guanosine residues at both ends into the corresponding macrocyclic tetramer 5.2.

Watson-Crick base pairing between the terminal cytidine and guanosine residues induces the exclusive formation of 5.2, which is thermodynamically very stable even in competitive media such as DMF [4]. The actual EM4 associated with the selfassembly can be estimated by relating the intrinsic stability of the guanosine-cytidine base pair Kainter with KT , the overall stability of 5.2. As the formation of 5.2 involves three consecutive oligomerization steps followed by cyclization, KT derives from mul
3 tiplying Kainter (oligomerization) with Kainter EMn (cyclization) to give equation (5.4).
4 KT = Kainter EM4

(5:4)

The value Kainter is determined independently by measuring the stability of a base pair formed from two monotopic building blocks, one containing a guanosine and the other a cytidine group. Entering this value together with KT into equation (5.4) yields EM4 values between 200 and 900 M for this system, depending on the solvent. These values range among the highest effective molarities observed for a synthetic system, consistent with the observed high fidelity of the respective self-assembly process. This efficiency is likely related to the hydrogen bonding pattern in the base pairs since the macrocyclization of a structurally related building block containing 2-aminoadenosine and uridine groups instead of the cytidine and guanosine residues in 5.1 is associated with an EM of only 0.1 M [5].

5.1 Self-assembly and template effects

253

Having thus seen that the EM values of the different species that are formed in a self-assembly process control the product ratio, we have to ask which factors determine the magnitude of the effective molarity. Information in this context is derived from the general form of equation (5.3), namely equation (5.5), by expressing the equilibrium constants in terms of ΔH 0 and ΔS0 . The rearrangement of the resulting equation (5.6) leads to (5.7), which shows that EMn is a product of two terms, the first accounting for the enthalpic and the second for the entropic differences of cyclization and chain elongation. EMn = Knintra = Kainter  .   1 1 exp − EMn = exp − ΔG0intra ΔG0inter RT RT      .  1
0 1
0 exp − = exp − ΔHintra − TΔS0intra ΔHinter − TΔS0inter RT RT       1
0 1
0 0 exp EMn = exp − ΔHintra − Hinter ΔSintra − ΔS0inter RT R

(5:5)

(5:6)

(5:7)

The enthalpic terms relate to the strength of the interactions between the growing 0 ) and those between the two end groups of chain and an additional monomer (ΔHinter 0 the chain (ΔHintra ). If the ring is strain-free, these enthalpic terms are equal because oligomerization and cyclization are mediated by interactions between the same 0 0 − ΔHinter = 0, so that the first term in equation functional groups. In this case, ΔHintra (5.7) has a value of 1, which is optimal. More common is, however, that cyclization is associated with the build-up of strain, causing cyclization to be enthalpically less 0 0 − ΔHinter thus ends up to be favorable than chain elongation. The difference ΔHintra positive, leading to a reduction of EMn . These considerations apply if the enthalpic driving force of cyclization and chain elongation derives only from the interaction of the end groups. If, however, one pathway leads to a product that is stabilized by more than the end group interactions, for example by interactions along the chain stabilizing a tightly folded conformation, ad0 0 − ΔHinter arise that could promote either chain elongation ditional effects on ΔHintra or cyclization. In this case, EMn can become solvent dependent if the solvent affects the conformational behavior of the acyclic and the cyclic product to different degrees. Note also the temperature dependence of the enthalpy term that causes the EMn to increase with temperature if the self-assembled rings are not free of strain. Entropically, chain elongation in each step has a disadvantage over cyclization because the molecule added to the chain has to give up degrees of translational and rotational freedom. The entropic component of EMn thus mainly derives from ΔS0inter (ΔS0inter < 0) but the intramolecular cyclization also has an entropy term, albeit typically a smaller one. In the absence of solvent effects, this cyclization entropy mainly reflects the restriction of torsional flexibility during ring closure, and

254

5 Assembling molecules

while this effect can be minimized by using rigid building blocks, it generally becomes more pronounced, the more subunits the self-assembled product contains. Hence, entropy drives the self-assembly in the direction of smaller aggregates, which can also be attributed to the fact that the formation of many copies of a product containing only a few building blocks is entropically more favorable than the formation of a smaller number of structurally more complex alternatives. The contribution of ΔS0intra therefore dictates the size of the products but does not principally prevent the formation of ordered aggregates. The effective molarity of each step of a self-assembly process in which chain elongation and cyclization compete thus depends on a balance of enthalpic and entropic factors, with enthalpy generally favoring chain elongation, unless the cyclic product is strain free, in which case there is no enthalpic difference between the intermolecular and intramolecular process. Entropy, on the other hand, favors cyclization, and the smaller the cyclic products, the stronger the preference. High EMn values are therefore expected for building blocks that are predisposed to afford the smallest possible unstrained ring and preorganized to further minimize the entropic penalty of cyclization. What is the difference between preorganization and predisposition?

It is important in this context to carefully distinguish the terms predisposition and preorganization. We saw in Section 3.5.1 that a receptor is preorganized if it undergoes only minimal structural changes when forming a complex. Analogously, self-assembling building blocks are preorganized if they are rigid, minimizing the loss of torsional flexibility when forming the product. In both cases, the term preorganization thus refers to the effects of entropy on the recognition process. Predisposition, on the other hand, means that a self-assembling building block has a strong conformational or structural preference to afford a particular product when incorporated into a larger structure [6]. This effect is thus associated with the enthalpic stability of the final product. Building blocks can, of course, be preorganized and predisposed if they are rigid and if their structure induces the formation of one product among several possible ones. The above principles were derived for the example shown in Figure 5.3 in which self-assembly leads to cyclic products. The same rules also apply to structurally more complex systems. Accordingly, self-assembly preferentially leads to products in which all binding sites are occupied to maximize the number of interactions, a criterion that is known as the principle of maximum site occupancy [7]. Any alternative product with vacant binding sites should be less stable and should therefore not form to a great extent. Unstrained products are moreover preferred over strained ones, and smaller products made up of fewer subunits over respective larger ones. Based on these rules, it is possible to devise building blocks and suitable conditions to efficiently access self-assembled products of high structural complexity.

5.1 Self-assembly and template effects

255

Important criteria for the design are the rigidity, size, and shape of the building blocks, and the spatial orientation of the interacting functional groups, all of which control the mutual orientation of the subunits in the final assembly and, hence, its overall structure. The actual interactions between the building blocks should moreover be efficient under the conditions employed and also directional to allow reliably predicting the outcome. As a consequence, only a few of the noncovalent interactions discussed in Section 3.1 are generally used for self-assembly, most importantly hydrogen bonding. In addition, coordinative interactions mediate metal-directed selfassembly. It should be noted that a crucial driving force for self-assembly in water is the hydrophobic effect. Because the building blocks involved in these processes usually do not engage in directed interactions, the structures of the products – examples are micelles or vesicles – are structurally less defined. There are, however, exceptions as we will see in Section 5.2. In the ideal case, self-assembly proceeds with high fidelity, affording mainly or even exclusively a single product. If, however, different products have similar thermodynamic stabilities, they coexist in the equilibrium. This situation arises, for example, if enthalpic and entropic factors work in opposite directions, with enthalpy favoring the larger unstrained product and entropy favoring the smaller but slightly strained analog. In this case, it is useful to take advantage of the dynamic nature of the self-assembly process by controlling product formation through the addition of suitable templates. How do templates work?

The concept of using templates to control the outcome of a reaction was first mentioned in Section 4.1, which also contains Busch’s definition of a template. We saw in this chapter that crown ether syntheses benefit from the presence of alkali metal ions in the reaction mixture that preorganize the linear precursor of the product, thus facilitating the cyclization. Let us look at this reaction again in more detail. The starting material is an oligoethylene glycol with a tosylate group at one end. This compound is treated with a base that induces the deprotonation of the free hydroxy group. The alkoxide thus generated now displaces the tosylate group at the respective carbon atom, thus affording an ether. Figure 5.5 shows in a schematic fashion that two competing pathways exist for such α,ω-difunctionalized precursors to react – an intermolecular one affording an acyclic oligomer and an intramolecular one that yields the macrocycle. Assuming that the ring is unstrained, the activation enthalpies of both pathways should be similar, but their activation entropies differ because the acyclic precursor has to adopt a cyclic conformation in which the two end groups are in close proximity. The associated unfavorable activation entropy renders the cyclization slower than the oligomerization as reflected in the energy profiles of the two reaction paths (Figure 5.5).

256

5 Assembling molecules

Template-stabilized transition state of the cyclization

Transition state of the cyclization

ΔG°

ΔΔG++ that determines how much faster the oligomerization proceeds with respect to the cyclization in the absence of the template ΔΔG++ that determines how much faster the template-mediated cyclization proceeds with respect to the oligomerization

Linear precursor

Macrocycle Oligomer Reaction coordinate Figure 5.5: Competing pathways of an irreversible reaction between functional groups at the ends of a linear precursor with the associated energy profiles. The presence of a template in the reaction mixture preorganizes the precursor for the intramolecular reaction, thus decreasing the associated ΔG‡ and causing cyclization to become the preferred reaction pathway.

Since the Williamson ether synthesis underlying crown ether formation is an irreversible reaction, the ratio of oligomerization and cyclization is determined by the difference of the Gibbs free energy of activation ΔΔG‡ of the two reactions. Changing the product ratio thus requires changing the rate of one reaction with respect to the other one, which is exactly how the alkali metal ions that serve as templates work. They interact with the linear precursor by inducing a folded conformation that is well suited for ring formation. Complex formation thus reduces the unfavorable activation entropy of cyclization, causing the crown ether to become the favored product. Since the template selectively increases the rate with which the desired product is formed, it acts as a kinetic template. We can thus conclude: Irreversible reactions that proceed along competing pathways are controlled by kinetic templates if the template selectively stabilizes the transition state of one reaction, thus increasing the rate with which the respective product is formed.

In the case of reversible reactions, the product ratio is determined by the thermodynamic stabilities of the products formed along the different pathways, that is, by how strongly the ΔG0 values of the products differ. This correlation is illustrated by the

5.1 Self-assembly and template effects

257

energy profile shown in Figure 5.6 for a reversible reaction yielding two products. An example of such a reaction could be the above-mentioned self-assembly process that either gives rise to the entropically favored cyclic trimer or the entropically disfavored but less strained tetrameric analog. In the absence of a template, the trimer dominates in the equilibrium according to Figure 5.6, and for the tetramer to become the favored product, it has to be thermodynamically stabilized. This can be achieved by adding to the reaction mixture a template that selectively interacts with the tetramer. The formation of the respective complex causes the removal of the tetramer from the equilibrium that has to adjust according to Le Chatelier’s principle by regenerating the tetramer at the expense of the cyclic trimer. For this effect to work, the template has to form a stable complex with the compound in the equilibrating mixture of products that should be selected. Since the template acts on the thermodynamics of the process, the reaction control is achieved by a thermodynamic template effect. We can thus generalize as follows:

ΔG°

The product distribution of reversible reactions that afford different products are controlled by using thermodynamic templates that selectively interact with one of the interconverting products, causing its enrichment in the equilibrium.

Tetramer

ΔΔG° that determines how strongly the trimer is favored over the tetramer in the absence of the template

Building blocks

ΔΔG° that determines how strongly the template complex of the tetramer is favored over the trimer

Trimer Template complex of the tetramer Reaction coordinate

Figure 5.6: Competing pathways of a thermodynamically controlled self-assembly process that affords a cyclic trimer and a cyclic tetramer. The trimer is favored in the absence of a template. A suitable template that selectively binds to the tetramer causes a thermodynamic stabilization of this compound and its enrichment in the equilibrium.

258

5 Assembling molecules

The metal-directed self-assembly shown in Figure 5.7 is a reversible process whose preferential outcome is controlled by the addition of thermodynamic templates. In this example, the coordination of (E)-1,2-di(pyridin-4-yl)ethene to the square planar palladium(II) complex [Pd(en)(NO3)2] (en = ethylenediamine) in D2O gives rise to a mixture of the trimetallic and tetrametallic metallacycles 5.3a,b [8]. The ratio of the two products depends on the concentration, with the trimer 5.3a dominating at low concentration (0.1 mM) and the tetramer 5.3b at higher concentrations (10 mM). Accordingly, 5.3b represents the enthalpically favored product, while the smaller 5.3a is entropically favored. The addition of p-dimethoxybenzene to a solution containing a mixture of 5.3a and 5.3b (2 mM) shifts the equilibrium toward the smaller macrocycle. By contrast, 5.3b dominates the equilibrium in the presence of 1,3adamantanedicarboxylic acid. These templates thus lead to the selective amplification of the respective size-complementary metallacycles. Similar coordination-driven self-assembly processes are presented in Section 5.5.

NH 2 n H 2 N Pd ONO2 ONO2 +

H2 N N Pd N H2 N

6+ H2N N Pd NH 2 N

8+

NH 2 H 2 N Pd N N

H2N N Pd NH 2 N

2n NO3

N

+

N H 2 N Pd N NH 2

N Pd NH 2 H2N

n

N

N

5.3a

N N Pd NH 2 H2N 5.3b COOH

MeO

OMe COOH

H2 N N Pd N H2 N

6+ H2N N Pd NH 2 N OMe

8+

NH 2 H 2 N Pd N N

H2N N Pd NH 2 N COOH

MeO N Pd NH 2 H2N N

COOH N H 2 N Pd N NH 2

N N Pd NH 2 H2N

Figure 5.7: Formation of the metallacycles 5.3a and 5.3b by coordination of (E)-1,2-di(pyridin-4-yl) ethene to [Pd(en)(NO3)2] and effects of suitable templates on the ratio of the products.

5.1 Self-assembly and template effects

259

While the template effect on the equilibrium shown in Figure 5.7 is easily rationalized, with each template favoring the formation of the metallamacrocycle with which it interacts best, it is important to emphasize that dynamic equilibria respond to changes in the conditions, such as the variation of the temperature or concentration, or also the presence of an additional binding partner, by assuming a new state that represents the thermodynamic minimum of the whole system. This system can be rather complex because it includes the solvent molecules and all possible solute–solvent and solute–solute combinations. Templates thus potentially also amplify structures that do not even exist in the absence of the template or that do not represent the optimal binding partners. We will see examples of such cases in Section 5.7. Can molecules be sorted?

Another way to increase the complexity of self-assembling systems is to use mixtures of building blocks that contain complementary functional groups, allowing several compounds to participate in product formation. Although self-assembly could in this case yield an intractable mixture of products, this outcome can often be avoided by the choice of the building blocks. As a consequence, order can prevail over chaos also in mixed systems. Possible scenarios are illustrated schematically in Figure 5.8.

Case A

4

+ 4

Case B

2

+ 2

+

Case C

2

+ 2

+

Case D

2

+ 2

+ 2

+

Figure 5.8: Self-assembly of mixtures of building blocks. In case A, the two products are composed of identical building blocks (narcissistic self-sorting), and in case B, the product is produced from different building blocks (social self-sorting). Both cases are completive as no unused building blocks remain. Case C shows the situation of a social but incomplete self-assembly. In case D, all building blocks are incorporated in the product, representing an integrative, social, and completive self-sorting.

260

5 Assembling molecules

Case A in Figure 5.8 involves two self-complementary compounds that preferentially interact with themselves. Since the actual interactions are the same for both compounds, involving hydrogen bonding, for example, the curved building block can also bind to the angular one. The corresponding complexes are, however, less stable and are therefore eventually outcompeted by the complexes containing the better complementary binding partners. We therefore expect self-assembly to afford only products composed of identical building blocks. As a consequence, a mixture of compounds autonomously ends up in a state in which no mixed products are present, explaining why the underlying process is termed self-sorting [9]. According to the general definition, mixtures undergo self-sorting if the individual molecules “recognize their mutual counterparts selectively so that specific pairs are formed rather that a library of all possible noncovalent complexes of the compounds present in the mixture” [10]. The other cases in Figure 5.8 therefore also represent self-sorting processes and to differentiate them, additional terms are used [11]. In case A, the self-sorting is narcissistic because only identical subunits like to bind to each other. It is moreover completive because no subunits remain unused. Case B also represents a completive self-assembly, but it does not involve selfcomplementary binding partners. The two building blocks thus form crossover species, leading to a product containing both of them in a specific arrangement. This scenario is termed social self-sorting. In case C, an additional building block is present in the mixture that can, however, only be integrated into structurally rather complex products. Since the formation of these products is entropically and potentially enthalpically disfavored, the third building block remains unused, rendering this self-assembly incomplete. Finally, case D represents a situation where all components are integrated into the same product and the respective selfsorting process is therefore social, completive, and integrative. The term integrative is also used for the social self-sorting in case B, while narcissistic self-sorting is never integrative. We will see several examples of self-sorting processes in the following chapters. This chapter cannot end, however, without referring to a fundamental study in selfsorting performed by Lyle Isaacs [9]. In this study, the nine compounds 5.4–5.12 shown in Figure 5.9, all of which contain hydrogen bond acceptors and donors, were mixed in chloroform together with barium picrate, and the composition of the resulting mixture was assessed. Independent studies had shown previously that, individually, all of these compounds produce characteristic self-assembled structures. We will come back to some of these systems in Section 5.3, and it should therefore suffice noting here that 5.4 forms a macrocyclic pentamer of which two rings stack in the presence of barium ions, 5.5 forms a similar assembly comprising a stack of four tetrameric rings, three molecules of 5.6 form a bis(rosette) together with six barbiturates 5.7, and all other compounds 5.8, 5.9, 5.10, 5.11, and 5.12 assemble

261

5.1 Self-assembly and template effects

NH2 NH2 N Si O

O

N

NH

N

O

O

N

Si O

O

O

O

R'

R'

NH HN

O O

O O

NHHN

R' NH

HN Ph HN

NH NH

O

N

N

N

5.10

N

N

R O

N

N

N

R R

O

O

O

O NH

5.11 (R = CO2 Et)

O

NH

R

O

5.7

NH Ph NH

N

Ph

O

N

N

N

N

R O

O

O

N

HN

N

O

Ph

Ph

Ph

NH

O

N

HN O

Ph

O

H N

5.9

O

N H

OPr

OPr

HN NH

NH

N N

5.6 H N

N

O

OR RO RO OR 5.8 (R = C10 H21 , R' = p-tolyl)

O

N NHHN

N

PrO PrO

O

Ph

NH2

NH2

N

5.5

R'

HN

N

N H

Ph

O

5.4

HN

NH

N

O

N

O

O

N

N

N

N

R R

O

R

O

5.12 (R = CO2 Et)

HN HN

O Ph

Figure 5.9: Structures of the nine compounds 5.4–5.12 investigated by Isaacs with regard to their self-sorting behavior when simultaneously present in solution.

into respective dimers. Perfect self-sorting should therefore lead to eight different products, while unspecific aggregation could afford a very complex product mixture. The fact that the mixture of all nine compounds gives rise to an 1H NMR spectrum that is almost a superposition of the spectra of the eight individual selfassembled products, provided that the concentration and ratio of the individual species are correctly chosen, demonstrates that almost perfect self-sorting occurs in this system. According to the influence of external parameters on the fidelity of self-sorting, the following general rules furthermore apply. – Increasing the temperature causes the dissociation of the self-assembled products, with the least stable aggregates dissociating first. Lowering the temperature, on the other hand, increases the amount of crossover species and thus decreases the fidelity of the self-assembly process if the exchange remains fully reversible.

262

5 Assembling molecules

– A very small difference between the stabilities of the self-sorted and the crossover products causes crossover products to be present in the product mixture to a substantial extent. – Similarly, the outcome of the self-assembly is also determined by the relative amounts of the components in the mixture. If the concentration of one component is too small, its homodimerization is less likely than the association with another species present at a higher concentration, which leads to social self-sorting. Narcissistic self-sorting dominates, on the other hand, if the required components are present in the correct ratios and at sufficiently high concentrations. – Finally, the presence of a large excess of a potential competitor in the mixture has an adverse effect on the self-sorting process. The authors concluded on the basis of these results that “self-sorting is neither the exception nor the rule” [9]. If systems become complex, their response to environmental changes depends on many parameters, which makes predicting the behavior difficult. Nevertheless, if the conditions and building blocks are properly chosen, even complex systems undergo self-assembly and self-sorting with high fidelity. These investigations thus provided valuable insight into the rules underlying selfassembly processes. The same rules also apply to natural systems and thus help to understand their behavior.

5.2 Self-assembly mediated by the hydrophobic effect The hydrophobic effect results from the tendency of water molecules to form an extended hydrogen-bonded network (Section 3.3). Transferring individually hydrated hydrophobic molecules from their cavities in the water matrix into a single cavity, where all of them are hydrated together, is generally associated with a gain in the Gibbs free energy of the entire system, and the hydrophobic effect therefore provides an important driving force for the self-assembly of hydrophobic molecules in water. It usually lacks a pronounced directional component, which makes it difficult to control the structure of the resulting assemblies, but many self-assembled systems are nevertheless formed in this way. Examples are lipid bilayers in which the hydrophobic chains of the individual components are buried in the interior of the membranes, or many protein complexes, including the viral coat proteins mentioned earlier. Although not strictly a self-assembling process, the folding of proteins is also often mediated to a substantial degree by the screening of hydrophobic residues from water. It should be noted in this context that protein folding and self-assembly share many common features. Both are often thermodynamically controlled, for example, thus proceeding autonomously. It is still useful to distinguish self-folding from self-assembly because the former is an intramolecular and the latter an intermolecular process.

5.2 Self-assembly mediated by the hydrophobic effect

263

Self-assembly driven by the hydrophobic effect can be used to generate micelles, vesicles, and larger multilamellar structures in water by using natural or synthetic amphiphiles. Although rules exist that allow correlating the structure, size, and composition of such assemblies with the structures of the underlying building blocks, these large and often structurally complex systems are not considered here. Instead, we focus on building blocks that lead to discrete assemblies by looking at three examples, of which two derive from the velcrands introduced in Section 4.1.9. Such velcrands preferentially adopt the kite conformation, which is prone to dimerize even in organic solvents, especially if substituents such as methyl groups are present in the 2-position of the resorcinol subunits. This tendency is even more pronounced for water-soluble velcrands such as 5.13 (Figure 5.10a) because of their large and difficult-to-hydrate hydrophobic surfaces (Figure 5.10b). These compounds therefore dimerize in water to shield the hydrophobic regions from the solvent, leaving only the residues that carry the solubilizing groups exposed. This behavior is illustrated in Figure 5.10c by using the calculated structure the dimer of 5.13 as an example. (a)

(b)

O NO

N N

N

N

O

O

O O

O

O R

N N

N O

O

(c)

O

O

R

R R

+ 5.13 R = Cl

–

N

Figure 5.10: Molecular structure of velcrand 5.13 (a) and calculated structure of its kite conformation (b), illustrating the extended hydrophobic surface that mediates the self-assembly in water. The calculated structure of the dimer of 5.13 is shown in (c). The solubilizing groups are replaced by methyl groups for reasons of clarity.

Velcrand 5.13 binds hydrophobic substrates such as linear long-chain alkanes and primary alcohols in water [12]. For complex formation to occur, the self-assembled dimer of 5.13 first has to dissociate and the resulting monomeric velcrand has to adopt the vase conformation to be able to accept the guest. The dimerization of 5.13 is therefore disadvantageous because it obstructs the guest uptake of this anyway poorly preorganized receptor. Gibb’s conformationally rigid octa acid 5.14 (Figure 5.11a) behaves more interestingly. This cavitand is also based on a central resorcinarene unit. The four resorcinol subunits are covalently linked via benzal groups whose aromatic rings are further

264

5 Assembling molecules

(a)

(b) O

–

O

O

O

O O

– O

O

– O

O

O

O

O

O

O O

O

–

O

O

O

O

OO

– O

O

– O

O O

O

– O

O

O

–

5.14

Figure 5.11: Molecular (a) and calculated (b) structure of octa acid 5.14. The surface in the calculated structure illustrates the region along the rim of the cavity that mediates self-assembly in water.

connected in their 3- and 5-positions with 3,5-dihydroxybenzoic acid moieties. 5.14 thus has a rigid structure with the eight aromatic subunits of the resorcinarene core and the benzal groups lining a hydrophobic binding pocket. The eight carboxylate groups ensure water solubility at pH > 7, but the aromatic rings surrounding the cavity render the rim relatively hydrophobic. These solvent-exposed aromatic surfaces thus mediate the dimerization of 5.14 in water (Figure 5.11b). As we have already seen in Section 4.1.9, the propensity to self-assemble depends on the type of guest that is bound. If carboxylic acids such as 1-adamantanecarboxylate occupy the cavity, their carboxylate group protrudes from the cavity opening, facilitating the hydration and thus preventing dimerization. The filling of the cavity with apolar substrates such as steroids or alkanes, on the other hand, enlarges the size of the hydrophobic surface at the cavity opening, thus promoting self-assembly and leading to a molecular capsule in which two molecules of 5.14 surround an encapsulated substrate [13]. Large substrates such as steroids have the perfect size to completely fill the void of the octa acid dimer. They therefore form 2:1 complexes in which one substrate molecule is surrounded by two octa acid subunits. Complex formation restricts the mobility of the substrate, which can only rotate along the long axis of the cavity. Each half of the capsule therefore hosts one part of the guest, which allows distinguishing both capsule halves by NMR spectroscopy. The complexation of linear alkanes by 5.14 varies with chain length. Starting from n-butane, short-chain alkanes are bound in the form of 2:2 complexes with two substrate molecules occupying the cavity. This stoichiometry is retained until n-octane that forms a mixture of 2:2 and 2:1 complexes. Longer alkanes up to n-heptadecane then form 2:1 complexes in which only one substrate is bound. These complexes

5.2 Self-assembly mediated by the hydrophobic effect

265

differ, however, in the conformation of the bound alkane. In the case of n-nonane and n-decane, there is enough room in the cavity for these alkanes to adopt fully extended conformations. Longer alkanes have to coil into helical conformations to fit. The ability of 5.14 to differentiate n-butane from n-propane by complexing two molecules of the former substrate, while the latter is too small to efficiently fill the cavity, moreover allows separating these alkanes by selectively scavenging n-butane from a gaseous mixture and transferring it into an aqueous receptor solution. Another application of 5.14 or structurally related derivatives is their use as nanoscale reactors to mediate photochemical reactions such as Norrish reactions or cycloadditions (Section 8.2) [13]. The third example of a building block that self-assembles in water is the benzenederived amphiphile 5.15 (Figure 5.12a), developed in the group of Mitsuhiko Shionoya [14]. Among several derivatives, 5.15 with two pyridinium and one pyridine unit forms the most stable aggregate. Product formation involves the self-assembly of six subunits of 5.15, resulting in a cube whose crystal structure is depicted in Figure 5.12b. This cube only forms in water. In methanol, 5.15 is monomeric, clearly demonstrating that self-assembly is due to the hydrophobic effect. Because of the amphiphilic nature of 5.15, the resulting cube can be regarded as a structurally defined micelle. In this cube, the six subunits of the gear-shaped 5.15 are arranged along the faces, with the aromatic rings interdigitating and the pyridine units intercalating at every corner between two pyridinium rings. The thus produced cavity is able to host two 2,4,6-tribromomesitylene molecules. Interestingly, a derivative of 5.15 with three N-methylpyridinium groups assembles into a mixture of hexameric and tetrameric aggregates in water. Self-assembly therefore proceeds in a less defined (a)

(b)

N

+

N N

+

2I

–

5.15

Figure 5.12: Molecular structure of the amphiphile 5.15 (a) and crystal structure of the corresponding hexameric aggregate containing two included 2,4,6-tribromomesitylene molecules (b). One subunit of 5.15 is shown in red and the hydrogen atoms of the amphiphiles are omitted for reasons of clarity.

266

5 Assembling molecules

fashion than that of 5.15, presumably because the interactions between three pyridinium groups at the vertices are weaker than those between pyridine and pyridinium groups. The equilibrium is shifted in the presence of suitable templates. Adamantane, for example, induces the exclusive formation of the tetramer [14]. These examples thus show that self-complementary structures with properly arranged hydrophobic surfaces assemble into ordered structures in water. For this concept to work, a good preorganization of the building blocks is beneficial. Moreover, the subunits should be predisposed to assemble into defined structures, and the aromatic subunits that come into contact in the aggregates should be electronically complementary to reinforce the stability of the products through attractive interactions. If these factors are met, high thermodynamic stability is achieved.

5.3 Self-assembly mediated by hydrogen bonds 5.3.1 Introduction Hydrogen bonds are especially attractive interactions to mediate self-assembly [15]. The reasons are the ease with which hydrogen bond donors and acceptors are introduced into molecular building blocks of varying structure and size, and the various possibilities available to predictably influence the strength and orientation of hydrogen bonds by using the principles outlined in Section 3.1.5. Another advantage is that hydrogen bonds are found in numerous natural systems where they control folding or assembly. Nature thus not only serves as blueprint for developing hydrogen-bonded self-assembled systems, but even offers a pool of building blocks in the form of peptides, nucleobases, or larger oligonucleotides from which to choose. Typical building blocks contain two or more subunits, each having a characteristic pattern of hydrogen bond donors and acceptors, connected by a suitable linker. When these compounds self-assemble as shown schematically in Figure 5.13, the structure, composition, and stability of the resulting products are controlled by various parameters, the most important of which are the number and pattern of the hydrogen bonds between the individual components, the structure and flexibility of the linking units, and the solvent. Number of hydrogen bonds: Interactions between molecules involving the formation of only one hydrogen bond between the connecting units are neither strong enough nor is the directionality of a single hydrogen bond sufficiently high to predictably control self-assembly. The connecting units must therefore be able to engage in at least two hydrogen bonds. Accordingly, heterocyclic systems such as nucleobases or related synthetic systems are attractive connecting units because they not only allow controlling the number of hydrogen bonds but also their arrangement.

267

5.3 Self-assembly mediated by hydrogen bonds

Connecting units with complementary patterns of hydrogen bond donors and acceptors

Pattern of hydrogen bonds

e.g R N H N

4

N N

H N H H

H

O N

N R

O Linking unit

Figure 5.13: Schematic illustration of a self-assembling process mediated by hydrogen bonds between the end groups of a ditopic building block, affording in this case a cyclic tetrameric product. A specific example of such a process is shown in Figure 5.4.

Hydrogen bonding pattern: We have seen in Section 3.1.5 that the stability of hydrogen-bonded duplexes formed between two heterocyclic systems not only depends on the total number of hydrogen bonds but also on the arrangement of the hydrogen donors and acceptors. The most stable complexes are formed between binding partners, of which one contains only hydrogen bond acceptors and the other the same number of donors, rendering all primary and secondary hydrogen bonds attractive. The inversion of one or more hydrogen bonds causes a decrease of stability because secondary interactions become repulsive. The intrinsic stability of a hydrogen-bonded complex thus depends sensitively on the exact pattern of hydrogen bonds holding the building blocks together. The chelate cooperativity operating in many self-assembling systems can still lead to high overall stability, even if repulsive secondary interactions within individual hydrogen bonding patterns cannot be avoided as in the rosettes discussed in the next chapter. Besides its influence on stability, the hydrogen bonding pattern also allows controlling the mutual arrangement of the binding partners. This aspect is illustrated by using the example of the bis(pyridones) 5.16 and 5.17 shown in Figure 5.14 [16]. The bis(pyridone) 5.16 is self-complementary (if both pyridones exist in the same tautomeric form) and two molecules thus interact to form a dimer stabilized by four hydrogen bonds. The C2v symmetric bis(pyridone) 5.17, on the other hand, cannot form a dimer, and since its rigidity and linearity also prevents the formation of larger self-assembled macrocycles, it assembles into a polymeric structure. Linker structure: The length and flexibility of the linking units connecting the hydrogen bonding sites have a decisive influence on the outcome of the selfassembly. If the linker is flexible, the self-assembly is associated with a restriction of conformational mobility, reducing the stability of the final product. Hence, the principle of preorganization usually causes the assembly of rigid components to be entropically less costly, leading to a more stable final product. Rigid linkers

268

5 Assembling molecules

2

n N

N

N

H

H

O

O

N H

O

H

5.16

N

N

H O

O

5.17

O

H

H

O

N

H

H

O

N

N

N

N O

O

O

H

H

N

H

H

O

O

N

N O

O

H

H

N

H O N

Figure 5.14: Molecular structures of the bis(pyridones) 5.16 and 5.17 and their modes of self-assembly. The hydrogen bonding patterns of the two pyridone subunits causes 5.16 to dimerize, whereas 5.17 polymerizes upon self-assembly.

orienting the interacting groups in a fashion that produces strain in the selfassembled product are, however, disadvantageous as illustrated by the different stabilities of the self-assembled dimers formed from the two closely related bis(nucleoside) derivatives 5.18 and 5.19 (Figure 5.15).

NH2 N N R

2

H

N H O N H N N R N ON R R NO N R N N H N N H N O H N

N N R N O NH

O O

R=

5.18

OAc

H2N N

N

OAc H N

N

N R 2

AcO

R

O

O R N O

N H

R

N

N

N

O H

N

N

O

N

O

O N

O

N H

N H

N H NH

N

R

O

R

N

5.19

Figure 5.15: Molecular structures of the ditopic bis(nucleosides) 5.18 and 5.19 and their modes of self-assembly. 5.18 dimerizes in a strain-free fashion, whereas the dimer of 5.19 is strained.

269

5.3 Self-assembly mediated by hydrogen bonds

Both bis(nucleosides) are fully dimerized in chloroform, but the progressive addition of DMSO affects the degree of dimerization of both systems to different extents [17]. While the dimer of 5.18 needs a DMSO content of 60 vol% to fully dissociate, only 25 vol% of DMSO cause the disassembly of the dimer of 5.19 because the strained arrangement of the nucleobases renders this duplex less stable than the one formed from 5.18. Finally, the linker structure also allows controlling the structure of the assembly. Compound 5.20, for example, arranges two complementary hydrogen bonding motifs along the two edges of the heterocyclic rings at an angle of 120°. The hydrogen bonding patterns and the angle at which the two units are oriented thus predispose 5.20 to form a cyclic hexamer as shown in Figure 5.16 [18]. The introduction of a pyrrole unit between the two recognition sites as in compound 5.21 reduces the angle to 90°, and 5.21 consequently prefers to form a cyclic tetramer (Figure 5.16) [19].

H

R N

6 H

R N

N N

H N 4 H

O N

O

H

N

H O R

N

H

N H

N H

R

O H N

N

O N

N R

N

N N

N H

H

H

N

H

N

H

H H

H

O

N

N

N

H N R

H N H N

N H

N

O

R

R N

HN

H N

H

O

N O H N

N

N

H

H

H

N H O

O

N H

N R

N

N H N N R NH

N O

H N

N

R

N

O H N H

H

H N H

H H

N H O

N N

H N

N

N

R N R

N

H

O

O

H N

O H

R N

N N

N R

H N

O

N

H

H

O H

H

O H N

O

H O

O

N

5.21 (R = C4H9)

N O

H N

N H

R N

N

N

N R

R N

N

H

5.20 (R = C8H17)

R N

H N H

Figure 5.16: Molecular structures of the Janus molecules 5.20 and 5.21 and structures of the hexameric assembly produced by 5.20 and of the tetrameric assembly produced by 5.21. The image shows a coin depicting the Roman god Janus [20].

270

5 Assembling molecules

Compounds 5.20 and 5.21 are examples of so-called Janus molecules, a term coined by Jean-Marie Lehn for heterocyclic compounds featuring hydrogen bonding arrays on opposite sides with which complementary binding partners interact [21]. The name refers to the Roman god Janus, who is usually depicted with two faces looking into opposite directions. Solvent: We have seen in Section 3.1.5 that the strength of hydrogen bonds strongly depends on the polarity of the medium. Hydrogen bonds that are strong in apolar solvents such as chloroform consequently become substantially weaker in polar aprotic or even protic solvents. The reason is that polar solvents not only increase the permittivity of the medium but usually also directly interfere in the interactions between the binding partners. DMSO, for example, is a strong hydrogen bond acceptor and its competitive binding to the NH groups of the bis(nucleosides) 5.18 and 5.19 explains why the corresponding dimers exist in chloroform but are fully dissociated in DMSO. Hydrogen-bonded assemblies are therefore only stable in polar solvents or even in water if the hydrogen bonds are screened from the surrounding medium as in the DNA double helix (whose stability in water is moreover due to the cooperativity of the multiple hydrogen bonds along the polynucleotide strands) and/or if the hydrophobic effect contributes to complex formation. Self-assembly mediated by hydrogen bonds is therefore often restricted to apolar solvents such as chloroform, dichloromethane, benzene, and toluene. These solvents do not efficiently solvate hydrogen bond donors and acceptors, rendering polar compounds that do not form discrete self-assembled products but rather interact with each other in a nonspecific fashion not well soluble. Defined hydrogenbonded assemblies, on the other hand, in which all binding sites are saturated and no vacant hydrogen bond donors or acceptors are exposed to the solvent, dissolve more easily, particularly if they contain additional peripheral solubilizing groups such as the alkyl substituents in the aggregates formed from 5.20 and 5.21 (Figure 5.16). The dissolution of a polar molecule in an apolar solvent therefore often qualitatively indicates the formation of a structurally defined self-assembled product. These products are characterized by using standard techniques. NMR spectroscopy, for example, provides information about the symmetry of the products and their composition if more than one component is involved in the selfassembly. Upfield shifts of signals in NMR spectra moreover show which protons are involved in hydrogen bond formation. Colligative techniques such as vapor pressure osmometry are used to determine the average molecular weight of the assemblies, thus also allowing a structural assignment. The ultimate proof that the predicted structure is correct is usually a crystal structure. In the following sections, we look at different classes of structures that are accessible through such self-assembly processes.

271

5.3 Self-assembly mediated by hydrogen bonds

5.3.2 Rosettes Macrocyclic assemblies such as those shown in Figure 5.16, comprising heterocyclic subunits interconnected by an array of hydrogen bonds, are called rosettes. Such rosettes are also formed when a 1:1 mixture of cyanuric acid and melamine cocrystallizes, leading to a highly stable layered structure that can be heated to 350 °C without decomposition. Each molecule of cyanuric acid interacts with three surrounding molecules of melamine through nine hydrogen bonds and vice versa (Figure 5.17), leading not only to the presence of the aforementioned cyclic rosettes in these layers, but also of linear and crinkled tapes.

H N

O H

N

N

H H

N H

N

N H

N H

H

N

N

N H

N

H N

N

O H

N H

N H

H H N

N

N H

H H

N H

N

N

N

N N H

H

N

N N H

H H N

N

N H

H H

H N

N H

N

N H

N

H N H

H

N H

H

N

N N H

H H

H H

H N N

N H

H

N N H

H N N

H

O N

H

O N H

N

H

N

N H

O N

H H

N

N

O

N N

N H O

H

N

N

H

O N

H H

O

H

H

O N

O

O N

N

N

H

O H

H

O N H

N

H

H N

N

N

N

N

O

O

H H

H

H

H

H

H

O

O N

H N

H

H N

O

O N H

H N

O N

N

N

H

O H

H

O N H

H N

H

N

N

N

N

N

N

O

H

H

O N

H N

O H

H

H

H

H

N

N

O

H

H

O

O

O N

N

N

O N H

H N

H

H N

N N

O H

H

H

N

O H

N

N

O H

H N

O

O

H

N H

H N

N

N H

H

Figure 5.17: Section of an infinite layer containing equimolar amounts of cyanuric acid and melamine. The three hydrogen-bonded submotifs, the rosette, the linear, and the crinkled tape, are highlighted in red, blue, and orange, respectively.

The self-assembly of cyanuric acid and melamine into infinite structures can be prevented by either blocking certain NH groups or by using analogs with a reduced number of hydrogen bond donors and/or acceptors such as barbiturates instead of cyanuric acid or pyrimidines instead of melamine. Examples of building blocks that self-assemble into linear or cyclic structures, thus potentially allowing the isolation of a specific submotif of the cyanuric acid-melamine network, are shown in Figure 5.18a.

272

(a)

5 Assembling molecules

Analogs

Parent compounds O

O H

H

N O H Cyanuric acid

O

H

N

N

O

N

N H

N

H

O

O

N

N

R

R

H

H

O

O

N

N

H O

5.7 N

N

H

H

N

N N N H H Melamine

N N R

H H

O

O H

N H

H

H

N

N

R

N

N H

H

H

N R

H

H

N N

N

N N R

H

H

N

H N

N

N

H

H R R 5.22a (R = CH3) 5.22b (R = COOCH 3) 5.22c (R = C(CH3)3)

Larger distance

(b) Smaller distance

Figure 5.18: Examples of cyanuric acid and melamine analogs with a reduced number of hydrogen bond donors and acceptors to restrict self-assembly to the formation of linear or cyclic structures (a). Compounds 5.7 and 5.22a–c were used to demonstrate the strategy of peripheral crowding that is schematically illustrated in (b).

Several strategies were developed in the group of George M. Whitesides to favor the formation of discrete rosettes over that of infinite tapes [22]. The first, termed peripheral crowding, makes use of the fact that large substituents in the interacting binding partners come closer in a linear self-assembled structure than in a cyclic one as illustrated in Figure 5.18b. Cyanuric acid and melamine derivatives carrying sufficiently large substituents are therefore expected to preferentially form rosettes, which was proven experimentally by studying the self-assembly of diethylbarbiturate 5.7

273

5.3 Self-assembly mediated by hydrogen bonds

with a series of N,N′-diphenylmelamine derivatives 5.22a–c structurally differing in the substituents at the 4-positions of the phenyl groups (Figure 5.18a). In this series of compounds, 5.22a with the smallest substituent assembles into linear tapes when interacting with 5.7, melamine 5.22b affords the crinkled tape, and the large substituents in 5.22c finally induce the formation of the rosette [23]. The second strategy is based on the concept of preorganization. By covalently connecting two or more building blocks in a suitable fashion, the number of molecules that have to come together for rosette formation becomes smaller, thus reducing the entropic disadvantage of self-assembly. In addition, the arrangement of the connecting units in the respective building blocks potentially also disfavors the formation of unwanted products enthalpically. When the tripodal hub 5.23 (Figure 5.19) interacts with mono N-alkylated cyanuric acid derivatives, for example, only four subunits come together to form a rosette, which is therefore more stable than a rosette formed

O

O

HN

O

H

O

N

N

H

N

Br

N

H N

N

N

H

N R1

N

N

N H

H 5.23 (R1 = CH2CH2C(CH3)3)

+ HN O

+

N

OC18H37

O O N C18H37

H N O

5.25

+

O NH

N R2

N R1

5.24 (R1 = CH2CH2C(CH3)3)

O

3

N

H

H N

H

O

O H

O N

O

O

= 2

R = CH2CH2C(CH3)3

3

HN O

NH N R3

O

= R 3 = CH

2 CH2C(C 6H 5 ) 3

Figure 5.19: Structures of hubs 5.23, 5.24, and 5.25 and schematic representations of the rosettes derived thereof.

274

5 Assembling molecules

from six individual subunits [24]. The stability of such rosettes furthermore depends on the flexibility of the subunits. The hubs 5.23 and 5.24, for example, both form rosettes with four subunits, but the rosette containing the more rigid and thus better preorganized 5.23 is more stable [25]. Combining 5.24 with the tris(cyanurate) 5.25 leads to an even more stable rosette because self-assembly in this case only involves two molecules, of which one is preorganized and the other flexible enough to adapt to the required binding geometry [26]. Based on these investigations, the Whitesides group demonstrated that assemblies containing up to three rosette motifs and comprising up to ten components can be assembled with high fidelity by using the above strategies [22]. David N. Reinhoudt and coworkers achieved the stabilization of the cyanuric acid-melamine rosette motif by using calix[4]arene derivatives such as 5.26 (Figure 5.20a) [27]. These calixarenes are characterized by a pinched cone conformation, induced by the propoxy groups along the narrower rim. The two melamine units along the wider rim allow three molecules to self-assemble together with six molecules of diethylbarbiturate 5.7, affording a bis(rosette) that is stabilized by 36 hydrogen bonds, 18 in each rosette motif (Figure 5.20b) [28]. The crystal structure of such a rosette illustrates the twin role of the calixarene units: they preorganize the melamine residues and ensure efficient peripheral crowding, thus favoring the formation of the cyclic assembly.

(a)

(b)

H

N

N H

N Bu

N

H

H

N

N NHHN

PrO PrO

OPr

N

H N

N

N Bu

H

OPr

5.26

Figure 5.20: Structure of the calix[4]arene derivative 5.26 (a), and crystal structure of the bis(rosette) formed from 3 equiv of a derivative of 5.26 with nitro groups in the aromatic subunits flanking the melamine units and 6 equiv of diethylbarbiturate 5.7 (b). Protons could not be located in the crystal structure and side chains were disordered and are therefore not shown.

5.3 Self-assembly mediated by hydrogen bonds

275

This bis(rosette) has a D3 symmetry in the solid state. It is therefore chiral and formed in the absence of chiral information as a mixture of enantiomers. The two rosette motifs are tightly stacked at a distance of ca. 3.3 Å, which is typical for π-stacked systems. The melamine rings in the calixarene subunits are directed away from each other, causing the two rosette motifs to assume a staggered arrangement as shown schematically in Figure 5.21. This D3 symmetric bis(rosette) is, however, not the only possible product. If the calixarene unit adopts a conformation with the melamine residues pointing into the same direction, two additional assemblies are possible whose rosette motifs are arranged in an eclipsed fashion. These assemblies differ in the orientation of the calixarene units. If all calixarenes have the same orientation, the resulting eclipsed bis(rosette) is C3h symmetric, whereas it has a Cs symmetry if one calixarene is oriented in the opposite direction with respect to the other two. H

N

N H

N Bu

H

H

N N

N

N NHHN

PrO PrO

OPr

H

H N N Bu

N

H

H

=

OPr

N

HH

N

H

N NN N H N NN NNH NH Bu Bu

PrO PrO

OPr

=

OPr

Cyanuric acid or = barbiturate derivative

Staggered (D3) P-enantiomer

Staggered (D3) M-enantiomer

Eclipsed (C3h) symmetrical

Eclipsed (Cs) unsymmetrical

Figure 5.21: Isomers of bis(rosettes) formed from 5.26 and cyanuric acid derivatives or barbiturates.

The different bis(rosette) isomers are distinguished by their 1H NMR spectra. The D3 and C3h symmetric bis(rosettes) give rise to two NH signals in the region of the spectra between 13 and 15 ppm, where the barbiturate or cyanurate NH protons absorb because the two edges of these building blocks have different environments but all six subunits are symmetry equivalent. The Cs symmetric bis(rosette) isomer, however, produces six signals because of the reduced symmetry. As a consequence, an 1 H NMR spectrum of a mixture of all possible bis(rosette) isomers features up to

276

5 Assembling molecules

10 NH signals if the subunit exchange is slow on the NMR timescale, which is indeed observed in most cases. The integration of the NH signals thus provides information about the ratio of the different bis(rosettes). Accordingly, the bis(rosettes) formed from 5.26 and an N-alkylated cyanurate with a –CH2CH2C(CH3)3 group exist in solution (1 mM in CDCl3) in a 5:2:3 (D3/C3h/Cs) ratio, for example [29]. The staggered and eclipsed isomers therefore form in equal amounts, and the ratio of the C3h and Cs isomers nearly corresponds to the expected statistical 1:3 ratio. Selfassembly thus proceeds practically nonselectively. In other cases, the exact correlation between the effects of the cyanuric acid substituents and the preferred structure of the product is complex, depending on a combination of steric and electronic parameters. Barbiturates exclusively lead to the formation of staggered bis(rosettes) because the alkyl groups in their 5-position, which are oriented perpendicular to the ring plane, sterically disfavor the formation of an eclipsed bis(rosette). In the absence of chiral information, the staggered D3 symmetrical assemblies are formed as racemic mixtures, but one of the two mirror images can also be produced selectively by using either chiral calixarene derivatives or chiral barbiturates. When combining 3 equiv of 5.27 containing two R-configured 1-phenylethylamine residues with 6 equiv of diethylbarbiturate 5.7, for example, the M-form of the staggered bis (rosette) is exclusively formed [30]. This enantiomer also results when using the achiral calixarene 5.28 and the R-configured barbiturate 5.29 (Figure 5.22) [31]. All chiral building blocks of the latter bis(rosette) can be exchanged by treating it with a slight excess (1.2 equiv per replaced 5.29) of the cyanurate 5.30 in benzene-d6. This exchange occurs because cyanurates form stronger hydrogen bonds than barbiturates, owing to their more acidic NH groups. According to circular dichroism spectroscopy, the resulting bis(rosette) still has the original configuration although it is fully deprived of chiral building blocks. Since the product thus seems to “remember” the structure of the starting material, the observed phenomenon is an example of a chiral memory effect. The reason for the lack of isomerization during the building block exchange is that the replacement of the chiral barbiturates in the original assembly by achiral cyanurates proceeds stepwise via short-lived intermediates. Only six hydrogen bonds have to be broken and reformed during each exchange reaction, rendering this reaction faster than any alternate one involving a more extensive bis(rosette) disruption. The intermediates moreover still contain 30 intact hydrogen bonds, rendering them sufficiently stable to retain their structure while they exist. Although the final cyanurate-containing product is initially formed with high fidelity, it slowly isomerizes because of its dynamic nature. The half-life of interconversion of the two D3 symmetrical assemblies amounts to ca. 4.5 d at room temperature, showing that the kinetic stability is significantly below that of a covalently assembled compound. According to mechanistic studies, the ratelimiting step of interconversion likely involves the dissociation of a calixarene from the intact bis(rosette) (requiring the dissociation of 12 hydrogen bonds), followed

5.3 Self-assembly mediated by hydrogen bonds

(a)

H

N

N Ph

N H

H

H

N

N

N

NHHN

N

H

O H

N N H

N

277

Ph

N

N

O

H O

5.7 OPr

R

OPr

R,

5.27

*

* *

R,R

PrO PrO

* *

R,R

(b)

H

N

N Bu

N N O N H2

H

H

N

N NHHN

N

H

O H

N N NON 2 H

Bu

N

N

O

R,R

H

O

O H

N

N

H

Ph O

5.29 PrO PrO

OPr

OPr

N Bu

O

5.30

5.28

R

R R

R R

Figure 5.22: Self-assembly of a bis(rosette) from the chiral calixarene derivative 5.27 and the achiral barbiturate 5.7 (a), and of a bis(rosette) from the achiral calixarene derivative 5.28 and the chiral barbiturate 5.29 (b). In both cases, the corresponding M-configured bis(rosette) is formed. The reaction scheme in (b) shows the replacement of the barbiturate units in the chiral bis(rosette) derived from 5.28 and 5.29 with achiral cyanurates 5.30 under retention of the bis(rosette) configuration.

by rearrangement and reassociation, which leads to a mixture of the respective bis (rosette) isomers. Although the crystal structure in Figure 5.20 indicates that there is not much space between the two rosette motifs to permit the binding of additional molecules, these bis(rosettes) nevertheless have an interesting host–guest chemistry. Alizarine 5.31 (Figure 5.23a), for example, forms a complex in chloroform with the bis(rosette) derived from 5.26 and diethylbarbiturate 5.7 [32]. The complexation equilibrium is slow on the NMR timescale, leading to a complex that is only fully formed when three molecules of 5.31 per bis(rosette) are present. The guest molecules intercalate between two rosette motifs, where they form a third hydrogen-bonded rosette as evidenced by a crystal structure (Figure 5.23b). Complex formation not only increases

278

5 Assembling molecules

(a)

(b)

O

OH OH

O 5.31

Figure 5.23: Molecular structure of alizarine 5.31 (a) and crystal structure of the complex between the bis(rosette) formed from 5.26, 5.7, and three molecules of 5.31 (b). Disordered side chains and protons in the bis(rosette) are not shown for reasons of clarity.

the distance between the two melamine-barbiturate rosettes, but also causes the staggered bis(rosette) to rearrange into the eclipsed one. The dynamic nature of these complexes allowed the Reinhoudt group to further investigate the effects of structural changes in the guest molecules and the rosette components on the preferred outcome of such self-assembly processes [33].

5.3.3 Capsules Among the receptors discussed in Section 4.1, molecular cages such as cryptophanes, hemicryptophanes, carcerands, and hemicarcerands have particularly intriguing properties since they have cavities in which the substrate is surrounded from all sides and where it is kinetically trapped. The synthesis of these receptors is, however, often challenging and low yielding. Fortunately, the use of covalently assembled cages is not the only way with which the full encapsulation of a substrate can be achieved. The alternative is self-assembly as we have seen when discussing viral capsid proteins. The appeal of this approach is that it proceeds under thermodynamic control and therefore potentially affords the target receptors in high yields, provided that they represent the thermodynamically favored species in the equilibrium. The corresponding assemblies are termed molecular capsules as opposed to molecular cages, the latter of which are obtained by covalent synthesis [15, 34].

5.3 Self-assembly mediated by hydrogen bonds

279

The assembly of capsules requires the design of suitable building blocks that surround a cavity of predictable size. A capsule with an approximately spherical shape can, for example, be obtained from two hemispherical compounds with functional groups along their edges to mediate dimerization, but there are many other shapes allowing the assembly of a sphere. The structural variability of building blocks that give rise to capsules is therefore large. One mainly has to make sure that the building blocks are curved and that they assemble with their concave faces directed toward the capsule interior. An early example of a capsule not assembled from hemispherical building blocks was developed in the group of Julius Rebek Jr. and derives from the bis(glycoluril) derivative 5.10 (Figure 5.24a). This C-shaped compound dimerizes by hydrogen bond formation between the glycoluril NH and C=O groups, leading to an assembly in which the two subunits are arranged in a perpendicular fashion (Figure 5.24b). The seam between the components of the capsule resembles the way in which the two halves of a tennis ball are connected, which explains why this capsule is known as the molecular tennis ball [35].

(a)

(b) O

O HN Ph HN

N

N

NH Ph NH

Ph

Ph

N

N O

(c)

5.10

O

Figure 5.24: Molecular structure of bis(glycoluril) 5.10 (a), and calculated structure of the molecular tennis ball formed from two molecules of 5.10 in which the substituents in the glycoluril residues are omitted for reasons of clarity (b). The picture in (c) illustrates that a tennis ball is assembled in a similar fashion.

With ca. 60 Å3, the cavity of the tennis ball is sufficiently spacious to allow the inclusion of guests such as methane, ethylene, chloroform, or dichloromethane. The binding of these substrates causes the shielding of their protons, so that the corresponding signals in the 1H NMR spectrum move upfield upon complex formation. Methane gives rise to a signal at −0.91 ppm when bound inside the tennis ball, for example, whereas free methane has a chemical shift of 0.23 ppm. Guest exchange proceeds via a gating mechanism similar to the one assumed for carcerands (Section 4.1.9), which involves the folding away of one glycoluril subunit to produce an opening that allows facile egress and entry of the guest. Only four

280

5 Assembling molecules

hydrogen bonds have to be broken for this process to occur as opposed to eight hydrogen bonds for complete tennis ball dissociation [36]. Although the selfassembly of 5.10 is restricted to apolar solvents such as chloroform, capsule formation is templated by suitable guests also in solvents such as DMF in which 5.10 alone does not dimerize. A number of other glycoluril derivatives can be used to obtain capsules with larger cavities. Examples are the extended bis(glycoluril) 5.32 and the tripodal tris (glycoluril) 5.33 (Figure 5.25a). Two molecules of 5.32 assemble in apolar solvents to afford a capsule with a similarly shaped seam of hydrogen bonds as the tennis ball but a significantly larger cavity. This so-called molecular softball is stabilized by a combination of hydrogen bonds between the terminal glycolurils and between the OH and C=O groups along the bridging units as shown in in Figure 5.25b. The size of the thus produced cavity allows the incorporation of guests such as ferrocene or adamantane derivatives, or a combination of two smaller molecules [37]. Guest binding is entropically driven in CDCl3 likely because of the entropically favorable release of solvent molecules from the cavity upon entry of the guest. This softball can also be used as a reaction chamber as we will see in Section 8.2.2.

(a)

H N

O

Ph NH

N O HN Ar HN

OH N

O

N N

Ar N O

O

OH

N

N N O

O

OH

O

NH Ar NH

Ar N OH

O HN Ph HN

O

(b)

Ph

N

O

Ph N

5.33

(c)

N

N

O

5.32 (Ar = 4-n-heptylphenyl)

Ph

N O

O NH

N H

Ph

Figure 5.25: Molecular structures of bis(glycoluril) 5.32 and tris(glycoluril) 5.33 (a), and calculated structures of the molecular softball formed from two molecules of 5.32 (b), and the molecular jelly doughnut formed from two molecules of 5.33 (c). The substituents in the glycoluril residues are omitted for reasons of clarity.

5.3 Self-assembly mediated by hydrogen bonds

281

The shape of the capsule derived from the tripodal building block 5.33 explains why it is termed molecular jelly doughnut (Figure 5.25c) [38]. This capsule binds diskshaped guests such as benzene or cyclohexane, demonstrating that the size and shape of the cavities of such capsules characteristically influence their selectivity. How much space does a substrate usually occupy in a receptor cavity?

With a series of capsules available, the Rebek group systematically evaluated the correlation between the size of the guest, the volume of the capsule, and the stability of the complex [39]. They defined a packing coefficient PC in this context that relates the computationally estimated inner volume of the capsule to the van der Waals volume of the guest. According to the results, neutral guests typically form stable complexes when the packing coefficients amount to ca. 0.55, that is, when about 55% of the available cavity volume is filled. For neutral guests, the correlation is general, applying not only to the complexes of self-assembled capsules but also to those of molecular cages such as cryptophanes or carcerands. Similar packing coefficients are found for organic solvents. In this case, they can be determined by relating the minimum space occupied by a certain number of solvent molecules, calculated by multiplying the number of molecules with their van der Waals volume, to the actual volume of the corresponding amount of solvent. Solvents, accordingly, have a lot of empty space, which is likely required to enable the individual molecules to move around. Solvent packing coefficients thus reflect a compromise between the translational freedom of molecules, which is determined by entropy, and a sufficient proximity to allow enthalpically favored intermolecular interactions. The agreement found between solvent packing coefficients and those determined for receptor substrate complexes indicates that it is beneficial for molecular recognition to offer the substrate similar room within a receptor cavity as in the bulk solvent. This conclusion, which is known as Rebek’s 55% rule, has been confirmed for many systems and has thus developed into an important tool to assess the binding properties of a receptor. Packing coefficients larger than 55% are possible and typically indicate that directional attractive interactions occur between the binding partners. The formation of capsules through self-assembly can also involve more than two building blocks. Compound 5.34 (Figure 5.26a), for example, containing a glycoluril and a cyclic sulfamide residue, forms a capsule in which four subunits of 5.34 are assembled in an up-down arrangement and held together by altogether 16 hydrogen bonds [40]. This capsule does not form in the absence of suitable templates as shown by the insolubility of 5.34 in apolar solvents such as CD2Cl2, CDCl3, or benzene-d6. Only when adamantane or adamantane derivatives are additionally present, dissolution and concomitant capsule formation occurs. The most stable

282

5 Assembling molecules

(a)

(b) O

HN Ar HN

H N

N

O S

Ar

O

N H

N O

5.34 (Ar = 4-n-heptylphenyl)

Figure 5.26: Molecular structure of 5.34 (a) and calculated structure of the tetrameric capsule formed from 5.34, containing an included adamantane-2,6-dione molecule. The substituents in the glycoluril residues are omitted for reasons of clarity.

complex is formed with adamantane-2,6-dione, likely because of stabilizing interactions between the carbonyl oxygen atoms of the guest and the glycoluril NH groups. The efficient filling of this tetrameric capsule, which is evidenced by the structure of the adamantane-2,6-dione complex shown in Figure 5.26b, is an important requirement for the encapsulation since the less bulky 1,4-cyclohexanedione does not induce capsule formation. Capsules with even larger cavities are accessible by using the deep cavitand 5.35 (Section 4.1.9), whose imide groups along the cavity edge mediate hydrogen bond formation (Figure 5.27a). Two molecules of 5.35 assemble in apolar solvents to give a cylindrical capsule that has a height of ca. 18 Å and a diameter of 10 Å (Figure 5.27b). The inner volume amounts to 425 Å3, sufficiently large to allow the inclusion of several guest molecules. The included molecules are often not able to move freely, but are restricted in their mobility by the shape and size of the cavity, which leads to properties not observed outside the capsule and unusual types of stereoisomerism [41].

(a) O

O N O

N

(b)

H N

H N O N

O H N N

O

O

NO

O R

18 Å O

O N N

N

10 Å

O O

O

O

H N

O

R

R R

5.35 (R = C11H23)

Figure 5.27: Molecular structure of the self-assembling deep cavitand 5.35 (a) and calculated structure and dimensions of the corresponding cylindrical capsule. The substituents in the resorcinarene ring are replaced by methyl groups for reasons of clarity.

5.3 Self-assembly mediated by hydrogen bonds

283

Capsule formation occurs in CDCl3, benzene-d6, or toluene-d8 but not in solvents comprising molecules too large to be encapsulated such as mesitylene-d12 [42]. In this solvent, capsule formation is, however, induced by suitable guests. Benzene and toluene lead to capsules, for example, containing two molecules of either benzene or toluene. Two molecules of p-xylene are too large to be bound. In the presence of a mixture of benzene and p-xylene, the capsule containing one molecule of benzene and one molecule of p-xylene forms preferentially. The same complex even dominates in the presence of a 2:1:1 mixture of toluene, benzene, and p-xylene, indicating that one benzene and one p-xylene molecule fill the available space more efficiently than two toluene molecules.

(a) 4-Picoline Free: δ(CH3) = 2.32 ppm Bound: δ(CH3) = –2.80 ppm

3-Picoline Free: δ(CH3) = 2.28 ppm Bound: δ(CH3) = –1.65 ppm

2-Picoline Free: δ(CH3) = 2.56 ppm Bound: δ(CH3) = 0.58 ppm

N

CH 3

CH 3

N

N

CH 3

N CH 3 H3 C

CH 3 H3 C

H3 C

N

N N

N N

>95%

N

H3 C

H3 C ca. 100%

(b)

Figure 5.28: Schematic illustration of the preferred orientations of 2-, 3-, and 4-picoline inside the capsule formed from the self-assembling deep cavitand 5.35 (a), and calculated structure of the 4-picoline complex of this capsule (b). The substituents in the resorcinarene ring are replaced by methyl groups for reasons of clarity.

The 1H NMR spectroscopic fingerprints of the complexes containing two molecules of 2-, 3-, or 4-picoline provide detailed structural information [43] (Figure 5.28). In the absence of 5.35, the methyl protons of the three picolines absorb at 2.56, 2.28, and 2.32 ppm, respectively. These resonances shift to 0.58, −1.65, and −2.80 ppm upon complex formation. The different positions of these signals indicate that the picolines cannot move freely within the capsule. If they would, averaged signals should result in the 1H NMR spectra at approximately the same chemical shift. The extent to

284

5 Assembling molecules

which the signals shift moreover provides information about the orientation of the guests within the cavity. The largest upfield shift of 5.12 ppm observed for 4-picoline demonstrates that the methyl groups of the two bound molecules are almost exclusively oriented in the complex near the ends of the capsule, where they reside in the resorcinarene bowls. The two 2-picoline molecules, on the other hand, preferentially (>95%) adopt orientations with their methyl groups oriented close to the equator, where the shielding effect is least pronounced. The arrangement of 3-picoline is less well defined, with the methyl groups somewhat more likely residing near the ends (60–70%) than near the center (40–30%) of the capsule. Complexes only differing in the mutual orientation of the included guest molecules are isomeric. In the case of 3-picoline, the interconversion of the different arrangements inside the capsule is too fast to allow distinguishing the corresponding isomers, but this situation changes for other guests or guest combinations. The 1H NMR spectrum of the capsule containing one molecule of chloroform and one molecule of 4-ethyltoluene, for example, contains two signals for the terminal protons of the ethyl group, one at ca. −3.8 ppm and one at −0.1 ppm. The tumbling of the 4-ethyltoluene is thus slow on the NMR timescale, allowing the complex in which the ethyl group is located at the end of the capsule to be distinguished from that in which it is located near the equator. The integrals of the two signals show that the latter complex is four times more abundant than the former (Figure 5.29). By enclosing molecules in the cavity of a capsule and restricting their mobility, it is thus possible to realize a type of stereoisomerism that does not exist in the absence of the capsule. The corresponding isomers differ in the mutual arrangements of the two guest molecules, which is why they are termed social isomers [44].

Cl

Cl

CH3 H

Cl

H3C

80%

Cl CH 3 H

H3C

Cl Cl

20%

Figure 5.29: Schematic representation of the social isomerism observed in the complex of the capsule formed from 5.35, containing one molecule of chloroform and one molecule of 4-ethyltoluene.

Complex formation also influences the preferred conformation of the included guest as the encapsulation of alkanes within the capsule formed from 5.35 illustrates [45]. The table in Figure 5.30a shows how the relative stabilities of alkane complexes, all of which contain only one guest molecule, vary with chain length. Undecane forms the most stable complex because it perfectly fits inside the capsule in an energetically favorable extended formation (Figure 5.30b). Shorter n-alkanes are also bound but fill the available space less efficiently, resulting in weaker interactions between

5.3 Self-assembly mediated by hydrogen bonds

(b)

285

(a)

Guest

Ka(rel)

n-C 9 H 20

0.3

n-C 10 H 22

16.9

n-C 11 H 24

100.0

n-C 12 H 26

24.4

n-C 13 H 28

1.0

n-C 14 H 30

0.008 (e)

(d)

(c)

18 Å

24 Å

30 Å

Figure 5.30: Relative stabilities of alkane complexes of the capsule formed from 5.35 (a), and calculated structures of the complexes of this capsule containing undecane in the extended conformation (b) and tetradecane in the coiled conformation (c). In addition, the complexes of tetradecane in the extended conformation with a capsule containing a belt of four glycoluril units and eicosane (C20H42) with a capsule containing two belts of eight glycoluril subunits are depicted in (d) and (e), respectively. The substituents in the resorcinarene ring are replaced by methyl groups and those in the glycoluril units are omitted for reasons of clarity.

the substrate and the inner cavity walls. Complex stability thus decreases with decreasing chain length. Alkanes longer than undecane are also incorporated into the capsule. For binding to occur, these alkanes have to adopt coiled conformations, however, which is energetically costly, explaining why the respective complexes are less stable than the undecane complex. Another reason for the drop of stability is the weakening of the hydrogen bonds stabilizing the capsule, caused by the internal pressure produced by the coiled guests. The longest alkane complexed by this capsule is tetradecane in its fully coiled conformation (Figure 5.30c).

286

5 Assembling molecules

The pressure generated by the included coiled alkanes is released by incorporating a belt of four glycoluril units between the capsule halves, leading to an extended capsule with an inner volume of 620 Å3. This capsule forms spontaneously when glycoluril is added to the tetradecane complex of 5.35 because it allows the included alkane to adopt the energetically favorable extended conformation (Figure 5.30d). In the absence of suitable guests, the extended capsule does not exist because its formation is entropically too costly. If the corresponding extended capsule contains a suitable alkane, this entropic term is compensated by the enthalpic stabilization that derives from the extensive van der Waals interactions between the guest and the cavity walls. The corresponding extended capsule also binds alkanes longer than tetradecane, which have to assume coiled conformations again. The internal pressure thus generated influences the rate with which the glycoluril units rearrange along the equator, that is, the rate with which this chiral capsule racemizes [46]. This rate increases with increasing chain length of the alkane because the internal pressure, which causes the weakening of the hydrogen bonds holding the assembly together, increases into the same direction. The extended capsule takes up alkanes up to nonadecane in their coiled conformations. The next longer alkane eicosane templates the formation of a capsule with two glycoluril belts made up of altogether 11 components (Figure 5.30e), and suitable guest molecules even stabilize a capsule with three belts [47]. Self-assembled capsules are also accessible from calix[4]arene tetraureas of the general structure 5.8 (Figure 5.31a) as reported independently and almost simultaneously by Julius Rebek Jr. and Volker Böhmer. These calixarenes self-assemble in apolar solvents, yielding dimers that are held together by 18 bifurcated hydrogen bonds [48]. The crystal structure in Figure 5.31b illustrates the shape complementarity of the two calixarene units in such capsules and the interdigitation of the side chains that enables the hydrogen bonding interactions between the urea groups [49]. Such capsules incorporate monocyclic aromatic or aliphatic guests into their 180 Å3 large cavity, and also structurally more complex guests such as terpenes or even cubane. Binding studies benefit from slow exchange on the NMR timescale, which allows observing different complexes simultaneously. The 1H NMR spectrum of 5.36b (Figure 5.31a) in a mixture of benzene-d6 and toluene-d7 contains two signal sets, for example, one belonging to the benzene and the other to the toluene complex [50]. The slow exchange also makes it possible to study the self-assembly of mixtures of different calixarene derivatives. Normally, a 1:1 ratio of two calixarene tetraureas affords the two homodimeric and the heterodimeric capsules in a statistical 1:2:1 ratio [51]. The combination of 5.36c and 5.36d, however, leads to the exclusive formation of the heterodimeric capsule. While the homodimeric capsules are S8 symmetric and therefore achiral, heterodimeric capsules are chiral because the urea

5.3 Self-assembly mediated by hydrogen bonds

(a)

287

(b)

R'

R' HN

O O

HN

RO

NH HN NHHN

RO

OR

R' O O

R' NH NH

OR

5.8 (R = C10H21, R' = p-tolyl) 5.36a (R = CH2CO2Et, R' = p-tolyl) 5.36b (R = benzyl, R' = phenyl) 5.36c (R = C10H21, R' = 4-n-heptylphenyl) 5.36d (R = C10H21, R' = SO2-p-tolyl)

Figure 5.31: Molecular structures of calix[4]arene tetraureas 5.8 and 5.36a-d that self-assemble into dimeric capsules (a), and side and top view of the crystal structure of the capsule derived from 5.36a (b). The substituents along the lower rim of the calixarene units are replaced by methyl groups for reasons of clarity.

units in a capsule comprising two different hemispheres can be arranged in a counterclockwise or clockwise fashion. These capsules are normally formed as a racemic mixture. If one component is made chiral by introducing substituents with a stereogenic center, however, the preferred formation of one of the two mirror images of the heterodimeric capsule is observed [52]. The two calix[4]arene derivatives 5.37 and 5.38 (Figure 5.32) also assemble into a heterodimeric capsule, but since the interactions between the amidinium and sulfonate groups involve a combination of hydrogen bonding and ion-pairing, self-assembly occurs even in polar solvents such as methanol or methanol/water mixtures in which the calixarene tetraureas do not dimerize [53]. The cavity thus produced is available for the binding of quaternary ammonium ions such as acetylcholine.

288

R

5 Assembling molecules

R

R

R

5.37 (R = SO3Na, R' = CH2CH2OC2H5) NH2Cl 5.38 (R =

R'O R'O

OR'

, R' = C3H7) NHC3H7

OR'

Figure 5.32: Molecular structures of calix[4]arene derivatives 5.37 and 5.38 that self-assemble in polar solvents such as water/methanol mixtures, affording the corresponding heterodimer.

Capsules featuring two hemispherical subunits can be assembled from resorcinarenes. The crystallization of resorcinarene 5.39a (Figure 5.33a) from aqueous methanol in the presence of tetraalkylammonium ions affords a dimeric capsule, for example, in which one cation occupies the cavity [54]. The two capsule halves are bridged by methanol and water molecules whose OH groups complete the hydrogen bonding pattern holding the subunits together. Much more interesting than these dimers are, however, the large hexameric capsules derived from

(a)

R HO HO

OH

R

HO

R OH

HO

OH R OH

R' R' 5.39a (R = H, R' = CH3) 5.39b (R = OH, R' = CH2CH(CH3)2) R'

(b)

R'

(c)

Figure 5.33: Molecular structures of resorcinarene 5.39a and pyrogallolarene 5.39b (a), and crystal structures of the hexameric capsules assembled from 5.39a (b) and 5.39b (c). The side chains in the pyrogallol subunits in (c) are truncated for reasons of clarity.

5.3 Self-assembly mediated by hydrogen bonds

289

resorcinarenes. The crystal structure of the octahydrate of 5.39a illustrates that the six resorcinarene subunits are positioned above the faces of a cube, surrounding a cavity with a volume of 1,375 Å3 (Figure 5.33b) [55]. Of the eight OH groups in each resorcinarene unit, four serve as hydrogen bond donors to interact with a hydroxy group in the same molecule, whereas the remaining four donors are involved in the intermolecular interactions. The eight water molecules in the corners of the cube are required to complete the seam of hydrogen bonds. These water molecules donate additional 12 hydrogen bonds (four water molecules donate two bonds each, and the other four only one), bringing the total number of hydrogen bonds in this capsule to sixty. A closely related assembly with a slightly larger volume of 1,510 Å3 is formed from the pyrogallolarene 5.39b (Figure 5.33c) [56]. Due to the presence of three OH groups in the aromatic subunits of this compound, the respective capsule does not require the participation of water molecules. It is stabilized by 48 intermolecular hydrogen bonds between the pyrogallolarene units, in addition to 24 intramolecular ones within the six rings. This capsule thus contains less components and more hydrogen bonds than the capsule formed form 5.39a, rendering it more stable. Also 2,6-dihydroxypyridine forms such hexameric capsules, although the situation is more complex in this case because of amide–iminol tautomerism [57]. Such capsules also exist in the gas phase and in solution. While resorcinarene-derived capsules only form in nonpolar solvents such as chloroform, the pyrogallolarene-containing analogs tolerate more competitive media such as acetone/ water, 1:1 (v/v). Such capsules include several guest molecules. The capsule derived from 5.39a binds eight benzene molecules, for example, three biphenyl molecules, various tetraalkylammonium salts, or combinations of different guest molecules [41]. These complexes thus differ with respect to structure and stoichiometry from the structural assignment originally made by Yasohiro Aoyama in the 1980s, who suggested that resorcinarenes bind dicarboxylic acids such as glutaric acid or certain monosaccharides in the form of 1:1 or 2:1 complexes, the latter featuring the guest sandwiched between two resorcinarene moieties (Section 4.1.9). By using modern NMR spectroscopic techniques not available to Aoyama, Rebek corrected this structural assignment in 2006, and showed that the binding of these guests also occurs inside of hexameric capsules, with one capsule either hosting six glutaric acid molecules or three monosaccharides [58]. The host/ guest ratio determined by Aoyama was therefore correct, but the total number of components was not. Resorcinarene-derived capsules belong to the largest synthetic noncovalently assembled hollow structures identified so far. They are not only able to host several guest molecules simultaneously, but also to mediate their transformation. More about these properties in Section 8.3.2.

290

5 Assembling molecules

5.3.4 Tubes A tube is a cylindrical object with a characteristic length and diameter. Building blocks from which tubular structures can be assembled result either from cutting a tube parallel or perpendicular to its axis. In the first case, one or more sheets are obtained that need to be wrapped up and joined together at appropriate edges to reassemble the tube. The second mode of disassembly affords rings that lead back to the tube when stacked. A third possibility involves cutting the tube in a helical fashion to yield a tape that can fold into the original tube. Typical examples of molecules folding in this way are α-helical peptides, which are stabilized by hydrogen bonds between amino acid residues located in adjacent helix turns. Some of the foldamers discussed in Section 4.1.13 also fall into this category. Since the interaction stabilizing such structures occur intramolecularly, the corresponding tubes are, however, not formed by self-assembly, and we therefore focus on only the two other ways of producing tubular structures here. The first strategy was realized by Stefan Matile by using p-octiphenyl derivatives containing short peptides in each aromatic subunit (Figure 5.34a) [59]. These compounds are predisposed to assemble into a circular structure because of the arrangement of the peptide residues that results from the nonplanar structure of the p-octiphenyl core. Self-assembly involves the interaction of two or more of these molecules by the interdigitation of the peripheral peptide strands, yielding antiparallel β-sheets. To favor β-sheet formation, the peptides have an alternating sequence of apolar and polar amino acids. Steric effects cause the side chains of the amino acids connected directly to the p-octiphenyl core to be oriented toward the exterior of the resulting tube (or barrel as the authors call the assembly), and since the peptides contain an odd number of amino acids, the side chains of the terminal amino acids are oriented in the same direction. The residues of the amino acids in even positions thus point toward the tube’s interior, allowing control over the properties of the inner surface by varying their structure. These compounds preferentially assemble into tetrameric circular structures consisting of four p-octiphenyl rods as staves and eight peptides as hoops (Figure 5.34b), but dimeric and hexameric assemblies are also possible. The stability depends on the length of the peptide residues and their sequence as well as the length of the staves: biphenyl or p-quaterphenyl derivatives with peptide side chains do not assemble, while longer rods afford tubes whose stability increases with increasing number of aromatic subunits and length of the peptides. The length of the staves moreover defines the height of the tubes, which is therefore well controlled, much better than the height of tubes derived from self-assembling rings that are discussed later in this chapter. This strategy of producing tubular structures is very versatile. The diameter of such tubes can, for example, be controlled to some degree by changing the length of the peptide residues. Moreover, water-soluble tubes can be generated that are polar on the outside and apolar on the inside by suitably distributing polar and

5.3 Self-assembly mediated by hydrogen bonds

(a)

291

Peptide O

O O

N H

O

O

O

H N I

N H

O

O

O

H N I

N H

NH2 O

O = Side chain residing outside the tube Peptide

O O

O O

I

= Side chain residing inside the tube

Peptide

(b) Peptide

O O

Peptide

O O

O O

Peptide

Peptide

O O

Peptide

O O

Figure 5.34: General structure of the p-octiphenyl-peptide conjugates that give rise to tubular structures by interdigitation and concomitant β-sheet formation of the peptide residues (a), and schematic structure of a corresponding tetrameric tube (or barrel) (b).

apolar amino acids along the peptide residues. Such tubes host hydrophobic substrates such as carotenoids. Conversely, tubes with a polar interior and an apolar exterior are accessible that partition into bilayer membranes, thus allowing the development of multifunctional pores to mediate membrane transport processes (Section 9.2.1). The alternate strategy of assembling tubes is based on macrocyclic compounds with a suitable sequence of complementary hydrogen bond donors and acceptors along the ring to mediate stacking. Several types of macrocyclic compounds are used for this purpose, the majority of which derive from cyclic peptides. The in this context probably most important cyclopeptides are composed of an even number of alternating L- and D-amino acids. That these compounds should be able to stack in a β-sheet-like arrangement was predicted as early as 1974 [60]. The experimental proof had to wait until 1993, however, when M. Reza

292

5 Assembling molecules

Ghadiri devised a way to control the assembly [61]. The respective strategy involves the use of cyclopeptides with acidic amino acid residues such as the glutamic acid subunits in cyclooctapeptide 5.40a (Figure 5.35a). These residues render the corresponding peptides soluble in basic media where assembly does not occur because of Coulomb repulsion between the negatively charged carboxylate groups in the side chains. The careful acidification of the respective solutions gradually removes this repulsion, allowing the peptides to stack in a controlled fashion that ultimately leads to the precipitation of microcrystalline fibers. These fibers contain tubes of peptides, which are stacked as schematically shown in Figure 5.35b. Cyclopeptide 5.40b provides insight into structural parameters of the assembly. This peptide contains N-methylated peptide bonds at every second amino acid subunit, which restrict the self-assembly to dimerization because the stacking of further rings is prevented. The crystal structure of the resulting dimer, shown in Figure 5.35c, confirms that the peptide bonds are oriented parallel to the main axis of the rings and alternately point into opposite directions so that every second NH bond is engaged in hydrogen bond formation [62]. The stability of this dimer amounts to 80 M−1 in CDCl3.

(a)

NH2 O

O HN D HN

(b)

O

O NH

O

L

L

O H HO H H OH H O N N N N N N R N N R OH O H H O HO

NH

O

D D

O

HN

NH

L

HO

L

O

D

N H

HO

R

N H

O

O

O 5.40a

O

H2N

O Ph O HN

Ph O

NH

N D

L

L

D

O

N O

L

NH

N N H OH H O

HO

R

D

O H H O

N

N N

OH

N R N N H O HO H

O H HO H H OH H O N N N N N N R N N R OH O H H O HO

N

O

Ph

(c)

OH

O H H O

N

N N H OH H O

N N

OH

N R N N H O HO H

L D

HN

N

O Ph

O 5.40b

Figure 5.35: Molecular structures of cyclopeptides 5.40a and 5.40b (a), schematic illustration of the β-sheet-like arrangement with which these peptides stack (b), and crystal structure of the dimer of 5.40b (c). Phenyl rings and protons except those on NH groups are omitted in the crystal structure for reasons of clarity. The spheres represent the N-methyl groups.

293

5.3 Self-assembly mediated by hydrogen bonds

The behavior and properties of these cyclopeptides can be varied in a wide range by changing the number, sequence, and nature of the amino acids from which they are composed. Appropriate cyclopeptides also assemble within bilayer membranes to afford pores whose diameter and height are determined by the size of the rings and the thickness of the membrane, respectively. These pores mediate the transport of appropriately sized ions or molecules. In the case of cyclic octapeptides, the pores have an internal diameter of 7 Å, for example, which is large enough to let sodium or potassium ions pass, whereas the 10 Å-wide pores of cyclic decapeptides allow the passage of small organic molecules such as glutamic acid or even glucose. These cyclopeptides also possess a pronounced in vitro antibacterial activity [63]. Cyclopeptides with an alternating sequence of D-α-amino acids and (1R,3S)-3aminoalkanecarboxylic acids (γ-amino acids), which were introduced by Juan R. Granja, also have conformations and arrangements of the peptide bonds that allow them to stack [64]. An example is cyclopeptide 5.41a containing (1R,3S)-3aminocyclohexanecarboxylic acid subunits (Figure 5.36a). In these cyclopeptides, one ring face contains only the α-amino acid and the other the γ-amino acid NH and C=O groups. Tube formation therefore either involves an antiparallel arrangement of the rings, so that they alternately interact through their α- and their γ-faces, or a parallel arrangement in which the α-faces interact with the γ-faces (Figure 5.36b). The α,α and their γ,γ interactions differ substantially in strength as demonstrated

(a)

(b)

O

O

CH 3

NCH 3 NO HN

O N

O O R1

R2 N

O

H

R1

N O

N

N R1

O

H

O H

O

O N N HN N N O N CH 3 CH 3 O O CH3 Dimer formed from 5.41b γ,γ-stacking O

N

N R O O R1 5.41a (R1 = R2 = H) 5.41b (R1 = CH3, R2 = H) 5.41c (R1 = H, R2 = CH3) 2

H

O

CH 3 N

N

CH 3

O NH

N CH 3 O

O N

N

N H

O

O

H

NO

O

CH 3

O

O N CH N3 NH O N H O

O

N

H

CH 3

O N CH N3 NH O N H O H

CH 3

O N H O

N O H

O N N HN N N O N CH 3 CH 3 O O CH 3 Dimer formed from 5.41b and 5.41c α,γ-stacking

N

CH 3

CH 3 Dimer formed from 5.41c α,α-stacking

O

Figure 5.36: Molecular structures of cyclopeptides 5.41a–c (a), and schematic illustration of the different modes with which the partially methylated derivatives 5.41b and 5.41c self-assemble (b). The interaction between two α-faces is more efficient than the interaction between two γ-faces. The most stable dimer is formed when an α- and a γ-face interact.

294

5 Assembling molecules

by the stability of dimers formed from suitable N-methylated cyclopeptides [65]. While 5.41b, in which the N-methylation of the α-amino acid NH groups restricts the interactions to the γ-face, forms a dimer in CDCl3 that has a log Ka of 2, the αface stacked dimer of 5.41c has a significantly larger log Ka of 6 in chloroform. When treating the more stable dimer of 5.41c with 5.41b, the heterodimer is formed, indicating that the parallel arrangement of the rings is the most favored arrangement (Figure 5.36b) [66]. Analogous tubes are formed from cyclopeptides containing cyclopentane instead of cyclohexane rings or from cyclooctapeptides, the latter of which afford tubes with larger inner diameters than the hexapeptides. Corresponding cyclic tetrapeptides are, however, unable to self-assemble because they cannot adopt the required disc-shaped conformation. Tubes assembled from such cyclopeptides contain the β-methylene groups of the cyclic amino acid residue within their lumen. By using peptides with γ-amino acids containing functional groups in β-position, it is thus possible to control the polarity of the inner surface of the respective tubes, which is not possible when using cyclopeptides containing alternating D and L α-amino acids [67]. The self-assembly of 5.40 and 5.41 leads to tubes in which the NH and C=O groups are oriented in both directions. A different arrangement is realized when using cyclopeptides derived from β-amino acids as building blocks. The respective hydrogen bonding pattern is shown in Figure 5.37 for cyclopeptide 5.42 containing four S-β3-homoleucine residues, which assembles into a tube with an inner diameter of ca. 2.6 Å [68]. In this tube, all C=O groups point into one direction and the NH groups into the opposite direction, resulting in a dipole moment that potentially influences ion transport along the interior.

O O

R O R

H N

H

R

O O

O

HN

R

R

N H

R

N H

NH

O

N

O O

R

O

N H O O

5.42 (R = CH2CH(CH3)3) R

N H

O O

N N H H O O

N N H H O O

N N H H O O

N N H H

N H N H N H N H

R

R μ R

R

Figure 5.37: Molecular structure of 5.42 and schematic illustration of the mode of assembly of this cyclopeptide. The arrow indicates the dipole moment of the tube.

5.4 Self-assembly mediated by halogen bonds

295

Another versatile approach to assemble tubes from macrocyclic building blocks relies on bis(ureas) [69]. These compounds, an example of which is 5.43 (Figure 5.38a), are rather rigid and almost planar, with the urea groups oriented perpendicular to the ring plane. The crystal structure of 5.43 (Figure 5.38b) illustrates that the urea groups mediate the self-assembly by similar bifurcated hydrogen bonds as found in the dimeric capsule of the calix[4]arene tetraurea 5.8 (Figure 5.31). Such macrocyclic bis(ureas) thus allow the fabrication of robust porous materials that may be used for gas storage, separations, or catalysis.

(a)

(b)

HN

NH O

O

HN

NH

5.43

Figure 5.38: Molecular structure of bis(urea) 5.43 (a), and crystal structure of the tubular assembly of this compound (b). Protons other than those in NH groups are omitted in the crystal structure for reasons of clarity.

5.4 Self-assembly mediated by halogen bonds 5.4.1 Introduction Noncovalent interactions must arrange individual building blocks in a predictable fashion to ensure that the overall self-assembly process proceeds with sufficient fidelity. Properly arranged patterns of hydrogen bonds are suitable for this purpose as discussed in the previous chapter, but halogen bonds are also an attractive option. The reason is the high directionality of halogen bonds, caused by the restriction of the σ-hole in the donor to a relatively small region at the terminus of the X–Hal bond (Section 3.1.6). The strongest electrostatic attraction therefore occurs

296

5 Assembling molecules

when the X–Hal and the halogen bond are arranged in a collinear fashion, while deviations from this optimal 180° angle cause a substantial weakening of binding. In spite of their promising properties, halogen bonds have so far not often been used to control self-assembly, however, and we therefore only briefly look at two examples.

5.4.2 Helices Orion B. Berryman investigated the interaction of the arylethynyl oligomer 5.44 (Figure 5.39a), containing three 4-iodo-1-methylpyridinium moieties, with iodide ions [70]. One potential mode of interactions involves wrapping a single chain of 5.44 around an anion to afford a foldamer complex similar to those we discussed in Section 4.1.13. This complex does not form, however, because it is not possible to arrange the iodine atoms along the chain of 5.44 in such a fashion that they simultaneously form three linear halogen bonds to a central iodide anion. Instead, selfassembly leads to the formation of a complex in which three subunits of 5.44 wrap in a helical fashion around two iodide anions.

(a)

(b)

I +N

I

N+

OMe

3 PF6 I

N+

5.44

OMe

Figure 5.39: Molecular structure of oligomer 5.44 (a), and crystal structure of the triple helicate in which three subunits of 5.44 wrap around two iodide ions (b). Protons are omitted in the crystal structure for reasons of clarity, and the backbones of the three ligands are shown in different shades of gray. The iodine atoms along the chain and the bound iodide anions are shown as small and large spheres, respectively.

To illustrate this arrangement, the corresponding crystal structure is shown in Figure 5.39b. The pronounced directionality of the halogen bonds thus controls the formation of a product that is structurally more complex than the entropically favored 1:1 assembly. This product also exists in DMF/acetonitrile and even in

5.5 Self-assembly mediated by coordination bonds

297

DMF/water mixtures. Although the complex shown in Figure 5.39b is structurally somewhat related to the foldamer complexes discussed in Section 4.1.13, 5.44 is not a foldamer because it only adopts a helical conformation in the presence of iodide anions. Complexes in which directed interactions between donor atoms along an oligomeric chain cause one or more of these ligands to wrap around suitable acceptors are called helicates. Depending on the number of ligands involved in helix formation, double and triple helicates are distinguished, with the complex in Figure 5.39b representing a triple helicate. While helicates stabilized by anions are rare, those between ligands and metal ions are more common. We will see a number of examples in Section 5.5.2.

5.4.3 Capsules François Diederich showed that the self-assembly of the two deep cavitands 5.45a and 5.45b (Figure 5.40a) affords a dimeric capsule held together by four linear halogen bonds between the iodine atoms of 5.45a and the pyridine nitrogen atoms of 5.45b as shown in Figure 5.40b [71]. Capsule formation requires the solvent to contain a small amount of an alcohol, which delivers the OH groups required for stabilizing the vase conformations of 5.45a and 5.45b. We discussed the stabilizing effect of water molecules on such cavitands in Section 4.1.9. The dimerization constant log Ka amounts to 3.7 in benzene-d6/DMSO-d6/methanol-d4, 70:30:1 (v/v) at 283 K, showing that the four halogen bonds cause a substantial stabilization. In solvents that are too large to enter the cavity (mesitylene-d12 with 2 vol% of 3,5-dimethylbenzylalcohol), the capsule between 5.45a and 5.45b hosts guest molecules such as two molecules of 1,4dioxane, one per capsule half.

5.5 Self-assembly mediated by coordination bonds 5.5.1 Introduction We now move away from noncovalent interactions and come to self-assembly processes mediated by the coordination of Lewis-basic ligands to metal ions. Coordinative interactions have a pronounced covalent character, rendering them stronger than the noncovalent interactions discussed above (60–200 kJ mol−1 vs. 1–40 kJ mol−1 of noncovalent interactions) (Section 3.2.1). As a consequence, self-assembly is not restricted to organic solvents as in the case of hydrogen bonds but occurs also in competitive media, including in water. At the same time, many metal–ligand bonds are labile, which is an important prerequisite for ensuring that product

298

5 Assembling molecules

(a)

H N

O O

R

F

O O

O

R R

O O

N H

5.45a (R = C11H23)

F F I

F

O

N

H N

N

I F

F I

F

O

O

F

N

O R

O

N

R R

N

F F

R

O

HN

F N

O N

N

F N

O

F F N F NH

H N

N

I

N

H N

F

N H

O

5.45b (R = C11H23)

(b)

Figure 5.40: Molecular structures of cavitands 5.45a and 5.45b (a), and calculated structure of the capsule formed from both compounds (b). Each cavitand of the capsule contains a 1,4-dioxane molecule, which is shown as a space-filling model. The cavitands are moreover stabilized in their vase conformations by methanol molecules bridging the benzimidazole units. The alkyl residues in the resorcinarenes are truncated for reasons of clarity.

formation proceeds under thermodynamic control. Frequently used metals are Cu+, Zn2+, Fe2+, Pd2+, Ga3+, and various others, not only because of the kinetic lability of their complexes and the possibility to realize different coordination geometries [tetrahedral with copper(I), octahedral with iron(II) and gallium(III), square planar with palladium(II)], but also because these metals are diamagnetic, facilitating product characterization by means of NMR spectroscopy. A major advantage of metal–ligand bonds is their pronounced directionality, resulting from their covalent nature, that renders product formation highly predictable. Parameters that need to be considered in this context are the number and arrangement of the Lewis-basic sites in the ligands and the preferred coordination number and geometry of the connecting metal centers. The structure of the product is then predicted by using the following concepts [72].

5.5 Self-assembly mediated by coordination bonds

299

Directional bonding approach: In this approach, the building blocks are divided into donors (the ligands) and acceptors (the metals). The donors are rigid organic compounds with two or more monodentate coordinating sites arranged at an angle between 0° and 180°. The acceptors are metal ions, metal complexes, or organometal compounds that serve to orient the incoming ligands in a predefined arrangement. To this end, they contain weakly bound groups that are replaced during selfassembly. Strongly bound ligands in the acceptors can serve to block specific sites and control the orientation of the incoming donors. Predictions about the most likely structure of the product produced from these building blocks are based on geometrical considerations. Bifunctional donors and acceptors give rise to macrocyclic compounds, for example, the size and symmetry of which depends on the angles at which the newly formed bonds are arranged as shown in Figure 5.41a. If three-dimensional assemblies are targeted, at least one of the components must be able to form three bonds. Figure 5.41b illustrates that a tetrahedron results, for example, from combining four tritopic precursors that arrange the newly formed bonds at 60° angles with six linear ones. A cube, on the other hand, requires a 2:3 ratio of a trifunctional precursor with 90° bond angles and a linear component. These predictions are only reliable if the precursors are sufficiently rigid. In case deviations from the optimal angles occur, the fidelity of the self-assembly process suffers.

(a)

Donor Acceptor

0°

60°

90°

108°

120°

180°

(b)

0° 60°

90°

108°

120°

180°

Figure 5.41: Correlation of the angles at which difunctional donors (ligands) and acceptors (metals) orient their binding sites with the structure of the preferred macrocyclic product (a), and assembly of a tetrahedron or a cube from suitable building blocks (b).

300

5 Assembling molecules

Paneling approach: The paneling approach is particularly useful for the construction of polyhedra, but it also allows accessing tubes, barrels, or bowls [73]. The term panel refers in this context to flat multidentate organic ligands that are joined together to form the faces of the final self-assembled product. In terms of coordination chemistry, this approach is closely related to the directional bonding approach in that the individual coordinating sites in the panels are usually monodentate, and the metals may contain additional ligands to control the number and orientation of the donors with which they interact. One parameter determining the structure of the product is the shape of the panels. A tetrahedron is produced from four triangles, for example, or a cube from six squares. However, many panels lead to more than one type of assembly. Triangular panels can give rise to tetrahedra, octahedra, icosahedra, or square pyramids, for instance, rendering the shape of the panels alone insufficient to control product formation. Equally important are the angles between the individual components, the number of linkages, and the arrangement of the donor atoms in the ligands. These aspects should be illustrated by using the tripodal ligands 5.46 and 5.47 and the palladium(II) complex [Pd(en)(NO3)2] 5.48 as examples (Figure 5.42). Since 5.48 accepts no more than two incoming ligands that replace the weakly bound nitrates, coordinatively saturated complexes derived from 5.48 and tripodal ligands must have the compositions M3L2, M6L4, M9L6, etc. Moreover, since the ethylene diamine ligand in 5.48 directs the incoming ligands into adjacent corners of a

(b)

(a) N

N NH2 Pd ONO2 H2N ONO2

N

=

N N

N 5.46

(c)

N

5.48 = N

N

=

N N

N 5.47

Figure 5.42: Molecular structures of the tripodal ligands 5.46 and 5.47 and the palladium(II) complex 5.48 (a), schematic illustration of the octahedral complex containing four panels of 5.46 (b), and of the square pyramidal complex containing four panels of 5.47 (c).

5.5 Self-assembly mediated by coordination bonds

301

square planar complex, the two panels at each metal center are arranged at a 90° angle. In the case of panel 5.46, the smallest complex where this is possible is the M6L4 complex, and since the donors in this panel are directed toward the corners, an octahedral complex results with four of the eight faces occupied by the ligands (Figure 5.42b). This complex forms with high fidelity. If two donor atoms point toward the edges as in panel 5.47, four triangles can only be joined in the form of a square pyramid that lacks the floor (Figure 5.42c). The result of a self-assembly process thus sensitively depends on the arrangement of the donor atoms in the panels. Many other coordination structures are accessible from triangular or rectangular panels, the latter often comprising porphyrin derivatives. Symmetry interaction approach: This approach allows making predictions about the structure of coordination compounds formed between chelating ligands and metal centers [74]. The analysis is based on symmetry considerations and involves three parameters, the coordinate vector, the chelate plane, and the approach angle. The coordinate vector connects the ligands with the metals. Bidentate chelating groups are, for example, bisected by the coordinate vector in the direction of the metal as shown in Figure 5.43a. The chelate plane contains all the coordinate vectors surrounding a metal and the approach angle is a measure of the arrangement of the ligands with respect to the symmetry axis of the metal, which is oriented orthogonally to the chelate plane (Figure 5.43b). The structure of the final assembly is determined by how the chelate planes of the metal centers are oriented relative to each other.

(a)

(b)

(c)

C3

(d)

C3

C3 70.6° Coordinate N vector N

N N

Figure 5.43: Construction of coordinate vector (a) and approach angle (b) for a bidentate ligand, and schematic illustrations of the structures of a triple M2L3 helicate (c), and a tetrahedral M4L6 cage (d), showing the arrangement of the coordinate vectors and the chelate planes at selected metal centers.

302

5 Assembling molecules

Two examples should illustrate this concept. The first involves the assembly of a triple helicate in which three ditopic ligands L are connected to two metal ions M, affording a complex with the composition M2L3 (Section 5.5.2). Such a complex is characterized by a C3 axis coinciding with the axis connecting the metals (Figure 5.43c). The threefold symmetry along this axis requires the coordinate vectors at each metal to be oriented at 120° angles. Pairs of coordinate vectors are connected by the ligands, and since each ligand must direct both of its chelating sites toward the metals, the chelate planes in such a helicate have to be oriented in a parallel fashion. In a tetrahedral M4L6 coordination cage containing the metals at the vertices and the ligands at the edges, the coordinate vectors are arranged in a similar fashion as in the triple helicate because each corner of the tetrahedron also has a threefold symmetry (Figure 5.43d). The corresponding chelate planes are, however, arranged at 70.6° angles. Ligands suitable for assembling triple helicates must therefore orient their chelating units in a parallel fashion, whereas those suitable for constructing tetrahedral coordination compounds must arrange the coordinate vectors at an angle of ca. 70° to predispose them for cage formation.

5.5.2 Helices Helicates with metal ions are more frequent than those containing anions such as the triple helicate discussed in Section 5.4.2. Such metal-containing helicates are formed from oligomeric ligands with suitable donor atoms along the chain. An example is the dicopper(I) complex of ligand 5.49a (Figure 5.44), which was first described by Jean-Marie Lehn [75]. The two 2,2′-bipyridine units in 5.49a cannot simultaneously coordinate to a metal ion. Copper coordination thus involves two ligand molecules to come together and arrange their 2,2′-bipyridine units in such a way that distorted tetrahedral coordination geometries are established at the metal centers. The ligands accordingly adopt helical conformations, leading to a product in which two helices are intertwined to afford a double helix. Both strands have the

N

N

N

N

N

N

O n

O 5.49a 5.49b 5.49c 5.49d

(n = 0) (n = 1) (n = 2) (n = 3)

Figure 5.44: Molecular structures of ligand 5.49a and calculated structure of the double helicate formed from two subunits of 5.49a and two copper(I) centers. Protons are omitted for reasons of clarity.

5.5 Self-assembly mediated by coordination bonds

303

same configuration; those in Figure 5.44 are, for example, P (plus) configured, meaning they both have a clockwise screw sense, rendering the corresponding helicate chiral. The equally probable enantiomer has the ligands arranged in a counterclockwise M (minus) fashion. Several parameters permit controlling the structure of such helicates [76]. One is the number of coordinating subunits in the ligands that determines how many metal centers the final helicate contains. Ligands 5.49b-d (Figure 5.44), deriving from 5.49a by chain elongation, hence afford double helicates with up to five metal centers. According to mechanistic studies, the formation of such polynuclear complexes typically involves several intermediates. In the case of the trinuclear double helicate derived from 5.49b, the most important ones are a complex ML in which one ligand binds to a single metal, presumably through its end groups, in addition to mono- and dinuclear complexes ML2 and M2L2 [77]. These incompletely formed complexes play only a minor role in the thermodynamic equilibrium, however, since they are orders of magnitude less stable than the trinuclear end product [78]. As a consequence, M3L2 forms with pronounced positive cooperativity, likely because of the strain-free arrangement of the ligands in the product. Positive cooperativity is, however, not guaranteed. If the complexation of the second or any subsequent metal ion produces strain in the ligand, for example, negative cooperativity results. Other factors that control the structure of helicates are the nature of the metal and the denticity of the chelating units. Double helicates result, for instance, by combining bidentate ligands with metals preferring a tetrahedral coordination geometry as in the helicate shown in Figure 5.44. In these structures, two ligands at each metal are arranged in an almost orthogonal fashion, which is optimal for inducing the chains to meander around the metals.

N N N Cu+ N N

N N Fe2+ N N N

N 2+N N Fe N N N

Figure 5.45: Schematic illustration of the binding modes that arrange two ligands in an orthogonal fashion, leading to double helicates, and the structure of an octahedral complex that affords a triple helicate.

A similar orthogonal arrangement is achieved by coordinating tridentate ligands such as terpyridines to metals that prefer an octahedral coordination geometry (Figure 5.45). When bidentate ligands interact with such metals, the resulting

304

5 Assembling molecules

complexes contain three ligands. These complexes thus give rise to triple helicates but only if the ligand permits the corresponding coordination geometry. The substitution pattern of the bipyridine subunits in 5.49a, for example, does not allow the formation of an octahedral complex, rendering this ligand unsuitable for assembling a triple helicate. Such a helicate can, however, be obtained from the ligands 5.50a or 5.50b in which the 2,2′-bipyridine units are linked via their 5,5′-positions (Figure 5.46) [79]. The structure of the corresponding dinickel(II) complex of 5.50a is shown in Figure 5.46. 5.50b forms an analogous trinuclear complex.

N N

N

N

n

N

N

5.50a (n = 0) 5.50b (n = 1)

Figure 5.46: Molecular structures of ligands 5.50a,b and calculated structure of the dinickel(II) triple helicate of 5.50a.

When iron(II) salts are used instead of nickel(II) salts, the trinuclear triple helicate of 5.50b forms initially in a kinetically controlled reaction. This product then rearranges upon heating to afford a thermodynamically favored circular helicate [80]. The exact structure of this product depends on the counterion of the iron(II) salt. In the presence of chloride ions, the pentanuclear circular helicate {Cl⊂[Fe55.50b5]}9+ is formed, whereas sulfate ions induce the formation of the hexanuclear helicate {SO4⊂[Fe65.50b6]}10+ (Figure 5.47). This influence of the anion on product formation is likely due to a template effect. In the case of the chloride complex, one anion is bound in the central cavity of the circular helicate according to a crystal structure. Assuming that sulfate binds in a similar fashion to the larger helicate, the size of the product is determined by the size of the anion occupying the cavity [81]. The structure of the ligand also controls stereochemical aspects of the assemblies. Ligands such as 5.49a or 5.50a that fold around the metals typically form racemic chiral helicates. Structural aspects in the ligand can prevent folding, however, leading to achiral meso-helicates with a symmetry plane bisecting the axis connecting the metal ions. It is also possible to produce one chiral helix form selectively by using ligands such as 5.51a,b (Figure 5.48) [82]. The two stereogenic centers in the central part of 5.51a,b are sufficient to afford one enantiomer of the corresponding helicate, with the two S-configurations in the ligand leading to the exclusive formation of the P-configured double helicate.

305

5.5 Self-assembly mediated by coordination bonds

N N

N

N

N Fe 2+ N N

N

N

N N N Fe 2+ N

N

Fe 2+ N

N

N

N N N

Fe 2+ N

N

N

N

N Fe 2+ N

N

N

N

N

N Fe 2+ N

N

N

N

N

N

N

N

Fe 2+

N

N

N

N

N Fe 2+ N

N

N N

N

N

N

N

N Fe 2+ N

N N

N

N

N Fe 2+ N

N

N

N Fe 2+ N N

N

Figure 5.47: Structures of the pentanuclear and hexanuclear circular triple helicates formed from 5.50b and FeCl2 and FeSO4, respectively.

N

N

O

N

N

O n

5.51a (n = 1) 5.51b (n = 3)

N

N

N

N

Cu + N N

O O

N

N

Cu + N N

O

N

O

Cu + N N

N

Figure 5.48: Molecular structure of ligands 5.51a,b and schematic illustration of the P-configured tricopper(I) double helicate derived from 5.51a.

These helicates generally form with high fidelity also if several different ligands or metals are simultaneously present in solution. The addition of copper(I) salts to a mixture of ligands 5.49a-d exclusively affords the four double helicates containing identical ligands once the thermodynamic equilibrium is reached, for example [83] (Figure 5.49a). This self-assembly is thus an example of a completive narcissistic self-sorting process (case A in Figure 5.8). Similarly, a mixture of 5.49b and 5.50b containing appropriate amounts of copper(I) and nickel(II) salts only affords the tricopper(I) double helicate of 5.49b and the trinickel(II) triple helicate of 5.50b through narcissistic self-sorting (Figure 5.49b), which is frequently observed in helicate chemistry.

306

5 Assembling molecules

(a) N N

N Cu +

N

N

O

O

N N O N N

N

N O

N N O N

N N

Cu +

N

O

N

O N

N

N

Cu +

O

O

N

N N

N

N

O

O

N

O N

N N

O N

N

O

O

N

N

(b)

N

O

O

Cu +

N

N

O

O

N

N

N Cu +

N

N

O

O

N Cu +

N

N

O

O

N

N Cu +

N N

N

N Cu + N

Cu + N

N

N

O

N

O

N

N

O

N

O

N

N

N

N

N

N

N

N

O N

N

N

O

N

Cu +

N

N

Cu +

N Cu +

N

N

N

N

Cu +

N

N

N

N

N

N

O

N Cu +

N

N

N

O

N Cu +

N

N O

N

N

N

O

O

N Cu +

N

N

N

N

N

Ni 2+

N

Cu +

N Cu +

N

N

O

O

N N

N Cu + N

N N Ni 2+ N N N N

N N N 2+ N Ni N N

N N Ni 2+ N N N N

Figure 5.49: Narcissistic self-sorting process involving ligands 5.49a-d and copper(I) ions that afford double helicates containing pairs of identical ligands (a). The self-sorting process in (b) shows that the ligands 5.49b and 5.50b afford a mixture of a double and a triple helicate when coordinating to copper(I) and nickel(II) ions.

Both reactions are consistent with the rules of self-assembly and self-sorting we derived earlier (Section 5.1). In the first reaction, two ligands of different lengths would lead to a helicate in which donor atoms of the longer ligand remain

5.5 Self-assembly mediated by coordination bonds

307

unused. The formation of these complexes violates the principle of maximum site occupancy and these complexes are therefore enthalpically disfavored. Using the dangling donors for further coordination is entropically unfavorable because of the large number of components that are involved in the formation of such a product. Only the completely self-sorted system leads to the maximum number of coordinative interactions (enthalpically favored) in combination with the maximum number of possible products (entropically favored). Similarly, the ligands in the second reaction are structurally programmed to either form a double helicate (5.49b) or a triple helicate (5.50b). Any mixed species can be assumed to be enthalpically disfavored as it should either be strained or should contain metal ions that are coordinatively unsaturated. Since these fundamental studies were performed by Lehn and coworkers, the chemistry of helicates has been extended to many other ligand types and metals [76]. One further example shall be presented because we discuss similar ligands again in Section 5.5.5. These complexes are formed from the biscatecholates 5.52a–c (Figure 5.50) that coordinate to gallium(III) ions in an octahedral fashion after deprotonation, leading to the corresponding triple helicates [84]. OH OH OH O

OH HN

O

OH HN

O

OH

OH HN

O

OH HN

O

NH NH O

OH

OH

OH

OH 5.52a

5.52b

5.52c

Figure 5.50: Molecular structure of the biscatecholates 5.52a–c and crystal structure of the triple helicate formed from 5.52b and gallium(III) ions.

In contrast to bipyridine-derived ligands, the biscatecholates are negatively charged and since these charges are not fully compensated by the gallium(III) ions, the resulting dinuclear helicates have six negative charges. If the three ligands 5.52a–c are used simultaneously in the reaction with the gallium(III) ions, narcissistic completive self-sorting is again observed because only ligands of identical lengths can be incorporated in a stable triple helicate (Figure 5.50).

308

5 Assembling molecules

5.5.3 Grids The coordination of bidentate chelate ligands to metals that prefer a tetrahedral coordination geometry leads to an orthogonal arrangement of the ligands at each

N

(a)

N N

n

5.53a (n = 1) 5.53b (n = 2) 5.53c (n = 4)

N

(b)

N N N Ag + N N Ag + N N N N N N Ag + N N Ag + N N N

(c) N N N N Ag + N N Ag + N N Ag + N N N N N N N N Ag + N N Ag + N N Ag + N N N N N N N N Ag + N N Ag + N N Ag + N N N N

N N N Ag + N N Ag + N N N N N N Ag + N N Ag + N N N N N N Ag + N N Ag + N N N

(d) N

N Ag + N N N Ag + N N Ag + N N N N N N N Ag + N N Ag + N N N N N N N Ag + N N Ag + N N N N N N Ag + N N Ag + N N N N N N N Ag + N N Ag + N N N

N N N Ag + N N N N N Ag + N N N N N Ag + N N N N N Ag + N N N N N Ag + N N

N N Ag + N N N N Ag + N N N N Ag + N N N N Ag + N N N N Ag + N N

NN N

N Ag +

N Ag +

N N N N N N N N Ag +

Ag +

NN NN

NN NN

Ag + Ag + N N N N N N N N Ag +

Ag + N NN

N

N

N

Ag + N

N

Figure 5.51: Molecular structures of ligands 5.53a–c and schematic illustration of the formation of the 2 × 2 and 3 × 3 grids from ligands 5.53a (a) and 5.53b (b), respectively. The 2 × 3 grid formed from a mixture of these ligands is shown in (c), and (d) depicts assemblies formed from 5.53c.

5.5 Self-assembly mediated by coordination bonds

309

metal center, affording double helicates if the ligands are sufficiently flexible as we have seen in the previous chapter. If, however, helicate formation is prevented by using rigid ligands, self-assembly leads to other molecular architectures. The rigid rod-like ligand 5.53a, for example, gives a tetranuclear 2 × 2 molecular grid (Figure 5.51a) [85]. Such grids are also obtained from ligands with tridentate coordination sites and metals that prefer an octahedral coordination geometry. The next larger ligand 5.53b accordingly gives rise to a 3 × 3 grid (Figure 5.51b), in this case with silver(I) ions [86]. A rectangular 2 × 3 grid preferentially results when 5.53a and 5.53b are simultaneously treated with a silver(I) salt (Figure 5.51c). Social self-sorting thus predominates over narcissistic self-sorting in this case, likely because of destabilizing effects of the central silver ion in the 3 × 3 grid [87]. The even longer ligand 5.53c accordingly does not afford the 5 × 5 grid [88]. Several reasons are responsible: the geometric fit of the ligand and the metals is not optimal in this system so that the ligand has to bend upon grid formation, which destabilizes the larger grid. In addition, 5.53c has to adopt a conformation in the grid with all nitrogen atoms arranged on the same side, which is normally disfavored because of repulsive interactions between the nitrogen lone pairs. If metal complexation is not sufficiently effective, the enthalpic cost for the ligand to adopt this conformation cannot be compensated. Instead of forming the 5 × 5 grid, 5.53c thus affords other complexes with silver(I) ions, one in which two rectangular 2 × 5 grids are arranged on opposite sides of five parallel ligands and a decametallic quadruple helicate (Figure 5.51d). Although the concept of grid formation is thus mainly limited to smaller grids, the general approach of using rigid rod-like ligands as building blocks for metaldirected self-assembly is very versatile. A variety of different molecular architectures are produced in a similar way, of which some are depicted in Figure 5.52 [89]. All of these complexes form with high fidelity because of the strict orthogonal arrangement of the ligands at each metal center, and because no other ligand arrangements are possible in which all donor atoms are used. Note that these complexes contain different ligands that predictably interact to form the final product. Another strategy to control the formation of such heteroleptic complexes is presented in Section 5.5.4.

5.5.4 Rings Figure 5.41 shows that the assembly of rings requires each metal to connect two ligands and vice versa. An instructive example of how to produce a metallacycle in this way is the molecular box described by Makoto Fujita [90]. This box is one of many coordination compounds developed by Fujita and others that make use of square planar palladium(II) complexes. We will see further examples in the next chapter.

310

5 Assembling molecules

Ph N N N Ru 2+ N N N N N N Ru 2+ N N N N N N Ru 2+ N N N

Rack

N N Cu + N N Ph

Ph N N Cu + N N Ph

Ph N N Cu + N N Ph

Ph N N Cu + N N Ph

Ph N N Cu + N N Ph

Ph N N Cu + N N Ph

Ladder

Ph Ph

N Ph Cu + N Ph N N N N + N N Cu N + Cu Ph N Ph N

Ph Ph

N Ph Cu + N Ph N N N N + N N Cu N + Cu Ph N Ph N

Ph Ph

N Ph Cu + N Ph N N N N + N N Cu N Cu + Ph N Ph N Multicellular cylinder

Figure 5.52: Examples of molecular architectures that are produced from rigid ligands by metal-directed self-assembly.

Palladium(II) ions accepts four ligands in a square planar coordination geometry. If two adjacent coordination sites are blocked with ethylenediamine as in 5.48, the replacement of the weakly bound counterions with other ligands arranges the newly formed coordinate bonds at a 90° angle. The cis-protected 5.48 is thus ideally predisposed to yield rectangular metallamacrocycles when treated with linear ligands. Indeed, the coordination 4,4′-bipyridine to 5.48 affords the corresponding almost perfectly square box 5.54 (Figure 5.53a) in quantitative yields. The length of the edges of 5.54, estimated from the distance of the two palladium ions, amounts to ca. 11 Å (Figure 5.53b). Accordingly, the cavity of this box roughly compares in size to that of γ-cyclodextrin (Table 4.4). 5.54 therefore hosts aromatic substrates in the aqueous solution in which it is assembled, with the stability constant log Ka of the 1,3,5-trimethoxybenzene complex amounting to 2.9, for example. The cavity diameter can be easily expanded by using rigid linear ligands with a larger distance between the two coordinating nitrogen atoms such as 1,4-di(pyridin4-yl)benzene 5.55 or 1,2-di(pyridin-4-yl)ethyne 5.56 (Figure 5.53c). More flexible ligands lead to the formation of a mixture of trinuclear and tetranuclear complexes as we have seen in Section 5.1 when discussing template effects (Figure 5.7). Depending on the bite angle in the ligand, other macrocycles such as the dinuclear

5.5 Self-assembly mediated by coordination bonds

(a)

311

(b) N NH 2 4 H 2N Pd ONO 2 + ONO 2

4

5.48

N 8 NO 3 H 2N N Pd NH 2 N

NH 2 H 2N Pd N N

8+

(c) N

N

N

N H 2N Pd N NH 2

5.55

N Pd NH 2 H 2N

N

5.54

(d)

N 5.56

F

F

F

F

H 2N N Pd NH 2

4+

6+

N H2 N N Pd N N H2

H2 N

N Pd N F

F

F

F 5.57

N H2

H2 N N Pd N N H2 N N Pd NH 2 H 2N 5.58

Figure 5.53: Formation of the molecular box 5.54 from 5.48 and 4,4′-bipyridine (a), and calculated structure of 5.54 (b). The structures in (b) show other linear ligands from which analogous boxes can be assembled, and those in (d) the structures of dinuclear and trinuclear metallamacrocycles that are prepared in a similar fashion.

and trinuclear metallacycles 5.57 and 5.58 (Figure 5.53d) are also accessible in this way [91]. Platinum(II) complexes are formed in a similar manner. Their formation usually requires high temperatures and longer times until equilibrium is reached, however, because the platinum–nitrogen bond is kinetically much less labile than the palladium–nitrogen bond. The advantage of platinum complexes is that they do not exchange the ligands at room temperature, which allows their isolation as stable and under ambient conditions inert compounds.

312

5 Assembling molecules

The approach to assemble macrocycles by using metal–ligand interactions is certainly attractive from the synthetic point of view because of the efficiency with which the product is obtained without having to initially synthesize a linear precursor in a stepwise fashion that is ultimately cyclized. One could nevertheless argue that this strategy is restricted to the synthesis of symmetrical macrocycles containing only identical subunits because of the lack of control over the sequence of the subunits along the ring. A mixture of ligands accordingly leads to the formation of a statistical mixture of products. This limitation can, however, be addressed by choosing metals and ligands whose interactions are limited to a certain subset of combinations as demonstrated by Michael Schmittel [92]. The corresponding approach builds on steric and electronic effects of the residues in 2,9-disubstituted phenanthroline derivatives on the structure of the corresponding tetrahedral metal complexes. Ligand 5.59 (Figure 5.54a), for example, is unable to form homoleptic complexes with copper(I) or zinc(II) ions because steric hindrance of the ortho-methyl groups in the mesityl substituents prevent two orthogonally arranged identical ligands from coming sufficiently close. 5.59 therefore requires another ligand for complex formation such as an unsubstituted phenanthroline. The stability of the corresponding heteroleptic complexes benefits

(a) Zn2+ N

N

N Zn2+

+ N

N

analogously with copper(II)

5.59

N

N

N

Zn2+

+ N

N

N N

(b)

N

N

N

+

Cu + N

N

Cu+ N

+

N

+

N

O

N N

N

+

+

O

N

Zn2+

+

Zn2+

5.59

O

N

N

N

O O

O O

N +

Cu

N N

N

N

+

O

N

2+

+ N

O

N

N

N

Zn N

O

N

5.60 O

O

Figure 5.54: Self-sorting processes involving a 1:1:1 mixture of 5.59, phenanthroline, and copper(I) or zinc(II) ions (a), or a 1:1:1:1:1:1 mixture of ligands 5.59, 5.60, phenanthroline, terpyridine, and both copper(I) and zinc(II) (b). The complexes in the boxes are preferentially formed.

313

5.5 Self-assembly mediated by coordination bonds

(a)

(b) N

N

N

+

Br

N

N

N

N

Br

Br

OC4H9 N

N

N

Br +

N

O

N

2

O

N N

H9C4O O

N

N

O

N

+

+ O C10H21

N N N

N

O

N N

O

N O

O H21C10 O

2 Cu+

1 Zn2+

2 Zn2+

N

O O

N

O O

2 Cu+

O

N NN + N Cu N

NO 2+

Zn N N N O

N N Cu+ N N

O Br

Br

H21C10O OC10H21 Br

N +

N Cu N N

H 9C 4O

O OC4H9

N N O N Zn2+ O N N

Br

O O

N N

O

Zn2+ N N N O

N + N Cu N N

O

Figure 5.55: Selective formation of a triangle (a) and a trapezoid (b) from appropriate ditopic ligands whose coordination to copper(I) and zinc(II) ions is restricted to the combinations in the products shown, while other metal–ligand combinations are much less stable and therefore do not contribute to product formation.

from the aromatic interactions between the two mesityl groups in 5.59 and the ligand sandwiched between them, rendering these complexes more stable than homoleptic complexes containing only unsubstituted phenanthrolines. An equimolar mixture of two different ligands and a metal ion therefore exclusively affords heteroleptic complexes.

314

5 Assembling molecules

The level of complexity can be further increased by using an equimolar mixture of the ligands 5.59, 5.60, phenanthroline, and terpyridine as well as both copper(I) and zinc(II) (Figure 5.54b). Again, only heteroleptic complexes are formed from these components. In addition, terpyridine preferentially coordinates to zinc(II) because copper(I) is unable to expand the coordination number beyond four. The corresponding complex therefore does not benefit from the additional donor atom, while that with zinc(II) does. Finally, of the two possible heteroleptic terpyridine-zinc(II) complexes, the one with 5.60 is more stable than the one with 5.59 because the oxygen atoms in 5.60 contribute to metal binding. Accordingly, this mixture undergoes social self-sorting, affording only two complexes as shown in Figure 5.54b. Based on these self-sorting principles, ditopic ligands can be designed that assemble in a predictable fashion, in turn allowing the control over the structure of the respective product. The three ligands shown in Figure 5.55a assemble in the presence of copper(I) and zinc(II) ions to form a triangle, for example, in which the position of the ligands and the metals are controlled by the rules derived above [93]. Another set of ditopic ligands affords the trapezoid in Figure 5.55b [94], and when including a further ligand in the system that exclusively forms a homoleptic complex, also a five-membered ring can be obtained [95].

5.5.5 Cages Coordination cages are hollow three-dimensional objects with metal centers distributed along the surface connecting the ligands. The construction of such cages requires one type of building block, either the metal or the ligand, to interact with three or more binding partners as illustrated in Figure 5.41b. These compounds are thus classified on the basis of the geometrical parameters of their components [96]. Figure 5.56 illustrates the principles. If two hemispherical components derived from macrocyclic compounds are connected to form a roughly spherical cage, the number of donor sites depends on the symmetry of the macrocycle. Cyclotriveratrylenes thus serve as trifunctional and calix[4]arenes or resorcinarenes as tetrafunctional ligands. The exact nature of the donor atoms in these building blocks determines which metal to use for the assembly. In the case of monodentate donors, metal centers with two vacant sites such as the palladium(II) complex 5.48 can serve as connecting units. Bidentate groups connect metals that form tetrahedral complexes and tridentate groups those that prefer an octahedral coordination geometry (Figure 5.45). Tetrahedral coordination cages usually contain the metals at the vertices. These metals either connect difunctional ligands arranged along the edges of the tetrahedron or trifunctional ones capping the faces. In both cases, self-assembly relies on the formation of octahedral complexes. Difunctional ligands connected through octahedral complexes alternatively also afford cubes or even larger cages but the tetrahedron is

5.5 Self-assembly mediated by coordination bonds

No. of ligands at each metal center

2

Coordination geometry Various at each metal center

3

3

3

2

Octahedral

Octahedral

Octahedral

Angled

315

Figure 5.56: General concepts for the construction of coordination cages of different shapes with the blue objects representing the ligands and the spheres the metals.

usually favored by entropy because it contains the smallest number of components. Cubes are obtained from tetrafunctional ligands arranged on the six faces and connected at the corners through octahedral metal complexes. In octahedra, four edges meet at the vertices, which is difficult to realize by metal coordination. By only connecting two triangular ligands via suitable metals, octahedra are, however, produced with only half of the faces occupied. The following examples of coordination cages should illustrate these concepts. The linking of macrocyclic ligands through metal centers is somewhat related to the formation of hydrogen-bonded capsules from cavitands or calixarenes (Section 5.3.3). The cavitand 5.61 containing four nitrile groups along the rim serves as a tetrafunctionalized ligand, for example. Bridging the nitrile groups via metal centers by treating 5.61 with the cis-protected palladium(II) complex 5.62 affords a tetranuclear cage with a cavity large enough to include anions, for instance one of the triflate anions released during cage formation (Figure 5.57) [97].

R

R

C 2

O

O

N

N

O

N O C

C

O

O

O R

N O CO

+ 4

O3SCF3 Pd O3SCF3

R

R

Ph Ph

Ph Ph

R

5.61 (R = C6H13)

– 8 F3CSO3–

N PhPh C Ph Ph O

R

O

O

CO O N Ph Ph C Ph Ph N Pd Pd N C Ph Ph N Ph Ph O CO O

O

N O C

O O

R

8+

O

C

N O Ph Ph C PhPh N Pd Pd

O

5.62

R

R

R R

Figure 5.57: Formation of a coordination cage from the tetrafunctionalized cavitand 5.61 and the palladium(II) complex 5.62.

316

5 Assembling molecules

Such cis-protected palladium(II) complex also serve to connect planar ligands that end up as panels in a coordination polyhedron. The ligands 5.46 and 5.47, for example, assemble into an octahedral and a square pyramidal complex, respectively, when combined with palladium complex(II) 5.48 as discussed in Section 5.5.1 (Figure 5.58a,b) [98]. Structurally slightly different ligands afford other structures. In the case of 5.63, two products are possible, both having the composition M8L4 but differing in the relative orientation of the four panels (Figure 5.58c). If the panels are arranged in a parallel fashion, the product has an open structure in the shape of a square pyramid with the tip cut off, while the antiparallel orientation (a) [(5.48) 6 5.46) ]

(b)

4

N N N N

N

N

N

N

N

5.46 N

[(5.48)6 5.47)4]

5.47

N

(c)

N

N

(d) N

N

N

N N

N N

N

N 5.64

5.63

(e) [(5.48)8 5.63)4] N

[(5.48)18 5.64)6]

N [(5.48)8 5.65)4]

N

(f)

N

N

N

5.65

N N

N

N 5.46 {[Pt(en)(NO3)2]6 5.46)2 ·(pyrazine)3}

Figure 5.58: Molecular structure of ligands 5.46, 5.47, and 5.63–5.65, and crystal structures of the multinuclear complexes resulting from these ligands upon coordination to 5.48 or the corresponding platinum(II) analog. The crystal structure of the pyramid-type complex of ligand 5.63 has not been reported. Hydrogen atoms are omitted for reasons of clarity.

317

5.5 Self-assembly mediated by coordination bonds

leads to a cage with a small cavity. Product formation is controlled by the addition of suitable templates that are bound within the cavities of these complexes. Large substrates such as dibenzoyl induce the formation of the larger pyramidal complex, while smaller ones such as CBr4 shift the equilibrium toward the closed cage. The ligand 5.64 with three pyrimidine groups leads to the formation of an M18L6 trigonal-bipyramidal structure upon treatment with 5.48 that has a volume of 900 Å3 and a closed surface so that included molecules cannot escape (Figure 5.58d). The tetrafunctional ligand 5.65 gives rise to a rectangular tubular assembly with two panels missing to close the cavity (Figure 5.58e). Two different types of ligands can also be combined in this approach, although some tricks are necessary. One is the use of platinum(II) instead of palladium(II) to obtain inert complexes. The second is to use appropriate templates to drive product formation. Accordingly, a mixture of ligand 5.46, pyrazine, and the platinum(II) analog of 5.48 in the presence of hexamethoxytriphenylene affords a complex in which the two tripodal panels are arranged in a parallel fashion and connected by three pyrazine pillars (Figure 5.58f). Due to the inert nature of the platinum-ligand bonds, the template can be removed without decomposition of the thus formed cage. Many of these cages have characteristic properties that relate to their respective structure. In general, these cages interact with anionic guests because of their multiple positive charges. In addition, hydrophobic compounds are also bound in the aqueous environment in which the cages are formed, with the driving force of complex formation coming from the hydrophobic effect. The platinum complex shown in Figure 5.58f binds 1,3-diketones and selectively stabilizes the enol form of these guests [99], for example. The platinum(II) analog of the pyramidal complex shown in Figure 5.58b mediates the folding of a designed nonapeptide with the

–

O

–

O

O O O

O NH

Hydrophobic Hydrophilic

NHNH O NH

O N H

O

O

NH NH

O

O NH HN

O

O NH

Hydrophobic Hydrophilic

OH

O

OH

NH O N H

H N O

NH

O N H

H N O

O N H

H N O

O N H

H N O

O N H

NH 2 O

Figure 5.59: Schematic illustration of the binding mode of the α-helical nonapeptide Ac-Trp-Ala-Glu -Ala-Ala-Ala-Glu-Ala-Trp-NH2 inside the cavity of the M6L4 complex formed from ligand 5.47 and the platinum(II) analog of 5.48.

318

5 Assembling molecules

amino acid sequence Ac-Trp-Ala-Glu-Ala-Ala-Ala-Glu-Ala-Trp-NH2 in aqueous solution [100]. This peptide has a random coil conformation in the absence of the cage. In its presence, the hydrophobic effect mediates the incorporation of the hydrophobic tryptophan side chains into the cavity. To allow both of these residues to bind simultaneously, a conformational reorganization of the peptide occurs, leading to an α-helix in which the tryptophan side chains are located on one side and the glutamic side chains on the opposite side (Figure 5.59). These polar groups are thus exposed to the solvent and help solvating the corresponding complex. In Section 8.2, we will see that these palladium cages are also used as reaction vessels to mediate Diels–Alder reactions or photochemical cycloadditions between two included substrate molecules [101]. In all of the above examples, the metal complex 5.48 serves as a difunctional connecting unit to orient two ligands at a 90° angle, while the ligands provide three or more donor atoms to extend the assembly into the third dimension. These roles are reversed by using ligands with only two donors and arranging them around palladium(II) ions as tetrafunctional connecting units. Due to the planar arrangement of the ligands at the metal centers, this strategy only affords discrete three-dimensional objects with a convex outer surface if curved ligands are used, with the structure of such cages depending sensitively on the shape of the ligand, that is, on the angle at which they orient their donor atoms. In the case of ligands with the two coordinating end groups arranged in a parallel fashion (at an angle of 0°), dimetallic complexes result. An example is the cage resulting from the coordination of four banana-shaped bis(pyridyl) ligands 5.66 to two palladium(II) ions (Figure 5.60a) [102]. This compound is rigid and features a cavity large enough to host aromatic disulfonates, which are mainly bound by electrostatic interactions. The assembly of the dinuclear palladium(II) complex of 5.67 (Figure 5.60b) follows similar principles [103]. In this case, the cavity is lined and almost completely closed by the anthracene panels, which thus strongly influence guest exchange. This cage hosts hydrophobic organic molecules whose binding in water is largely driven by the hydrophobic effect. If rigid ligands are used for self-assembly in which the two donor atoms are oriented at an angle > 0°, the planes of individual palladium complexes are tilted against each other and the formation of dimetallic complexes is no longer possible. The outcome of the self-assembly then depends on the exact structure of the ligand [104]. Ligands such as 5.68, in which the angle between the two coordinating pyridyl units amounts to 90°, form an M6L12 cube. Extending the angle to 127° as in 5.69 leads to an M12L24 cuboctahedron, and further to 149° as in 5.70 to an M24L48 rhombicuboctahedron (Figure 5.61a).

319

5.5 Self-assembly mediated by coordination bonds

N

O

N

N O

O O

F 3C O

CF3 O O

O N

N

O

N

5.66

5.67

Figure 5.60: Molecular structures of ligands 5.66 and 5.67 and crystal structures of their dinuclear palladium(II) complexes. Protons in the crystal structures are omitted for reasons of clarity. One ligand in each structure is shown in red.

(a)

O O

N

N O

127°

O

5.69 N

S

N

N

90° 5.68

5.70

(b)

N

O

N

N

127° 127°

M 12 L 24

N

149°

N H

135° 131° 134°

N

N

N

N

N

147°

S

N

149° M24 L 48

Figure 5.61: Molecular structures of ligands 5.68–5.70 and crystal structures of the polyhedra resulting after coordination to palladium(II) ions (a). The correlation of the bite angle in such difunctional ligands with the structure of the product is shown in (b). Protons in the crystal structures are omitted for reasons of clarity. One ligand in each structure is shown in red.

149°

320

5 Assembling molecules

The selectivity with which these cages form is very sensitive to the bite angle of the ligands. If a 8:2 mixture of 5.69/5.70 is used for the self-assembly, corresponding to an average bite angle of 131°, only the cuboctahedron is formed, while a 7:3 mixture with an average angle of 134° exclusively leads to the rhombicuboctahedron. This correlation thus allows reliably predicting which of the ligands shown in Figure 5.61b forms the cuboctahedron and which the rhombicuboctahedron when interacting with palladium(II) ions. The structure of these cages and, in turn, their properties can easily be varied by changing the structures of the building blocks, with substituents in the convex region of the ligands being arranged on the outside of the corresponding cages and those in the concave region inside the cavities (Figure 5.62a). Since the nature of these substituents can moreover be varied in a wide range (alkanes, fluorinated alkanes, oligoethylene glycol residues, oligosaccharides, peptides, etc.), it is possible to adapt the properties of the cages to different applications. An M12L24 cage has even been prepared that is large enough to host ubiquitin, an 8.6 kDa globular protein with a diameter of 3–4 nm (Figure 5.62b) [105]. This cage can be assembled

(a)

N N

N

N

(b)

+ NMe3 5.71a

23 + 5.71b

N O

O

D2O/CD3CN (1:1), 45 °C, 3 h

O

N

N

N

12 equiv Pd(NO3)2

N S

Ubiquitin

Figure 5.62: General structure of difunctional ligands that give rise to M12L24 cages with additional substituents on the outside and the inside (a), and molecular structures of ligands 5.71a,b that afford a cage containing an included ubiquitin molecule (b).

5.5 Self-assembly mediated by coordination bonds

321

from 23 subunits of ligand 5.71a and one subunit of 5.71b containing the covalently bound ubiquitin molecule. The final class of cages that should be discussed features octahedral metal complexes at the vertices. An example is the coordination cage developed by Kenneth N. Raymond [106], which is formed from the deprotonated form of bis(catechol) 5.72 (Figure 5.63a) and gallium(III) ions. Cage formation thus involves a similar coordination mode as the formation of the helicates from ligands 5.52a–c (Figure 5.50), only that 5.72 is predisposed to afford a cage rather than a helicate according to the symmetry interaction approach. The six gallium(III) centers are located in this cage at the vertices of a tetrahedron with the ligands arranged along the edges. Each metal center has an octahedral coordination geometry, with all four centers having either the Δ- or the Λ-configuration (Figure 5.63b). Cage formation thus proceeds in a diastereoselective fashion, affording the racemate of the ΔΔΔΔ and the ΛΛΛΛ cage in the absence of chiral induction. The (b)

(a)

OH

12–

O

3+

– Ga HN O

NH

HN

O–

O

OH

Ga3+

O

O 3+

Ga –O

HO 5.72

NH O

Ga3+

HO

(c) B + H+

B H+

pKa (H2O) A

C

H+

log Keff ( B H +)

B

log Kass (B) ~ 0

pKeff B H+

B D

Figure 5.63: Molecular structure of ligand 5.72 and the corresponding M4L6 coordination cage in which six deprotonated ligands coordinate to gallium(III) ions (a). The crystal structure of this cage is shown in (b). In (c), the thermodynamic cycle is depicted that allows estimating the shift of the effective basicity of amines caused by their incorporation into the cage. Protons in the crystal structure are omitted for reasons of clarity. One ligand is shown in red.

322

5 Assembling molecules

volume of the cavity ranges between 350 and 500 Å3, which is large enough for the incorporation of small guest molecules. Since the cage is overall 12-fold negatively charged, it binds preferentially to cationic substrates such as protonated amines or quaternary ammonium ions. The stability log Ka of the tetramethylammonium ion complex amounts to 4.4 in water, for example, with the driving force of complex formation originating mainly from the release of water molecules and the associated gain in entropy. The binding of cations is so efficient that protonated amines are stabilized within the cage even if the pH value of the surrounding medium would normally lead to deprotonation. Complex formation thus results in a shift of the effective basicity of the amine, which is quantified by using the thermodynamic cycle shown in Figure 5.63c [107]. Equilibrium A in this figure shows the protonation of the amine outside the cage, which is quantitatively specified in water by the corresponding pKa value (with the letter “a” standing for acid and not association in this case). Equilibrium B shows the binding of the ammonium ion by the cage with the corresponding equilibrium constant log Keff that specifies the stability of the complex. Equilibrium C depicts the binding of the neutral amine. This equilibrium can be neglected because the affinity of the cage for uncharged guests is significantly smaller than for charged ones, which allows calculating the pKeff value of the amine inside the cage (equilibrium D) by adding pKa and log Keff (note that Ka actually quantifies the extent to which the deprotonation of the acidic ammonium ion occurs, but since the pKa value is the negative decadic logarithm of Ka , the addition of pKa and log Keff correctly affords pKeff ). In the case of diisopropylamine, for example, the pKa amounts to 10.8 and the log Keff of the corresponding ammonium complex to 3.4, yielding an effective pKeff of 14.2. Thus, 50% of this amine exists in the protonated form even at a pH of 14. The cage thus shifts the pKa of the amine by more than three orders of magnitude and similarly pronounced effects are observed for other amines. This efficient stabilization of cations gives rise to interesting catalytic properties as we will see in Section 8.2. The final family of cages that should be discussed, also featuring octahedral metal complexes at the vertices, was developed in the group of Jonathan R. Nitschke. These cages are prepared by using a strategy known as subcomponent self-assembly, which involves combining two reversible reactions in the self-assembly process, namely the coordination of the ligands to the metal center and the generation of the actual ligands from smaller building blocks through a reversible covalent reaction [108]. This concept is illustrated in Figure 5.64 by using the example of a mononuclear complex. In this reaction, the two starting materials, 2-aminoethanesulfonic acid (taurine) and 2-formylpyridine, and the simultaneously present copper(I) ions afford a tetrahedral complex in which two chelating ligands, formed by imine formation from the amine and the aldehyde, coordinate to the metal center. This complex

5.5 Self-assembly mediated by coordination bonds

pK a = 3.2 Cu + O 2

pK a = 9.1

–O 3 S N + 2

NH 3+

2 H+ 2 H 2O

SO 3–

–O 3S

2

–O 3 S

323

NH 3+ N

N

N Cu + N N SO 3–

N Cu + N N

–O 3 S 2 NH 3+

SO 3–

Figure 5.64: Subcomponent self-assembly of a copper(I) complex from 2-aminoethanesulfonic acid and 2-formylpyridine and subsequent replacement of the aliphatic amine from this complex by an aromatic amine.

represents the thermodynamically most stable species of all possible products, which is why it dominates the final equilibrium. Figure 5.64 indicates that the structure of the ligands in this approach can be easily varied by just selecting different subcomponents without having to synthesize and isolate the ligand prior to coordinating it to the metal. Structurally changing the ligand is even possible after the assembly of the complexes because the imine bonds remain dynamic despite the stabilization by metal coordination. Accordingly, it is possible to exchange parts of the ligands at any time by just adding another subcomponent if a thermodynamically more stable state is thus reached. In the case of the complex shown in Figure 5.64, for example, the aliphatic amine can quantitatively be replaced with more acidic 4-aminobenzenesulfonic acid residues (pKa of the ammonium group 3.2 vs. 9.1 of the ammonium group of the aliphatic amine). The equilibrium thus favors a state in which the neutral form of the aromatic amine is incorporated in the complex, while the aliphatic amine is protonated and not involved in imine formation. The ability of such complexes to sensitively respond to the external conditions or the presence of new components allows the development of stimuli-responsive systems. Subcomponent self-assembly is useful to prepare helicates or cages, but we mainly focus on the cages here [109]. A prototypic example of such a cage is shown in Figure 5.65a. It contains the diimine 5.73 as chelating ligand, deriving from 2-formylpyridine and 4,4′-diaminobiphenyl-2,2′-disulfonic acid. Six of these ligands are arranged at the edges of a tetrahedron and are held together by iron(II) centers occupying the vertices. Each metal center is thus surrounded by the nitrogen atoms of three ligands in an octahedral coordination geometry. The twelve sulfonate groups ensure that this cage is soluble in water, where it incorporates neutral apolar guest molecules such as cyclohexane in its hydrophobic cavity. These guests are released by using different stimuli. One involves decreasing the pH, which causes the hydrolysis of the imine bonds holding the cage together

324

5 Assembling molecules

(a)

NH3+ O SO3–

6 – O3S

N

+ 12

–O S 3

12 H + 12 H2O

N

N

N

N

6

–O S 3

aqueous NH3

5.73

(b)

NH3+ N

Fe 2+

N 4 FeSO4 SO3–

–O S 3 Fe 2+

N Fe 2+

Fe 2+

N

Figure 5.65: Subcomponent self-assembly of a coordination cage from 2-formylpyridine, 4,4′-diaminobiphenyl-2,2′-disulfonic acid and an iron(II) salt (a) and crystal structure of the P4 complex of the cage (b). Counterions and hydrogen atoms are omitted for reasons of clarity.

and the protonation of the thus formed amines (Figure 5.66a). This process is reversed by another change of the pH, so that the system can be switched back and forth from the cage that hosts a guest to the fully disassembled state by repeated pH changes. Alternatively, the cage is also disassembled by adding tris(2-ethylamino) amine to the solution (Figure 5.66b). This ligand scavenges the metal ions from the cage, thus also causing guest release. The shift of the equilibrium is in this case partly due to entropy because the disassembly of the cage and the concomitant formation of the mononuclear iron(II) complex is associated with an increase of the number of molecules from 5 (1 filled cage and 4 tris(2-ethylamino)amine molecules) to 11 (6 difunctional ligands, 4 iron(II) complexes, and 1 cyclohexane molecule). Both of the above stimuli rely on the reversibility of the imine bonds, similar to the equilibrium shown in Figure 5.64. This cage hosts a variety of other guest molecules. One is pyrophoric white phosphorous that is stabilized within the cage over a period of months (Figure 5.65b) [110]. The addition of benzene to the solution causes the expulsion of the P4 molecule, thus initiating its oxidative decomposition to phosphoric acid. Another possible guest is furan, which is unable to participate in a Diels–Alder reaction when included in the cage (Figure 5.67). The reaction with maleimide present in the outside solution only starts after the included furan molecule is replaced with benzene and thus released.

5.5 Self-assembly mediated by coordination bonds

325

(a) N

SO3–

H 3O +

Fe 2+

N

6

+

NH3+

H3N –O S 3 +

SO3–

O

–O S 3

12 N N

Fe 2+ Fe 2+

+

Fe 2+ OH–

N

(b) N

SO3–

Fe 2+

N

NH2

NH2

6 H2N

NH2

N

–O S 3

H2N

SO3–

+ N

–O S 3 Fe2+

N Fe2+

4 Fe2+

N N Fe 2+ N N N

N

N

Figure 5.66: Disassembly of the cage formed from ligand 5.73 and subsequent release of its cargo by changing the pH (a) or adding a scavenging ligand (b).

The concept of constructing cages through subcomponent self-assembly is very versatile, giving access to coordination cages of widely varying structure and size [109]. Figure 5.68 shows two further examples to illustrate the scope. Tripodal ligands produced from 2-formylpyridine and a triamine afford tetrahedral cages upon coordination to iron(II) in which the ligands cap the four faces (Figure 5.68a). By changing the structure of the triamine, cages with volumes ranging between 31 and 823 Å3 are accessible. Using the porphyrin-derived tetraamine and 2-formylpyridine leads to a cube that contains iron(II) ions in the corners and nickel(II) ions bound to the six porphyrin subunits capping the faces (Figure 5.68b). These cubes have cavities large enough to host buckminsterfullerenes.

326

5 Assembling molecules

N

N

Fe 2+ N

Fe 2+ N

Fe 2+

SO 3–

SO3– O

–O S 3 N Fe 2+

–O 3S N

Fe 2+

Fe 2+ O

N O

N H

Fe 2+

Fe 2+

N

O

O O NH O

Figure 5.67: Schematic illustration of the initiation of a Diels–Alder reaction between furan and maleimide by replacing the bound and thus protected furan molecule from the cage with benzene (b).

5.6 Self-assembly mediated by covalent bonds 5.6.1 Introduction A distinction is made in the introduction of this chapter between covalent synthesis and noncovalent self-assembly to explain the differences between kinetically and thermodynamically controlled processes – the first involving irreversible reactions and the second relying on the continuous interchange of molecular building blocks. The principles of self-assembly do not only apply to noncovalent syntheses, however, but to any process in which the interactions between two or more building blocks are fully reversible. Covalent syntheses relying on reversible instead of irreversible reactions are no exception. One could question whether the term self-assembly is appropriate for covalent syntheses under thermodynamic control – the more apt term is, in fact, dynamic covalent chemistry [111]. Another question is how such reactions relate to supramolecular chemistry since the formation of a covalent bond is hardly a molecular recognition process. However, syntheses involving reversible reactions display so many characteristics of genuine self-assembly processes that it is warranted to discuss them here.

5.6 Self-assembly mediated by covalent bonds

327

(a) NH2 4 Fe(OTf)2 O 4

12

H2N

N

NH2

(b) NH2 8 Fe(OTf)2

N

N NH2

Ni 2+

6 H2N N

N O 24

N

NH2

Figure 5.68: Subcomponents required for the formation of a face-capped tetrahedral cage (a) and a face-capped cubic cage (b), together with crystal structures of the corresponding assemblies. Counterions and hydrogen atoms are omitted for reasons of clarity. One ligand in each structure is shown in red.

One important analogy is that, like in self-assembly processes where complex molecular architectures are produced from suitable building blocks through multiple noncovalent interactions, reversible reactions permit the synthesis of structurally defined, higher molecular weight compounds by connecting all of the individual subunits in a one-pot reaction. Macrocycles are formed directly from their monomers, for instance, without having to first synthesize the linear precursor in a stepwise fashion and then cyclize it. The reversibility of bond formation, which allows incorrectly formed bonds to dissociate in the course of the reaction, eventually affording the thermodynamic product, moreover causes such syntheses to proceed with a similarly high fidelity as noncovalent syntheses. Finally, the nature of the product eventually dominating in the equilibrium is determined by factors analogous to those controlling self-assembly. An enthalpically favorable situation arises, for example, if all functional groups available for product formation are used (principle of maximum site occupancy). Thermodynamics furthermore favors the formation of the smallest

328

5 Assembling molecules

possible unstrained product, and the product distribution is also affected by stabilizing intramolecular interactions occurring in one product but not in the starting materials or competing products. Such syntheses are particularly effective and high yielding if only a single product remains in the equilibrium. If, however, several products have a similar thermodynamic stability, a mixture of products results. In this case, there are still means to shift the equilibrium toward a single species. One is the use of templates, an attractive strategy discussed in the next chapter. Alternatively, it is sometimes possible to make use of kinetic traps. Choosing a solvent for the synthesis in which the desired product is insoluble, for example, causes it to precipitate during the reaction and the equilibrium to continuously adapt to produce more of this compound. A major difference between noncovalent and covalent syntheses is that the latter typically afford robust products, which can be isolated, characterized, and further processed if the exchange can be stopped. Figure 5.69 lists a selection of reversible reactions where this is possible. In reactions (a) through (f), two different functional groups are coupled, liberating one equivalent of water in each case. This water is required to mediate the

O (a)

R1

OH

+

R1 B

R1

+

O R2

H+

O

H 2 N R2

O R1

+ H O

(f)

R1 H

R2

O R1

H+

H2N

H N

H R1

H (e)

R2

O R1 B

HO

O R1

R2

O

R2

+ H

O R1

HO HO

O

(d)

H+

+ OH

(c)

R2

HO

OH (b)

HO

H+ R2

+ H 2 N O R2

(g)

R 1 SH

+

HS R 2

(h)

R1

+

R2

H R1

H+

R2

N H N

N

R2

H R1

O2/RS R1 Catalyst R1

N S

O

S

R2

R1 + 1 S R2 + 2 S R2 R R S S

R2 + 1 R

R1 + 2 R

R2

Figure 5.69: Selection of important reversible reactions used in dynamic covalent chemistry.

5.6 Self-assembly mediated by covalent bonds

329

respective back reaction, thus ensuring reversibility. The ester formation in reaction (a) is catalyzed by acids but usually requires long reaction times and elevated temperatures. An alternative is the transesterification between two esters that is mediated by catalytic amounts of alkoxides. Cyclic boronic acid esters form rapidly even at neutral pH as we have seen in Section 4.2.4. These esters are, however, usually much less stable than the esters of carboxylic acids. Acetals (reaction (c)) and imines (reaction (d)) are formed conveniently from aldehydes under relatively mild acidic conditions upon treatment with alcohols or primary amines, respectively. Both functional groups are stable in the absence of acids, allowing the reaction to be switched off by a change of the pH. Imines are, however, very prone to hydrolysis in the presence of water. The isolation of the respective products is therefore often only possible after postfunctionalization, for example by reduction to the corresponding amines. Hydrazones (reaction (e)) and oximes (reaction (f)) do not have this disadvantage. Their formation typically requires more strongly acidic conditions than that of imines, but in the absence of acids, they are stable even in water. Disulfide formation (reaction (g)) and olefin metathesis (reaction (h)) are examples of reversible reactions between two identical functional groups. As a consequence, two different thiols or olefins lead to three products, namely the heterocoupling product and the two homocoupling products. Disulfides are formed from thiols by oxidation, with the subsequent shuffling of the substituents at the disulfide bond requiring the presence of small amounts of thiolate. The reaction thus proceeds in basic media and is switched off by decreasing the pH. A major advantage of disulfide exchange is that it also works in aqueous solvent mixtures and in water. Olefin metathesis requires a transition metal catalyst that mediates the coupling of two terminal olefins with the simultaneous release of ethylene. Since the catalyst also accepts internal olefins as substrates, the initially released ethylene is not required for the exchange to work. The scope of olefin metathesis is often limited by the lifetime of the catalyst, which must remain active until the reaction has reached the thermodynamic equilibrium. In the remainder of this chapter, we will see how these reactions allow the preparation of structurally complex products from simple building blocks. In this context, we mainly concentrate on syntheses of rings and cages to demonstrate that supramolecular and dynamic covalent chemistry share more than just conceptual links. Many of the compounds that are produced by reversible reactions are directly relevant for molecular recognition and self-assembly. Dynamic covalent chemistry under the influence of templates even allows developing new receptors as discussed in Section 5.7, and can give rise to interlocked molecules, which are the topic of Chapter 6. This strategy is thus extremely useful to address many synthetic challenges in supramolecular chemistry.

330

5 Assembling molecules

5.6.2 Rings Before referring to specific syntheses, it is worth noting that several of the receptors discussed in Section 4.1 are produced by using reversible reactions. Examples are cyclotriveratrylenes (Section 4.1.6), calixarenes (Section 4.1.7), resorcinarenes (Section 4.1.9), pillararenes (Section 4.1.10), and cucurbiturils (Section 4.1.11), that is, all aldehyde-derived macrocycles. Hallmarks of the thermodynamic control are the possibility of making use of template effects, which play a role in calixarene syntheses, and the fact that the composition of the reaction mixtures changes with time, eventually converging onto a single thermodynamically favored product as observed in resorcinarene and cucurbituril syntheses. Some of the most important receptors in supramolecular chemistry thus owe the efficiency with which they are prepared to dynamic covalent chemistry. Moreover, since these syntheses require relatively harsh acidic or basic conditions, the respective products are inert after isolation and can thus be conveniently used in binding studies or for further transformations. An instructive example of a macrocyclization involving a transesterification reaction is the cyclotrimerization of the cinchona alkaloid derivatives 5.74a and 5.74b (Figure 5.70). These esters afford the corresponding cyclic trimers in > 90% yields when heated in toluene together with catalytic amounts of KOMe/18-crown-6 (the crown ether serves to complex the potassium ions, thus mediating the dissolution of KOMe in toluene) (Figure 5.70a) [112]. The reaction conditions suggest that the selectivity and efficiency of this reaction is due to thermodynamic control but it cannot be excluded that the formation of the cyclic trimer is a kinetic effect and that no exchange of the building blocks occurs once the product is formed. Reference experiments are therefore required to demonstrate the reversibility of the underlying reaction. One important aspect that needs to be demonstrated is that kinetic control produces another product distribution. Indeed, the cyclization of the acid chloride 5.74c at room temperature gives rise to a mixture of products, with the cyclic trimer still dominating (37%) but the tetramer and larger macrocycles also being present in significant amounts (Figure 5.70b). If, however, the linear dimer 5.74d is treated with 5% KOMe/18-crown-6 in refluxing toluene, the product distribution is the same as that resulting from monomer 5.74b (Figure 5.70a). Thus, the preferential formation of the trimer by transesterification cannot be a kinetic effect because products other than the trimer are present when the reaction is performed under kinetic control. The product distribution is, however, independent of which oligomer of the starting material is used, indicating that the synthesis involves continuous assembly and disassembly reactions. Further proof that the system is dynamic is obtained from experiments involving mixtures of starting materials. Treating either a 1:1 ratio of the two monomers 5.74a and 5.74b, a 1:1 ratio of the two cyclic homotrimers, or a 2:1 ratio of 5.74a and

331

5.6 Self-assembly mediated by covalent bonds

N N

N

R1

CO2Me OH

R2

N

COR 3

O

OH 5.74b (R 3 = OMe) 5.74d (R 3 = Cl)

(a)

N 5% KOMe/18-crown-6 toluene, reflux

or

N

R2 =

N

5.74c

MeO

O

R2 5.74b

R1 =

O

O

R2

OH 5.74a

OMe

R2

R2

O

O O

>90%

N

5.74c O

N

O

R2

(b)

N N

O

R2

5.74d

DMAP, CH2Cl 2 25°C

O

R2

O

O

O

R2

O

R2 O

O

N

+

N

N

O

O

O

N

R2

O

R2

O R2

O

N

37%

23% +40% higher oligomers

(c)

N

or 1 equiv of the cyclic trimer of 5.74a + 1 equiv of the cyclic trimer of 5.74b

O

O

1 equiv 5.74a + 1 equiv 5.74b

N

O

R1

R1

O

O

O

R1

O N

N

O

N

5% KOMe 18-crown-6 toluene, reflux

O

N

O

R1

3 N

O

R1

O

O

O

R2

1 N

or

O

R2 R2

O 2 equiv 5.74a + 1 equiv 5.74c

O

R1

O

O

R2

O N O

N

O

R2 3

N O

N

O

R2 1

Figure 5.70: Molecular structures of building blocks 5.74a-d and cyclization of 5.74b and 5.74c under thermodynamic (a) and of 5.74d under kinetic control (b). The outcome of reactions using mixtures of starting materials is shown in (c).

332

5 Assembling molecules

5.74d under the above conditions yields in all cases identical product distributions, comprising the expected mixed macrocycles in the statistical 1:3:3:1 ratio (Figure 5.70c). Even the cyclic products thus shuffle their subunits, which conclusively shows that the transesterification is reversible under the reaction conditions and that product formation proceeds under thermodynamic control. That the formation of the cyclic trimer is so strongly preferred indicates that it is thermodynamically the most stable product because the respective monomers are predisposed to preferentially form this ring. Another useful reversible reaction for the preparation of macrocycles is imine formation, with the resulting macrocyclic Schiff bases or the corresponding polyamines obtained after reduction often serving as ligands for transition metal ions. In this context, it is frequently observed that macrocyclic imines undergo exchange reactions in solution to produce mixtures of rings varying in size. The dissolution of the tetraimine 5.75c (Figure 5.71) in chloroform affords a mixture of the three macrocycles 5.75a–c, for example, in a reaction mediated by traces of HCl. The presence of the three products in the equilibrium suggests that they have comparable thermodynamic stability [113]. The corresponding oxygen analogs 5.76a–c undergo a similar exchange reaction [114]. Thermodynamic control is reflected in this case in the concentration dependence of the equilibrium, which shifts from the preferred formation of the dimer 5.76a at low concentrations (0.006 M: 5.76a:5.76b:5.76c = 72:23:5) to the formation of the larger macrocycles at higher concentrations (0.2 M: 5.76a:5.76b:5.76c = 44:31:25) as a result of entropic factors.

X X

N

X

N

X

N

N X N X

5.75a (X = NH) 5.76a (X = O)

5.75b (X = NH) 5.76b (X = O)

N X

N

N

X N

X

5.75c (X = NH) 5.76c (X = O)

Figure 5.71: Molecular structures of the macrocyclic imines 5.75a–c and 5.76a–c.

The fidelity of such macrocyclizations strongly depends on the structure of the subunits. Combining (R,R)-1,2-diaminocyclohexane with terephthalaldehyde leads to the exclusive formation of the cyclic hexaimine 5.77 (Figure 5.72a), for example, with E-configurations at all imine bonds. This compound is surprisingly robust for

333

5.6 Self-assembly mediated by covalent bonds

a hexaimine and can moreover be structurally varied in a wide range by changing the nature of the aromatic subunits. Using biphenyl-4,4′-dicarbaldehyde instead of terephthalaldehyde affords a macrocycle with a larger diameter, for example, and additional substituents in the aromatic subunits either serve for intramolecular conformational control or as binding sites for potential substrates. Hexaimine 5.77 thus represents the prototype of a larger family of macrocyclic hexaimines known as trianglimines [115]. When using isophthalaldehyde derivatives in the macrocyclization reaction, the imine formation proceeds somewhat less selectively with the hexamine only representing the kinetically controlled product that slowly reequilibrates to afford the thermodynamically favored smaller ring. Pyridine-2,6-dicarbaldehydes yield a mixture of rings of different sizes. Some of these compounds possess interesting binding and/or self-assembly properties, but the general disadvantage of trianglimines is that they rapidly hydrolyze under acidic aqueous conditions. Hydrazones do not suffer this drawback and, indeed, the macrocycle 5.78 (Figure 5.72b) can even be prepared

N N

CHO NH2

N

(a)

+ NH2

N CHO

N N 5.77

N

CHO

N

N

(b)

NH2 HN O HN NH2

Aqueous TFA

N

N

HN

NH

+

O

O

HN

NH N

N

N CHO

N

N

5.78 Figure 5.72: Synthesis of the macrocyclic hexaamine 5.77, the prototype of a trianglimine (a), and of the macrocyclic hydrazone 5.78 (b).

334

5 Assembling molecules

in aqueous solution from the respective dialdehyde and carbonohydrazide under acidic conditions [116]. Figure 5.73 illustrates that it is also possible to combine different reversible reactions in a macrocyclization, in this case imine and boronic ester formation [117]. Since the formed bonds originate from different functional groups – an aldehyde and an amine in one case and a boronic acid and a diol in the other – the corresponding reactions proceed independently without giving rise to crossover products. When performing the condensation reaction in a Dean-Stark trap, substantial amounts of polymeric side products are also observed, which indicates that the removal of water could prevent the reaction from reaching the thermodynamic equilibrium. Intimately mixing the required components in the presence of a small amount of solvent in a ball mill often affords better yields [118]. O

O

O

O

B

B

HO 4 OHC

B

OH

HO +

OH

2

+ HO

OH

NH2

N

N

NH2

N

N

2

O

O

O

O

B

B

Figure 5.73: Formation of a macrocycle by the simultaneous formation of imine and boronic ester bonds.

An impressive example of a macrocycle obtained by disulfide exchange, which was described by Sijbren Otto, is shown in Figure 5.74. This compound forms spontaneously within ca. 16 days when equilibrating building block 5.79 at a concentration of 0.5 mM in borate buffer [119]. Under similar conditions, 1,3-benzenethiols normally afford cyclic trimers and tetramers. These small rings indeed form preferentially when the concentration of 5.79 amounts to only 0.05 mM, indicating that the enthalpic stabilization of the large 15-mer at higher concentrations overcompensates the entropic advantage of forming smaller rings. This macrocycle is tightly folded and stabilized by an intricate pattern of intramolecular hydrogen bonds as illustrated by the crystal structure in Figure 5.74. The burial of hydrophobic residues within this structure moreover suggests the importance of the hydrophobic effect for the folding and the resulting thermodynamic stabilization. Indeed, performing disulfide exchange in the presence of salts that reinforce the hydrophobic effect favors the formation of the 15-mer, while this compound does not form in solvent

5.6 Self-assembly mediated by covalent bonds

335

N

H2N

N N

N

NH O O

HS

CO2H NH

SH

5.79

Figure 5.74: Molecular structure of building block 5.79 and crystal structure of the macrocyclic 15-mer derived thereof.

mixtures in which a substantial fraction of an organic component suppresses the thermodynamic gain associated with the desolvation of hydrophobic residues.

5.6.3 Cages The most frequently used reversible reaction to prepare cages is imine formation, which is why we focus in this chapter mainly on this approach. Again, an important class of receptors, namely the polyazacryptands discussed in Section 4.1.2, are synthesized in this way. Reacting isophthalaldehyde and tris(2-aminoethyl) amine in a 3:2 ratio in methanol affords the respective hexaimine in yields up to 90%, for example (Figure 5.75) [120]. The precipitation of the product from the reaction mixture, which allows isolation by simple filtration, drives this reaction almost to completion. Treating the product with sodium borohydride then affords the corresponding cryptand. Other polyazacryptands are synthesized in a similar fashion from tris(2-aminoethyl) amine or other triamines and a wide variety dialdehydes [121]. Experiments using mixtures of starting materials or involving the transformation of a thermodynamically less favorable into a more favorable cage demonstrate the reversibility of the reaction. The condensation between the two triamines and two aldehydes shown in Figure 5.76a leads to the exclusive formation of two self-sorted products, for example, which thus represent the thermodynamically favored species. Similarly, the hexaimine shown in Figure 5.76b is quantitatively converted into a more

336

5 Assembling molecules

H2N NH2

N

2

N

N

N

NH2 +

N

NaBH4

HN

NH

N

N

N

N

HN

NH

N

N

HN

NH

3 OHC

CHO

Figure 5.75: Example of the synthesis of a polyazacryptand via the corresponding hexaimine.

stable cage by replacing the terminal tris(2-aminoethyl)amine with tris(aminomethyl)benzene residues. (a)

N N

N

N

CHO

H2N N

N

NH2

CHO

N N

NH2 + NH2 NH2

CHO N

H2N

CHO N

(b) N N

N

N N

NH2 N

N N

NH2 N

N

N N

H2N

H2N N N

N

N N

NH2 N

NH2

N N

Figure 5.76: Self-sorting leads to the exclusive formation of two cages from a mixture of two triamines and two dialdehydes (a), and the exchange of the terminal triamine residues converts the original hexaimine into a less strained and therefore thermodynamically more stable analog (b).

5.6 Self-assembly mediated by covalent bonds

337

Imine formation also allows preparing analogs of Cram’s carcerands and hemicarcerands from the resorcinarene-derived tetraaldehyde 5.80 and aromatic diamines (Figure 5.77) [122]. The cage containing 1,3-diaminobenzene linkers, for example, is obtained in almost quantitative yields by treating a solution containing the components in the proper ratio with trifluoroacetic acid. Dissolving this cage in chloroform together with four equivalents of a 5-substituted 1,3-diaminobenzene derivative causes the stepwise replacement of the unsubstituted for the substituted linkers, which demonstrates that imine formation remains reversible even after the cage has been formed. Since the replacement of the linkers has to proceed via short-lived intermediates in which a bridge is opened or completely removed, the acid-induced linker exchange also facilitates the release of a guest molecule entrapped within the cavity of this cage. R

R

OHC 2

O

O OHC O R

O

O

O

+ R

R

NH2

O OHC O

R

5.80 (R = CH2CH2Ph)

TFA CHCl3

R

O

O

O

N

NO

O

O O N N

N

NO

O

N

O

CHO

R

4 NH2

O

O

O

O R

N O O

R

R R

Figure 5.77: Formation of a carcerand-type imine cage from a resorcinarene-derived tetraaldehyde and 1,3-diaminobenzene.

The outcome of the reaction shown in Figure 5.77 is largely determined by the structure of the diamine. 1,3-Diaminobenzene arranges two cavitands in a converging fashion upon imine formation, thus giving rise to the formation of the entropically favored octaimine. Short aliphatic diamines such as 1,2-diaminoethane cannot be integrated into such a structure in a strain-free fashion, which is why they induce the formation of larger cages. In chloroform, the trifluoroacetic acidcatalyzed condensation of the cavitand tetraaldehyde with 1,2-diaminoethane affords an octahedral cage containing six macrocyclic units connected through 12 linkers (Figure 5.78) [123]. Product formation sensitively depends on the solvent since this cage does not form in tetrahydrofuran where the same reactants yield a tetrahedral cage containing four cavitands. In dichloromethane, a square antiprismatic cage is preferentially formed in which the cavitands occupy the eight corners.

338

5 Assembling molecules

Tetraaldehyde OHC

O O OHC O R

Diamine

Linker

CHO

O

O

O

O OHC O

=

H2N

R

R

NH 2

N

N

=

R

[R = (CH2)5CH3]

In CHCl3 from 6 tetraaldehydes and 12 diamines

In THF from 4 tetraaldehydes and 8 diamines

In CH2Cl2 from 8 tetraaldehydes and 16 diamines

Figure 5.78: Schematic illustration of the cages formed from a resorcinarene-derived tetraaldehyde and 1,2-diaminoethane in chloroform, tetrahydrofuran, and dichloromethane.

Large covalently assembled cages are thus accessible in a relatively straightforward fashion. Further examples are shown in Figure 5.79. Ralf Warmuth used the trifluoroacetic acid-mediated reaction between the cyclotriveratrylene-derived trialdehyde and 1,4-phenylenediamine to obtain a homochiral cube with a molecular diameter of approximately 3.7 nm, for example, in which the cyclotriveratrylene units occupy the corners and the linkers the edges (Figure 5.79a) [124]. Tetrahedral cages were prepared in the group of Andrew I. Cooper from 1,3,5-triformylbenzene with 1,2-ethylenediamine or other vicinal diamines (Figure 5.79b) [125], and an endo-functionalized adamantoid cage was obtained in the group of Michael Mastalerz by reacting a triptycene-derived triamine with salicylic dialdehyde (Figure 5.79c) [126]. Can a liquid be porous?

In the solid state, the latter two cages afford porous microcrystalline materials with high surface areas and promising gas adsorption properties. Dissolving the smaller cage in an appropriate solvent also allows creating holes in a liquid [127]. Two factors are crucial to ensure that the dissolved cage molecules remain empty. First, they need to contain solubilizing residues in the periphery that are unable to enter and thus block the cavity. This is realized as shown in Figure 5.79b by introducing crown ether

5.6 Self-assembly mediated by covalent bonds

339

(a) OHC

OCH3 NH2

CHO

H 3 CO OHC

NH2

OCH3

8

(b)

16 O

O

O CHO

O O

O NH2

H2N OHC

CHO 4

6

(c) H2N

OHC

CHO OH 6

H2N 4

NH2

Figure 5.79: Building blocks and structures of polyimine cages formed from a cyclotriveratrylenederived trialdehyde and 1,4-phenylenediamine (a), 1,3,5-triformylbenzene and 1,2-ethylenediamine (b), and salicylic dialdehyde and a triptycene-derived triamine (c). The crown ether in (b) allows preparing a soluble empty cage for a porous liquid. The structure in (a) was calculated, while the structures in (b) and (c) are crystal structures. Hydrogen atoms are omitted for reasons of clarity.

residues into the ligands. Second, the solvent molecules must be so large that they are unable to enter the cavity, with 15-crown-5 fulfilling this requirement. The solution of the crown ether-decorated cage in 15-crown-5 thus has a permanent porosity and a high capacity for absorbing gases such as methane. Bound gas molecules are expelled from the cages by adding a better binding guest, which causes the solution to visibly start bubbling. The reliable and facile way with which such cages are prepared by

340

5 Assembling molecules

using dynamic covalent chemistry thus makes it possible to explore their uses in novel applications. Although imine formation is likely the most frequently used reversible reaction to create covalently assembled cages, there are alternatives. The formation of boronates from boronic acids and diols is one of them. Florian Beuerle used design principles closely related to those underlying the formation of the cube shown in Figure 5.79a, for example, to prepare an analogous boronic ester-derived cube from the tribenzotriquinacene derivative 5.81 and the bis (boronic) acid 5.82 (Figure 5.80a) [128]. The synthesis is very efficient, affording the product in 94% yield by just mixing 5.81 and 1.5 equivalents of 5.82 in THF in the presence of molecular sieves. The reason for this high fidelity is the rigidity of 5.81 and its predisposition to form a cube, resulting from the arrangement of the aromatic residues at angles of almost exactly 90°. When positioning these building blocks at the vertices of a cube, the aromatic residues are pointing into the direction of the edges, which allows them to be linked by the linear components through boronate formation. The calculated structure of the resulting product is shown in Figure 5.80b. By combining other bis(boronic) acids and diols, various other cage architectures can be realized [129]. (a)

HO

(b)

OH

HO

OH HO

5.81

8

OH

B(OH)2 H9C4 C4H9 B(OH)2 5.82

16

Figure 5.80: Molecular structures of the tribenzotriquinacene 5.81 and the bis(boronic acid) 5.82 (a) and calculated structure of the boronate-derived cage formed from both components (b). Protons are not shown in the calculated structure, the side chains in the tribenzotriquinacene moieties are truncated, and those in the bis(boronic acid) moieties are omitted for reasons of clarity.

5.7 Dynamic combinatorial chemistry

341

5.7 Dynamic combinatorial chemistry 5.7.1 Introduction The motivation of using dynamic covalent chemistry for synthetic purposes is often to efficiently prepare a compound that would be much more difficult if not impossible to obtain by using other strategies. Since more than one covalent bond is generally formed during such syntheses, the desired product is, however, never the only possible one. Sometimes, one product is so much more stable than potential other ones that it dominates the equilibrium. In this case, the other products are conceivable but not actually observed, and they may thus be termed virtual [130]. Sometimes, the outcome of a synthesis under thermodynamic control is indeed a mixture as we have seen in Section 5.6.2. In contrast to conventional syntheses, the components in this mixture are not structurally fixed, however, but constantly exchange their components, which is a feature that has interesting implications for supramolecular chemistry as shown in this chapter. Let us start by looking at the possible strategies for receptor development. The traditional approach would be to design a receptor prototype, perhaps with the help of computational methods, then prepare this compound, and finally determine its binding properties. Since it is unlikely that the first product already has all the desired properties, the results of these binding studies usually lead to a redesign, which initiates a new cycle of receptor optimization. After several rounds of iteration, the receptor is finally available, or maybe not. This approach is thus timeconsuming and not necessarily successful. An alternative could be to develop receptors in a combinatorial fashion. This approach is faster than the traditional one because many receptors are synthesized and screened more or less simultaneously, but it is restricted to the receptors that are actually present in the investigated library. The chances are high that a very good receptor can thus be identified, but there is always the possibility that better receptors may exist that were absent in the library screened. The development of receptors by conventional combinatorial chemistry therefore also has disadvantages. A third strategy of receptor development is based on the mixtures of interconverting molecules produced by using reversible reactions. This strategy also involves the use of receptor libraries, but it differs from conventional combinatorial chemistry in several ways. First, the receptors are not synthesized as stable compounds but as dynamically interconverting species. The structural diversity of the library is thus not restricted to the compounds that have been synthesized but that can potentially be produced. Even a single building block can, for example, lead to acyclic and cyclic oligomers of varying sizes, and the structural diversity becomes even larger if mixtures of building blocks are used for library generation. Some control over the properties of the target compounds is exerted by choosing building blocks with binding sites that could induce receptor properties, but it is not

342

5 Assembling molecules

necessary before setting up the libraries to have a clear idea about how the product should look like. The possibility of accessing all possible building block combinations, even those that are not physically present but only virtual, makes it very likely that among all receptors accessible, the best one will be identified. The second difference between conventional and dynamic combinatorial chemistry is that receptor assembly and screening are performed simultaneously by analyzing the effects of templates on library composition. Adding a substrate to a dynamic receptor library, for example, should shift the product distribution observed in the absence of the template toward those library members to which the substrate binds. The formation of the corresponding complexes causes the equilibrium to adapt by forming more of the selected receptors at the expense of other compounds that do not interact with the template. Another way to look at this effect is that the template changes the thermodynamic landscape of the whole system of interconverting species so that receptors are eventually enriched in the equilibrium whose interactions with the template lead to a pronounced thermodynamic stabilization (Figure 5.81). The extent of stabilization correlates with the stability of the corresponding complexes, causing the best receptors to be amplified the strongest under ideal conditions. Analyzing how a template changes the library composition thus allows identifying the compounds to which it binds. Experimentally, this is often achieved even for libraries containing many different members by using modern HPLC-MS techniques. A final advantage of dynamic covalent chemistry over conventional combinatorial chemistry is the ease with which the receptor, once identified, can be synthesized. This synthesis is performed under the same conditions used for generating the dynamic library by again making use of the template effect of the substrate but

ΔG°

Figure 5.81: Schematic illustration of the energetic landscape of a dynamic library of interconverting molecules prior and after the addition of a template. The template causes the thermodynamic stabilization of the receptor complementary to the template, concomitantly shifting the relative proportions of the library members as symbolized by the black circles residing in the minima of the curve.

5.7 Dynamic combinatorial chemistry

343

restricting the building blocks to only those that end up in the selected product. To be able to isolate this product as a stable compound, it is only necessary to stop the interconversion of the library members by, for example, changing the reaction conditions. The concept of dynamic combinatorial chemistry, which was simultaneously and independently developed in the groups of Jean-Marie Lehn and Jeremy K. M. Sanders, is not restricted to the identification of a receptor for a substrate but is also used for other purposes [131]. The possible applications differ in the roles of the template and the library members, but the underlying principles are the same in all cases. If the participating compounds can be classified as receptors (converging binding sites) and substrates (diverging binding sites), two strategies are distinguished – molding and casting. Molding involves forming the receptor like a mold around the shape of the substrate (Figure 5.82a). The inverse strategy is termed casting because

(a)

(b)

(c)

(d)

Figure 5.82: Possible applications of dynamic combinatorial chemistry, involving the identification of a receptor for a given substrate (a), the identification of a substrate for a receptor (b), the selection of binding partners in self-assembling systems (c), and the control over the sequence and length of oligomers of different subunits by intramolecular interactions (d).

344

5 Assembling molecules

a cast is formed of the receptor cavity (Figure 5.82b). If all library members have diverging binding sites, they can interact by self-assembly. In this case, one compound in the dynamic library can act as a template to produce more of the compounds with which it interacts (Figure 5.82c). When including intramolecular recognition processes in this classification, a fourth possibility is the selection of the preferred sequence and length of oligomers by intramolecular interactions between their building blocks (Figure 5.82d). At this point the question could arise about the difference between dynamic covalent chemistry and dynamic combinatorial chemistry. When focusing on product formation, is it correct to say that any process involving reversible reactions that leads to a single product belongs to dynamic covalent chemistry, while processes that yield mixtures encompass dynamic combinatorial chemistry? Probably not because whether a reaction affords a single product or a mixture can change depending on the reaction conditions. Moreover, if we include virtual products into the considerations, there is no reaction that does not have a combinatorial element. Another criterion for differentiation could be that the combinatorial aspect of dynamic combinatorial chemistry refers to building blocks used to generate the libraries. If mixtures of building blocks are used, the approach is a combinatorial one and if not, it belongs to dynamic covalent chemistry. While the former argument is undeniably true, what about reactions where a single building block leads to mixtures (libraries) of different products? Does this reaction not have a combinatorial aspect, too? A convenient approach would be to make the differentiation with regard to the use of templates, with dynamic combinatorial chemistry involving template effects and dynamic covalent chemistry not, but this classification makes it difficult to classify the process shown in Figure 5.82d. Can an intramolecular interaction really be regarded as templation, which would render the approach a dynamic combinatorial one, or would it not be equally true to state that the product selected by the intramolecular interactions is simply the most stable one, rendering the term dynamic covalent chemistry more appropriate? There is thus no simple way to make a clear-cut distinction between dynamic covalent and dynamic combinatorial chemistry. Maybe it is best to conclude that dynamic combinatorial chemistry is an overarching concept that is based on dynamic covalent chemistry as the tool. As in dynamic covalent chemistry, the key to using dynamic combinatorial chemistry is the reversible reaction that mediates the interconversion of the library members. This reaction must meet several requirements for the concept to work. – The reaction needs to be sufficiently fast to allow generating a dynamic library and following the effects of templates on library composition on a reasonable timescale, typically within hours or days. – No side reactions between the functional groups involved in the reversible reaction and those that mediate the interactions between the library members and the template should occur.

345

5.7 Dynamic combinatorial chemistry

– Since the equilibration and templation proceed simultaneously, the conditions required for the exchange to work should be sufficiently mild to not impair the interactions between the template and the library members. – All library members and their complexes should remain in solution during the equilibration to avoid kinetic traps. – Ideally, all library members should have a similar thermodynamic stability because it is energetically costly to shift the equilibrium if the library is strongly biased toward the formation of a few or even only one product in the absence of the template. – Finally, it must be possible to freeze the reaction and to thus convert the library members into stable products to allow their isolation and characterization. In light of these requirements, some of the reversible reactions shown in Figure 5.69 are not ideal for dynamic combinatorial chemistry. Ester formation or transesterification usually requires relatively harsh conditions, for example, which are incompatible with the molecular recognition processes required for template effects to work. Boronic esters, on the other hand, are often too labile to allow further processing. The most frequently used reversible reactions in dynamic combinatorial chemistry are imine formation (usually followed by reduction and isolation of the corresponding amines), hydrazone formation, and disulfide formation, which are shown in Figure 5.83 in the form of the corresponding exchange processes. A few other reactions are possible but have been used less frequently [131].

H (a)

R1

H N

H (b)

R1

N

H N

R2

+

R3

N H

R2

+

R3

H

H+ R4

N

H N

R1

R1 R1

(c)

R1

S

S

R2

+

R3

S

S

R4

N H

H+ R4

H

RS R1 R3

N S S S

R4

H N

S S S

+

R3

N H

R4

+

R3

R4

+

R3

R1

+

R2

R3

+

R4

N S S S

R2

H N

S S S

R2

R2 R2 R4

Figure 5.83: Most frequently used exchange reactions in dynamic combinatorial chemistry.

Does a template always amplify the best binder?

A crucial question is whether it is really always the best binder that is chosen by the template. Maybe surprisingly, the answer is no. The reason is that the dynamic

346

5 Assembling molecules

library responds to the presence of the template by adopting the thermodynamically most favorable state of the whole system [132]. As a consequence, situations could arise, in which a template leads to the amplification of a good but not necessarily the best binding partner as illustrated by the following examples. Consider a dynamic library containing macrocyclic receptors formed from a single building block (Figure 5.84a). In this library, the cyclic hexamer should form the most stable complex with the template, but the cyclic trimers should also exhibit affinity. Only one cyclic hexamer but two cyclic trimers are formed from the same number of building blocks, and which of these macrocycles is amplified the strongest depends on whether the formation of one complex of the hexamer or two complexes of the trimers leads to an overall more favorable thermodynamic situation. If the ΔG0hexamer associated with the formation of the hexamer complex is significantly more negative than twice the ΔG0trimer , the better receptor wins. If, however, 2 × ΔG0trimer < ΔG0hexamer , the trimer is preferentially amplified and not the receptor that forms the more stable complex.

(a)

High template concentration

T

Low template concentration

(c)

T

T

+

+

T

T

T T

(b)

(d) T

T

T

T

T

T

Figure 5.84: Scenarios illustrating the competition of different receptors in a dynamic library differing in structure and affinity for the binding to the template. The receptors in (a) comprise macrocycles of different sizes of which the hexamer should have a larger affinity than the trimer. In (b), cyclic trimers containing different subunits interconvert, with the homotrimer having a larger affinity than one of the heterotrimers. The equilibria in (c) and (d) show how these situations change when the template concentration is reduced.

347

5.7 Dynamic combinatorial chemistry

A similar situation arises in libraries in which different building blocks form macrocyclic receptors of the same size. Two building blocks, for example, yield four macrocyclic trimers in a statistical 1:3:3:1 ratio (Figure 5.84b). One of the homotrimers should be the best receptor in this dynamic library, but the receptor containing two building blocks of this homotrimer also interacts with the template, albeit with a weaker affinity. Which macrocycle is selected by the template depends again on whether complex formation of the homotrimer has an overall lower ΔG0 or the formation of several complexes of the mixed receptor. These considerations thus show that if compounds exist in a dynamic library that bind to the template and are present in larger amounts than the best receptor, these compounds can be amplified the strongest. There are, however, experimental ways to avoid this unwanted outcome. The above examples show that the weaker binders only contribute to the overall ΔG0 of the reaction if enough template molecules are present. If the number of template molecules in solution is too small to involve the weaker binders in the selection process as shown in Figures 5.84c and 5.84d, a direct competition arises between receptors with different affinities. In this case, the best receptor wins. The smaller the concentration of the template, the more likely it should therefore be that the best receptor is identified. That this assumption is indeed correct has been verified computationally by predicting the responses of simulated dynamic combinatorial libraries to the presence of templates [133]. The approach involved the generation of a 322-member library of interconverting compounds from a set of virtual building blocks in which each library member is present in amounts that depend on the equilibrium constants of the exchange reactions, statistical factors, and building block concentrations. Binding constants for the template complexes of (b)

100

100

10

10

1

1

AF

AF

(a)

0.1

0.1

0.01

0.01

0.001

0.001

–30

–20

–10

0

10

ΔG° (kJ mol–1)

–30

–20

–10

0

10

ΔG° (kJ mol–1)

Figure 5.85: Relationship between the amplification factor AF, describing the change of the concentration of a library member induced by the presence of the template, and the Gibbs free energy ΔG0 of the complex between the template and the corresponding library member in simulated dynamic libraries differing only in the concentration of the template. The template concentration amounts to 10 mM in (a) and to 1 mM in (b).

348

5 Assembling molecules

the library members were randomly assigned, and the change in library composition, specified in terms of amplification factors that denote the ratio of the concentrations of each library member prior and after template addition, was then calculated. Figure 5.85a shows the result of such a calculation for a template concentration of 10 mM. In this case, the library strongly responds to the presence of the template by enriching a number of receptors and depleting many others. Notably, the receptor whose binding to the template is associated with the most negative ΔG0 is not amplified at all, whereas the receptor that forms the third most stable complex is amplified the strongest. If the template concentration is reduced to 1 mM, the distribution of amplification factors for the same library looks completely different (Figure 5.85b). In this case, the concentrations of most library members are not or only weakly affected by the presence of the template but the best receptor is amplified the strongest, which confirms the above considerations. Restricting the template concentration is not without negative consequences, as it also affects the absolute amount to which a receptor can be amplified. It could therefore be that a good receptor remains undetected at low template concentrations because its concentration is below the detection limit. Consequently, it is usually advisable to perform a series of experiments with different template concentrations to increase the chances that the best binder for a given template is indeed identified. The following examples show how dynamic combinatorial chemistry is used to select substrates for a given receptor (casting), receptors for a given substrate (molding), and complementary components in a self-assembly process. A selection process that involves internal templation is the formation of the macrocyclic 15-mer discussed at the end of Section 5.6.2 (Figure 5.74).

5.7.2 Casting Casting usually involves the identification of a binder for an enzyme whose active site serves as a negative from which a cast is produced. Of the range of available reversible reactions, only relatively few are suitable in this context because the exchange has to proceed under conditions where the enzyme is stable. The acidic conditions required for hydrazone exchange, for example, are often incompatible with sensitive biomolecules. The most frequently used reactions are imine exchange, coupled with the subsequent reduction of the imines to amines, and disulfide exchange. An early study in this context, performed in the Lehn group, targeted the identification of an inhibitor for the enzyme carbonic anhydrase, which converts carbon dioxide into carbonic acid [134]. The three aldehydes and four amines chosen as building blocks in this case give rise to a dynamic library of twelve imines (Figure 5.86). Equilibrating this mixture together with sodium cyanoborohydride in

349

5.7 Dynamic combinatorial chemistry

Ar

CHO

+

R

H2N

Ar

Dynamic library containing 12 imines

– Ar

H2N

CHO =

R =

N

O

O OOC

CHO

NH 2

H2N

Ar

N H

R

Carbonic anhydrase –

O3S

NaCNBH3

R

CHO

H N

H2N

O S

CHO

– H N

H2N

O

O

COO

H2N

H2N

O O

O H2 N + Selected product

O

O S

NH 2

O S

NH 2

O O Hexyl 4-sulfamoylbenzoate

Figure 5.86: Selection of an inhibitor for carbonic anhydrase from a dynamic library of 12 imines that are reduced in the course of the reaction to yield the respective amines. Performing the equilibration in the presence of the known inhibitor hexyl 4-sulfamoylbenzoate reduces the amplification of the selected amine.

aqueous phosphate buffer at pH 6 in the absence of carbonic anhydrase affords a mixture of amines with a characteristic product distribution. When the same reaction is performed in the presence of the enzyme, the distribution changes, with the enzyme causing the amplification of an amine whose structure resembles that of a known inhibitor of carbonic anhydrase, namely hexyl 4-sulfamoylbenzoate. If this inhibitor is also present during the selection process, the amount of the selected amine decreases, demonstrating that product amplification is controlled by the enzyme’s active site. A second example is shown in Figure 5.87, involving a dynamic library of 21 interconverting disulfides containing monosaccharide residues [135]. A bis(mannoside) disulfide is amplified if the carbohydrate-binding protein concanavalin A is present during this exchange reaction, consistent with the selectivity of concanavalin A for oligosaccharides with terminal α-D-mannosyl residues. The scrambling is reasonably fast at physiological pH but stops when lowering the pH to 4. In these examples, the selection of the substrate and the freezing of the exchange reaction proceed independently. Both processes can be combined if the enzyme not only serves as template but also mediates the irreversible stabilization of the selected library components. An application of this concept is the reaction shown in Figure 5.88. Here, a library of nitroaldols is generated by equilibrating a

350

5 Assembling molecules

HO HO HO

O HO

O

S

X

S

X

OH O O

OH

O

X

S

S

X

O OH OH

O Dynamic library containing 21 disulfides

OH

OH

Concanavalin A HO HO HO

OH O

HO

OH

HO O

O

HO

O

O

HO

S

S

S 2

2

O

αMan(C3) HO HO HO

O

HO

HO O

OH

βGal(C3)

βGal(C2)

HO O

O

O

HO

HO O

HO HO

O

HO

S

O S

2

O

βGlc(C2)

O HO

S

2

2

O

2

O βXyl(C2)

βAra(C2)

Figure 5.87: Selection of a binder for concanavalin A from a dynamic library of 21 monosaccharidederived disulfides.

O O

CHO +

R

OH

NO 2

O O2N

O

NO 2

Major product (99% e.e.)

Cl NO 2

O

R

O

Dynamic libary containing 10 nitroaldols (five racemic products)

NO 2 Pseudominas cepacia lipase

F3C Minor product (99% e.e.)

O2N

CHO R

CHO

CHO

F

= F3C

Cl

Cl CHO

CHO

CHO Cl

Figure 5.88: Selection of enantiopure acetylated nitroaldols by equilibrating a dynamic library generated from five aromatic aldehydes and a nitroalkane in the presence of a lipase and an acyl donor.

5.7 Dynamic combinatorial chemistry

351

set of aldehydes with a nitroalkane in the presence of a base [136]. The products structurally differ in the residue originating from the aldehyde and also in the absolute configuration at the newly formed stereogenic center, with the selection process addressing both structural features. The structure of the product preferentially formed when performing the reaction in the presence of a lipase and an acyl donor therefore not only reflects which substitution pattern of the aldehyde is optimal for the interactions with the enzyme, but also which product enantiomer is most rapidly acetylated. This strategy thus adds a combinatorial element to the separation of enantiomers by dynamic kinetic resolution, which is why it has been termed dynamic combinatorial resolution.

5.7.3 Molding The majority of receptors in supramolecular chemistry is macrocyclic (Section 4.1), which is why the identification of receptors by dynamic combinatorial chemistry is often based on building blocks with functional groups that allow oligomerization followed by cyclization once the chain is long enough to form a strain-free ring. In addition, the building blocks usually feature structural elements of known receptors to increase the chances that the corresponding products interact with the substrate. Groundbreaking studies in this context were performed in the group of Jeremy K. M. Sanders. An example is compound 5.83 that contains an acetal group at the aromatic subunit and a hydrazide group at the opposite end of the chain [137]. After cleavage of the acetal, the released aldehydes can react with hydrazide groups of other molecules to form rings of different sizes, whose cyclically arranged aromatic subunits and carbonyl groups are expected to induce an affinity for cations. When treating 5.83 with trifluoroacetic acid in chloroform, the formation of a range of cyclic acyl hydrazones from the dimer to the 15-mer is indeed observed. Product distribution changes with time until the thermodynamic equilibrium is reached after two days, when only the cyclic dimer (88 %) and cyclic trimer (11 %) are still present. The entropically favored formation of the smaller ring is consistent with the principles of self-assembly described earlier. The subsequent addition of quaternary ammonium ions to this dynamic library induces a pronounced change of product distribution. N-Methylquinuclidinium iodide and acetylcholine chloride both shift the equilibrium in favor of the cyclic trimer, indicating that these cations template the formation of the macrocycle to which they bind. Performing the same reaction on a larger scale and interrupting the exchange by neutralizing the reaction mixture after the equilibrium is reached allows the isolation of the selected cyclic trimer. According to binding studies, this trimer indeed interacts with both templates. The higher stability of the acetylcholine complex (230 M−1) in comparison to the Nmethylquinuclidinium complex in CDCl3/CD3OD, 95:5 (v/v) (150 M−1) is consistent

352

5 Assembling molecules

with the fact that acetylcholine more efficiently shifts the equilibrium in the dynamic library (Figure 5.89).

N

O

H 3 CO H 3 CO

H N

N O

TFA CHCl 3

O O

N NH

NH 2

O

N H

N

O

HN N

N

O

N

O

O

HN N

+

O

O

N 5.83

N

NH

N

O Equilibrium state in the absence of template

88%

11%

Equilibrium state in the presence of

41%

56%

13%

86%

O

Equilibrium state in the presence of

N + + N

O

Figure 5.89: Molecular structure of building block 5.83 and structures of the corresponding cyclic dimer and trimer. The extents to which the library composition shift in the presence of N-methylquinuclidinium iodide and acetylcholine chloride are also specified.

A significantly more efficient acetylcholine receptor is obtained from building block 5.84 whose macrocyclization proceeds in a similar fashion as that of 5.83 [138]. In this case, the resulting library composition is complex, comprising mainly the cyclic dimer, trimer, and tetramer, with small amounts of the pentamer, hexamer, and higher oligomers also being present. The addition of acetylcholine causes the appearance of a product not present in the original library, which eventually comprises ca. 70% of the mixture. This receptor is made up of six subunits, which are, however, not part of a single ring, but of two interlocked cyclic trimers. The best receptor in this library, only virtually present in the absence of the template, is therefore a catenane that binds to acetylcholine in CHCl3/DMSO, 95:5 (v/v) with an impressive log Ka of 7.1. Due to the complexity of its structure, the exact binding mode with which this catenane interacts with acetylcholine is not entirely clear, although it is likely that the binding pocket is located between the two rings (Figure 5.90). Catenanes and related structures are the topic of Chapter 6. They have unusual structures and are often synthesized efficiently by using the concepts of supramolecular chemistry. Faced with the task of designing a receptor for acetylcholine, it is unlikely, however, that the first choice of a supramolecular chemist

353

5.7 Dynamic combinatorial chemistry

O

H N

N O

H N

O

O N

O

H N

N O

H 3CO

TFA CHCl 3/DMSO, 95:5 ( v/v)

O

N H

N HN

NH 2 HN

HN

3

5.84

N O Acetylcholine

N O

O

O O N N H N HN

O

NH N

NH

O N H

NH

O O

N

N O

O OCH

O

O O

N

H N N

O N

O

Catenane

Figure 5.90: Molecular structure of building block 5.84 and schematic structure of the catenane consisting of two interlocked cyclic trimers amplified by acetylcholine chloride.

would be to prepare a catenane. The above example therefore demonstrates the power of dynamic combinatorial chemistry, which delivers completely unexpected receptors if they are accessible in a dynamic library from the chosen building blocks and if their interaction with the template leads to a thermodynamically favored state [139]. Besides allowing the de novo development of receptors, dynamic combinatorial chemistry also permits assessing and potentially improving existing receptors. In Section 4.1.5 we saw that cyclophane 4.25 has a pronounced cation affinity in 10 mM borate buffer [140]. Is it possible to amplify a structurally related receptor from a dynamic library of interconverting macrocycles? Mixing the three thiols 5.85, 5.86, and 5.87 in water at pH 8–9 affords a library of interconverting cyclic disulfides in which 5.88, the closest analog of 4.25, is indeed present, but only in minor amounts (Figure 5.91) [141]. The overall composition of this library is complex, comprising more than 45 compounds also because 5.87 is chiral and the corresponding macrocycles are therefore formed as mixtures of stereoisomers. The addition of N-methylisoquinolinium iodide causes the amplification of one of these macrocycles (also as a mixture of stereoisomers) whose structure differs, however, from that of 5.88 by the number of subunits. This receptor binds N-methylisoquinolinium iodide in 10 mM borate buffer with a log Ka of 5.4. The corresponding complex of 4.25 has a log Ka of 7.2 under the same conditions [140], showing that 4.25 is still the better receptor for this cation, although 5.89 wins among the disulfide analogs.

354

5 Assembling molecules

COO

OOC

OOC

O

O

COO

S

S

S

S

S

S O

O OOC

S

S

OOC

COO

4.25

COO COO

5.88

OOC

S S

N 5.89

S S

S S

COO OOC HS

COO

N

SH

OOC

5.85 O OH

+ COO

S

COO

S

S S

OH OOC OOC

HS

5.86

SH

+ OOC

S

COO S

OOC

COO OOC

SH

S S

COO

HS

COO

RR

SS 5.87 (Racemate)

OOC

N S S

S S RR

OOC

COO

SS

S S

OOC

COO

Figure 5.91: Molecular structure of building blocks 5.85, 5.86, and 5.87 and of the disulfide analog of receptor 4.25, compound 5.88. The receptor selected by N-methylisoquinolinium iodide is 5.89. The structures of macrocyclic receptors amplified by other cationic templates are also shown.

5.7 Dynamic combinatorial chemistry

355

The amplification of 5.89 by N-methylisoquinolinium iodide in the dynamic library produced from 5.85, 5.86, and 5.87 is highly specific. Using other cationic templates such as a morphine derivative or tetramethylammonium iodide leads to the amplification of other macrocycles. In the case of the tetramethylammonium cation, the tetrameric receptor forms stereoselectively because only the selected diastereomer is able to fold into a compact structure with a cavity that is suitable to host the tetramethylammonium cation [142]. A final example of the molding strategy involves the orthoester-derived cryptands introduced by Max von Delius (Section 4.1.2). The synthesis of these compounds also proceeds under thermodynamic control and involves the use of template effects. The cryptand 5.90a is obtained, for example, by treating 1,1,1-trimethoxyethane with three equivalents of diethylene glycol in chloroform under the influence of catalytic amounts of trifluoroacetic acid and one equivalent of a sodium salt containing a large lipophilic anion (Figure 5.92) [143]. Under these conditions, the three alkoxy groups at the orthoester carbon are shuffled, giving rise to a library of different products. Of these products, 5.90a forms predominantly in 67% yield because this cryptand is thermodynamically stabilized by the sodium ion.

2

OMe OMe OMe

Na+ +

3 HO

OH

O

CDCl3, 5 mol% CF3COOH, 4 Å molecular sieves

O O

O

2

2

OMe OMe OMe

OMe OMe OMe

2 HO

K+

OH

O

+ 1 HO

O

O

1 HO

OH

O

O

O O

OH

CDCl3, 30 mol% CF3COOH, 5 Å molecular sieves

OMe OMe OMe

+

3 HO

O

O

O O O

O O

CDCl3, 30 mol% CF3COOH, 5 Å molecular sieves

O

O Rb+ O O O O O O O O O

OH

5.90b

O

Cs+ 2

5.90a

O

O

+ 2 HO

O O

K+

O

Rb+

OH

O

C2Cl4, 30 mol% C2F5COOH, 20 mbar

O

Na+ O O

O O

5.90c

O +

O

O Cs O O O O O O O

5.90d

Figure 5.92: Formation of orthoester-derived cryptands 5.90a-d from 1,1,1-trimethoxyethane and different diols in the presence of suitable metal ions as templates. Product formation correlates with the size of the metal if diethylene glycol and triethylene glycol are present in the required ratios.

356

5 Assembling molecules

Allowing orthoester exchange to proceed in the presence of a mixture of diols adds a combinatorial component to this synthesis [144]. In this case, cryptands of different sizes are formed, and the control over product formation depends on the template effects of suitable metal ions. When reacting 1,1,1-trimethoxyethane with two equivalents of diethylene glycol and one equivalent of triethylene glycol in the presence of a potassium salt, the cryptand 5.90b with a slightly larger cavity than 5.90a forms preferentially (Figure 5.92). Inverting the diethylene glycol and triethylene glycol ratio and using a rubidium salt as template affords 5.90c as the main product, and the largest cryptand 5.90d is obtained when reacting 1,1,1-trimethoxyethane with three equivalents of triethylene glycol in the presence of a cesium salt. Accordingly, product formation correlates with the ionic radius of the templating metal ion.

5.7.4 Self-assembly A remarkable system in which self-assembly is responsible for the amplification of members of a dynamic combinatorial library was reported by Sijbren Otto. This system is based on building block 5.91, consisting of a 1,3-benzenedithiol core that contains in 5-position a short pentapeptide chain with the sequence Gly-Leu-LysLeu-Lys (Figure 5.93) [145]. The alternating sequence of polar and apolar side +

O HS

H N

N H

O

O N H

+ NH 3

NH 3

H N O

O N H

COO

–

=

SH 5.91

(b) Shaking Cyclic trimer O2 borate buffer pH 8 +

Cyclic tetramer

(c) Stirring

(a) No agitation

Figure 5.93: Molecular structure of building block 5.91 and schematic illustration of the products formed in solutions that are left standing (a), or are shaken (b), or stirred (c).

5.7 Dynamic combinatorial chemistry

357

chains predisposes this peptide to aggregate into amyloid-like β-sheets. Hence, macrocycles derived from 5.91 by disulfide formation should be prone to stack. Which product exactly results when equilibrating 5.91 in borate buffer at pH 8 strongly depends on the conditions. Letting the solution stand preferentially affords cyclic trimers and tetramers, that is, the macrocycles that are normally formed when cyclizing 3-benzenedithiol derivatives under thermodynamic control. When the solution is shaken, the cyclic hexamer appears with time and is the dominating product after ca. 20 day. When the solution is stirred, the cyclic heptamer eventually dominates the equilibrium. The concentrations of both products increase exponentially with time, indicating that they autocatalytically mediate their own formation once present. 1,3-Benzenedithiol derivatives without the peripheral substituents or with substituents that are unable to form β-sheets do not exhibit this behavior, showing that the ability to self-assemble into tubular structures is crucial for the formation of the larger rings. The stability of the self-assembled tubes correlates with the number of peptide residues. Three or four peptides in the cyclic trimer and tetramer are insufficient to lead to stable aggregates, but the larger macrocycles with more side chains stack under the conditions of the experiments as demonstrated by the presence of fibrillar structures in cryo-transmission electron microscopy images of the corresponding solutions. Thus, if these large rings are formed by a random event, they are trapped by selfassembly. The formation of stacks alone does not explain the observed autocatalysis, however, since tubes can only grow at both ends, which should cause the concentrations of the larger macrocycles to increase linearly and not exponentially. Exponential self-replication only occurs if the stacks are continuously broken to create new ends to which further rings can attach. This breaking of the aggregates requires the solution to be agitated. If just left standing, stacks that might form in solution from larger macrocycles do not break and therefore grow too slowly to precipitate before the building blocks are again consumed in the exchange reaction. Mechanical agitation, however, induces breaking of the stacks, thus explaining autocatalysis. The force produced by shaking is just sufficient to break the hexamer stacks whose concentration therefore rises exponentially. Increasing the mechanical stress by stirring causes breaks also in the heptamer stacks, which are now produced preferentially because they self-assemble more efficiently. It is important to stress that the amplification of the large macrocycles is not a thermodynamic effect. Intrinsically, these macrocycles are less stable than the smaller ones as demonstrated by their absence in the nonagitated solutions. However, by kinetically trapping the large rings within the self-assembled stacks, they are prevented from further participating in the exchange reaction, demonstrating that kinetic products can result from a process in which all steps are reversible. The relevance of this work thus extends beyond dynamic combinatorial chemistry, showing that out-of-equilibrium states, which must have played a role in the evolution of life, can emerge spontaneously.

358

5 Assembling molecules

5.8 Systems chemistry Systems theory is an interdisciplinary scientific field that deals with the behavior of complex systems whose many components do not behave independently, but interact in multiple ways by following certain rules. Our world is full of such systems. Examples are the climate, the economy, the World Wide Web, or biological ecosystems. These systems are adaptive. Removing a food source from an ecosystem or adding a predator to it influences the network of the interacting species at many different levels, for example. Moreover, small changes at one level can have massive effects at another level. This feature is generally associated with the term “butterfly effect,” which implies that a drastic effect in one part of a complex system, for example a tornado in one part of the world, could have been triggered weeks earlier and in a completely different location by something as harmless as the flapping of the wings of a butterfly. Complex systems are intensively studied in many scientific disciplines, including biology, physics, mathematics, and computer science, but have long been ignored in chemistry. Why is that so? Are mixtures of molecules always messy?

One reason could be that the synthetic targets of synthetic chemistry generally need to be pure to allow their use as, for example, pharmaceuticals. Reactions that lead to mixtures of products are therefore usually avoided. Another factor is that analyzing complex mixtures of compounds is technically challenging and has only become possible more recently with the development of sufficiently sensitive analytical techniques. These developments had a strong impact on the rapid development of dynamic combinatorial chemistry, with the work carried out in this context soon showing that networks of interconverting compounds can react unexpectedly to an external stimulus. Just recall how widely the amplification factors vary when a certain amount of a template is added to a dynamic library and that it is not the best receptor that is necessarily amplified (Figure 5.85a). Another example is shown in Figure 5.94 [146]. This simulated dynamic library is again made up of building blocks that combine into 322 different oligomeric species ranging from dimers to tetramers. The affinities of the template toward the library members are attributed randomly from a log-normal distribution, so that most library members only have a weak affinity but a few members bind strongly. Figure 5.94a shows the result of one simulation, expectedly illustrating that the presence of the template leads to the amplification of the best binders. Conversely, the concentrations of most library members change much less strongly, with some slightly increasing in concentration while most others decrease in concentration. Certain combinations of binding affinities, however, cause the outcome of such a simulation to be drastically different. An example

359

5.8 Systems chemistry

(a)

(b) 10

AF

AF

10

1

0.1

–30

1

0.1

–20

–10

0

10

ΔG° (kJ mol–1)

–30

–20

–10

0

10

ΔG° (kJ mol–1)

Figure 5.94: Correlation between the amplification factor AF and the affinity for the template in terms of ΔG0 of the components of a simulated dynamic combinatorial library of which each member interacts with the template with a randomly assigned binding constant. The graph in (b) shows the emergence of patterns under certain conditions. The concentration of the template amounts to 1 mM in both cases.

is shown in Figure 5.94b, where the amplification factors cluster in lines. Patterns thus emerge in this library, which is a hallmark of complex systems. The reason for this pattern is that the incorporation of certain building blocks into the library members penalizes their amplification in the presence of the template by a relatively large and constant factor. The upper line in Figure 5.94b thus describes the amplification of the library members containing none of these building blocks, the members on the next line contain one of them, then two, and so on. Another and perhaps more intuitive example is the behavior of a dynamic library produced from five different building blocks A to E that can only form dimeric species. Compound AB should bind strongly to template T1 and CD should bind with the same affinity to template T2. Adding both templates to the dynamic library thus leads to the amplification of AB and CD, and since there are no building blocks left with which E can react, it must form the homodimer EE. As a consequence, the concentration of EE strongly increases upon the addition of T1 and T2, although this compound does not bind to any of the templates. The behavior of dynamic libraries thus could hold many surprises. One of the most fascinating aspects of complex systems is the emergence of properties that transcend those of the individual components, with the paramount example in this context undoubtedly being the emergence of life from simple molecules. We will never know exactly how life on earth came about because the exact conditions are unknown and therefore cannot be reproduced. It is, however, reasonable to assume that external conditions and the presence of networks of interacting molecules on the primitive earth favored the appearance of early complex chemical systems that exhibited some characteristics of life such as organization, replication, metabolism, and evolution. By studying how mixtures of molecules behave under various stimuli, we

360

5 Assembling molecules

can try to mimic such systems and learn about the fundamental mechanisms controlling their behavior. Thus, we could reach a point one day, where we understand much better how life began. The young research field dealing with work in this area is called systems chemistry [147]. It emerged from supramolecular chemistry and dynamic covalent chemistry and promises to deliver fascinating insights into the behavior and possible functions of complex molecular networks. Stay tuned!

Bibliography [1] [2] [3] [4]

[5]

[6]

[7]

[8] [9] [10] [11] [12] [13] [14]

[15] [16]

Fraenkel-Conrat H, Williams RC. Reconstitution of active tobacco mosaic virus from its inactive protein and nucleic acid components. Proc. Natl. Acad. Sci. U. S. A. 1955, 41, 690–8. Bhyravbhatla B, Watowich SJ, Caspar DL. Refined atomic model of the four-layer aggregate of the tobacco mosaic virus coat protein at 2.4-Å resolution. Biophys. J. 1998, 74, 604–15. Klug A. The tobacco mosaic virus particle: structure and assembly. Philos. Trans. R. Soc., B 1999, 354, 531–5. Montoro-García C, Camacho-García J, López-Pérez AM, Bilbao N, Romero-Pérez S, Mayoral MJ, González-Rodríguez D. High-fidelity noncovalent synthesis of hydrogen-bonded macrocyclic assemblies. Angew. Chem. Int. Ed. 2015, 54, 6780–4. Montoro-García C, Camacho-García J, López-Pérez AM, Mayoral MJ, Bilbao N, González-Rodríguez D. Role of the symmetry of multipoint hydrogen bonding on chelate cooperativity in supramolecular macrocyclization processes. Angew. Chem. Int. Ed. 2016, 55, 223–7. Rowan SJ, Hamilton DG, Brady PA, Sanders JKM. Automated recognition, sorting, and covalent self-assembly by predisposed building blocks in a mixture. J. Am. Chem. Soc. 1997, 119, 2578–9. Kramer R, Lehn JM, Marquis-Rigault A. Self-recognition in helicate self-assembly: spontaneous formation of helical metal complexes from mixtures of ligands and metal ions. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 5394–8. Lee SB, Hwang S, Chung DS, Yun H, Hong JI. Guest-induced reorganization of a selfassembled Pd(II) complex. Tetrahedron Lett. 1998, 39, 873–6. Wu A, Isaacs L. Self-sorting: the exception or the rule? J. Am. Chem. Soc. 2003, 125, 4831–5. He Z, Jiang W, Schalley CA. Integrative self-sorting: a versatile strategy for the construction of complex supramolecular architecture. Chem. Soc. Rev. 2015, 44, 779–89. Safont-Sempere MM, Fernández G, Würthner F. Self-sorting phenomena in complex supramolecular systems. Chem. Rev. 2011, 111, 5784–814. Zhang KD, Ajami D, Gavette JV, Rebek Jr. J. Alkyl groups fold to fit within a water-soluble cavitand. J. Am. Chem. Soc. 2014, 136, 5264–6. Jordan JH, Gibb BC. Molecular containers assembled through the hydrophobic effect. Chem. Soc. Rev. 2015, 44, 547–85. Hiraoka S, Nakamura T, Shiro M, Shionoya M. In-water truly monodisperse aggregation of gear-shaped amphiphiles based on hydrophobic surface engineering. J. Am. Chem. Soc. 2010, 132, 13223–5. Prins LJ, Reinhoudt DN, Timmerman P. Noncovalent synthesis using hydrogen bonding. Angew. Chem. Int. Ed. 2001, 40, 2382–426. Ducharme Y, Wuest JD. Use of hydrogen bonds to control molecular aggregation. Extensive, self-complementary arrays of donors and acceptors. J. Org. Chem. 1988, 53, 5787–9.

Bibliography

[17] [18]

[19]

[20] [21]

[22]

[23]

[24] [25]

[26]

[27] [28]

[29]

[30] [31] [32]

[33]

[34] [35]

361

Sessler JL, Wang R. Design, synthesis, and self-assembly of “artificial dinucleotide duplexes”. J. Org. Chem.1998, 63, 4079–91. Marsh A, Silvestri M, Lehn JM. Self-complementary hydrogen bonding heterocycles designed for the enforced self-assembly into supramolecular macrocycles. Chem. Commun. 1996, 1527–8. Asadi A, Patrick BO, Perrin DM. G^C Quartet – a DNA-inspired Janus-GC heterocycle: synthesis, structural analysis, and self-organization. J. Am. Chem. Soc. 2008, 130, 12860–1. Original image taken by an anonymous photographer and published under the Public Domain Mark 1.0 license. Marsh A, Nolen EG, Gardinier KM, Lehn JM. Janus molecules: synthesis of double-headed heterocycles containing two identical hydrogen bonding arrays. Tetrahedron Letters 1994, 35, 397–400. Whitesides GM, Simanek EE, Mathias JP, Seto CT, Chin DN, Mammen M, Gordon DM. Noncovalent synthesis: using physical-organic chemistry to make aggregates. Acc. Chem. Res. 1995, 28, 37–44. Zerkowski JA, Seto CT, Whitesides GM. Solid-state structures of rosette and crinkled tape motifs derived from the cyanuric acid melamine lattice. J. Am. Chem. Soc. 1992, 114, 5473–75. Seto CT, Whitesides GM. Self-assembly based on the cyanuric acid-melamine lattice. J. Am. Chem. Soc. 1990, 112, 6409–11. Seto CT, Whitesides GM. Molecular self-assembly through hydrogen bonding: supramolecular aggregates based on the cyanuric acid-melamine lattice. J. Am. Chem. Soc. 1993, 115, 905–16. Seto CT, Whitesides GM. Synthesis characterization and thermodynamic analysis of a 1 + 1 self-assembling structure based on the cyanuric acid.melamine lattice. J. Am. Chem. Soc. 1993, 115, 1330–40. Prins LJ, Timmerman P, Reinhoudt DN. Non-covalent synthesis of organic nanostructures. Pure Appl. Chem. 1998, 70, 1459–68. Timmerman P, Vreekamp RH, Hulst R, Verboom W, Reinhoudt DN, Rissanen K, Udachin KA, Ripmeester J. Noncovalent assembly of functional groups on calix[4]arene molecular boxes. Chem. Eur. J. 1997, 3, 1823–32. Prins LJ, Jolliffe KA, Hulst R, Timmerman P, Reinhoudt DN. Control of structural isomerism in noncovalent hydrogen-bonded assemblies using peripheral chiral information. J. Am. Chem. Soc. 2000, 122, 3617–27. Prins LJ, Huskens J, de Jong F, Timmerman P, Reinhoudt DN. Complete asymmetric induction of supramolecular chirality in a hydrogen-bonded assembly. Nature 1999, 398, 498–502. Prins LJ, Huskens J, de Jong F, Timmerman P, Reinhoudt DN. An enantiomerically pure hydrogen-bonded assembly. Nature 2000, 408, 181–4. Kerckhoffs JMCA, van Leeuwen FWB, Spek AL, Kooijman H, Crego-Calama M, Reinhoudt DN. Regulatory strategies in the complexation and release of a noncovalent guest trimer by a self-assembled molecular cage. Angew. Chem. Int. Ed. 2003, 46, 5717–22. Kerckhoffs JMCA, Mateos-Timoneda MA, Reinhoudt DN, Crego-Calama M. Dynamic combinatorial libraries based on hydrogen-bonded molecular boxes. Chem. Eur. J. 2006, 13, 2377–85. Conn MM, Rebek Jr. J. Self-assembling capsules. Chem. Rev. 1997, 97, 1647–68. Wyler R, de Mendoza J, Rebek Jr. J. A synthetic cavity assembles through self‐complementary hydrogen bonds. Angew. Chem. Int. Ed. Engl. 1993, 32, 1699–701.

362

5 Assembling molecules

[36] Szabo T, Hilmersson G, Rebek Jr. J. Dynamics of assembly and guest exchange in the tennis ball. J. Am. Chem. Soc. 1998, 120, 6193–4. [37] Kang J, Rebek Jr. J. Entropically driven binding in a self-assembling molecular capsule. Nature 1996, 382, 239–41. [38] Grotzfeld RM, Branda N, Rebek Jr. J. Reversible Encapsulation of disc-shaped guests by a synthetic, self-assembled host. Science 1996, 271, 487–9. [39] Mecozzi S, Rebek Jr. J. The 55 % solution: a formula for molecular recognition in the liquid state. Chem. Eur. J. 1998, 4, 1016–22. [40] Martı́n T, Obst U, Rebek Jr. J. Molecular assembly and encapsulation directed by hydrogenbonding preferences and the filling of space. Science 1998, 281, 1842–45. [41] Rebek Jr. J. Simultaneous encapsulation: molecules held at close range. Angew. Chem. Int. Ed. 2005, 44, 2068–78. [42] Heinz T, Rudkevich DM, Rebek Jr. J. Pairwise selection of guests in a cylindrical molecular capsule of nanometre dimensions. Nature 1998, 394, 764–6. [43] Ajami D, Theodorakopoulos G, Petsalakis ID, Rebek Jr. J. Social Isomers of picolines in a small space. Chem. Eur. J. 2013, 19, 17092–6. [44] Shivanyuk A, Rebek Jr. J. Social isomers in encapsulation complexes. J. Am. Chem. Soc. 2002, 124, 12074–5. [45] Ajami D, Liu L, Rebek Jr. J. Soft templates in encapsulation complexes. Chem. Soc. Rev. 2015, 44, 490–9. [46] Ajami D, Rebek Jr. J. Compressed alkanes in reversible encapsulation complexes. Nat. Chem. 2009, 1, 87–90. [47] Ajami D, Rebek Jr. J. Longer guests drive the reversible assembly of hyperextended capsules. Angew. Chem. Int. Ed. 2007, 46, 9283–86. [48] Rebek Jr. J. Host-guest chemistry of calixarene capsules. Chem. Commun. 2000, 637–43. [49] Mogck O, Paulus EF, Böhmer W, Thondorf I, Vogt W. Hydrogen-bonded dimers of tetraurea calix[4]arenes: unambiguous proof by single crystal X-ray analysis. Chem. Commun. 1996, 2533–4. [50] Shimuzu KD, Rebek Jr. J. Synthesis and assembly of self-complementary calix[4]arenes. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 12403–7. [51] Mogck O, Böhmer V, Vogt W. Hydrogen bonded homo- and heterodimers of tetra urea derivatives of calix[4]arenes. Tetrahedron 1996, 52, 8489–96. [52] Castellano RK, Kim BH, Rebek Jr. J. Chiral capsules: asymmetric binding in calixarene-based dimers. J. Am. Chem. Soc. 1997, 119, 12671–2. [53] Corbellini F, Fiammengo R, Timmerman P, Crego-Calama M, Versluis K, Heck AJR, Luyten I, Reinhoudt DN. Guest encapsulation and self-assembly of molecular capsules in polar solvents via multiple ionic interactions. J. Am. Chem. Soc. 2002, 124, 6569–75. [54] Mansikkamäki H, Nissinen M, Schalley CA, Rissanen K. New. J. Chem. 2003, 27, 88–97. [55] MacGillivray LR, Atwood JL. A chiral spherical molecular assembly held together by 60 hydrogen bonds. Nature 1997, 389, 469–72. [56] Atwood JL, Barbour LJ, Jerga A. Chem. Commun. 2001, 2376–7. [57] Kiesilä A, Beyeh NK, Moilanen JO, Puttreddy R, Götz S, Rissanen K, Barran P, Lützen A, Kalenius E. Thermodynamically driven self-assembly of pyridinearene to hexameric capsules. Org. Biomol. Chem. 2019, 17, 6980–4. [58] Evan-Salem T, Baruch I, Avram L, Cohen Y, Palmer LC, Rebek Jr. J. Resorcinarenes are hexameric capsules in solution. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 12296–300. [59] Sakai N, Matile S. Synthetic multifunctional pores: lessons from rigid-rod β-barrels. Chem. Commun. 2003, 2514–23.

Bibliography

363

[60] De Santis P, Morosetti S, Rizzo R. Conformational analysis of regular enantiomeric sequences. Macromolecules, 1974, 7, 52–8. [61] Bong DT, Clark TD, Granja JR, Ghadiri MR. Self-assembling organic nanotubes. Angew. Chem. Int. Ed. 2001, 40, 988–1011. [62] Ghadiri MR, Kobayashi K, Granja JR, Chadha RK, McRee DE. The structural and thermodynamic basis for the formation of self‐assembled peptide nanotubes. Angew. Chem. Int. Ed. Engl. 1995, 34, 93–5. [63] Brea RJ, Reiriz C, Granja JR. Towards functional bionanomaterials based on self-assembling cyclic peptide nanotubes. Chem. Soc. Rev. 2010, 39, 1448–56. [64] Montenegro J, Ghadiri MR, Granja JR. Ion channel models based on self-assembling cyclic peptide nanotubes. Acc. Chem. Res. 2013, 46, 2955–65. [65] Amorín M, Castedo L, Granja JR. New cyclic peptide assemblies with hydrophobic cavities: the structural and thermodynamic basis of a new class of peptide nanotubes. J. Am. Chem. Soc. 2003, 125, 2844–45. [66] Calvelo M, Lamas A, Guerra A, Amorín M, Garcia-Fandino R, Granja JR. Parallel versus antiparallel β-sheet structure in cyclic peptide hybrids containing γ- or δ-cyclic amino acids. Chem. Eur. J. 2020, 26, 5846–58. [67] Rodríguez-Vázquez N, Amorín M, Granja JR. Recent advances in controlling the internal and external properties of self-assembling cyclic peptide nanotubes and dimers. Org. Biomol. Chem. 2017, 15, 4490–505. [68] Clark TD, Buehler LK, Ghadiri MR. Self-assembling cyclic β3-peptide nanotubes as artificial transmembrane ion channels. J. Am. Chem. Soc. 1998, 120, 651–6. [69] Shimizu LS, Salpage SR, Korous AA. Functional materials from self-assembled bis-urea macrocycles. Acc. Chem. Res. 2014, 47, 2116–27. [70] Massena CJ, Wageling NB, Decato DA, Rodriguez EM, Rose AM, Berryman OB. A halogenbond-induced triple helicate encapsulates iodide. Angew. Chem. Int. Ed. 2016, 55, 12398–402. [71] Dumele O, Trapp N, Diederich F. Halogen bonding molecular capsules. Angew. Chem. Int. Ed. 2015, 54, 12339–44. [72] Chakrabarty R, Mukherjee PS, Stang PJ. Supramolecular coordination: self-assembly of finite two- and three-dimensional ensembles. Chem. Rev. 2011, 111, 6810–918. [73] Fujita M, Umemoto K, Yoshizawa M, Fujita N, Kusukawaa T, Biradha K. Molecular paneling via coordination. Chem. Commun. 2001, 509–18. [74] Caulder DL, Raymond KN. Supermolecules by design. Acc. Chem. Res. 1999, 32, 975–82. [75] Lehn JM, Rigault A, Siegel J, Harrowfield J, Chevrier B, Moras D. Spontaneous assembly of double-stranded helicates from oligobipyridine ligands and copper(I) cations: structure of an inorganic double helix. Proc. Natl. Acad. Sci. U.S.A. 1997, 84, 2565–9. [76] Albrecht M. “Let’s twist again” – Double-stranded, triple-stranded, and circular helicates. Chem. Rev. 2001, 101, 3457–98. [77] Fatin-Rouge N, Blanc S, Pfeil A, Rigault A, Albrecht-Gary AM, Lehn JM. Self-assembly of tricuprous double helicates: thermodynamics, kinetics, and mechanism. Helv. Chim. Acta 2001, 84, 1694–711. [78] Hamacek J, Borkovec M, Piguet C. Simple thermodynamics for unravelling sophisticated self-assembly processes. Dalton Trans. 2006, 1473–90. [79] Krämer R, Lehn JM, De Cian A, Fischer J. Self-assembly, structure, and spontaneous resolution of a trinuclear triple helix from an oligobipyridine ligand and NIII ions. Amgew. Chem. Int. Ed. Engl. 1993, 32, 703–6.

364

5 Assembling molecules

[80] Hasenknopf B, Lehn JM, Boumediene N, Leize E, Van Dorsselaer A. Kinetic and thermodynamic control in self-assembly: sequential formation of linear and circular helicates. Angew. Chem. Int. Ed. 1998, 37, 3265–8. [81] Hasenknopf B, Lehn JM, Boumediene N, Dupont-Gervais A, Van Dorsselaer A, Kneisel B, Fenske D. Self-assembly of tetra- and hexanuclear circular helicates. J. Am. Chem. Soc. 1997, 119, 10956–62. [82] Zarges W, Hall J, Lehn JM, Bolm C. Helicity Induction in helicate self-organisation from chiral tris(bipyridine) ligand strands. Helv. Chim. Acta 1991, 74, 1843–52. [83] Kramer R, Lehn JM, Marquis-Rigault A. Self-recognition in helicate self-assembly: spontaneous formation of helical metal complexes from mixtures of ligands and metal ions. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 5394–8. [84] Caulder DL, Raymond KN. Supramolecular self-recognition and self-assembly in gallium(III) catecholamide triple helices. Angew. Chem. Int. Ed. Engl. 1997, 36, 1440–2. [85] Youinou T, Rahmouni N, Fischer F, Osborn JA. Self-Assembly of a Cu4 complex with coplanar copper(I) ions: synthesis, structure, and electrochemical properties. Angew. Chem. Int. Ed. Engl. 1992, 31, 733–5. [86] Baxter PNW, Lehn JM, Fischer J, Youinou MT. Self-assembly and structure of a 3 × 3 inorganic grid from nine silver ions and six ligand components. Angew. Chem. Int. Ed. Engl. 1994, 33, 2284–7. [87] Baxter PNW, Lehn JM, Kneisel BO, Fenske D. Multicomponent self-assembly: preferential generation of a rectangular [2×3]G grid by mixed-ligand recognition. Angew. Chem. Int. Ed. Engl. 1997, 36, 1978–81. [88] Baxter PNW, Lehn JM, Baum G, Fenske D. Self-assembly and structure of interconverting multinuclear inorganic arrays: a [4×5]-AgI20 grid and an AgI10 quadruple helicate. Chem. Eur. J. 2000, 6, 4510–7. [89] Swiegers GF, Malefetse TJ. New self-assembled structural motifs in coordination chemistry. Chem. Rev. 2000, 100, 3483–538. [90] Fujita M, Yazaki J, Ogura K. Preparation of a macrocyclic polynuclear complex, [(en)Pd(4,4ʹbpy)]4(NO3)8, which recognizes an organic molecule in aqueous media. J. Am. Chem. Soc. 1990, 112, 5645–7. [91] Fujita M. Metal-directed self-assembly of two- and three-dimensional synthetic receptors. Chem. Soc. Rev. 1998, 27, 417–25. [92] Saha ML, Neogi S, Schmittel M. Dynamic heteroleptic metal-phenanthroline complexes: from structure to function. Dalton Trans. 2014, 43, 3815–34. [93] Schmittel M, Mahata K. A fully dynamic five-component triangle via self-sorting. Chem. Commun. 2010, 46, 4163–5. [94] Mahata K, Schmittel M. From 2-fold completive to integrative self-sorting: a five-component supramolecular trapezoid. J. Am. Chem. Soc. 2009, 131, 16544–54. [95] Saha ML, Mittal N, Bats JW, Schmittel M. A six-component metallosupramolecular pentagon via self-sorting. Chem. Commun. 2014, 50, 12189–92. [96] MacGillivray LR, Atwood JL. Structural classification and general principles for the design of spherical molecular hosts. Angew. Chem. Int. Ed. 1999, 38, 1018–33. [97] Fochi F, Jacopozzi P, Wegelius E, Rissanen K, Cozzini P, Marastoni E, Fisicaro E, Manini P, Fokkens R, Dalcanale E. Self-assembly and anion encapsulation properties of cavitand-based coordination cages. J. Am. Chem. Soc. 2001, 123, 7539–52. [98] Fujita M, Tominaga M, Hori A, Therrien B. Coordination assemblies from a Pd(II)-cornered square complex. Acc. Chem. Res. 2005, 38, 371–80.

Bibliography

365

[99] Tashiro S, Tominaga M, Yamaguchi Y, Kato K, Fujita M. Folding a de novo designed peptide into an α-helix through hydrophobic binding by a bowl-shaped host. Angew. Chem. Int. Ed. 2006, 45, 241–4. [100] Kumazawa K, Biradha K, Kusukawa T, Okano T, Fujita M. Multicomponent assembly of a pyrazine-pillared coordination cage that selectively binds planar guests by intercalation. Angew. Chem. Int. Ed. 2009, 48, 3418–38. [101] Yoshizawa M, Klosterman JK, Fujita M. Functional molecular flasks: new properties and reactions within discrete, self-assembled hosts. Angew. Chem. Int. Ed. 2003, 42, 3909–13. [102] Han M, Engelhard DM, Clever GH. Self-assembled coordination cages based on bananashaped ligands. Chem. Soc. Rev. 2014, 43, 1848–60. [103] Yoshizawa M, Catti L. Bent anthracene dimers as versatile building blocks for supramolecular capsules. Acc. Chem. Res. 2019, 52, 2392–404. [104] Harris K, Fujita D, Fujita M. Giant hollow MnL2n spherical complexes: structure, functionalisation and applications. Chem. Commun. 2013, 49, 6703–12. [105] Fujita D, Suzuki K, Sato S, Yagi-Utsumi M, Yamaguchi Y, Mizuno N, Kumasaka T, Takata M, Noda M, Uchiyama S, Kato K, Fujita M. Protein encapsulation within synthetic molecular hosts. Nat. Commun. 2012, 3, 1093. [106] Pluth MD, Bergman RG, Raymond KN. Proton-mediated chemistry and catalysis in a selfassembled supramolecular host. Acc. Chem. Res. 2009, 42, 1650–9. [107] Pluth MD, Bergman RG, Raymond KN. Making amines strong bases: thermodynamic stabilization of protonated guests in a highly-charged supramolecular host. J. Am. Chem. Soc. 2007, 129, 11459–67. [108] Nitschke JR. Construction substitution and sorting of metallo-organic structures via subcomponent self-assembly. Acc. Chem. Res. 2007, 40, 103–12. [109] Ronson TK, Zarra S, Black SP, Nitschke JR. Metal-organic container molecules through subcomponent self-assembly. Chem. Commun. 2013, 49, 2476–90. [110] Mal P, Breiner B, Rissanen K, Nitschke JR. White phosphorus is air-stable within a selfassembled tetrahedral capsule. Science 2009, 324, 1697–9. [111] Rowan SJ, Cantrill SJ, Cousins GRL, Sanders JKM, Stoddart JF. Dynamic covalent chemistry. Angew. Chem. Int. Ed. 2002, 41, 898–952. [112] Rowan SJ, Sanders JKM. Macrocycles derived from cinchona alkaloids: a thermodynamic vs kinetic study. J. Org. Chem. 1998, 63, 1536–46. [113] Gugger PA, Hockless DCR, Swiegers GF, Wild SB. Acid-catalyzed rearrangements between cyclic oligomers of 3,4-dihydro-2H-1,5-benzodiazocine. Crystal and molecular structures of mono- and dinuclear copper(II) complexes of tetraaza (16-membered), hexaaza (24membered), and octaaza (32-membered) macrocycles. Inorg. Chem. 1994, 33, 5671–7. [114] Hockless DCR, Lindoy LF, Swiegers GF, Wild SB. Acid-catalysed rearrangements between 3,4dihydro-2H-1,5-benzooxazocine and -benzothiazocine and their macrocyclic (16-membered, 24-membered and 32-membered) oligomers. J. Chem. Soc., Perkin Trans. 1 1998, 117–22. [115] Kwit M, Grajewski J, Skowronek P, Zgorzelak M, Gawroński J. One-step construction of the shape persistent, chiral but symmetrical polyimine macrocycles. Chem. Rec. 2019, 19, 213–37. [116] Zhang Y, Zheng X, Cao N, Yang C, Li H. A kinetically stable macrocycle self-assembled in water. Org. Lett. 2018, 20, 2356–9. [117] Christinat N, Scopelliti R, Severin K. Multicomponent assembly of boronic acid based macrocycles and cages. Angew. Chem. Int. Ed. 2008, 47, 1848–52. [118] Içli B, Christinat N, Tönnemann J, Schüttler C, Scopelliti R, Severin K. Synthesis of molecular nanostructures by multicomponent condensation reactions in a ball mill. J. Am. Chem. Soc. 2009, 131, 3154–5.

366

5 Assembling molecules

[119] Liu B, Pappas CG, Zangrando E, Demitri N, Chmielewski PJ, Otto S. Complex molecules that fold like proteins can emerge spontaneously. J. Am. Chem. Soc. 2019, 141, 1685–9. [120] Harding CJ, Lu Q, Malone JF, Marrs DJ, Martin N, McKee V, Nelson J. Hydrolytically-sensitive hexaimino and hydrolytically-inert octaamino-cryptand hosts for dicopper. J. Chem. Soc., Dalton Trans., 1995, 1739–47. [121] Acharyya K, Mukherjee PS. Organic imine cages: molecular marriage and applications. Angew. Chem. Int. Ed. 2019, 58, 8640–53. [122] Stuart SR, Rowan J, Pease AR, Cram DJ, Stoddart JF. Dynamic hemicarcerands and hemicarceplexes. Org. Lett. 2000, 21, 2411–4. [123] Liu X, Warmuth R. Solvent effects in thermodynamically controlled multicomponent nanocage syntheses. J. Am. Chem. Soc. 2006, 128, 14120–14127. [124] Xu D, Warmuth R. Edge-directed dynamic covalent synthesis of a chiral nanocube. J. Am. Chem. Soc. 2008, 130, 7520–1. [125] Tozawa T, Jones JTA, Swamy SI, Jiang S, Adams DJ, Shakespeare S, Clowes R, Bradshaw D, Hasell T, Chong SY, Tang C, Thompson S, Parker J, Trewin A, Bacsa J, Slawin AMZ, Steiner A, Cooper AI. Porous organic cages. Nat. Mat. 2009, 8, 973–8. [126] Mastalerz M. One-pot synthesis of a shape-persistent endo-functionalised nano-sized adamantoid compound. Chem. Commun. 2008, 4756–8. [127] Giri N, Del Pópolo MG, Melaugh G, Greenaway RL, Rätzke K, Koschine T, Pison L, Costa Gomes MF, Cooper AI, James SL. Liquids with permanent porosity. Nature 2015, 527, 216–20. [128] Klotzbach S, Scherpf T, Beuerle F. Dynamic covalent assembly of tribenzotriquinacenes into molecular cubes. Chem. Commun. 2014, 50, 12454–7. [129] Beuerle F, Gole B. Covalent organic frameworks and cage compounds: design and applications of polymeric and discrete organic scaffolds. Angew. Chem. Int. Ed. 2018, 57, 4850–78. [130] Lehn JM. Dynamic combinatorial chemistry and virtual combinatorial libraries. Chem. Eur. J. 1999, 5, 2455–63. [131] Corbett PT, Leclaire J, Vial L, West KR, Wietor JL, Sanders JKM, Otto S. Dynamic combinatorial chemistry. Chem. Rev. 2006, 106, 3652–3711. [132] Severin K. The advantage of being virtual – target-induced adaption and selection in dynamic combinatorial libraries. Chem. Eur. J. 2004, 10, 2565–80. [133] Corbett PT, Otto S, Sanders JKM. Correlation between host-guest binding and host amplification in simulated dynamic combinatorial libraries. Chem. Eur. J. 2004, 10, 3139–43. [134] Huc I, Lehn JM. Virtual combinatorial libraries: dynamic generation of molecular and supramolecular diversity by self-assembly. Proc. Natl. Acad. Sci. 1997, 94, 2106–10. [135] Ramström O. Lehn JM. In situ generation and screening of a dynamic combinatorial carbohydrate library against concanavalin A. ChemBioChem 2000, 1, 41–8. [136] Vongvilai P, Angelin M, Larsson R, Ramström O. Dynamic Combinatorial resolution: direct asymmetric lipase-mediated screening of a dynamic nitroaldol library. Angew. Chem. Int. Ed. 2007, 46, 948–50. [137] Cousins GRL, Furlan RLE, Ng YF, Redman JE, Sanders JKM. Identification and isolation of a receptor for N-methyl alkylammonium salts: molecular amplification in a pseudo-peptide dynamic combinatorial library. Angew. Chem. Int. Ed. 2001, 40, 423–8. [138] Lam RTS, Belenguer A, Roberts SL, Naumann C, Jarrosson T, Otto S, Sanders JKM. Amplification of acetylcholine-binding catenanes from dynamic combinatorial libraries. Science 2005, 308, 667–9.

Bibliography

367

[139] Cougnon FBL, Sanders JKM. Evolution of dynamic combinatorial chemistry. Acc. Chem. Res. 2012, 45, 2211–21. [140] Kearney PC, Mizoue LS, Kumpf RA, Forman JE, McCurdy A, Dougherty DA. Molecular recognition in aqueous media. New binding studies provide further insights into the cation–π interaction and related phenomena. J. Am. Chem. Soc. 1993, 115, 9907–19. [141] Otto S, Furlan RLE, Sanders JKM. Selection and amplification of hosts from dynamic combinatorial libraries of macrocyclic disulfides. Science 2002, 297, 590–3. [142] Corbett PT, Tong LH, Sanders JKM, Otto S. Diastereoselective amplification of an induced-fit receptor from a dynamic combinatorial library. J. Am. Chem. Soc. 2005, 127, 8902–3. [143] Brachvogel RC, Hampel H, von Delius M. Self-assembly of dynamic orthoester cryptates. Nat. Commun. 2015, 6, 7129. [144] Shyshov O, Brachvogel RC, Bachmann T, Srikantharajah R, Segets D, Hampel F, Puchta R, von Delius M. Adaptive Behavior of dynamic orthoester cryptands. Angew. Chem. Int. Ed. 2017, 56, 776–81. [145] Carnall JMA, Waudby CA, Belenguer AM, Stuart MCA, Peyralans JJP, Otto S. Mechanosensitive self-replication driven by self-organization. Science 2010, 327, 1502–6. [146] Corbett PT, Sanders JKM, Otto S. Systems chemistry: pattern formation in random dynamic combinatorial libraries. Angew. Chem. Int. Ed. 2007, 46, 8858–61. [147] Ludlow RF, Otto S. Systems chemistry. Chem. Soc. Rev. 2008, 37, 101–8.

6 Threading molecules CONSPECTUS: Have you ever wondered whether it is possible to make a knot in a molecule, to connect cyclic molecules like chain links, or to thread them onto a linear molecule like pearls on a necklace? In this chapter we will see that all of this can be done rather efficiently by using the concepts of supramolecular chemistry. In the first part of the chapter we focus on the structural and stereochemical aspects of knots and molecules containing two or more interlocked components. We then look at general strategies to prepare such entangled molecules and subsequently at a number of specific examples of syntheses, classified according to the noncovalent interactions that mediate product formation. The topics discussed in these contexts serve as the basis for the next chapter that shows how mechanically interlocked molecules are used to create functional systems.

6.1 Molecular topology Molecules are structurally characterized at different levels. The most fundamental parameter is the composition, which specifies how many atoms of which type a molecule contains. The constitution of a molecule then describes how the atoms are connected and the configuration how they are oriented in space. Closely associated with these terms is the concept of isomerism: two molecules are structural (or constitutional) isomers if they have the same composition but differ in the connectivity of the atoms. Configurational isomers, on the other hand, have the same composition and connectivity but differ in how the atoms are spatially arranged. While these concepts are frequently used, structurally differentiating molecules on the basis of their topology is less common. What is the difference between topology and structure?

Topology is a term that has its origin in mathematics where it refers to the geometric properties of three-dimensional objects. The central aspect relevant in the context discussed here is that topologically equivalent objects can be transformed into each other by continuous deformations such as stretching, twisting, crumpling, or bending. Accordingly, an empty crumpled balloon is topologically equivalent to a filled one, but even objects that at first sight have different shapes can be topologically equivalent. An example is a doughnut and the crown depicted on the cover of this book. When transferring the concept of topology to molecules, one must allow them to be totally flexible as if the atoms were connected by infinitely stretchable rubber bands. The molecules may thus be deformed into any shape by changing the distance of the atoms or their arrangement in space as long as the connectivity remains the same. Whether two molecules are topologically equivalent is then determined by https://doi.org/10.1515/9783110595611-006

370

6 Threading molecules

constructing graphs with all atoms residing in the same plane. If no bonds cross in these graphs they are planar, while graphs containing crossings are not. The number of crossings determines the topological properties of the respective molecule, with molecules that have the same number of crossings being topologically equivalent. Let us perform this analysis for the molecules depicted in Figure 6.1. The chair conformation of cyclohexane can be easily drawn as a regular hexagon in a twodimensional projection in which no bonds cross. Bicyclo[2.2.2]octane can be similarly flattened, while buckminsterfullerene is more difficult to draw in a two-dimensional fashion. A way to do this involves pulling a five-membered ring widely apart with the concomitant flattening of the whole molecule. The resulting graph also does not contain crossed bonds, rendering all three molecules, structurally diverse as they may initially seem, topologically equivalent.

Threedimensional representation

Twodimensional projection

Figure 6.1: Analysis of the topological properties of cyclohexane, bicyclo[2.2.2]octane, and buckminsterfullerene.

The crossing of bonds in such graphs is unavoidable in some of the interlocked molecules discussed in this chapter, of which examples are shown in Figure 6.2. The structure in Figure 6.2a is a catenane made up of two interlocked macrocycles that are connected without being linked by an actual bond. The schematic illustration of the two interlocked rings clearly shows that they cross twice, and this molecule can therefore only be drawn as a nonplanar graph. The structure in Figure 6.2b is a single macrocycle whose folding into a trefoil knot leads to three crossings, thus also affording a nonplanar graph when it is drawn in a two-dimensional fashion. The rotaxane in Figure 6.2c consists of a ring through which a linear compound, the socalled axle, is threaded. Bulky end groups at the ends of the axle prevent the ring from escaping. In spite of the interlocking of the two components, this compound is

6.1 Molecular topology

371

topologically equivalent to those in Figure 6.1 because the ring is allowed to be indefinitely flexible when applying the concepts of topology so that it can slip over the end groups of the axle. A rotaxane can thus be drawn as a planar graph of the two separated components. All three molecules shown in Figure 6.2 are therefore topologically distinct with the rotaxane having the same topological properties as the compounds in Figure 6.1, while the catenane and the trefoil knot have two and three crossings, respectively.

(a)

O O O N N

O

O

O

N

O

O O O O

N

O

(b)

O

O

O O

O

O

O N

N

N

N

N

N

N N O

O O

O O

(c)

O

O

RO

OR O

O N

N

N

N

R=

O

O O

O O

O

Figure 6.2: Examples of a catenane (a), a knot (b), and a rotaxane (c) together with the corresponding schematic illustration of these mechanically interlocked compounds.

372

6 Threading molecules

Let us now integrate the concept of topology into the structural classification of molecules with which this chapter began [1, 2]. At the first level of structural classification, the topology of a molecule is defined by the connectivity of the atoms. If two molecules have a different composition or constitution, they are never topologically identical because there is no way they can be interconverted by mere deformations. Different compounds can, however, be topologically equivalent. Consider in this context a cycloalkane with the composition (CH2)40, a ring with only half the number of carbon atoms (CH2)20 and a catenane containing two of the smaller macrocycles (Figure 6.3a). The large and the small ring have different topologies because they differ in composition but they are topologically equivalent because they can both be represented as planar graphs. The large macrocycle and the catenane are constitutional isomers whose topologies not only differ because of the different connectivity of the atoms but also because one molecule can be represented as a planar graph while the other cannot. Moving on to stereoisomers, we realize that most stereoisomers we learned to distinguish in basic stereochemistry classes are in fact topologically equivalent. The Eand Z-isomer of an olefin, for example, or the two enantiomers of a chiral compound can mostly be represented in the form of planar graphs, rendering them indistinguishable in terms of topology. Topological isomerism only emerges when two isomers cannot be converted into each other by bond deformations because bonds cross. An example is the macrocycle in Figure 6.3b and the corresponding trefoil knot. These compounds are topological diastereomers. The chirality of the knot moreover causes

(a)

(CH2)40

(CH2)20

(b)

[(CH2)20]2

(c)

(CH2)40

(CH2)40

Topological diastereomers

(CH2)40

(CH2)40

Topological enantiomers

Figure 6.3: Classification of topological isomers. The structural isomers in (a) have different topologies, independent of whether they are topologically equivalent or not. The ring and the knot in (b) are topological diastereomers, while the two knots in (c) are topological enantiomers.

6.1 Molecular topology

373

it to exist in two enantiomeric forms (Figure 6.3c), which have the same number of crossings in their nonplanar graphs but cannot be interconverted just by bond deformations, rendering them topologically distinct. Describing molecules in terms of topology therefore requires a carefully structural analysis. It is important to emphasize in this context that the term topology is not a synonym for structure and its use in this sense can lead to wrong conclusions. Moreover, stating that two different molecules have different topologies is not wrong, but if topological aspects are irrelevant for the discussion, using the term topology may not be necessary and can even be confusing. What is a mechanical bond?

Disentangling the interlocked structures shown in Figure 6.2 requires the breaking of bonds. In the case of the catenane, the cleavage of one ring affords a pseudorotaxane in which an acyclic component is threaded through an intact ring. Since the acyclic subunit does not contain bulky end groups, it is easily dethreaded, affording the individual components. To separate the components of the rotaxane, one could either cleave the ring or the axle, while the knot is unknotted by opening the ring, followed by a conformational reorganization of the resulting acyclic product. The structural integrity of these molecules is therefore closely associated with the structural integrity of their components. If the latter is lost, the entanglement is also no longer guaranteed. The entanglement is therefore merely a mechanical effect which does not necessarily require any special interactions in the regions of the molecules where the crossings occur. It nevertheless holds the components of interlocked molecules tightly together and therefore qualifies as a bond, the so-called mechanical bond. Such a mechanical bond does not involve orbital or electrostatic interactions, but still leads to a substantial degree of stabilization because it only disappears after covalent bonds have been broken [3, 4]. The structures in Figure 6.2 are examples of large and structurally diverse families of mechanically interlocked molecules. All three subgroups come in different flavors. Catenanes, for instance, can be composed of two or more rings of widely differing structures and sizes. Rotaxanes can also contain more than one axle and/ or ring. How many components are actually connected is usually specified in square brackets in front of the family name. Accordingly, a [3]catenane consists of three interlocked rings. Knots are characterized not only by the number of components but also by the number of crossings in the planar graph. The combination of these parameters leads to numerous theoretically possible knots, which are treated mathematically in an overarching theory. Of these knots, several have been realized in a molecular version, not only the trefoil knot shown in Figure 6.2 but also more complex knots as we will see later.

374

6 Threading molecules

First ideas to synthesize mechanically interlocked molecules date back to the first half of the twentieth century [5]. Initial work was mainly curiosity driven, likely motivated by questions similar to those posed in the outline of this chapter. Only later it turned out that, as so often, Nature had solved the problem already. There are a number of biological processes that involve catenated or knotted single- or double-stranded DNA, including DNA recombination and replication [6]. A selection of structures demonstrating the structural versatility of interlocked DNA is shown in Figure 6.4. Related knotted structures have also been identified in proteins where they serve to improve (thermal) stability or to precisely arrange several subunits in space [7]. In some cases, the entanglements in biomolecules translate into complex functions like in ATP synthase, an archetypical natural molecular machine. Mechanically interlocked molecules are therefore more than just laboratory curiosities and with the efficient strategies with which they are available today, devising functional systems has become possible. The most important concepts used for the preparation of such molecules are presented in the following parts of this chapter.

Trefoil 3

Trefoil 3

Torus 5

Trefoil 3

Trefoil 3

Granny 6

Figure 6.4: Examples of knotted DNA structures. The images are electron micrographs of knots formed by treating a circular DNA with 13S condensin, a protein involved in the structural maintenance of chromosomes, and a topoisomerase. The number of crossings and the type of knot are indicated below each image. Images adapted with permission from [8]. Copyright Elsevier, 1999.

6.2 Synthetic strategies 6.2.1 Molecular strategies The first report of a successful catenane synthesis was published in 1960 by Edel Wasserman [9] (although the validity of Wasserman’s claims has been questioned [10]). The idea behind this synthesis was that a mixture of an acyclic and a macrocyclic molecule should transiently contain interlocked species, comprising the acyclic component threaded through the cyclic one. If this pseudorotaxane could be trapped by macrocyclization of the acyclic component, the corresponding [2]catenane should result. Wasserman realized this concept by first synthesizing the 34-membered

375

6.2 Synthetic strategies

deuterated cycloalkane 6.1 from the long-chain α,ω-dicarboxylic acid ester 6.2 by acyloin condensation and subsequent Clemmensen reduction using Zn/DCl (Figure 6.5). 6.1 and 6.2 were then mixed and subjected to the conditions of another acyloin condensation. After the chromatographic removal of the alkane 6.1, the obtained macrocyclic acyloin 6.3 was characterized by IR spectroscopy. The corresponding spectrum provided evidence for the presence of C–D bonds, indicating that the isolated material not only contained 6.3 but also the catenane 6.4. Indeed, the treatment of this material with hydrogen peroxide afforded a small amount (ca. 1%) of 6.1 whose presence was attributed to the cleavage of the acyloin-containing ring in 6.4 followed by dethreading of the diacid 6.5. Wasserman later reported the isolation of a few milligrams of 6.4 in a 0.0001% yield [11].

O O

O

D

1. Na/xylene 2. Zn/DCI

D

D

D

O

O O 32 Na/xylene

D

O O

O

6.2

6.1

O

HO

HO +

D D

HO

O 32

OH

6.5 (ca.99%)

H2O2/OH–

6.4

6.3 O

D D D

D D

+

D D D

6.1 (ca.1%)

Figure 6.5: Synthesis of catenane 6.4 by using a statistical approach that involves the acyloin condensation of the diester 6.2 in the presence of the deuterated macrocycle 6.1. The synthesis started with the preparation of 6.1. The proof of the presence of the catenane was based on the release of 6.1 upon treatment of the product mixture with hydrogen peroxide.

376

6 Threading molecules

Independent of whether Wasserman’s approach was actually successful or not, it suggests that catenanes should be accessible from preformed macrocycles by threading an acyclic molecule through the annulus and then cyclizing it. Statistical approaches based on the probability (or hope) of capturing pseudorotaxanes in solution are, however, clearly inefficient if the components are not held together prior to forming the mechanical bond. Based on this understanding, Gottfried Schill and Arthur Lüttringhaus developed a number of template-directed syntheses of mechanically interlocked molecules in which the preorganization of the subunits was achieved by covalent bonds that eventually had to be cleaved in order to release the free interlocked components [10]. Figure 6.6 shows the central steps of their first [2]catenane synthesis, described in 1964 [12], to illustrate the strategy used to achieve the interlocking of the two rings. This synthesis starts with the preparation of the ansa-catechol derivative 6.6 in a linear ten-step sequence. This catechol is condensed with an aliphatic dichloroketone to obtain the ketal 6.7 in which the sp3 hybridization of the ketal carbon atom causes the two side chains to be oriented orthogonally to the plane of the aromatic ring. As a consequence, the double ansacompound 6.8, formed from 6.7 by nitration, reduction, and two intramolecular nucleophilic substitutions, already has the topology of the final catenane with the rings still covalently connected. The separation of both rings involves cleaving the ketal group, oxidizing the resulting catechol, and hydrolyzing the enamine to afford 6.9. The further transformation of 6.9 leads to 6.10 in which only the acetylated amino group provides evidence for the covalent template that served to direct catenane formation. The corresponding substituted catechole ring was cleaved by ozonolysis of a hydroxybenzoquinone-containing intermediate [13]. Overall, the reaction sequence involves 26 steps. An overall yield was not reported. By using conceptually similar approaches, often based on the use of ketals as covalent templates, Schill and Lüttringhaus independently continued pursuing the synthesis of mechanically interlocked molecules with remarkable success [10]. They were the first to prepare a [2]catenane consisting of two cycloalkanes, a [3]catenane, a rotaxane, and a precursor for a knot. With these elegant syntheses, they laid much of the conceptual groundwork in the field and inspired modern approaches to using covalent templates for the synthesis of mechanically interlocked molecules [14]. Despite these pioneering achievements, mechanically interlocked molecules only became the popular study objects they represent today when supramolecular syntheses became available. The main reasons are that supramolecular syntheses are typically much shorter than the methods described by Schill and Lüttringhaus, afford the products in larger quantities, and allow more facile structural variations. The products moreover usually contain built-in binding sites that conveniently allow the development of functional systems such as those described in Chapter 7.

377

6.2 Synthetic strategies

Cl HO HO

CHO

O Cl

HO

10 steps

O

Cl 12

12

O

HO CHO 6.6

6.7

Cl

Cl

O

1. Nitration 2. Reduction

O

O

Double macrocyclization

NH 2

O

N

6.8

Cl

O

O HO

O

Hydrolysis

Oxidation N

HO

N

O

O

O

HO 1. Hydrolysis 2. Tautomerization

O

HO NH

O

4 Steps

O NAc

O

6.9 O O Ozonolysis

O

3 Steps NAc

O

NAc

6.10

Figure 6.6: First catenane synthesis described by Lüttringhaus and Schill in which the ketal 6.7 serves as a covalent template to direct the topology of the precatenane 6.8.

Before we come to these syntheses, we look at another and conceptually different approach to access mechanically interlocked molecules. This approach is based on the peculiar properties of Möbius strips. Such strips are easily obtained from a rectangular band of paper by gluing the two narrow ends together to form a loop.

378

6 Threading molecules

Depending on how often the band is twisted by 180° before fixing the loop, strips with different topologies result. No twist affords a simple ribbon that has an outer and an inner surface (Figure 6.7a). Bisecting this band leads to two rings with the same diameter as the original one. A Möbius strip is obtained by performing one half-twist prior to connecting the ends (Figure 6.7b). In this strip, every point on the surface can be reached without having to cross the edges. Its bisection leads to a single ring with twice the diameter of the rings obtained by bisecting the untwisted ribbon. A strip with two half-twists affords a catenane after bisection (Figure 6.7c), and one with three half-twists a trefoil knot (Figure 6.7d).

(a)

(b)

(c)

(d)

Figure 6.7: Structures of strips differing in the number of half-twists and products resulting from these strips upon bisection. The bands have no (a), one (b), two (c), and three (d) half twists. The band in (b) is a Möbius strip.

Molecules with similar topologies are therefore attractive precursors for accessing interlocked systems. This concept was pursued by David M. Walba who based his approach on the polycyclic crown ether 6.11 (Figure 6.8) [15]. The macrocyclization of this compound affords two products in an almost 1:1 ratio, one representing the nontwisted ribbon 6.12 and the other the Möbius ladder 6.13 (the latter as a pair of enantiomers). Cleaving the double bonds in these ladders by ozonolysis gives two equivalents of the small macrocycle 6.14 in the case of 6.12 and one equivalent of 6.15 in the case of 6.13. The doubly twisted strip that should give rise to the catenane could, unfortunately, not be detected in the reaction mixture, presumably because it would be too strained.

6.2 Synthetic strategies

HO

O

O

O

O

O

O

O

O

OTs

HO

O

O

O

O

O

O

O

O

OTs

379

6.11 Double macrocyclization

O O O

O

O

O

O

O O O

O

O

+

O

O

O

ca. 1:1

O

O

O

O

O

O

O

O

O

O

O3 O O

O

O O

O O

O

O

O

6.13

O

O

O

O

O3 O

O

O

O

6.12

O

O O

OO O

O

O

O O

OO

O

O

O

O

O

O O

O

Enantiomers

O

O O

O

O

O

O

O

O

O

O

O

O

O O

O

O O

O

O

O

O

O O O

O

O

O O

O 6.14

O O O

O

O O

O

O

O

6.15

Figure 6.8: Attempted synthesis of a catenane by using the Möbius approach. The ribbon 6.12 and the racemate of the Möbius ladder 6.13 were formed in the macrocyclization of 6.11 in almost equal amounts and could be transformed into the corresponding rings by ozonolysis. The doubly twisted product was not observed.

380

6 Threading molecules

6.2.2 Supramolecular strategies Supramolecular syntheses of mechanically interlocked molecules are based on noncovalent or reversible coordinative interactions that hold the precursors together or fix them in an interlocked conformation. Accordingly, such syntheses require the presence of suitable subunits in the building blocks that interact intramolecularly or intermolecularly with an external template. If these interactions cause the targeted interlocked product to become thermodynamically more stable than potential other products, the synthesis can be performed under thermodynamic control by using reversible reactions such as those discussed in Section 5.6. An example is the acetylcholine-mediated catenane synthesis described already in Section 5.7.3. Similar thermodynamically controlled syntheses also allow the preparation of rotaxanes and have proven particularly useful to access complex knotted molecular architectures as we will see. Alternatively, suitable interactions facilitate the formation of the correct bonds on the way from the precursor to the mechanically interlocked product by appropriately arranging the reacting groups in space. The reaction leading to the desired product thus becomes faster than other competing reaction pathways. Conceptually, this approach is closely related to template-directed crown ether syntheses (Section 5.1) or the catenane synthesis shown in Figure 6.6. In the latter case, the arrangement of the two side chains in the intermediate 6.7 is responsible for the formation of a product with the topology of a [2]catenane. The covalent template in 6.7 thus preorganizes the side chains, kinetically favoring the formation of the desired product and disfavoring that of unwanted side products. Supramolecular catenane syntheses rely on similar strategies but make use of noncovalent or reversible coordinative interactions for the preorganization of the precursors instead of covalent bonds. The actual bond forming reactions are irreversible and preorganization therefore lowers the Gibbs free energy of activation associated with the formation of the desired product. Possible approaches are shown in Figure 6.9. Route (a) in Figure 6.9 involves arranging two acyclic precursors in an orthogonal fashion, which is achieved, for example, by coordinating appropriate divalent ligands to a metal ion that favors a tetrahedral coordination geometry (Section 6.3.1). The cyclization of both precursors and, if necessary, the subsequent removal of the template then leads to the [2]catenane. Alternatively, catenane formation can also start from a cyclic and an acyclic precursor, both of which form a pseudorotaxane (route (b)). Intermolecular interactions between the two pseudorotaxane components stabilize the complex and therefore favor catenane formation. Finally, it is also possible that the first ring mediates the formation of the second interlocked ring (route (c)). This strategy underlies the active metal template synthesis of catenanes discussed in Section 6.3.1.

6.2 Synthetic strategies

381

Route (a)

Route (b)

Route (c)

Figure 6.9: Schematic illustration of synthetic strategies that give rise to [2]catenanes, all relying on the proper preorganization of the precursors by reversible interactions.

Syntheses of knots require the subunits in the corresponding precursors to stabilize a knotted conformation that is ultimately fixed by closing the ring (Figure 6.10). The formation of a trefoil knot, for example, involves the use of a linear oligomer with subunits along the chain whose intramolecular interactions or interactions with suitable external templates induce a conformation in which the chain ends pass through two loops. Connecting these ends stabilizes the knot (route (a)). Alternatively, a trefoil knot is also obtained when connecting proper chain ends in a double helicate in which the two precursors cross three times (route (b)). Structurally, more complex knots are prepared by using similar strategies. Thermodynamically controlled reactions are useful if the products contain several intricately interlocked components.

Route (a)

Route (b) =

Figure 6.10: Possible routes for the synthesis of a trefoil knot, starting from precursors whose conformations and mutual orientations are stabilized by suitable interactions. That the two central structures have the same topology is illustrated in Figure 6.19.

382

6 Threading molecules

Rotaxanes are synthesized in a number of ways. The most straightforward approach is to slip the ring over the fully formed axle that already contains the two stoppers at the ends (“slipping”) (route (a) in Figure 6.11). This strategy requires a fine-tuned size match of the ring and the stopper groups so that the activation energy required

Route (a) "slipping"

Route (b) "clipping"

Route (c) "stoppering"

Route (d) "trapping"

Figure 6.11: Rotaxane syntheses involving the slipping of the ring over the stoppers (a), the clipping of a ring around a preformed axle (b), the stoppering of the axle of a pseudorotaxane (c), and the capture (or trapping) of the two components of the axle by the ring (d).

for the ring to bypass the stoppers is just overcome under certain conditions, for example at elevated temperatures. At lower temperatures, the product must be kinetically inert. No bonds are formed in this synthesis, which is why it proceeds under thermodynamic control, with the yield correlating with the extent to which the rotaxane is more stable than the free components under the slipping conditions. A variation of this approach is the use of a large ring whose diameter allows it to easily slip over the end groups of the axle. A chemical reaction that reduces the effective diameter of the ring then leads to the stabilization of the rotaxane (“shrinking”). Conversely, a rotaxane with small stopper groups that do not significantly impair the movement of the ring is stabilized by enlarging the size of the end groups (“swelling”). The most frequently used approaches to preparing rotaxanes rely on the covalent synthesis of one rotaxane component in the presence of the preformed other component. The formation of the ring in the presence of the axle is termed “clipping” (route (b) in Figure 6.11). In this case, the axle templates the macrocyclization of the acyclic

6.3 Syntheses using metal coordination

383

precursor by stabilizing a conformation in which the end groups come into close proximity. The two other synthetic strategies involve the assembly of the axle, which is either achieved by attaching the bulky end groups to the chains ends of the acyclic component in a pseudorotaxane (“stoppering”) (route (c) in Figure 6.11) or by joining the two dumbbell halves of the axle (“trapping”) (route (d) in Figure 6.11). In the latter approach, the ring usually plays an active role in bond formation and thus remains trapped in the product. All of these kinetically controlled approaches require bond formation to be irreversible.

6.3 Syntheses using metal coordination 6.3.1 Catenanes Metal–ligand interactions formed the basis of the first successful supramolecular synthesis of a [2]catenane, and have since then be used in numerous other syntheses of mechanically interlocked molecules. Due to their covalent character, these interactions tightly hold the precursors together in a well-defined arrangement, similarly to the covalent bonds used by Schill and Lüttringhaus for the synthesis of interlocked molecules. However, while covalent templates require additional synthetic steps to separate the components, the reversibility of metal-ligand interactions renders the removal of the metal template far easier. In 1980, Jean-Pierre Sauvage described the first example of a metal-directed catenane synthesis [16]. The corresponding short communication, published in the journal Tetrahedron Letters and written in French, represents a landmark in the field, demonstrating for the first time how mechanically interlocked molecules can be efficiently prepared in only a few steps by using the concepts of supramolecular chemistry. Sauvage subsequently developed many other interlocked molecules of increasing structural complexity, thereby establishing his reputation as the “master of chemical topology” [17] and laying the foundation for the Nobel Prize in Chemistry that he received in 2016 (Section 7.1). Conceptually, Sauvage’s catenane synthesis relies on the orthogonal arrangement of two ligands in the complex of a metal that prefers a tetrahedral coordination geometry (Figure 6.12). By covalently bridging the end groups of each ligand via linkers of suitable length, a metal complex with the topology of a [2]catenane results. To indicate that the transition metal is still an intrinsic part of the structure, Sauvage proposed the term catenate for such a complex. Removing the templating metal ion affords the corresponding catenand. The same catenand also results when starting from a preformed ring through which an acyclic precursor is threaded upon metal coordination. Since four bonds are formed in the former so-called entwining approach and only two in the alternative threading strategy, the latter usually proceeds more efficiently.

384

6 Threading molecules

Entwining

Threading

+

2

M

M Catenate 2 equiv

M

1 equiv M

M

M

Demetalation

Catenand

Figure 6.12: Catenane syntheses involving the use of metal templates. In the entwining strategy, two ligands are arranged by the metal ion in an orthogonal fashion. Bridging their end groups with linkers of suitable length affords a catenate, which is transferred into the corresponding catenand by the removal of the metal template. The alternative threading strategy proceeds via a metal complex in which a linear ligand is threaded through a cyclic one. In this strategy, only two covalent bonds are formed in the second step instead of four in the entwining strategy.

The actual catenane synthesis described by Sauvage is shown in Figure 6.13. Accordingly, catenate formation involves Williamson ether syntheses between the ligand 6.16 and the ethylene glycol derived diiodide 6.17. The respective catenate 6.18 is isolated in 27% yield when using the entwining strategy and a remarkable 42% when using the alternative and clearly more efficient threading strategy [18]. Note that the addition of 1 equiv of a copper(I) source to a 1:1 mixture of 6.16 and the corresponding crown ether in the threading strategy leads to the exclusive formation of the heteroleptic complex with two different ligands. The reason is that the macrocyclic component alone is unable to form a complex with copper(I) in which two ligand molecules are arranged in an orthogonal fashion. Phenanthroline 6.16 can form such a homoleptic complex but its formation only requires half of the available copper(I) ions. The rest can bind to the phenanthroline units in the macrocyclic component but remain coordinatively unsaturated. The formation of the homoleptic complex thus violates the principle of maximum site occupancy. As a consequence, the formation of the heteroleptic complex leads to the thermodynamically most favorable situation. We have seen another strategy to induce the formation of heteroleptic complexes in Section 5.5.4.

6.3 Syntheses using metal coordination

O

O

OH O N N

O O

O I

N

O

Cs2CO3/DMF

N

O

6.17

I

6.16

385

O O

OH Cu+

Cu+ OH

O

HO

O

HO N Cu+ N

N

N

N

N

Cu+

O

N N

O

HO

HO OH

27%

O O

42% O

Cs2CO3/DMF

O

O I

O 6.17

I

O O

O

O 6.18

N

O O

N

O

O

N Cu+ N

O

O

O O NBu4CN

NBu4[Cu(CN)4] O

N

O O O O O

N

O

6.19

O O O O

N N

O

Figure 6.13: [2]Catenane syntheses developed by Sauvage that involve the entwining of the ligands in the copper(I) complex of 6.16 upon reaction with the diiodide 6.17 (a), or the macrocyclization of the acyclic ligand in the copper(I) complex containing a preformed ring (b). The yields of the corresponding Williamson ether syntheses show that the threading strategy is more efficient. The crystal structures demonstrate the interlocked topologies of 6.18 and 6.19. Hydrogen atoms are omitted in the crystal structures for reasons of clarity.

386

6 Threading molecules

The treatment of 6.18 with an excess of cyanide salts that compete in copper(I)phenanthroline coordination by inducing the formation of the stable [Cu(CN)4]3− complex causes demetalation and the release of the two interlocked rings. The synthesis of catenand 6.19 thus involves either a three (entwining) or four (threading) step reaction sequence, both of which afford the product in substantial absolute amounts. The crystal structures in Figure 6.13 unambiguously demonstrate the interlocked structures of 6.18 and 6.19. While the orientation of the two phenanthroline units is fixed by the coordination to the metal ion in 6.18, 6.19 exhibits a co-conformation in the solid state in which the bulky phenanthroline units are oriented away from each other, likely for steric and electronic (repulsion of the nitrogen lone pairs) reasons. What is a co-conformation?

The term co-conformation used in the previous paragraph was introduced by J. Fraser Stoddart to describe the relative spatial arrangement of different components in an interlocked molecule [4]. Co-conformations are interconvertible by translation, pirouetting, or rocking motions. In the case of 6.19, the conversion of the co-conformation found in the crystal structure into the less favorable coconformation with two facing phenanthroline units is achieved, for example, by changing the relative orientations of both rings. Besides crystal structures, other important methods to structurally characterize mechanically interlocked molecules are NMR spectroscopy and mass spectrometry. The spatial proximity of the components of an interlocked molecule, which is enforced by the mechanical bond, produces characteristic effects on the proton resonances, for example. 1H NMR spectroscopy thus allows assigning the preferred coconformation in solution. Temperature-dependent NMR measurements are moreover useful to characterize dynamic processes such as the rate at which different co-conformations interconvert. In the case of mass spectrometry, interlocked structures give rise to characteristic fragmentation patterns. The chemical ionization (CI) mass spectrum of catenand 6.19 contains a peak at an m/z ratio of 1133, for example, which corresponds to the mass of the monoprotonated form of 6.19. An isomer of 6.19 consisting of a single macrocycle should exhibit in its mass spectrum the same peak along with a range of other peaks at lower m/z ratios corresponding to ions formed after ring opening and fragmentation of the linear product. In the case of the catenane, however, the opening of one ring causes the separation of the two components. As a consequence, the next smaller fragment in the mass spectrum of 6.19 has the m/z ratio of a single ring and additional signals appear only at smaller m/z ratios, accounting for further fragmentation. The separation of the components of catenanes induced by fragmentation thus leads to characteristic fingerprints in their mass spectra that allow differentiating them from structural isomers lacking the mechanical bond.

387

6.3 Syntheses using metal coordination

By adapting the strategy shown in Figure 6.12 to other metal complexes, catenanes of widely varying structures and compositions are accessible. The cyclodimerization of the linear component in a metal complex containing an acyclic ligand threaded through a cyclic one gives rise to a [3]catenane after demetalation, for example (Figure 6.14a). This concept was realized by using the copper(I) complex 6.20 in which a 1,10-diphenylphenanthroline derivative containing two terminal alkyne groups is threaded through the same macrocyclic ligand that is also used in the synthesis of 6.19 [19]. Treating 6.20 under the conditions of a Glaser coupling (CuCl, CuCl2, air) leads to the formation of catenate 6.21 containing three interlocked rings (Figure 6.14b). The remarkable efficiency of this reaction, which results in an overall yield of 58%, allows the isolation of 6.21 in gram quantities. As a side product, the [4] catenane is formed by cyclotrimerization in a yield of 22%. 6.21 can again be transformed into the corresponding catenand by treatment with a cyanide salt.

(a) 2

M

M

M M

(b)

O

O

O

H

O

O

N

O

N

CuCl / CuCl2 air / DMF

N

O

N

O

N

58% O

O

N

N

Cu +

Cu + N

N O

O H

O

O O

6.21 O O

N

O

O

O

N

N N

O

O

6.20

KCN

O

O

O

Cu + N

O

O

O

N N

O

O

O O

O

O

O

N N

O O

O

N N

O O

Figure 6.14: Strategy for the synthesis of a [3]catenane by cyclodimerization of the acyclic component in a metal complex comprising a cyclic and an acyclic ligand (a) and realization of this strategy by linking two equiv of the copper(I) complex 6.20 by means of a Glaser coupling (b).

Mechanically interlocked molecules are also accessible by making use of other metal complexes than those between copper(I) ions and bidentate ligands as long as the metals induce the appropriate arrangement of the ligands. We have seen in Section 5.5.1 about metal-directed self-assembly that an orthogonal arrangement of

388

6 Threading molecules

two ligands also results when coordinating tridentate ligands to metals that prefers an octahedral coordination geometry. Alternatively, suitable linear or square planar metal complexes also allow constructing precursors for catenanes. A [2]catenane such as 6.19 is termed Hopf link in knot theory. It represents the simplest link containing two interlocked rings. More complex links are the Solomon link comprising four crossings between two rings, the Star of David catenane comprising six crossings between two rings, or Borromean rings comprising six crossings between three rings (Figure 6.15). These interlocked structures are also accessible from appropriate coordination complexes. We look at the synthesis of a Solomon link in the next chapter because it is closely related to the synthesis of a trefoil knot that will be discussed there. Synthetic strategies to access the other catenanes are outlined in the paragraphs that follow.

(a)

(b)

(c)

(d)

Figure 6.15: Schematic illustrations of a Hopf link (a), a Solomon link (b), a Star of David catenane (c), and of Borromean rings (d).

A Star of David catenane consists of two rings that are similarly entwined as the two triangles in the flag of the state of Israel. David A. Leigh showed that the synthesis of such a catenane can be based on the circular triple helicate discussed in Section 5.5.2 (Figure 5.47) [20]. When using the tris(bipyridine) 6.22 as ligand, a hexanuclear circular helicate results in the presence of iron(II) sulfate, in which pairs of terminal alkene groups are oriented at the six corners (Figure 6.16). These double bonds come sufficiently close to allow covalently connecting them by ringclosing metathesis, thus affording the respective catenate in up to 92% yield. The crystal structure of the product confirms the expected topology. By removing the metal ions as their ethylenediaminetetraacetate (EDTA) complex, the metal-free form of this catenate is obtained. Borromean rings are links made up of three rings that are interlocked without being catenated. Thus, opening one ring leaves the remaining two unconnected. Historically, Borromean rings have been used as symbols in many different contexts. The name derives from a family of merchants and bankers in Renaissance Italy, the Borromeo family, whose crest features three similarly entangled rings.

6.3 Syntheses using metal coordination

N

N

N

N

N

6.22 Fe2+

389

N

as FeSO4

N N N Fe2+ N N

N

N

N N Fe2+ N N

N N N Fe2+ N

N

N N 2+ N N Fe N

N

N

N

N

N Fe2+ N

N

N

N

N N N Fe2+ N N

N

Figure 6.16: Molecular structures of ligand 6.22 and of the circular triple iron(II) helicate derived from 6.22, and crystal structure of the product obtained by treating 6.22 under the conditions of a ring-closing metathesis. Protons in the crystal structure are omitted for reasons of clarity.

Several attempts to synthesize such links remained unsuccessful before the Stoddart group showed that Borromean rings can be prepared in a surprisingly simple fashion in a one-step procedure by mixing equimolar amounts of the 2,2ʹbipyridine derived diamine 6.23, 2,6-diformylpyridine, and zinc(II) acetate in isopropanol (Figure 6.17) [21]. These components react by imine formation and by coordination of the zinc(II) ions to the available nitrogen donors. Since both of these reactions are reversible, product formation proceeds under thermodynamic control in contrast to several other syntheses discussed so far. The rules of self-assembly derived in Section 5.1 thus control the formation of the thermodynamically favored product, which happens to be in this case a zinc complex containing three interlocked macrocyclic ligands as confirmed by a crystal structure. This synthesis, which affords 6.24 in almost quantitative yield, is in fact so reliable that it can even be performed in undergraduate teaching laboratories [22]. The reduction of the imine groups and the removal of the zinc(II) ions in the form of their EDTA complexes give rise to the metalfree product.

390

6 Threading molecules

N

N

ni6.23

+

O

O

H2N

NH2 N

OHC

CHO

Zn2+

N

N

N

tri-

as Zn(OAc)2

N N

Zn2+

N

N N

O O

O O

N

N

O

NO

N N

O

N Zn2+

tas

N

Zn2+ Zn2+ N N N O N N O N O O

unitas

N N

6.24

O N N N Zn2+ Zn2+ N N N N

Figure 6.17: Self-assembly of ligand 6.23, 2,6-diformylpyridine, and zinc(II) acetate to afford the Borromean rings 6.24 and crystal structure of 6.24. Protons in the crystal structure are omitted for reasons of clarity and the three rings are shown in different colors to illustrate their arrangement. The icon shows the symbol of the Christian Trinity that also has the topology of Borromean rings [23].

6.3.2 Knots Trefoil knots are obtained by arranging an acyclic oligomeric ligand around a metal center in such a way that the coordination geometry induces a knotted topology of the ligand chain. This concept was pioneered by Christopher A. Hunter who showed that ligand 6.25 is predisposed to adopt the form of an open trefoil knot when coordinating to a zinc(II) ion (Figure 6.18a) [24]. This structure is stabilized by linking the chain ends through esterification or ring-closing metathesis [25]. Certain tris(2,6-pyridinedicarboxamides) form similar open knot complexes with lanthanide ions [26]. One of these complexes was used by David A. Leigh to tie a knot into the axle of a rotaxane (Figure 6.18b) [27]. The respective synthesis makes use of 6.26 as the axle, which contains a central dibenzylamine unit flanked by a bulky tetraphenyl moiety and a chain containing three 2,6-pyridinedicarboxamide

6.3 Syntheses using metal coordination

(a)

HO

O

N

N

N

N

391

O

O

6.25 O N

(b)

O

O

O

N NH

HN

NH

OH

N

O

O

HN

NH

N

O HN

6.26 tBuO

O

O

O

O

O

O

O

N NH

O

O

O

N NH

HN

O

O NH

HN

O

O

N NN

O O

O

O

O O

O

O

O

O

O

O

O

N

+ N H2

O

HN O O + O N H 2 O O O O

O tBuO

O

O

O

O

O

O H N O

O

Lu3+ O

HN O O

O

O

N NN

O

Lu3+

H N O

N

O N

O

O

O

HN

O

N

O O

NH

as Lu(CH3SO3 )3

NH

OtBu N NN

O O + O N O H2 O O O

O O

O

O

Figure 6.18: Molecular structure of the oligomeric ligand 6.25 containing three 2,2ʹ-bipyridine units that folds around a zinc(II) ion in the form of an open trefoil knot (a), and rotaxane formation that involves stoppering the axle by tying a knot (b). The axle 6.26 of this rotaxane contains a tris(2,6pyridinedicarboxamide) unit at one end, a dibenzylamine unit in the center, and a tetraphenylmethane moiety at the other chain end. The rotaxane is formed by threading the tris (2,6-pyridinedicarboxamide) subunit through dibenzo-24-crown-8, followed by tying it into a knot upon metal coordination.

subunits. This axle forms a pseudorotaxane with dibenzo-24-crown-8 in which the crown ether ring preferentially resides close to the ammonium group of the dibenzylammonium unit with which it interacts by hydrogen bonding. Since the tetraphenyl

392

6 Threading molecules

moiety is too large to slip through the ring, pseudorotaxane formation must involve the threading of the three 2,6-pyridinedicarboxamide units. The subsequent addition of Lu(CH3SO3)3 to a solution of this pseudorotaxane induces this end of the axle to fold into a knot whose steric bulk prevents the ring from leaving the axle. Untying the knot by removing the metal ion causes the release of the ring. This synthesis impressively illustrates the extent of structural control that is possible by cleverly combining suitable structural motifs and interactions. Seminal contributions to the metal-templated synthesis of molecular knots also came from the Sauvage group. They were based on the realization that a double helicate in which the end groups of the ligands are pairwise joined in the correct fashion has the topology of a trefoil knot. This structural relationship is illustrated schematically in Figure 6.19a. The synthesis of such a knot thus follows route (b) in Figure 6.10, making use of metal ions to stabilize the double helical intermediate (Figure 6.19b). The synthetic strategy developed by Sauvage in this context is closely related to the catenane synthesis depicted in Figure 6.13. Accordingly, the dicopper(I) double helicate derived from the bis(phenanthroline) ligand 6.27 is reacted under high dilution conditions with an ethylene glycol-derived diiodide of suitable length under the conditions of a Williamson ether synthesis (Figure 6.19c) [28]. A number of products are formed in this step, with the yield of the desired double helicate 6.28 containing linkages between the correct pairs of OH groups (a–c and b–d) amounting only to 3%. Major side products are macrocycles derived from the monocopper(I) complex of 6.27 in which both phenanthroline units coordinate to one copper ion, or from incorrectly formed linkages between OH groups in the double helicate. Connecting the OH groups a–d and b–c in the double helicate leads to the formation of a macrocycle, for example, while connecting a–b and c–d leads to a catenane. The crystal structure confirms the knotted topology of 6.28 (Figure 6.19d) and demetalation of 6.28 again affords the metal-free analog. The chirality of the product can be demonstrated by using NMR spectroscopy in the presence of chiral shift reagents. The efficiency of knot formation significantly improves when changing the ligand structure and the synthetic method to connect the ligands. The dicopper(I) complex of a ligand with a 1,3-phenylene unit instead of the tetramethylene bridge between the two phenanthroline units affords the corresponding trefoil knot in an isolated yield of 74%, for example, when establishing the linkages by ruthenium(II)-catalyzed ring-closing metathesis. Only seven synthetic steps are required for the synthesis of the product from commercially available 1,10-phenanthroline, with the overall yield amounting to 35% [29]. The analogous stabilization of a trinuclear double helicate involves the coordination of the tris(phenanthroline) ligand 6.29 to lithium ions (Figure 6.20), followed by linking the side chains in the resulting helicate by ring-closing metathesis [30]. Since the two ligands thus end up in different macrocycles, this approach affords a link

393

6.3 Syntheses using metal coordination

(a) =

=

=

=

(b) 2 equiv M

M

M

M

M

(c) N

Cu+

N

b OH

d HO

OH

N

2

N

Cu+ N

6.27

N

N

HO

HO

O O

O

O O I

Cu+ N

N OH c

(d)

O

O O

O O

a

N

N

N

O

O N

N

I

Cu+

Cu+ Cs2CO3/DMF

N

N

N

N

N

N O

O O

O O

O

O

6.28

Figure 6.19: Structural relationship of a macrocyclic double helicate and a trefoil knot (a) and general strategy for synthesizing such a knot from a suitable helicate precursor (b). An actual synthesis is shown in (c). The crystal structure of 6.28 in (d) confirms the knotted topology. Protons in the crystal structure are omitted for reasons of clarity.

rather than a knot. The two rings cross four times, rendering them topologically equivalent to a Solomon link. Sauvage’s strategy of covalently stabilizing helicates thus allows obtaining molecular interlocked molecules of varying topologies, with helicates containing an even number of metal centers giving rise to knots and those with an odd number to links. This approach becomes progressively more difficult with increasing distance of the end groups in the helicates. Complex knots are therefore often more efficiently obtained by using circular helicates, with helicates that have an even number of metal

394

6 Threading molecules

N

N

N

2

N

N

N

O

O O

O O

O

Li+

O

O

O

O

O O

O O N

Li+ N

N

N Li+

N

N

N

N N

O

O

N

N

N

N

Li+

O

O O

O

O

N O

O

Ring-closing metathesis

N

O

O O

Li+

N

O

O

O O

N

O

O

6.29

Li+

N

N

N

N Li+

N

N

N O

O O

O O

O

O

O

Figure 6.20: Synthesis of a Solomon link that involves covalently linking the side chains in the trilithium double helicate of ligand 6.29.

ions giving rise to links, as we have seen in Figure 6.16. Conversely, those with an odd number afford knots. An example of the latter case is the use of a pentanuclear circular helicate for the synthesis of a pentafoil knot. The corresponding synthesis involves the thermodynamically controlled subcomponent self-assembly of the 2,2ʹ-bipyridine derivative 6.30, the diamine 6.31, and iron(II) chloride, [31]. Recall in this context that iron(II) chloride templates the assembly of pentanuclear circular helicates, while the corresponding sulfate salt leads to hexanuclear helicates, of which a suitably functionalized derivative forms the basis of the synthesis the star of David catenane presented in Section 6.3.1.

6.3.3 Rotaxanes Rotaxanes are obtained by using metal templates in a very similar way as catenanes. An approach closely related to the catenane synthesis shown in Figure 6.13, which uses a stoppering approach to prevent the ring from leaving the axle, was developed by Harry W. Gibson [32]. The stoppers are introduced by alkylation of the two hydroxy groups in the copper(I) complex formed from 6.16 and a macrocyclic phenanthroline

6.3 Syntheses using metal coordination

OHC

N

N

N

O

H2N

Fe 2+

O

N

N

6.30

NH 2 6.31

O

O

CHO

N

+

395

as FeCl 2

N Fe 2+ N N O O

N

N

N

N

N N N

N

N Fe 2+ N N N

2+ N N Fe N N

O

O

N

Fe 2+ N

N

N

N

N

Fe 2+

O

N N

N O

O

O

Figure 6.21: Synthesis of a pentafoil knot by subcomponent self-assembly of the dialdehyde 6.30, the diamine 6.31, and iron(II) chloride. The crystal structure of the product confirms the knotted topology. Protons in the crystal structure are omitted for reasons of clarity.

derivative (Figure 6.22). The subsequent removal of the template upon treatment with a cyanide salt affords the metal-free rotaxane 6.32 in 42% yield over both steps. When is a template passive and when is it active?

In all syntheses discussed so far, the metal templates serve structural purposes. They preorganize the precursors of the product in an interlocked arrangement or contribute to product formation by thermodynamically stabilizing the product, but they are not involved in the actual bond forming reactions. Transition metals can, however, also play an active role in this step. The respective concept of combining the structural and catalytic properties of metal ions to prepare mechanically interlocked molecules was pioneered by David A. Leigh, who introduced the term active metal template synthesis for the underlying strategy as opposed to the passive metal template syntheses discussed so far [33]. Active metal template syntheses allow syntheses of catenanes and rotaxanes but we focus only on the more frequently used synthesis of rotaxanes here. The general strategy is shown in Figure 6.23. It involves the use of a macrocycle with a coordination site that points toward the interior of the ring. This site serves to bind the metal, arranging it within the macrocyclic cavity. The interactions of the

396

6 Threading molecules

O

O

N

N

N

O

6.16

O

O

Cu +

O N

+ N

HO

O

O

O

Cu +

O

N

O

OH

N

HO

N OH

O 1. K 2 CO 3 N

N

I 2 equiv O 2. Cyanide salt

N

O O

42% O

O

N

O O

6.32

O

Figure 6.22: Synthesis of rotaxane 6.32 that involves the stoppering of the axle in the copper(I) complex between 6.16 and a macrocyclic ligand.

Figure 6.23: General strategy of a rotaxane synthesis that involves the use of an active metal template. The respective metal ion initially binds to a coordination site oriented toward the interior of a macrocycle. The subsequent interaction of the metal with the axle components leads to a rotaxane-like arrangement that is stabilized by the metal-induced linking of the axle components. The rotaxane thus formed releases the metal ion, which subsequently mediates further reactions.

metal with the axle components then preorganize the reaction partners in the manner required for producing the final interlocked structure. In addition to this structural stabilization, the metal actively mediates the coupling of the axle components, initially

397

6.3 Syntheses using metal coordination

affording a rotaxane in which the axle is fully formed, with the metal still being present. If the metal is only weakly bound, it can now dissociate to find another empty ring and induce further conversions. This approach has several attractive features, the most important of which is the possibility to use the metal in only catalytic amounts. The product is thus obtained directly in the metal-free form without having to remove stoichiometric amounts of metal ions in a separate step. Another advantage is that the axle components only need to contain the functional groups involved in the coupling reaction, while additional chelating units that are required in passive metal template syntheses to hold the rotaxane precursors together do not have to be present. As a consequence, the structural variability of the rotaxane precursors is high. Figure 6.24 shows a rotaxane synthesis that involves an active metal template [34]. In this case, the metal is copper(I), which is introduced into the reaction as the PF6– N O Cu(I) O L L L

R=

O

O

O

PF6–

[6.33.CuL3]+

OR

N O Cu(I) O L N– + N N

O RO

+L PF6–

RO

N O Cu(I) O L L L N N N OR O O

L HPF6 OR

–3 L

–L [CuL4]PF6

N N N RO

N3 6.34 L3Cu

RO

O

O

N

OR 6.35

[CuL4]PF6

O

RO +L + HPF6

N O Cu(I) O L L N N N O

+3 L OR O

L Cu(I) L OR

O The coordination of another macrocyclic ligand to the second copper center in this bimetallic intermediate leads to the formation of a [3]rotaxane

[CuL4]PF6

Figure 6.24: Active metal template synthesis of a [2]rotaxane by copper(I)-catalyzed azide-alkyne cycloaddition. The macrocycle 6.33 serves as the ligand for the copper(I) center, which, in turn, mediates the coupling of the azide 6.34 and the alkyne 6.35. The central intermediate is a bimetallic complex in which one copper ion is bound to 6.33.

398

6 Threading molecules

acetonitrile complex [Cu(CH3CN)4]PF6 to catalyze the azide-alkyne cycloaddition between the axle components. The other components are the macrocycle 6.33, containing the inwardly directed nitrogen donor of a pyridine unit, as well as the azide 6.34 and the terminal alkyne 6.35, both containing bulky end groups. Mixing all four components in equimolar amounts affords the expected [2]rotaxane in 57% yield after demetalation along with 41% of the noninterlocked axle. The use of an excess of 6.34 and 6.35 (5 equiv with respect to 6.33) leads to the incorporation of 94% of the originally present macrocycle into the rotaxane. Substoichiometric amounts of the copper(I) source cause the yield to drop, likely because turnover is prevented by the tight binding of copper(I) to the product. This problem is circumvented by adding 3 equiv of pyridine as a competing ligand. In this case, the product is obtained in a yield of 82% if 1 equiv of 6.33 and 5 equiv of each 6.34 and 6.35 are reacted in the presence of 0.2 equiv of the copper catalyst. The central intermediate of this transformation is a dicopper(I) complex, consistent with the generally accepted mechanism of the copper(I)-catalyzed azide–alkyne cycloaddition [35]. If this bimetallic intermediate contains one macrocyclic ligand, the expected [2]rotaxane is formed, while a bimetallic intermediate with two macrocycles should give rise to a [3]rotaxane. This [3]rotaxane indeed becomes a prominent side product when the concentration of the macrocyclic ligand is increased, lending support to the proposed mechanism. Active metal template syntheses of rotaxanes or catenanes can be based on a number of other transition metal-catalyzed transformations. Example are alkyne–alkyne heterocouplings (Cadiot–Chodkiewicz reaction), oxidative Heck reactions, and palladium(II)-mediated Michael additions [33]. The versatility and possibility of using the metals in catalytic amounts render this approach an important tool in the development of mechanically interlocked molecules.

6.4 Syntheses using charge-transfer interactions 6.4.1 Catenanes Charge-transfer interactions between electron-rich and electron-poor aromatic moieties are a characteristic feature of a large family of mechanically interlocked molecules that was introduced by J. Fraser Stoddart. Stoddart not only developed reliable methods to prepare these compounds but also systematically characterized dynamic processes occurring in them such as the movement of one component in relation to others [36]. With this work he laid the foundation for the development of switchable systems and somewhat later also for that of molecular machines, for which he was awarded the Nobel Prize in Chemistry in 2016 together with Jean-Pierre Sauvage and Bernard (Ben) L. Feringa. We focus on synthetic strategies here and look at the other aspects in the following chapter.

399

6.4 Syntheses using charge-transfer interactions

Many of the mechanically interlocked molecules developed in the Stoddart group are based on cyclophane 6.36 (Figure 6.25a), the so-called blue box (Section 4.1.5). This macrocyclic receptor preferentially incorporates electron-rich aromatic guests such as hydroquinone derivatives into the cavity, causing the solution to change its color from almost colorless to deep orange as a consequence of the charge-transfer interactions between the binding partners. The complexes thus formed have a pseudorotaxane-like structure with the substituents of the guests protruding from the cavity openings, rendering them available for the construction of catenanes and rotaxanes.

(a)

+ N

+ N "Blue box" 6.36

O

O

O N

N +

+

(b)

2 PF6 +

N

N

2 PF 6

–

+ Br

+

N

N

O

O

+ O

– + N Br

(c) Br O

O

O

O O N +

N +

O

O O

O

N

N

O O

O 6.38

Br N

O O

+

6.37

O

–

O

N

N 2 PF6 Br

Br

–

–

O

O

+ 70% after anion exchange

O N +

N O +

O

O

O O

N

N + 4 PF6

+

–

6.39

Figure 6.25: Molecular structure of the blue box 6.36 (a), formation of a [2]catenane by clipping the blue box around one of the hydroquinone units of crown ether 6.38 (b), and crystal structure of the product 6.39 (c). Protons and counterions in the crystal structure are omitted for reasons of clarity.

These interlocked systems are usually accessed by synthetizing 6.36 in the presence of appropriate templates. Reacting, for instance, equimolar amounts of the bis(pyridylpyridinium) precursor 6.37 and 1,4-bis(bromomethyl)benzene in the presence of 2.5 equiv of the hydroquinone-containing crown ether 6.38 affords, after purification and ion exchange, the hexafluorophosphate salt of the [2]catenane 6.39 in an impressive yield of 70% (Figure 6.25b) [37]. The crystal structure of 6.39 confirms the interlocked arrangement of the two rings.

400

6 Threading molecules

Catenane formation likely involves the initial chain elongation of 6.37 to afford a tricationic intermediate. This compound folds around one hydroquinone moiety of 6.38 to maximize the charge-transfer interactions between the aromatic subunits. In the corresponding complex, the end groups of the linear component are arranged in close proximity, thus facilitating their reaction to close the ring. The blue box is thus “clipped” around the macrocycle that acts as a template. Based on this general strategy, a variety of structurally more complex catenanes are accessible by using other macrocycles or blue box derivatives. The use of a large crown ether containing four hydroquinone moieties allows synthesizing a [3] catenane, for example in which two blue boxes are clipped onto the central ring (Figure 6.26a). Conversely, the macrocyclization of a blue box precursor with biphenyl residues in the presence of 6.38 leads to a [3]catenane in which two crown ethers are connected by an enlarged blue box (Figure 6.26b).

(a)

O

O O

O

O O

O

+ N 2 PF6–

O

O N

Br

N

+

+

O

O

O

O

N +

N

N +

O

O O

Br O

O O

O O

O N +

+

O

O O

N

O

O

O O

O N+

8 PF6– O

O O

+ N

O

Br

N

+ N

(b)

+

O

O

O

O O O

O

2 PF6–

O

O

O O

O O

N +

O

O

O O

N

N +

O N

O Br

O

+N

O

O O

O

4 PF6–

O

O

O N +

O

O

O

O O N

N +

+

O

O

O O

Figure 6.26: Formation of a [3]catenane by clipping two blue boxes onto a crown ether (a), or by connecting two crown ethers through an enlarged blue box (b).

+

401

6.4 Syntheses using charge-transfer interactions

This approach also allows combining two rings of crown ether 6.40 and the enlarged blue box in a [3]catenane (Figure 6.27). The two crown ethers in this product are large enough to allow the clipping of two additional blue boxes, one onto each ring, to give rise to the [5]catenane 6.41. This catenane has the topology of the five interlocked rings that form the symbol of the Olympic games, which is why it has been termed olympiadane [38]. The precise understanding of the principles underlying the formation of these catenanes thus allows the synthesis of products of high structural complexity.

+

N

O

+

N

O

2 PF6–

O

O

O

O

O N Br

O

O O

N

+

O

O

O N

O

N +

O

O

O

O O N

N +

O O

OO O O

O

O

O

O

O

O O

O

Br

O

O

O

O O

O O

O O O

O O

6.40 N

N O

2 PF6–

O

Br

+

O O N

O N

O

N

O

O O N

O

Br O

N +

OO + O N O

O O +

+ N O O

O O N

O

O

N +

N +

O O + O N

+ N O O

O O +

N

N +

6.41

Figure 6.27: Synthetic strategy of obtaining the olympiadane 6.41. A [3]catenane is produced initially from the large crown ether 6.40 and the enlarged blue box to which two additional blue boxes are clipped in the following step.

[2]Catenanes combining electron-rich and electron-poor aromatic residues also result from clipping a macrocycle with 1,4,5,8-naphthalenetetracarboxylic diimides (NDI) moieties around a crown ether that contains 1,5-dioxynaphthalene groups.

402

6 Threading molecules

The corresponding synthesis, developed in the group of Jeremy K. M. Sanders, involves the macrocyclization of the dialkyne 6.42 under the conditions of a Glaser coupling in the presence of the crown ether 6.43 (Figure 6.28) [39]. In this way, the [2]catenane 6.44 is obtained in 52% yield. This catenane exhibits a strong band in the UV–vis spectrum at ca. 460 nm that is not observed for either of the building blocks alone, reflecting the charge-transfer interaction between the NDI (electronpoor, acceptor, A) and 1,5-dioxynaphthalene (electron-rich, donor, D) units. Catenane 6.44 structurally differs from Stoddart’s catenanes by the alternating A-DA-D arrangement of the aromatic subunits and the fact that that it is neutral, whereas catenanes derived from the blue box 6.36 always feature the positive charges of the paraquat units.

O

O

N

O

N

O

O

O

O

O

O

O

52% N

O

O

O O

+

O

N O

CuCl/CuCl2 air/DMF

O

O O

O

O OO

N O

6.43

O O O

O

O 6.42

O N

O

O

6.44

Figure 6.28: Synthesis of the [2]catenane 6.44 containing an alternating arrangement of electronpoor 1,4,5,8-naphthalenetetracarboxylic diimide and electron-rich 1,5-dioxynaphthalene residues by clipping the NDI-containing ring around the crown ether.

The alternative thermodynamically controlled preparation of catenanes from naphthalene and NDI derivatives involves equilibrating the cysteine-derived building blocks 6.45 and 6.46 in water under conditions that mediate disulfide formation and exchange (Figure 6.29) [40]. Although many different oligomeric and macrocyclic products could potentially result in this reaction, it only affords a series of macrocycles together with the [2]catenane 6.47. This catenane does not feature the alternating arrangement of electron-rich and electron-poor residues, in which the charge-transfer interactions should be strongest, but a D-A-A-D stacking sequence. The pronounced dependence of the extent to which 6.47 is present in the equilibrium on the salt concentration, with 1 M NaNO3 leading to a six-fold increase in the amount of 6.47 with respect to the NaNO3 free conditions, strongly suggests that a major contribution to catenane formation comes from the hydrophobic effect.

403

6.4 Syntheses using charge-transfer interactions

HS

SH COOH O

N

O

COOH

O

NH

HN HOOC

O

O

O2 HOOC water pH 8

O O N

O

O

HOOC

HN

HOOC SH

HOOC

6.45

SH 6.46

S

S SS

+

O

O

O O

O

N

N

O O

O O

N HOOC O NH O

HN COOH COOH SS S S COOH

N O

COOH NH

O O O 6.47

Figure 6.29: Synthesis of the [2]catenane 6.47 from the dithiols 6.45 and 6.46 under thermodynamic control. The product is shown in the preferred co-conformation with a D-A-A-D stacking sequence of the aromatic moieties.

Charge-transfer interactions thus seem to play a minor role in the selection of 6.47, whose preferential formation likely results from a compromise between optimizing the number of charge-transfer interactions and minimizing the strain and size of the solvent-exposed surface in the product. Depending on the building block structure and the conditions of equilibration, the formation of other catenanes is also possible and some building blocks even give rise to more complex links and knots. Since the formation of these interlocked systems is mainly mediated by the hydrophobic effect, this work is presented in Section 6.7.1.

6.4.2 Rotaxanes The clipping approach used by Stoddart for preparing catenanes also allows the synthesis of rotaxanes. An example is shown in Figure 6.30a. In this reaction, the axle 6.48, which contains a central hydroquinone unit and terminal triisopropylsilyloxy groups, is treated with 6.37 and 1,4-bis(bromomethyl)benzene to afford the corresponding [2]rotaxane 6.49 in 14% yield [41]. The same [2]rotaxane is obtained by stoppering the end groups in the pseudorotaxane in which the diol 6.50 is threaded through the blue box 6.36 (Figure 6.30b) [41]. This stoppering strategy affords the product in a yield of 22%, slightly better than the yield associated with clipping. An example of a [2]rotaxane containing the electron-poor aromatic units in the axle and the electron-rich ones in the ring is 6.51, which is obtained by slipping the crown ether 6.38 over the end groups of the corresponding acyclic component (Figure 6.31) [42].

404

6 Threading molecules

+ N 2 PF6–

(a)

Br

N +

N +

N

Br

6.37 Si O

O

O

O

O

O

O

O Si +

6.48

N N

6.49

Si O

O

O

O +

O

O

O

N

O +

HO

O

O

O

O

O

O

O Si

+ Si OTf

+

O

O

O

OH

N N

(b)

O

2,6-Lutidine CH 3 CN N

HO

O

N N

+

+

+

OH +

+ N

N

6.50

N

N +

+ 6.37

Figure 6.30: Synthesis of the [2]rotaxane 6.49 by clipping of 6.36 around the axle 6.48 (a), or by stoppering the pseudorotaxane containing 6.50 and 6.36 (b).

6.5 Syntheses using hydrogen bonds 6.5.1 Catenanes Good starting points for the hydrogen bond-templated synthesis of catenanes and rotaxanes are again pseudorotaxanes in which functional groups in the axle components allow cyclization or the modification with bulky end groups. Suitable precursors are, for example, the complexes between crown ethers and protonated dialkylamines in which the ammonium NH groups are hydrogen bonded to the crown ether oxygen atoms. We encountered such a complex in Section 6.3.2 in the form of the dibenzo-24crown-8-containing rotaxane stabilized by a knot at one end of the threaded chain

6.5 Syntheses using hydrogen bonds

+ N

O

O

+ N

O

O O

O CH3CN, 60 °C 10 days, 52%

O

O

O

O

O

O

O O

O

+ N O

O O

O

O

O 6.38

+ N

O

O

O O

O

405

O

O

O

O

O

O

6.51

Figure 6.31: Synthesis of the [2]rotaxane 6.51 by slipping the crown ether 6.38 with two hydroquinone moieties over the end groups of an axle containing a paraquat unit.

(Figure 6.18a). 21-Crown-7 is also suitable to form such pseudorotaxanes but 18-crown -6 is too small, allowing ammonium ions to bind only in a perching arrangement as shown in Section 4.1. Alternative precursors for the synthesis of interlocked molecules are the complexes of macrocyclic lactams in which NH and C=O groups serve as hydrogen bond donors and acceptors, respectively. Two such catenanes are shown in Figure 6.32, both of which are prepared from simple starting materials in a surprisingly facile and efficient fashion. Catenane 6.52 precipitates spontaneously, for instance, when adding at room temperature dilute solutions of isophthaloyl dichloride and 1,4-bis(aminomethyl)benzene to a solution of triethylamine in chloroform. The product can thus be isolated by simple filtration in 20% yield (Figure 6.32a) [43]. Product formation likely involves the initial oligomerization of the starting materials, which gives rise to products that dimerize by hydrogen bond formation. The subsequent macrocyclization then traps these complexes in the interlocked arrangement.

406

6 Threading molecules

(a)

O

O NH

COCl

HN N H

O

+ H2N

NH2

HN

NH HN

O

NH

COCl

HN

O

O

O

O 6.52 (20%)

(b)

O

O N H

COCl

O

+ COCl

H2N

NH2 6.54

H N

O

N H H N

N H H N

O

O

O N H

N H

H N

H N

N H H N

O

O

O

6.53 (34%)

+

O

O

6.55 (51%)

Figure 6.32: Syntheses of the [2]catenanes 6.52 and 6.53 by the reaction between isophthaloyl dichloride with 1,4-bis(aminomethyl)benzene (a) or the diamine 6.54 (b).

The [2]catenane 6.53 is prepared in a similar way from isophthaloyl dichloride and the diamine 6.54 in 34% yield along with 51% of the noninterlocked macrocycle 6.55 (Figure 6.32b). The synthesis was initially performed by Christopher A. Hunter with the aim to prepare a benzoquinone receptor as mentioned in Section 4.1.5. The synthetic conditions did not only yield the desired macrocycle, however, but also the corresponding [2]catenane [44]. After this serendipitous discovery, systematic work was performed by Fritz Vögtle to elucidate structural aspects of such catenanes and the mechanism of their formation [45]. This work benefitted from the rigidity of 6.53 imposed by the bulky cyclohexyl residues, which are too large to pass the cavity openings and therefore fix the two rings in a defined relative orientation. As a consequence, derivatives of 6.53 in which one isophthaloyl residue in each ring contains an additional substituent exist in three noninterconverting and thus separable co-conformations, in/in, in/out, and out/out (Figure 6.33). The

407

6.5 Syntheses using hydrogen bonds

R O

O

O

O

HN

NH

O

O

HN

NH

HN

NH

R

R O

O NH

O

HN

O NH

HN

NH

R

NH O

O

O

HN

O NH

HN

HN

NH O

O

HN

O

O

R NH

HN

O

NH O

O

HN

NH O

HN

O

O R

R

R

in/in

R

R

R

in/out

R

out/out

Figure 6.33: Possible co-conformations of a derivative of 6.53 in which each ring contains an additional substituent in one isophthaloyl subunit.

selectivity with which these isomers are obtained from suitable building blocks provides information about the mechanism of catenane formation. If, for example, the isophthaloyl dichloride derivative 6.56a with an additional methoxy group is reacted with diamine 6.57a, only the in/out and the in/in isomers are obtained (Figure 6.34a,b). If, on the other hand, the additional methoxy group is located in the central ring of the diamine 6.57b, the in/out and out/out isomers result (Figure 6.34c). This selectivity indicates that catenane formation involves the initial formation of a macrocycle. In the next step, a pseudorotaxane is formed but since the diamine containing the cyclohexyl residues is too large to be threaded through the ring, only the isophthaloyl dichloride acts as axle component in this step. The thus formed complexes are stabilized by hydrogen bonding interactions between the C=O groups of the bound isophthaloyl chloride and NH groups along the ring. They contain the isophthaloyl group either close to the unsubstituted or to the substituted subunit. Trapping these orientations in the final step of catenane formation thus leads to two products whose substitution patterns depend on the choice of the starting materials, with the substituted isophthaloyl dichloride leading to the in/out and the in/in isomer and the substituted diamine to the in/out and out/out isomers as illustrated in Figure 6.34.

408

6 Threading molecules

R

(a) R

O

R O

O NH

HN

NH 2

H2N

R

O Cl

Cl

6.56a (R = OCH3)

6.56b (R = H) 6.57b (R = OCH3)

6.57a (R = H)

(b)

(c) 6.57a

6.57b

6.56a R

6.56b

R

R

R

R

R

+

R

R

R

+

R

R

R

R

R

R

in/out

R

in/in

R

R

out/out

R

R

in/out

Figure 6.34: Structures of unsubstituted and substituted precursors used for the syntheses of the dimethoxy analogs of 6.53 (a), and mechanisms of catenane formation that explain the selectivity with which the different co-conformations of the product are formed. The scheme in (b) shows the reaction with the methoxy group residing in the isophthaloyl dichloride 6.56a and that in (c) the case in which the substituent is located in the diamine 6.57b.

A further level of selectivity is achieved when using an analog of 6.57 in which one amide group is replaced by a sulfonamide. Since sulfonamide NH groups are better hydrogen bond donors than amide NH groups, the isophthaloyl dichloride prefers to be located close to the sulfonamide in the intermediate pseudorotaxane. As a consequence, the second macrocyclization exclusively leads to the in/out catenane (Figure 6.35) [46]. This catenane is, however, formed as a mixture of two topological enantiomers because the orientation with which the diamine reacts with the pseudorotaxane is not controlled. The topological enantiomerism is in this case a consequence of the presence of two different functional groups in the rings, both of which have a defined directionality.

409

6.5 Syntheses using hydrogen bonds

O O S N H

NH2 =

H N

NH2

O 6.56a

6.56a

R

R R

R

R

R

R

Topological enantiomers

R

R

in/out Figure 6.35: Catenane formation between 6.56a and an analog of 6.57a in which one amide group is replaced by a sulfonamide group. Only the corresponding in/out isomer of the catenane is formed, in this case as a mixture of two topological enantiomers.

The with respect to amide NH groups higher acidity of sulfonamide NH groups allows selectively connecting two sulfonamide groups in such a catenane through appropriate linkers. The product thus obtained has the topology of a pretzel, which is why it is termed pretzelane (Figure 6.36) [47]. Another class of catenanes whose formation is templated by hydrogen bonds between NH and C=O groups derives from the self-assembling calix[4]arene tetraureas introduced in Section 5.3.3. The interdigitation of these calixarenes causes the urea groups to be alternately oriented along the seam of the corresponding capsules. Covalently linking groups oriented in the same direction in a pairwise fashion consequently affords interlocked rings. This concept was realized by Volker Böhmer by using the calix[4]arene derivative 6.58 (Figure 6.37a) with terminal double bonds [49]. Assembling the corresponding dimeric capsule in dichloromethane/benzene,

410

6 Threading molecules

95:5 (v/v), treating it with Grubbs catalyst under high-dilution conditions, and hydrogenating the double bonds formed during the ring-closing metathesis, affords three isomeric products that differ in their topologies. Figure 6.37b shows that linking adjacent substituents (a–b, c–d, e–f, and g–h) leads to a topologically trivial product that can be represented in the form of a planar graph, comprising the two calixarene moieties connected through four linkers. The other extreme is a product in which every second substituent is linked (a–c, b–d, e–g, f–h). In this product, neighboring substituents that point into the same direction form macrocycles through which substituents oriented in the opposite direction are threaded. An interlocked structure thus results with the topology of a bis([2]catenane). A third product contains two interlocked rings and two direct linkages.

O

O O

N H H O S N O

N H OH S N

O

O

N H

N H OCH3 H N

H N

O OCH3

R

R

OCH3

R

R

O

O

O

I I K2CO3, DMF

O

O O

N H

N H O S N

O S N O

O O

O

N H

N H OCH3 H N

H N

O

O

O

Figure 6.36: Formation of a pretzelane by covalently linking the two sulfonamide NH groups in a [2] catenane. The image of the pretzel should illustrate the structural relationship [48].

6.5 Syntheses using hydrogen bonds

R' R' NH O O O O NH NHHN HN R' = R'

HN

411

R'

NH HN

O R = C5H11 6.58

RO

RO

OR

OR a–b / c–d / e–g / f–h

a–c / b–d / e–g / f‒h b

a–b / c–d / e–f / g–h

c d

a

+

+

e h g

f

Bis([2]catenane) 5–12%

Doubly-bridged [2]catenane 26–32%

Quadruply-bridged dimer 10–15%

Figure 6.37: Molecular structure of the substituted calix[4]arene tetraurea 6.58 (a), and schematic structures of the linked products resulting from 6.58 upon ring-closing metathesis between the marked double bonds followed by hydrogenation (b).

Because of the low yield of the above statistical bis([2]catenane) synthesis, the Böhmer group also developed a rational synthetic approach involving the calixarene derivatives 6.59 and 6.60 as precursors (Figure 6.38) [50]. 6.59 is an analog of 6.58 with shorter side chains, whereas 6.60 derives from 6.59 in that the two rings are already closed. This calixarene is unable to homodimerize because the interdigitation of the substituents is prevented by the rings. 6.60 does form a capsule with 6.59, however, in which two substituents of 6.59 are threaded through the rings of 6.60. This capsule is perfectly preorganized to furnish, after ring-closing metathesis and hydrogenation, the corresponding bis([2]catenane). The 65% yield with which this product is isolated demonstrates the superiority of this approach with respect to the statistical one. Note that in contrast to 6.58, 6.59 alone is unable to form a bis([2]catenane) because the linkers are too short. Moreover, since the ring-closing metathesis in the intermediate calixarene dimer proceeds in two different ways, two topologically equivalent but enantiomeric bis([2]catenanes) are obtained, which can be separated chromatographically on a chiral stationary phase. A whole family of related mechanically interlocked molecules has been developed by using similar approaches [51]. The final example of a catenane syntheses involving hydrogen bonding interactions illustrates that the preorganization of the precursors does not necessarily require direct interactions between the building blocks but can also be mediated by an external anionic binding partner. Important work in this context comes from the

412

6 Threading molecules

(CH2)10 O

O

H

O N

N

H

O

N

N H

O

R O OR RO O R

R O OR RO O R

O

H N

N H

N H

O

H N

H

N H

N

N

O

H N

H

O

O

N

H

H

N

N

H

O N

H

H

O

O

O

O

O (CH2)10

6.59 (R = C5H11)

6.60 (R = C5H11)

and

and 65%

Topological enantiomers

Figure 6.38: Rational approach to calix[4]arene-derived bis([2]catenanes) that is based on the selfassembly of 6.59 and 6.60 followed by ring-closing metathesis and hydrogenation.

group of Paul D. Beer [52]. Note that this strategy is complementary to the use of cationic metal templates described in Section 6.3. Beer’s syntheses are based on the interlocked chloride complexes between N,Ndialkylated isophthalamides and corresponding pyridinium derivatives. The isophthalamide 6.61 and the pyridinium ion 6.62 (Figure 6.39) bind to chloride with a Ka of 260 M−1 in CD2Cl2, for example, with the interactions in the corresponding complex involving hydrogen bonding interactions between the NH groups and the anion, charge-transfer interactions between the coplanar electron-poor and electron-rich aromatic units, C–H···O hydrogen bonds between the oxygen atoms and the protons of the pyridinium methyl group, and electrostatic interactions [53]. The two binding partners are arranged in an orthogonal fashion like the bidentate ligands in a copper(I) complex, rendering this complex a suitable precursor to prepare mechanically interlocked molecules.

413

6.5 Syntheses using hydrogen bonds

O O

O

O

HN

O

O

O

HN O

6.61 + O NH

O Cl–

NH O

O H O+ N H O H O

HN

O

Cl– O NH

HN

O

O

+ N –

PF 6

NH O 6.62

Figure 6.39: Molecular structures of 6.61 and 6.62 and schematic illustration of the arrangement of these compounds in their chloride complex.

The synthesis of a [2]catenane is achieved by using either a threading or an entwining approach. An example of the first strategy involves the chloride-mediated threading of the disubstituted N-methylpyridinium precursor 6.63 containing two terminal double bonds through the opening of cyclophane 6.64 followed by ringclosing metathesis. In this way, the [2]catenane 6.65 is obtained in a yield of 45% (Figure 6.40a). The pyridinium derivative 6.66, on the other hand, allows the preparation of the [2]catenane 6.67 by entwining (Figure 6.40b). The yield is in this case strongly anion-dependent, with the hexafluorophosphate salt of 6.66 giving rise to a yield of 16%, whereas equimolar amounts of the chloride and the hexafluorophosphate salt lead to a yield of 78%, clearly demonstrating the template effect of the chloride anion [52]. Since these catenanes contain a converging arrangement of NH donors, they are able to serve as receptors for appropriate guests, most importantly for the templates present during their synthesis [52]. The catenane 6.65 as the hexafluorophosphate salt, which is obtained from the corresponding chloride salt by anion exchange, interacts in CDCl3/CD3OD, 1:1 (v/v) with chloride, dihydrogenposphate, and acetate, for example, exhibiting the highest affinity among these anions for chloride (Ka = 730 M−1). The thread 6.63 alone binds to the same anions but with a significantly different selectivity, strongly favoring acetate over chloride. Thus, the interlocking of the two macrocycles in 6.65 leads to characteristic binding properties that differ from those of the individual components.

414

(a)

6 Threading molecules

(b)

O NH

O

O

O

O

+ N Cl

NH

O

NH

O

O

O

O

O

O

O

O

O

O

O

O

O

+ N

NH

O

O

O

O

PF6

6.63·Cl– +

O

O

6.66·PF6– +

O NH

O

O

O

O NH

O

O

NH

O

NH

O

+ N

O

Cl

O

–

O

–

6.64

6.66·Cl

O

O NH O O HN Cl– O HN NH O

O H O+ O N H O H O

O

O

O

O

O

O

O O

O

O

O

O

O

H O

PF6 N + HH O O O

O 45%

Ring closing metathesis

78%

O

Ring closing metathesis

O NH O O HN Cl– O HN NH O

O H O+ O H N O H O

NH O O HN Cl– O HN NH O

O H O+ N H O H O

O

O

O

O

O

O

O

O O

NH O

O H O+ N H O H O

O HN

Cl– O HN NH O

O

O

O

O

H O

N PF6 + HH O O O

O 6.65·Cl–

6.67·Cl–·PF6–

Figure 6.40: Anion-templated syntheses of [2]catenanes by the threading (a) and the entwining (b) approach.

6.5.2 Knots In the course of their work about the tetralactam-derived [2]catenanes, the Vögtle group found unexpectedly that the reaction between the diamine 6.57a and pyridine2,6-dicarbonyl dichloride not only affords the expected [1 + 1] and [2 + 2] macrocycles, but also a [3 + 3] condensation product with the topology of a trefoil knot (Figure 6.41) [54]. Conclusive evidence for knot formation came from a crystal structure. The two enantiomers of this knot can be separated chromatographically by using a chiral stationary phase and their absolute configurations assigned by circular dichroism spectroscopy [55]. Knot formation is likely due to the propensity of the oligomeric intermediates, formed from the starting materials on the way to the product, to adopt tightly folded interlocked conformations, which are eventually trapped in the final macrocyclization step [56].

6.5 Syntheses using hydrogen bonds

O

O

O

N Cl

415

O NH

HN

NH2

H2N

+

Cl NEt3 CH2Cl2

20%

O

6.57a

O

N NH

HN

O

H N

N H

O

O HN O

HN O

NH O

N

H N O

NH

O

NH

HN

O

N

H N O

Figure 6.41: Reaction between the diamine 6.57a and pyridine-2,6-dicarbonyl dichloride that leads to a trefoil knot as demonstrated by the corresponding crystal structure. Protons in the crystal structure are omitted for reasons of clarity.

A similar knot is obtained from 6.68 containing an L-valine and aminodeoxycholanic acid subunit. The synthesis in this case involves the stepwise chain elongation up to the stage of the trimer whose completely deprotected form is then cyclized, giving rise to the formation of the knotted cyclic hexamer in 21% yield (Figure 6.42) [57]. In contrast to Vögtle’s synthesis, knot formation proceeds enantioselectively because of the chirality of the subunits. Mechanistically, the formation of the knot likely also involves the tight folding of the acyclic knot precursor into an interlocked conformation. Certain foldamers such as oligomeric amides or peptides are therefore potentially predisposed to adopt knotted conformations, suggesting a possible pathway of knot formation in Nature.

416

6 Threading molecules

O H N

N H

Boc

HO

O

6.68

O

H N O O

HN

O

H N

NH

HO O HN NH

O

OH

O

NH

O O

O

NH

HN

HN

O O

O

NH

HO

N H

Figure 6.42: Structure of building block 6.68 and schematic structure as well as crystal structure of the knotted cyclic hexamer. Protons in the crystal structure are omitted for reasons of clarity.

6.5.3 Rotaxanes Pseudorotaxanes employed for the syntheses of catenanes are converted into rotaxanes by introducing bulky end groups into the threaded acyclic subunits. In Section 6.6.1, we will look at how the anion-templated strategy introduced by Beer is used in this context. Here, we concentrate on rotaxanes derived from the tetralactams used by Hunter and Vögtle for the synthesis of mechanically interlocked molecules. These macrocycles form complexes with dicarboxylic acid dichlorides that are trapped as the axles of rotaxanes upon treatment with bulky amines. An example involving the pseudorotaxane formed from isophthaloyl dichloride and 6.55 is shown in Figure 6.43 [45]. Based on these tetralactams, Vögtle also developed a conceptually different approach for rotaxane syntheses. This strategy builds on the affinity of 6.55 for anions such as phenolates. In the corresponding complexes, the amide NH groups form hydrogen bonds to the phenolate oxygen atom, resulting in the steric shielding of this oxygen atom from almost all directions except the ring opening (Figure 6.44). As a result, an electrophile has to approach the phenolate from the opposite side of the ring, causing the newly formed bond in the respective Williamon ether synthesis to be threaded through the cavity.

417

6.5 Syntheses using hydrogen bonds

O

O NH HN

O

O N H

N H

H N

H N

O

Cl

Cl O

O Cl

CH2Cl2

Cl O

O

O

O

O NH HN

NH HN

O

O

6.55 H2N

H N

H N O

O NH HN

11% O

O

Figure 6.43: Formation of a [2]rotaxane by stoppering the pseudorotaxane formed from cyclophane 6.55 and isophthaloyl dichloride.

An application of this concept in shown in Figure 6.44. It involves the use of 6.69 as the phenol component and 6.70 as the electrophile, which are both reacted with 6.55 in dichloromethane in the presence of solid potassium carbonate. Under these conditions, the corresponding rotaxane is formed efficiently in a remarkable yield of 95% [58]. Rotaxane formation likely involves the initial SN2 reaction between 6.69 and 6.70 to produce one half of the axle. The subsequent reaction of this intermediate with the complex between 6.55 and the deprotonated form of 6.69 then leads to the formation of the [2]rotaxane. Since the macrocyclic component is captured during the second SN2 reaction, this synthetic strategy is associated with the term trapping [59]. It is somewhat related to active metal template syntheses because a reactive intermediate bound to the ring component is involved in rotaxane formation. The reaction proceeds in a stoichiometric fashion, however, and not catalytically.

418

6 Threading molecules

HO

Br

Br

O

6.69

+

K2CO3, CH2Cl2

6.70

Br

O

O NH H N

O

O N H

N H

H N

H N

O

HO

6.69

– O

K2CO3, CH2Cl2 O O

N H HN

O O

NH H N

6.55

O

– O

Br

N H HN

O O

O O

Overall yield of this one-pot reaction 95%

O NH H N

O

O

N H HN O

O

Figure 6.44: Example of a rotaxane synthesis that involves the use of the trapping strategy. In the first step, the phenol 6.69 and the dibromide 6.70 react under the influence of a base to produce one half of the axle. The second SN2 reaction involves the reaction between the thus formed product and the phenolate complex of 6.55 to afford the corresponding [2]rotaxane.

6.6 Syntheses using halogen bonds 6.6.1 Rotaxanes The anion-mediated strategy developed by Beer to prepare mechanically interlocked molecules also allows the use of halogen bonds to preorganize the components. We focus on rotaxanes here although catenanes can be produced in a similar manner. The

419

6.6 Syntheses using halogen bonds

respective synthetic approach is closely related to the one introduced in Section 6.5.1 with the difference that the isophthalamide moiety in one component of the final product, typically the ring, is combined with a pyridinium derivative containing two flanking 1,4-disubstituted 5-iodo-1,2,3-triazole moieties. An example is the bis(triazole) 6.71a containing permethylated β-cyclodextrin stoppers to mediate the water solubility of the product. A ring is clipped around the chloride salt of 6.71a by treating the diamine 6.72 with pyridine-3,5-dicarbonyl dichloride in dichloromethane in the presence of 6.71a and triethylamine (Figure 6.45) [60]. The methylation of the pyridine moiety in the initially obtained product and subsequent ion exchange affords the corresponding [2]rotaxane 6.73a in 33% yield over the three steps. Rotaxane formation is mediated by the preorganization of the two components in their complex with a chloride anion, which is stabilized by a combination of halogen and hydrogen bonding interactions with the C–I groups of 6.71a and the NH groups of 6.72, respectively. Rotaxane 6.73a has a pronounced iodide affinity in water (Ka = 2,200 M−1), significantly higher than the corresponding prototriazole derivative 6.73b that is obtained in 25% yield by clipping the same ring around 6.71b. The lower yield with which 6.73b is obtained and its lower iodide affinity with respect to 6.73a demonstrates the beneficial effects of halogen bonding interactions in these systems. N O

H2N

NH2

N Cl

R N

X

O

R

X

O

N

N N N

+ O

O

O

O

Cl

R N

Cl

OX

XO

N N

N N O

O N

+

N

OX

XO

N

O O

N N N

O O

+ O

2 NO3

O

O

O O O

O

–

O 6.73a X = I (33%) 6.73b X = H (25%)

O

O O

O

N O

O

R N

O

O

R=

HN

O O

O

O NH

R

+ O

O

6.72

O

R N

N

CH2Cl2, NEt3

+

6.71a X = I 6.71b X = H

1. CH3I 2. NH4NO3

O Cl

N

HN

NH

O

N

O

O O O

O O O O

O

O

O

O O

O

Figure 6.45: Formation of the [2]rotaxanes 6.73a,b by clipping a ring around an axle containing either 1,4-disubstituted 5-iodo-1,2,3-triazole groups (6.71a) or 1,2,3-triazole groups (6.71b). In the first case, rotaxane formation is mediated by a combination of halogen and hydrogen bonding, and in the second case, only by hydrogen bonding.

420

6 Threading molecules

Appending electron-withdrawing substituents to the triazole moieties even enhances this effect by increasing the positive potential of the iodide σ-holes. Accordingly, clipping a ring around the axle 6.74a in which two tetrafluorinated 1,4-phenylene moieties connect the triazole units to the cyclodextrin residues affords the corresponding rotaxane 6.75a in 91% yield, while the rotaxane 6.75b with nonfluorinated aromatic residues is formed in an analogous reaction in a yield of only 45% (Figure 6.46) [61]. Because of its superior halogen bonding ability, the hexafluorophosphate salt of 6.75a also has a higher anion affinity than 6.75b. N R HN

X

X

X

X X

Cl

O

X

N

N

N

N

X

+

O

O

O

O

O X

N O

O Cl

+

N

N N

H2N

O

X

−

NH2

R NH

O

O

R HN X

X X

Cl

O

O

N

X

X N N N

O

X

HN

NH

CH2Cl2, NEt3

N N

O

N N

O

+ O

N

PF6

O O

O O

R=

O O O

O

O O

OO O

O

O

O O

O O

O –

N

+ O

O

X

N

X

N N

O

X

X

O

O 6.75a X = F (91%) 6.75b X = H (45%)

R NH

O

O

X X

O

X

N

6.72

O

HN X

O

N

NH4PF6

Cl−

O

6.74a X = F 6.74b X = H R HN

NH

R NH

O X

O OO

O O

O

O

O O O

O O

O

Figure 6.46: Influence of electron-withdrawing substituents on the efficiency of rotaxane formation by using an axle containing two 1,4-disubstituted 5-iodo-1,2,3-triazole moieties. The clipping of the ring around the axle 6.74a, in which the electron-withdrawing nature of the tetrafluorinated 1,4phenylene substituents enhances the positive electrostatic potential of the iodine σ-holes, proceeds significantly more efficiently than when using the axle 6.74b.

6.7 Syntheses using the hydrophobic effect 6.7.1 Knots In the course of their work on the formation of catenanes from cysteine derivatives with electron-rich 1,5-dioxynaphthalene or electron-poor 1,4,5,8-naphthalenetetracarboxylic diimide (NDI) groups, the Sanders group found that the NDI derivatives

421

6.7 Syntheses using the hydrophobic effect

afford knotted structures when reacting by themselves through disulfide formation. Equilibrating the dithiol 6.76 with three NDI subunits at pH 8.0 in water not only affords a macrocyclic monomer and dimer, for example, but also a trimer with the topology of a trefoil knot (Figure 6.47) [62]. At a 5 mM starting concentration of 6.76, these three products coexist in the thermodynamic equilibrium in almost equal amounts but the additional presence of NaNO3 (1 M) leads to the almost exclusive formation of the knot, indicating that a major driving force of knot formation comes from the hydrophobic effect. Among the three observed products, the knot indeed represents that with the smallest solvent accessible outer surface due to the burial of the hydrophobic units in the interior. The tight folding of the acyclic trimer of 6.76 into a knotted structure likely explains the efficiency with which this product is formed. Knot formation moreover proceeds in an enantioselective fashion because of the chirality of the starting material. R

HS O

N

O –

OOC S

O O

N R O

O N

–

O H2O pH 8.0

–

O

N

OOC O NO

O

O

– OOC

O R

N O

N

N

O –

O

N R

O

N OOC

O

O

O

N

COO N O O COO – O N O O N O N

O

O

O O N N

O O

ON – OOC

–

O

OO

O S O S

COO

O N

N

O

S O

OOC

O

N

O

O N

N

O COO –

O N O

O S S COO

–

COO

–

6.76 (R = COOH)

SH

Figure 6.47: Molecular structure of dithiol 6.76 and structure of the trefoil knot derived thereof by disulfide formation.

Equilibrating the dithiol 6.77 with two NDI residues in water at pH 8.0 results in the formation of several different mechanically interlocked structures, the most abundant of which are a Solomon link (60%) and a figure eight knot (18%) (Figure 6.48a) [63].

422

6 Threading molecules

(a)

O

O

N

N

O

O

R HS

O

O

N

N

O

O

SH

6.77 (R = COOH)

R

H2O pH 8.0 –

OOC

O N O

– S OOC O N S O O S N – OOC O S –

O N OOC O

O N O

O N O O N O

O N O O N O

O N O O N O

O N O

–

O COO N O – O COO S N O S O S

–

OOC O

– OOC O O S N N S O O N

O O

N

+

N S – O COO O N – O COO

–

O

60%

–

(b)

S

O

O

OOC

N

N

O

O NO O O

O O N O N ON O O OOC S

COO

O

N

–

S

S

COO O O N NO

O O N

O O S N N S O – O COO

–

O COO –

18%

Topologically chiral Solomon link

90° rotation

Both knots are interconverted by a 90° rotation, showing that they are identical and therefore topologically achiral

Homochiral figure eight knots tied into one and into the opposite direction

90° rotation

The 90° rotation of one knot illustrates that both knots are mirror images and therefore topologically chiral

Meso figure eight knots tied into one and into the opposite direction

Figure 6.48: Molecular structure of dithiol 6.77 and structures of the Solomon link (left) and the figure eight knot (right) formed from this building block (a). In terms of topology, the Solomon link containing only L-cysteine is chiral, the corresponding figure eight knot achiral, and the knot derived from racemic 6.77 again chiral as illustrated in (b). The red and black lines in the meso knot symbolize that the respective subunits contain cysteine residues with opposite absolute configurations.

6.7 Syntheses using the hydrophobic effect

423

Again, knot formation is primarily driven by the hydrophobic effect, that is, the screening of hydrophobic surfaces from the solvent. When using the racemate of 6.77 instead of the homochiral compound, the figure eight knot is practically the only product, indicating that the knot containing cysteine units of opposite absolute configuration is thermodynamically more stable than that formed from a single enantiomer of 6.77. The three products formed from homochiral and racemic 6.77 give rise to interesting cases of stereoisomerism. The homochiral products formed from only one enantiomer of 6.77 are stereochemically chiral because they contain only cysteine units with the same absolute configuration. The Solomon link is also topologically chiral, thus existing in two enantiomeric forms. The figure eight knot is, however, topologically achiral because it is S4 symmetric. A way to confirm this involves comparing the structures of the two knots that are tied in opposite directions, which are interconvertible by a simple rotation (Figure 6.48b). In terms of stereochemistry, the figure eight knot formed from racemic 6.77 is achiral because it contains an equal number of R- and S-residues, rendering it a meso form. Topologically, it is chiral, however, because the combination of two different building blocks causes this knot to have a lower symmetry (C2) than if all cysteine residues have the same absolute configuration. This chirality is again illustrated by comparing the two knots tied in opposite directions. In this case, the rotation of one form by 90° leads to a knot that is the exact mirror image of the other form. Both forms can interconvert, however, by conformational interchange.

6.7.2 Rotaxanes We have seen many examples in the previous chapters illustrating that macrocyclic compounds through which slim acyclic molecules are threaded to yield pseudorotaxanes represent convenient starting points for the preparation of mechanically interlocked molecules. Appropriate rings should have a roughly cylindrical shape with two openings of similar size so that, among the macrocyclic receptors described in Section 4.1, crown ethers, cyclodextrins, diphenylmethane or paraquat-derived cyclophanes, pillararenes, or cucurbiturils are appropriate building blocks for the synthesis of such interlocked molecules. The cone shaped cyclotriveratrylenes, calix[4]arenes, calix[5]arenes, or resorcinarenes, on the other hand, are not suitable because one of their cavity openings is too narrow to allow threading. In the case of calixarenes, both cavity openings become wide enough to insert another molecule only if the ring contains six or more aromatic subunits. Interlocked molecules derived from receptors that engage in attractive electrostatic interactions with an included substrate have already been presented. Examples are the crown ethers in Section 6.5 and the paraquat-derived blue box in Section 6.4. Pillararenes also belong to this category because they bind positively charged guests

424

6 Threading molecules

through a combination of charge-transfer, hydrogen bonding, and cation–π interactions [64]. In the case of water-soluble receptors, substrate binding in the aqueous medium benefits to a large extent from the hydrophobic effect. Rotaxanes are thus accessible by threading suitable α,ω-difunctionalized substrates through the ring and stoppering the chain ends with large groups. Figure 6.49 shows examples of such rotaxanes that have thus been prepared from an α-cyclodextrin, a cucurbit[6]uril, or a cyclophane. (a) O N +

Fe

– SO3

N H

α-CD

(b) O O H N O2 N

NO2

N N NN

N H2 N N + N O N O N N N O N O

(c)

N

N O N N O N O N + N N N HO N O N N O N

– O2N

OH MeO

NO2 +

N H – 2 Cl

O 3S

–

N

O

MeO

O

OMe

N N

N N

MeO

O3S

– SO3

OMe

MeO O

OMe

O

N+

OMe HO + 2 Na

– SO3

Figure 6.49: Examples of rotaxanes containing an α-cyclodextrin (a), a cucurbit[6]uril (b), and a positively charged cyclophane (c).

Since the exact syntheses of these rotaxanes does not involve aspects beyond those already discussed, we will not go into more detail here. Of the vast number of mechanically interlocked molecules known today, we could anyway touch only a fraction in this chapter. The interested reader is referred to the monograph by Carson J. Bruns and J. Fraser Stoddart entitled “The Nature of the Mechanical Bond” [4] that provides an excellent and much broader overview. We, at this point, move away from the structural aspects of mechanically interlocked molecules and the concepts of their preparation and discuss in the next chapter one of the most fascinating applications of these compounds, namely their use to control molecular motion and to thus gain access to molecular machines.

Bibliography [1] [2] [3]

Walba DM. Topological stereochemistry. Tetrahedron 1985, 41, 3161–212. Mitchell DK, Chambron JC. Chemical topology. J. Chem. Edu. 1995, 72, 1059–64. Stoddart JF. The chemistry of the mechanical bond. Chem. Soc. Rev. 2009, 38, 1802–20.

Bibliography

[4] [5] [6] [7] [8]

[9] [10] [11] [12] [13] [14]

[15]

[16] [17] [18] [19] [20] [21] [22]

[23] [24] [25] [26]

[27]

425

Bruns CJ, Stoddart JF. The nature of the chemical bond. John Wiley & Sons: Hoboken, 2017. Frisch HL, Wasserman E. Chemical topology. J. Am. Chem. Soc. 1961, 83, 3789–95. Forgan RS, Sauvage JP, Stoddart JF. Chemical topology: complex molecular knots, links, and entanglements. Chem. Rev. 2011, 111, 5434–64. Dabrowski-Tumanski P, Sulkowska JI. Topological knots and links in proteins. Proc. Natl. Acad. Sci. 2017, 114, 3415–20. Kimura K, Rybenkov VV, Crisona NJ, Hirano T, Cozzarelli NR. 13S Condensin actively reconfigures DNA by introducing global positive writhe: implications for chromosome condensation. Cell 1999, 98, 239–48. Wasserman E. The preparation of interlocking rings: a catenane. J. Am. Chem. Soc. 1960, 82, 4433–4. Brückner R. Pioneering work on catenanes, rotaxanes, and a knotane in the University of Freiburg 1558–1988. Eur. J. Org. Chem. 2019, 3289–319. Wasserman E. Chemical topology. Sci. Am. 1962, 207, 94–102. Schill G, Lüttringhaus A. The preparation of catena compounds by directed synthesis. Angew. Chem. Int. Ed. Engl. 1964, 3, 546–7. Schill G, Logemann E, Vetter W. Ozonolytic degradation of a catenane. Angew. Chem. Int. Ed. Engl. 1072, 11, 1089–90. Cornelissen MD, Pilon S, Steemers L, Wanner MJ, Frölke S, Zuidinga E, Jørgensen SI, van der Vlugt JI, van Maarseveen JH. A Covalent and modular synthesis of homo- and hetero[n] rotaxanes. J. Org. Chem. 2020, 85, 3146–59. Walba DM, Homan TC, Richards RM, Haltiwanger RC. Synthesis and cutting “in half” of a molecular Möbius strip – applications of low dimensional topology in chemistry. New. J. Chem. 1993, 17, 661–81. Dietrich-Buchecker CO, Sauvage JP, Kintzinger JP. Une nouvelle famille de molecules: les metallo-catenanes. Tetrahedron Lett. 1983, 24, 5095–8. Stoddart JF. The master of chemical topology. Chem. Soc. Rev. 2009, 38, 1521–9. Dietrich-Buchecker CO, Sauvage JP. Interlocking of molecular threads: from the statistical approach to the templated synthesis of catenands. Chem. Rev. 1987, 87, 795–810. Dietrich-Buchecker CO, Khemiss A, Sauvage JP. High-yield synthesis of multiring copper(I) catenates by acetylenic oxidative coupling. J. Chem. Soc., Chem. Commun. 1986, 1376–8. Leigh DA, Pritchard RG, Stephens AJ. A Star of David catenane. Nat. Chem. 2014, 6, 978–82. Chichak KS, Cantrill SJ, Pease AR, Chiu SH, Cave GWV, Atwood JL, Stoddart JF. Molecular Borromean rings. Science 2004, 304, 1308–12. Pentecost CD, Tangchaivang N, Cantrill SJ, Chichak KS, Peters AJ, Stoddart JF. Making molecular Borromean Rings. A gram-scale synthetic procedure for the undergraduate organic lab. J. Chem. Educ. 2007, 84, 855–9. Original image created by Anon Moos and published under the Public Domain Mark 1.0 license. Adams H, Ashworth E, Breault GA, Guo J, Hunter CA, Mayers PC. Knot tied around an octahedral metal centre. Nature 2001, 411, 763. Guo J, Mayers PC, Breault GA, Hunter CA. Synthesis of a molecular trefoil knot by folding and closing on an octahedral coordination template. Nat. Chem. 2010, 2, 218–2. Gil-Ramírez G, Hoekman S, Kitching MO, Leigh DA, Vitorica-Yrezabal IJ, Zhang G. Tying a molecular overhand knot of single handedness and asymmetric catalysis with the corresponding pseudo-D3-symmetric trefoil knot. J. Am. Chem. Soc. 2016, 138, 13159–62. Leigh DA, Pirvu L, Schaufelberger F, Tetlow DJ, Zhang L. Securing a supramolecular architecture by tying a stopper knot. Angew. Chem. Int. Ed. 2018, 57, 10484–8.

426

6 Threading molecules

[28] Dietrich-Buchecker CO, Sauvage JP. A synthetic molecular trefoil knot. Angew. Chem. Int. Ed. Engl. 1989, 28, 189–92. [29] Dietrich-Buchecker CO, Rapenne G, Sauvage JP. Efficient synthesis of a molecular knot by copper(I)-induced formation of the precursor followed by ruthenium(II)-catalysed ring-closing metathesis. Chem. Commun. 1997, 2053–4. [30] Dietrich-Buchecker CO, Sauvage JP. Lithium templated synthesis of catenanes: efficient synthesis of doubly interlocked [2]-catenanes. Chem. Commun. 1999, 615–6. [31] Ayme JF, Beves JE, Leigh DA, McBurney RT, Rissanen K, Schultz D. A synthetic molecular pentafoil knot. Nat. Chem. 2012, 4, 15–20. [32] Wu C, Lecavalier PR, Shen YX, Gibson HW. Synthesis of a rotaxane via the template method. Chem. Mater. 1991, 3, 569–72. [33] Crowley JD, Goldup SM, Lee AL, Leigh DA, McBurney RT. Active metal template synthesis of rotaxanes, catenanes and molecular shuttles. Chem. Soc. Rev. 2009, 38, 1530–41. [34] Aucagne V, Hänni KD, Leigh DA, Lusby PJ, Walker DB. Catalytic “click” rotaxanes: a substoichiometric metal-template pathway to mechanically interlocked architectures. J. Am. Chem. Soc. 2006, 128, 2186–7. [35] Rodionov VO, Fokin VV, Finn MG. Mechanism of the ligand-free CuI-catalyzed azide-alkyne cycloaddition reaction. Angew. Chem. Int. Ed. 2005, 44, 2210–5. [36] Philp D, Stoddart JF. Self-assembly in natural and unnatural systems. Angew. Chem. Int. Ed. Engl. 1996, 35, 1154–96. [37] Ashton PR, Goodnow TT, Kaifer AE, Reddington MV, Slawin AMZ, Spencer N, Stoddart JF, Vicent C, Williams DJ. A [2]catenane made to order. Angew. Chem. Int. Ed. Engl. 1989, 28, 1396–9. [38] Amabilino DB, Ashton PR, Reder AS, Spencer N, Stoddart JF. Olympiadane. Angew. Chem. Int. Ed. Engl. 1994, 33, 1286–90. [39] Hamilton DG, Davies JE, Prodi L, Sanders JKM. Synthesis, structure and photophysics of neutral π-associated [2]catenanes. Chem. Eur. J. 1998, 4, 608–20. [40] Au-Yeung HY, Dan Pantos G, Sanders JKM. Dynamic combinatorial synthesis of a catenane based on donor–acceptor interactions in water. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 10466–70. [41] Amabilino DB, Ashton PR, Brown CL, Cordova E, Godinez LA, Goodnow TT, Kaifer AE, Newton SP, Pietraszkiewicz M, Prodi L, Reddington MV, Slawin AMZ, Spencer N, Stoddart JF, Vicent C, Williams DJ. Molecular meccano. 2. Self-assembly of [n]catenanes. J. Am. Chem. Soc. 1995, 117, 1271–93. [42] Ashton PR, Bělohradský M, Philp D, Stoddart JF. Slippage – an alternative method for assembling [2]rotaxanes. J. Chem. Soc., Chem. Commun. 1993, 1269–74. [43] Johnston AG, Leigh DA, Pritchard RJ, Deegan MD. Facile synthesis and solid-state structure of a benzylic amide [2]catenane. Angew. Chem. Int. Ed. Engl. 1995, 34, 1209–12. [44] Hunter CA. Synthesis and structure elucidation of a new [2]-catenane. J. Am. Chem. Soc. 1992, 114, 5303–11. [45] Jäger R, Vögtle F. A new synthetic strategy towards molecules with mechanical bonds: nonionic template synthesis of amide-linked catenanes and rotaxanes. Angew. Chem. Int. Ed. Engl. 1997, 36, 930–944. [46] Ottens-Hildebrandt S, Schmidt T, Harren J, Vögtle F. Sulfonamide-based catenanes – regioselective template synthesis. Liebigs Ann. 1995, 1855–60. [47] Jäger R, Schmidt T, Karbach D, Vögtle F: The first pretzel-shaped molecules – via catenane precursors. Synlett 1996, 723–5. [48] Original image taken by an anonymous photographer and published under the Pixabay license that allows free use.

Bibliography

427

[49] Vysotsky MO, Bolte M, Thondorf I, Böhmer V. New molecular topologies by fourfold metathesis reactions within a hydrogen-bonded calix[4]arene dimer. Chem. Eur. J. 2003, 9, 3375–82. [50] Bogdan A, Vysotsky MO, Ikai T, Okamoto Y, Böhmer V. Rational synthesis of multicyclic bis[2] catenanes. Chem. Eur. J. 2004, 10, 3324–30. [51] Böhmer V, Rudzevich Y, Rudzevich V. Rotaxanes and catenanes derived from tetra-urea calix [4]arenes. Macromol. Symp. 2010, 287, 42–50. [52] Lankshear MD, Beer PD. Interweaving anion templation. Acc. Chem. Res. 2007, 40, 657–68. [53] Wisner JA, Beer PD, Drew MGB, Sambrook MR. Anion-templated rotaxane formation. J. Am. Chem. Soc. 2002, 124, 12469–76. [54] Safarowsky O, Nieger M, Fröhlich R, Vögtle F. A molecular knot with twelve amide groups – one-step synthesis, crystal structure, chirality. Angew. Chem. Int. Ed. 2000, 39, 1616–8. [55] Vögtle F, Hünten A, Vogel E, Buschbeck S, Safarowsky O, Recker J, Parham AH, Knott M, Müller WM, Müller U, Okamoto Y, Kubota T, Lindner W, Francotte E, Grimme S. Novel amide-based molecular knots: complete enantiomeric separation, chiroptical properties, and absolute configuration. Angew. Chem. Int. Ed. 2001, 40, 2468–71. [56] Brüggemann J, Bitter S, Müller S, Müller WM, Müller U, Maier NM, Lindner W, Vögtle F. Spontaneous knotting – from oligoamide threads to trefoil knots. Angew. Chem. Int. Ed. 2007, 46, 254–9. [57] Feigel M, Ladberg R, Engels S, Herbst-Irmer R, Fröhlich R. A trefoil knot made of amino acids and steroids. Angew. Chem. Int. Ed. 2006, 45, 5698–702. [58] Hübner GM, Gläser J, Seel C, Vögtle F. High-yielding rotaxane synthesis with an anion template. Angew. Chem. Int. Ed. 1999, 38, 383–6. [59] Seel C, Vögtle F. Templates, “wheeled reagents”, and a new route to rotaxanes by anion complexation: the trapping method. Chem. Eur. J. 2000, 6, 21–4. [60] Langton MJ, Robinson SW, Marques I, Félix V, Beer PD. Halogen bonding in water results in enhanced anion recognition in acyclic and rotaxane hosts. Nat. Chem. 2014, 6, 1039–43. [61] Bunchuay T, Docker A, Martinez‐Martinez AJ, Beer PD. A potent halogen-bonding donor motif for anion recognition and anion template mechanical bond synthesis. Angew. Chem. Int. Ed. 2019, 58, 13823–27. [62] Ponnuswamy N, Cougnon FBL, Clough JM, Pantoş GD, Sanders JKM. Discovery of an organic trefoil knot. Science 2012, 338, 783–5. [63] Ponnuswamy N, Cougnon FBL, Pantoş D, Sanders JKM. Homochiral and meso figure eight knots and a solomon link. J. Am. Chem. Soc. 2014, 136, 8243–51. [64] Ogoshi T, Yamagishi TA, Nakamoto Y. Pillar-shaped macrocyclic hosts pillar[n]arenes: new key players for supramolecular chemistry. Chem. Rev. 2016, 116, 7937–8002.

7 Controlling molecular motion CONSPECTUS: A machine is a device that performs a certain task as a result of the controlled movement of the components from which it is assembled. In this chapter, we will see that also molecules exhibit machine-like behavior if the permanent and unavoidable Brownian motion that they and their subunits undergo is overcompensated by a directional component. The mechanically interlocked molecules discussed in the previous chapter are particularly attractive (but not the only) starting points for the development of such machines. The chapter begins with a look at the dynamic behavior of interlocked molecules and then introduces methods how to control it. In the two following sections, we look at specific examples of interlocked molecules in which the components move relative to each other either in a lateral back-and-forth manner or circularly. The latter systems ultimately give rise to molecular motors, and since molecules lacking mechanical bonds can also be used for this purpose, they will be considered, too.

7.1 Introduction Molecules are constantly in motion. Not only do bond lengths and angles continuously change, molecules in solution also move around all the time. Triggered by external stimuli such as light or the protonation of a functional group, these random events can be superimposed by additional directional motions such as changes in the distance or the orientation of certain subunits. An example is the photochemical E,Z-isomerization of alkenes. These motions sometimes resemble movements found in mechanical devices, suggesting that molecules that change their structure in response to a particular stimulus may be able to accomplish tasks at the molecular level, similar to those of analogous macroscopic machines. Although molecular and macroscopic machines can indeed behave surprisingly similar, one has to consider that molecular machines never actually rest, even in the absence of an energy supply, because of the unavoidable Brownian motion. As a consequence, they (have to) function in an extremely dynamic environment. What is a molecular machine?

In this chapter, we discuss various strategies to control the relative motion of two molecules or two subunits in a single molecule. Maybe not surprisingly, the most ingenious solutions in this respect are found in Nature, and we therefore start our discussion with an example of a natural molecular machine, namely the protein complex that allows certain bacteria to actively move. These bacteria carry on their outer surface rotating filaments – hollow tubular structures consisting of many copies of the protein flagellin – that push them forward (Figure 7.1a). A curved subunit just outside

https://doi.org/10.1515/9783110595611-007

430

7 Controlling molecular motion

the cell wall, the so-called hook, connects the filament to a group of proteins spanning the cell membrane, which terminates in the rings of the basal protein body. These rings are able to rotate and are surrounded inside the cell membrane by static proteins, the stator. The overall assembly thus resembles the structure of a rotaxane, with the stator representing the ring through which a rod-shaped protein is threaded. This axle component contains the filament at one end and the basal protein body at the opposite end. The unidirectional rotation of the whole assembly is driven by an inward directed flux of protons across the cell membrane that causes the basal protein body to move relative to the stator. (a)

Filament

2 μm

Hook

Ring of stator proteins

Outer membrane

(b) R

Motor

Inner membrane

Basal protein body

7.1

Figure 7.1: Schematic illustration of the proteins constituting a bacterial flagellum (a), and molecular structure of the triptycene derivative 7.1 (b). The image shows a flagellated Vibrio vulnificus bacterium with the site of the attached filament marked [1].

Another important motor protein whose mode of action involves the protondriven rotation of a rotaxane-like assembly is the ATP synthase. Natural machines in which the components move in a lateral manner are myosin, which is responsible for muscle contraction, or kinesin, which transports cargo inside cells. Fascinating movies are available on YouTube that illustrate how these machines work [2]. Although small in absolute dimensions, natural molecular machines are still sufficiently large to behave almost like conventional machines. This situation changes when going to small isolated molecules such as 7.1 (Figure 7.1b) in which the propeller-like triptycene group rotates around the residue R. In this case, many copies of 7.1 tumble around freely in solution without being attached to a larger structure that serves as an external reference point. It is therefore only possible to discuss motions in such a molecule in relative terms but not in absolute ones. The rotation of the triptycene unit around the bond that carries R is moreover largely random. However,

7.1 Introduction

431

even in this small molecule, it can still be controlled to some extent as demonstrated by the work of T. Ross Kelly [3]. An example is the triptycene derivative 7.2 (Figure 7.2) that acts as a molecular brake because metal coordination to the bipyridine unit almost completely arrests the rotation around the central bond [4]. In its metal-free form, 7.2 is dynamic according to the 1H NMR spectrum, which contains a single set of four sharp signals for the protons in the triptycene group. These signals are broadened in the spectrum of the Hg2+ complex of 7.2, indicating that the metal-induced conformational rigidification of the bipyridyl unit has a strong effect on the rate with which the two subunits rotate around each other. The 1H NMR spectrum recorded at −30 °C contains two sharp signal sets for the triptycene group, consistent with a rigid structure in which two aromatic rings of the triptycene group flank an intercalated bipyridyl residue. The brake is released by adding ethylenediaminetetraacetic acid (EDTA), which binds to Hg2+ more strongly than 7.2. OMe

OMe

N

Hg 2+

N

EDTA

N Hg 2+ N

OMe

MeO

7.2 Brake off

Brake on

Figure 7.2: Mode of action of the molecular brake 7.2. In the metal-free form, the triptycene group rotates freely, whereas mercury coordination induces a conformation with the terminal pyridyl group intercalated between two triptycene blades, thus arresting the rotation.

The Kelly group also introduced the molecular ratchet 7.3 (Figure 7.3a) in which the chiral helicene unit renders the way to the transition state of the conformational interconversion to be steeper for one sense of rotation than for the other (Figure 7.3b) [5]. 1 H NMR spectroscopy showed that the three triptycene blades of 7.3 indeed reside in different environments as expected for a ratchet, but also that the rotation of the triptycene group occurs with equal rates in both directions. This behavior is in accordance with the principle of microscopic reversibility, which implies that the rates of reactions are only governed by the energy distance between the transition state and the ground state and not by how the transition state is reached. The asymmetry of the energy profile in Figure 7.3b therefore does not allow the rotation to be unidirectional. The absence of unidirectional rotation in 7.3 is consistent with the outcome of a famous thought experiment described by Richard Feynman [6]. Feynman proposed a device consisting of an axle with a ratchet and pawl system on one end that resides in a gas-filled container with a temperature T1 (Figure 7.3c). The other end of the axle is connected to a paddle wheel in a second container held at a temperature

432

7 Controlling molecular motion

(a)

(b)

(c)

ΔHf (kcal mol–1)

240 235 230 225 T1

220

T2

7.3

215 0

30 60 90 120 Dihedral angle (degrees)

Figure 7.3: Molecular structure of the molecular ratchet (a), calculated potential energy profile of the clockwise rotation of the triptycene residue in 7.3 around the central bond (b), and schematic illustration of Feynman’s ratchet and pawl (c).

T2. Feynman then posed the question whether the ratchet in this system would be able to convert the random movement of the paddle wheel induced by colliding gas molecules into a unidirectional rotation that could eventually be used to perform work. A unidirectional rotation is indeed possible for T1 < T2. However, in the absence of a temperature difference between the two reservoirs, the paddle wheel moves in a random manner as in Kelly’s ratchet 7.3. The reason is that turning the ratchet into the seemingly easier direction requires overcoming a relatively long way by random movements, whereas the ratchet only has to turn by a small fraction in the other direction upon disengaging the pawl. As a consequence, both rotations proceed with the same probability. In accordance with the second law of thermodynamics, Feynman’s device and Kelly’s ratchet therefore cannot generate work just by absorbing heat from a single thermal reservoir. Unidirectional rotation requires an additional energy input, which was realized by Kelly with the development of triptycene derivative 7.4 (Figure 7.4) [7]. This system uses phosgene as a chemical fuel, which converts the amino group into an isocyanate group. Random rotations of the subunits in the thus activated product bring this group close enough to the OH group to undergo carbamate formation. This reaction locks the molecule in a high energy conformation that can only relax by a clockwise rotation (if viewed as in Figure 7.4) to yield a structure that is thermodynamically more stable. The cleavage of the carbamate group then affords an atropisomer of 7.4 in which the amino group resides in another position than in the isomer activated by phosgene. The 120° rotation from one into the other atropisomer thus proceeds in a controlled unidirectional manner, which would have not been possible without the chemical reaction that prevents the intermediate from turning back.

433

7.1 Introduction

O NH 2

O N C Rotation

Cl NEt 3

O 7.4

Cl

O

O

OH

OH

Carbamate formation

N

C

O

OH

Clockwise rotation

Hydrolysis HN

H2N O

O

O O OH

O

NH O O

Figure 7.4: Reaction scheme illustrating the steps that mediate the clockwise rotation of the triptycene group in the molecular motor 7.4. The free amine in the initial atropisomer reacts with phosgene in the first step to give an isocyanate. The intramolecular carbamate formation then leads to an energetically unfavorable conformation that can only relax by a clockwise rotation. The subsequent hydrolysis of the carbamate releases the new atropisomer that has been formed from the initial one by a unidirectional motion.

Although 7.4 is unable to perform a complete or even continuous unidirectional rotation, the behavior of this and the other triptycene derivatives allows deriving guidelines pertinent to the development of molecular machines and motors. We concentrate on these general aspects here, and then come back to related motors and their uses in Section 7.4. An important aspect that needs to be considered when designing a molecular machine is that the components that are supposed to move relative to each other must be connected, either covalently or through a mechanical bond. Individual molecules that are not linked also perform movements comparable to those of a machine – the incorporation of a substrate into the cavity of a macrocyclic receptor resembles the movement of a piston in a cylinder, for example. Such systems fall apart, however, when the components dissociate, which is generally not desired in a machine. Another aspect is that the individual components of a molecular machine need to be able to communicate, either sterically or electronically. This aspect finds an analogy in the function of gears in a macroscopic machine, which only transmit their motion if they interdigitate. In the molecular brake 7.2, the triptycene residue is in direct contact with the bipyridine unit, for example, which allows it to sense whether a metal ion is bound or not.

434

7 Controlling molecular motion

Finally, controlling movements beyond the Brownian motion requires that the electronic properties or the structure of a particular subunit can be altered, thus triggering the positional change of this subunit relative to another one. The trigger in 7.2 is metal coordination but many other stimuli such as protonation, lighttriggered conformational changes, redox reactions, and others are possible. Combining all these aspects in a single system is often more straightforward when using mechanically interlocked molecules than covalently assembled ones [8]. One reason is that rotaxanes and catenanes are obtained in a modular fashion by first independently preparing the individual components before combining them in a single structure. Thus, different subunits with complementary reactivities can be combined in a product of relatively high structural complexity. In the final product, the individual components are held together by the mechanical bond but are still able to move rather freely relative to each other, with the possible motions depending on the type of the interlocked molecule. Rotaxanes rotate around their longitudinal axes, for example, and the rings in rotaxanes or catenanes around their centers (Figure 7.5a). Since these motions do not lead to substantial structural changes, lateral movements of rings along the axle of a rotaxane in a back-and-forth manner (Figure 7.5b), or the closely related movements of one ring in a catenane along the perimeter of the other one are more important in the context of molecular machines (Figure 7.5c). These movements

(a)

(b)

(c)

(d)

Figure 7.5: Dynamics in mechanically interlocked molecules. The rotations of rings around their centers or of axles around their main axes are not of primary importance in the context of molecular machines (a). More relevant are the lateral movement of the ring along the axle of a rotaxane in a back-and-forth motion (b), and the nondirectional change of the position of one ring of a catenane with respect to another one (c). The unidirectional rotation in a catenane gives rise to a molecular motor that potentially performs work (d).

7.1 Introduction

435

involve uncontrolled oscillations or the deliberate switching between different positions, but both processes cannot be used to perform work because any mechanical effect associated with the movement into one direction is undone when the direction is reversed (unless the back-and-forth movement of the ring in a rotaxane drives the unidirectional rotational motion of an attached ring like a piston in a combustion engine, which would be a really cool molecular machine that has, however, not been realized yet). Work is only performed if the movement proceeds continuously along a certain trajectory, which is the case when a ring in a catenane returns to its original position through a 360° unidirectional rotation (Figure 7.5d). As we have seen earlier, such molecular motors require the input of energy in order to work. Fundamental work on using mechanically interlocked molecules for the development of molecular machines started in the 1980s in the group of Jean-Pierre Sauvage. A bit later, J. Fraser Stoddart joined the field. In 2016, both researchers, together with Bernard (Ben) L. Feringa, who invented a light-driven molecular motor in which two subunits perform a continuous unidirectional rotation (Section 7.4), were awarded the Nobel Prize in chemistry in recognition of their pioneering achievements. This decision was, in part, based on the expectation that molecular machines have the potential to lead to “new materials, sensors and energy storage systems [9].” Indeed, we will see in Section 11.6 that a rotaxane-based memory storage device has already been realized, but before we come to applications, it is instructive to take a closer look at the dynamics of these systems and see how their behavior is controlled. A system particularly well studied with respect to the motions of its components is Stoddart’s [2]catenane 7.5 (Figure 7.6a) [10]. In this compound, three dynamic processes occur at different timescales. The fastest one is a rocking motion that involves changes in the angle at which the two rings are arranged without substantially affecting the interactions between them (Figure 7.6b). This motion is slowed down by decreasing the temperature, but it occurs rapidly at 25 °C, where it has a rate constant of 1.6 × 106 s−1. The pirouetting rotation of the outer hydroquinone ring around the blue box is significantly slower, proceeding at 7,000 times per second because it involves the loss of one charge-transfer interaction in the transition state (Figure 7.6c). The slowest process is the rotation of the blue box around its center (Figure 7.6d), which takes place only 22 times per second at 25 °C because both charge-transfer interactions have to be sacrificed for it to proceed. Similar dynamic processes occur in the [2]catenane 7.6 in which the blue box is threaded onto a crown ether with four hydroquinone moieties (Figure 7.7a) [11]. In this case, the pirouetting motion occurs 2.8 × 104 times per second, significantly faster than in 7.5 because of the larger conformational flexibility of the crown ether. A further process in this catenane is the shuttling of the blue box from one hydroquinone unit to the next. With a rate constant of 300 s−1, this process is slow because the blue box has to completely give up the interactions with one hydroquinone ring to reach another one. Co-conformations with the blue box residing in the

436

7 Controlling molecular motion

region of the oligo(ethylene glycol) chains of the crown ether are only transiently populated.

(b)

O

+ N

O

O

O

O

O

O

k = 1.6 × 10 6 s

(c)

O O O

O N +

N O +

O

O

N +

+

O k = 7,000 s–1

O N + B

N O +

O

O 7.5

O

O

A

O O

N +

O

O

O N

O

O O

O

+ N N + O

O

1

O

+ N

O

O

(a)

O

O

N +

N +

O

O

N +

O

O

N

+

N +

N +

O N +

O N

B

A

N O +

O

O

O

O O

O

(d)

O

k = 22 s

O

O O

O N +

N O +

A

O O

O O

+

N +

B

O

1

O

O

N

O

O N +

+ N O +

B

O

A

N

+

N +

O

O O

Figure 7.6: Molecular structure (a) and dynamic processes occurring in the [2]catenane 7.5. The fastest one is the rocking motion of the rings (b). The pirouetting movement of the crown ether is significantly slower (c), and the slowest process involves the rotation of the blue box around its center (d).

Similar positional changes also occur in the rotaxane 7.7 (Figure 7.7b) in which the blue box shuttles between the two hydroquinone stations 1,800 times per second at 25 °C [12]. At −50 °C, the shuttling stops, which allows distinguishing the complexed and uncomplexed hydroquinone unit by 1H NMR spectroscopy.

7.1 Introduction

(a) O

+ ON

O O

+

O

O

O

O

O

O

O

O

O O

7.6 N

O

+

O

...and further to the other stations... k = 300 s

O 7.7

O

O

+

1

N N

Si O

O

O

+ (b)

O

N +

ON

437

+

O

O

O

O

O

O

O

O

O Si

N N

k = 1,800 s

1

+

Figure 7.7: Molecular structures of the catenane-derived molecular train 7.6 (a) and the rotaxanederived molecular shuttle 7.7 (b). The shuttling motions in these systems are indicated together with the corresponding rate constants at 25 °C.

What is the difference between a shuttle and a switch?

Dynamic processes thus occur in these mechanically interlocked molecules that resemble lateral motions in simple machines. The shuttling motions of the blue box in the molecular train 7.6 or the molecular shuttle 7.7 lack a directional component, however, only proceeding randomly between thermodynamically degenerate states (Figure 7.8a). Accordingly, both molecules cannot be considered to exhibit machine-like behavior because the relative movement of their components is not controlled. Based on the energy profile associated with the shuttling motion, it is, however, possible to develop concepts with which this control is achieved. A crucial aspect in this context is to break the symmetry of the track on which the blue box moves and to thus thermodynamically favor one co-conformation over other ones (Figure 7.8b). In rotaxane 7.8, for example, the blue box prefers the side of the axle where the benzidine group is located because of its better π-donor properties with respect to the 4,4ʹ-biphenol unit (Figure 7.8c) [13]. Once this imbalance in the co-conformational equilibrium is established, the directions into which the ring moves become clearly defined. All that is now needed is a stimulus that triggers a place change by destabilizing the initial co-conformation, leaving the ring no other choice but to move to the next station (Figure 7.8b). In rotaxane 7.8, this destabilization is achieved by protonation or oxidation of the benzidine

438

7 Controlling molecular motion

unit. In both cases, positive charges develop close to the likewise positively charged blue box, which avoids the ensuing repulsive electrostatic interactions by moving away. Both stimuli thus cause the co-conformation of 7.8 to become the thermodynamically favored state in which the blue box binds to the 4,4ʹ-biphenol unit. This bias is reversed by bringing the benzidine unit back to its initial state. Identical stations

(a)

(b)

Green station has the higher affinity for the ring

Stimulus E

E

E

(c)

+

N

+

N Si O

O

O HN

NH O

+

84%

Red station has the higher affinity for the ring

O

N

O

O

O

+

Protonation + H+

+

O

O

O Si

N N

O H2N

O Si

+

Deprotonation H+

O

O

7.8 N

Si O

O

+

+ NH 2 O

O >98%

O

O

+

O

N N

+

Figure 7.8: Energy profiles associated with a random and a directed movement of a ring along the axle of a rotaxane. The rotaxane in (a) contains two identical subunits in the axle while that in (b) contains two different subunits whose affinity to the ring can be controlled. In (c), the switching between the co-conformations of rotaxane 7.8, triggered by the protonation and deprotonation of the benzidine unit, is shown. Switching by oxidation and reduction of same unit proceeds analogously.

The rotaxane 7.8 thus behaves like a molecular switch whose states are deliberately controlled by altering the electronic properties of one subunit in the axle. The switching results in a directional movement of the ring from one side of the axle to the other because it affects the thermodynamic stabilities of the two co-conformations of 7.8. Note that unlike a real switch that stays on or off after it has been pressed, such molecular switches need to be permanently activated or deactivated to remain in the corresponding positions.

7.2 Rotaxane-derived machines

439

Based in these fundamental concepts, many molecular switches have been developed whose modes of action involve different types of interactions between the subunits or various stimuli to induce switching. A few further examples are presented in the following chapters, but these chapters mainly serve to illustrate the wide variety of processes that have been realized with mechanically interlocked molecules.

7.2 Rotaxane-derived machines Molecular switches The translocation of the ring in a rotaxane cannot only be induced by changing properties of individual subunits along the axle, like we have seen for 7.8, but can also be triggered by an external binding partner. A classic example is compound 7.9 that was developed in the Sauvage group (Figure 7.9) [14]. This rotaxane, which contains a phenanthroline unit in the ring and both a phenanthroline and a terpyridine unit in the axle, binds to copper ions, with the position of the ring depending on the oxidation state of the metal. Copper(I) prefers to be tetracoordinated and therefore connects the two phenanthroline units of 7.9. When oxidized to copper(II), the ring and the metal ion move to the terpyridine unit because the pentacoordinated complex geometry is now favored. 7.9 is thus switched electrochemically between two distinct co-conformations.

N

O

RO

N

Cu + N 7.9

N

N

OR

N

O O

O O

RO

N

O O

R=

Reduction Oxidation + e– –e–

O

N

N

Cu 2+ N

N

N

N

OR

N

O

O

O

O O

O

Figure 7.9: Switching between two co-conformations of 7.9 by electrochemically changing the oxidation state of the copper ions. Copper(I) prefers the tetrahedral coordination geometry in the complex with two phenanthroline units, whereas copper(II) preferentially binds to a phenanthroline and a terpyridine ligand.

440

7 Controlling molecular motion

In 7.10, metal coordination and the movement of the ring are also linked, although the metal does not participate in the interactions of the rotaxane components. In its metal-free form, 7.10 contains the ring preferentially at the succinamide unit, where hydrogen bonding to the carbonyl groups is stronger than that to the amide and ester C=O groups at the opposite end of the axle (Figure 7.10a) [15]. When cadmium(II) binds to the dipicolylamine moiety, steric effects destabilize the hydrogen bonding interactions, resulting in the translocation of the ring. A variety of structurally related switches have been developed in the Leigh group, containing the same tetralactam as in 7.10 but differing in the subunits along the axle [8]. In 7.11, the axle contains two amide groups on one side and a cinnamic acidderived residue on the other side (Figure 7.10b). Switching is achieved by changing the protonation state of the phenolic OH group, with the ring preferring to reside close to the amides if the phenol is protonated, and on the other side of the axle if it can bind by charge-assisted hydrogen bonds to the deprotonated phenolate group [16]. Rotaxane 7.12 is switched photochemically by isomerizing the double bond [17]. The fumaramide group with the E-configured double bond has two hydrogen bond acceptors and is therefore the preferred binding site for the tetralactam. Photochemically isomerizing the double bonds leads to a maleamide unit in which one carbonyl groups is involved in an intramolecular hydrogen bond. As a consequence, the ring moves to the other side of the axle where it finds better accessible hydrogen bond acceptors (Figure 7.10c). The additional introduction of a blocking group into these rotaxanes allows kinetically trapping them in their thermodynamically disfavored states. An example is 7.13, which is isolated in the synthesis with the ring residing exclusively on one side of the axle (Figure 7.11a) [18]. Because of the silyl ether, the ring cannot reach the other side of the axle where it would find an equally suitable binding partner. When the silyl group is cleaved, this movement becomes possible, leading to a 1:1 ratio of the two co-conformers. Since the two stations are identical, it is not possible to further influence the co-conformational equilibrium once it has been reached. This is different in 7.14 in which the combination of a fumaramide and a succinamide group gives rise to a further level of control (Figure 7.11b) [18]. In the absence of the silyl group, 7.14 behaves like 7.12, with the ring favoring the succinamide subunit only if the double bond is in the Z-configuration. If the silyl group is present, the translocation of the ring induced by the isomerization of the double bond is no longer possible. Accordingly, the ring is trapped on one side of the axle even if the other side contains the better binding side. If the tetralactam is located close to the unsaturated residue, for example, switching this residue to the Z-configuration only allows the ring to move if the silyl ether is cleaved before or after the isomerization. After the positional change, the ring can be prevented from moving back by reintroducing the silyl group. The blocking group in this rotaxane therefore allows freezing it in specific states even after the stimulus that controls the position of the ring has been turned off. 7.14 thus more closely resembles a conventional switch than the other rotaxane-based switches described so far.

7.2 Rotaxane-derived machines

(a)

441

O O NH O

N H

N N

H N

N 7.10

O

NH O

H N

Cd 2+ N N

(b)

N H Ph O H N

NH O Ph

N H

H N

7.11

+ H+

(c)

H N

O

NH O

O

H N H N

O HN

h.ν

NH O

H N 7

Ph

NH

Ph

O

O

O

Ph

O

O

O

O Ph

Ph

O O

H N

N H

7

7.12

Ph

O

N H

N H

Ph

HN

H N

O

Ph

H N

Ph

– H+

– N H NH O

6

O

O

Ph

O

O

N H

H N O

O

O

Ph

O

6

O HN

O Ph

O

HO H N

N H

Ph

Ph

O HN

O

O O

Cd(NO3)2

N H

7

O

Ph

O

NaCN O O

N

O

N H

7

O HN

O

O

Ph

O

H N

N H H N

N H H N

Ph Ph

O HN O

O

Figure 7.10: Control over the position of the ring in rotaxanes containing a tetralactam as ring component by metal coordination (a), deprotonation/protonation (b), and E/Z isomerization (c).

442

7 Controlling molecular motion

(a)

O O NH O Ph Ph

Si O

N H H N

N H H N

7.13

(b)

NH O Ph Ph

O

N Ph

N

N Ph

N

O

Si O

N H H N

N H

7.14

O HN

O

O O

O

H N

H N

O HN

O O

O

O

H N

Desilylation, E/Z isomerization (h. ν ), resilylation or E/Z isomerization (h. ν ), desilylation, resilylation

Desilylation, E/Z isomerization ( Δ), resilylation or E/Z isomerization (Δ ), desilylation, resilylation

O Si O

N H

H N O Ph Ph

NH

H N NH O

O O

H N

O O HN N Ph

N

O

Figure 7.11: Effect of blocking groups in the axle of a molecular switch on the co-conformational equilibrium. After the silyl ether has been cleaved in 7.13, the rotaxane affords a 1:1 distribution of both co-conformations (a). In 7.14, the silyl group prevents the ring from reaching the favored binding site after E/Z isomerization (b). The redistribution of the co-conformations therefore takes place only after the cleavage of the silyl group.

Molecular muscles: The expansion and contraction of a muscle is based on the lateral sliding of actin and myosin filaments past each other. Sauvage suggested that interlocked rotaxanes, so-called molecular daisy chains, behave in a similar pattern when changing the distance between the two macrocycles that hold the individual subunits together. Each subunit contains a chain terminated at one end by a macrocycle and at the other end by a bulky residue that prevents dissociation. The interlocking of two of these molecules is achieved by threading the chains through the ring of the corresponding other component, leading to a

443

7.2 Rotaxane-derived machines

doubly threaded product in which the two subunits can perform an antiparallel sliding motion (Figure 7.12a). The incorporation of suitable stations in the linear parts of these molecules enables switching [19]. (a)

Expansion

Contraction

(b)

N

RO

N N Cu+ N N

N

O

O

O

7.15

O

O O

O O

O O

O

N

O

O

N N Cu+ N N

2+

N

N N

O

N

N

R=

N

Zn RO

O

1. KCN 2. 2 equiv. Zn(NO3)2

1. KCN 2. 2 equiv. CuPF6(CH3CN)2

N

O

O

O

N

N

N

O

O

O

O O

O

O

O O O N

O

N

O

O N

N

Zn2+ N

(c)

N

O

OR

N

+ N O O N O +H 2O O O

O

+ N

O O H2 + O N O O

N

+ 2 H+

+ 7.16 O O O+ O N O O

H N N

– 2 H+

+ +

N N H

N

+

O O + N O O O O

Figure 7.12: General structure and mode of action of a molecular muscle (a), and structures and modes of action of molecular muscles 7.15, whose expansion and contraction relies on Cu+/Zn2+ exchange (b), and 7.16 that is switched by changing the degree of protonation.

OR

444

7 Controlling molecular motion

Sauvage’s molecular muscle 7.15 contains a combination of a phenanthroline and a terpyridine unit in each chain segment, rendering its mode of switching somewhat related to the behavior of 7.9 (Figure 7.12b) [20]. The dicopper(I) complex of 7.15 in which the metal ions connect the phenanthroline units represents the expanded state. This state is contracted by demetalation followed by the introduction of zinc(II) ions, which connect the phenanthroline units in the rings with the terpyridine groups in the axles. This sequence causes the bulky end groups of the axles to move together. The reverse motion, leading back to the expanded state, is induced by another metal exchange. Expansion and contraction of such daisy chain molecules can also rely on other stimuli. In 7.16, for example, the crown ethers preferentially bind to the paraquat subunits if the amino groups are deprotonated but to the ammonium groups after their protonation (Figure 7.12c) [21]. Thus, the muscle adopts the expanded state in acidic media and contracts upon adding a base. Molecular elevators: Molecular elevators are switches in which large platforms with several macrocyclic subunits move on scaffolds containing the corresponding number of axles. An example is the tripodal system 7.17 introduced by Stoddart in which switching is based on the protonation states of the three amino groups as in 7.16 (Figure 7.13) [22]. O

O O O

O O O

O + NH 2O

O

NH

O

+

NH

O

O

O

O

+

O

NH2

O N

O H NO 2

+

O N

+

O

O

+

+ 3 H+

O

O

3 H+

+

HN

+

N

N O

O

O

O N

+ N + N

+

O

O O

O

O

O + N

O

O

O +

O

O O

O

7.17 N

+ N

+ N

O O

O +

O

N

O

O O

Figure 7.13: Structure and states of the molecular elevator 7.17 in which the movement of the platform between the “upper floor” and the “ground floor” is controlled by changing the protonation states of the amino groups in the axles.

If all three amino groups are protonated, the platform is located on the “upper floor” where it is bound with high fidelity. The downward motion of the platform is

7.2 Rotaxane-derived machines

445

induced by deprotonating the amino groups, which in their neutral form bind less strongly to the crown ethers than the positively charged paraquat units. This motion concludes after the deprotonation is complete, but since the crown ethers in 7.17 are flexible, each deprotonation step causes only one subunit to move. The platform thus reaches the “ground floor” in a stepwise manner, which is why the authors concluded that “molecular elevators are more reminiscent of a legged animal” and less suited to transport freights [22]. Molecular valve: Immobilized switchable rotaxanes act as valves if they block the entry and egress of molecules into the pores of the solid support in one state while allowing it in another. These valves are typically obtained by covalently attaching rotaxanes to mesoporous silica nanoparticle via end groups in the axles, with the position of the ring determining whether the valve is opened or closed. The mode of operation is shown schematically in Figure 7.14a. In the open state of such valves, the rings in the rotaxanes are located sufficiently far away from the surface to allow guest molecules such as dyes to be loaded by diffusion into the pores of the support. When using 7.18, for example, rhodamine B can be incorporated into the pores if the blue boxes bind to the electron-poor tetrathiafulvalene units (Figure 7.14b) [23]. The valves are closed by moving the rings to a station closer to the surface, where they block the pores and thus trap the included dye molecules. In the case of 7.18, this translocation is achieved by the iron (III)-mediated oxidation of the tetrathiafulvalenes. Once the valves are closed, the nanoparticles can be isolated and washed to remove the excess of the dye. Switching the rotaxanes back to the open position, which is achieved in the case of 7.18 by adding ascorbic acid, is accompanied by dye release. Another example of such a valve is the pseudorotaxane 7.19 with a dialkylammonium derivative as the axle (Figure 7.14c) [24]. Porous silica particles containing this ammonium ion on their surface can be loaded with coumarin 460 molecules and the dye-loaded particles subsequently capped with dibenzo24-crown-8. The trapped molecules are released by deprotonation of the ammonium group. Although the crown ether is lost when opening this valve, the silica particles containing the immobilized axles can be reused. Molecular cable car: A rotaxane containing an 18-crown-6 moiety in the moving ring has been shown by He Tian and his group to mediate the transport of cations across lipid membranes. The respective rotaxane 7.20 consists of a symmetric axle with two secondary amino groups close to the end groups and a central triazolium unit that serves as a weak intermediate station (Figure 7.15) [25]. The length of the axle is comparable to the thickness of a phospholipid bilayer, allowing the incorporation of 7.20 into the membrane of a vesicle. If 7.20 is present, a pH gradient across the membrane – resulting from the addition of KOH to the external solution – rapidly disappears, which suggests that the potassium ions are shuttled by the moving crown

446

7 Controlling molecular motion

(a)

Valve opened

Valve closed

(b) N +

O O Si O O

N H

O

O O

O

O

O

O

O

N

O

3

O

S

S

S

S

O N

7.18

+

+

O

O

+

N

(c) O O O O O Si O O

N H

O

O

+ O N H2 O

F

O O 7.19

Figure 7.14: Schematic illustration of the mode of operation of a rotaxane-based molecular valve immobilized onto a mesoporous silica nanoparticle (a), and structures of the redox-activated bistable rotaxane 7.18 (b), and the pH-driven rotaxane 7.19 (c). The rings of these rotaxanes block the pores of the support onto which they are immobilized when located close to the particle surface. Switching the rotaxanes to states in which the rings are located away from the surface induces the release of molecules trapped inside the pores.

ether from the outside into the interior of the vesicle with the concomitant transport of hydroxide ions in the same direction or protons in the opposite direction. The decrease of the transport rate when the triazolium unit in the center of 7.20 is replaced by a less efficient intermediate station or the loss of transport activity resulting from covalently modifying the amino groups in 7.20 suggests that potassium transport indeed results from the shuttling motion. Note that the rotaxane itself is not

447

7.2 Rotaxane-derived machines

O O

O

O

3

O

O O

O

O

H2 O N + O

O O

O O

4

O

N + N N

4

H2 N +

O

O 3

O

7.20

Potassium ions O O

Shuttling O

O O

O O

Lipid bilayer

O

Figure 7.15: Structure of the molecular cable car 7.20 and schematic illustration of the potassium transport mediated by 7.20 from one side of a bilayer membrane to the other side.

switchable in this case but that its shuttling motion is controlled to some extent by the external pH gradient. Molecular pump: In the rotaxane-derived switches discussed so far, the rings move in a back-and-forth manner, depending on the states and positions of the subunits in the axle. In a molecular pump, the rings move in a defined direction. Compound 7.21 (Figure 7.16) should serve to illustrate the concept [26]. When adding the blue box to a solution of 7.21 in acetonitrile, no interaction takes place initially because both the ring and the paraquat unit in 7.21 are positively charged. Once both components are reduced to the respective radical cations, however, they start to interact through stabilizing radical–radical interactions, with the formation of the corresponding complex requiring the blue box to squeeze past the two methyl groups of the terminal pyridinium ring. When subsequently oxidizing the radical cations, the blue box is forced to move away from the paraquat unit and does so by slipping over the isopropyl group in the central ring so that it ends up on the alkyl chain. The alternative movement in the other direction is associated with a higher activation energy and is therefore disfavored. The reduction–oxidation sequence thus pumps the blue box into a specific direction to afford a [2]rotaxane. Free rings

448

7 Controlling molecular motion

N

+

+

N

+ N

+

O N

N

+

N

+

+

O

N N N

O

+ 7.21

Zn

N

Zn2+

N

+

N N N

N

+ N

+

O N

N

N

+ +

N N

+

+

N

N

O N

N

+ N

+

N N N

O

NOPF6

N

NO

N

+

+

N

+ N

O N

+

N

+

N N N

O N

+

+

N

Figure 7.16: Structure of the molecular pump 7.21 and reduction-oxidation sequence that pumps a blue box across the paraquat unit onto the alkyl chain.

remaining in solution after the first stroke of the pump can be threaded onto this [2] rotaxane to yield a [3]rotaxane by repeating the reduction–oxidation sequence. The back movement of the blue box from the alkyl chain to the paraquat unit is prevented in the reduced form of 7.21 by the isopropyl group. This pump is thus able to perform work repetitively for two cycles, driving the rings from a low concentration in solution toward a higher local concentration on the chain. Peptide synthetase: Another example of a rotaxane in which the ring moves into a specific direction while at the same time performing a task is 7.22. This rotaxane mediates the sequence-specific synthesis of an oligopeptide, with each component having a characteristic function (Figure 7.17) [27]. The axle serves as the donor of the amino acids that are sequentially transferred onto the growing product, with the rigid subunits ensuring that this transfer proceeds in the correct sequence. The ring is the acceptor, featuring a tripeptide unit with a cysteine residue whose thiol group mediates the actual transfer reaction.

449

7.2 Rotaxane-derived machines

Compound 7.22 is obtained in three steps, of which the first comprises an active metal template synthesis to assemble a precursor through copper(I)-catalyzed azide–alkyne cycloaddition (Figure 7.17). In the next step, the aldehyde group in the ring of the product serves to anchor a tripeptide hydrazide containing an S-trityl protected cysteine residue and a tert-butyloxycarbonyl (Boc) group at the Nterminus. The last step involves the cleavage of the protecting groups, affording 7.22 after deprotonation, which performs the peptide synthesis autonomously.

CHO

BocHN

O

O

N3

O O

O N H

O

O Phe O

H N O

PivHN

BocHN

Ph

O

H N

N H

N N= N

O O Leu

O

N H

N N= N

O Ala

H N O

NHPiv

N Cu(CH3CN)4PF6

CHO BocHN O

O

O

O

N H

N N N

O

O O

H N O

PivHN

BocHN

Ph

O Phe

O

H N

N H

N N= N

O O Leu

O

N H

N N= N

O Ala

H N O

NHPiv

N O H N O HN N

O

H N

BocHN O

O

O

O

N H

N N N

O

O STrt

NHBoc

N H

O STrt

O

H N

N H

NHBoc

PivHN

BocHN

Ph

O

O Phe O

H N O

N H

N H

N N= N

O O Leu O

H N O

N H

N N= N

O Ala

H N O

NHPiv

N

O HN N

1. CF3COOH 2. N,N-Diisopropylethylamine

O

H N

NH2

N H

O SH

H 2N O

O O

O N N N N

O N H

O O

H N O

PivHN

H2N

Ph

O Phe N H

N N N

O O Leu O

H N O

N N N

N H

O Ala

H N O

7.22

Figure 7.17: Synthesis of 7.22, involving the assembly of the rotaxane by active metal template synthesis, followed by the introduction of the accepting peptide unit and deprotection (Trt, triphenylmethyl, trityl; Boc, tert-butyloxycarbonyl; Piv, pivaloyl, trimethylmethylcarbonyl).

NHPiv

450

7 Controlling molecular motion

O HN N

O

H N O SH

NH 2

N H H2N O

O HN N

HN N

O

O

O

O

O

NH 2

N H

S

O

O

PivNH

O

H N

O

H2N

Ph

H2N

Ph

H2N

OH

O

O

H N O

N H

Ph H N

PivNH

O

O

O

NH 2

O

SH

H2N OH

PivNH

O

O

O

O

Ph O

Two further amino acid transfers and dissociation

HN N

O

H N O

N H

O

H N

N H

O

H N

NHPiv O

SH

+

OH

OH

OH

Figure 7.18: Mode of operation of the synthetic peptide synthetase 7.22. The thiol group undergoes a transacylation followed by an S–N–acyl transfer reaction, resulting in the transfer of the amino acid from the respective tyrosine moiety in the axle to the N-terminus of the accepting peptide. After each step, the ring moves to the next amino acid, where it reacts in a similar fashion. After all three amino acids have been transferred, the rotaxane dissociates. The cleavage of the peptide from the ring affords the product.

7.3 Catenane-derived machines

451

In this rotaxane, the ring moves freely between the adjoining blocking groups but cannot pass the first amino acid. The thiol group of the peptide therefore selectively reacts with this amino acid by transesterification to yield the corresponding thioether. The subsequent S–N transfer reaction shifts the amino acid to the N-terminus of the peptide, simultaneously regenerating the thiol. With the first amino acid now cleaved from the axle, the ring moves one position further to undergo the second and subsequently the third chain elongation. Once the last amino acid is transferred onto the product, no blocking group is left that prevents the ring from leaving the axle so that the rotaxane components dissociate. Since many molecules of 7.22 (ca. ~1018 in the actual experiment) work in parallel in this reaction, the peptide is isolated in milligram quantities after cleavage from the ring. Rotaxane 7.22 thus demonstrates that molecular machines can perform iterative tasks in synthesis that are closely related to those observed in natural systems.

7.3 Catenane-derived machines Catenanes allow the development of molecular machines in which one ring performs a continuous directed motion along the perimeter of another ring, potentially giving rise to molecular motors. Many catenanes operate, however, merely like switches. The two catenanes 7.23 and 7.24 (Figure 7.19), for example, find their direct counterparts in the rotaxanes 7.8 and 7.9. In 7.23, the arrangement of the rings depends on whether the amino groups are protonated or not [28], while the position of the rings in 7.24 is controlled electrochemically by changing the oxidation state of the copper ion [29]. Although the switching between the possible co-conformations in both cases involves one ring to move along the perimeter of the other one, the direction of this motion is not controlled and thus occurs randomly, clockwise or counterclockwise. Possible strategies to realize directionality are outlined in this chapter. O O

O

O

O

HNN +

O

N

O +

N

O

N

O

N

N Cu +

N

N O +

HN O

O

+

N N N

O O

O

O 7.23

7.24

Figure 7.19: Examples of switchable catenanes. The arrangement of the rings in catenane 7.23 depends on whether the 1,5-diaminonaphthalene unit is protonated or not, and that in 7.24 on the oxidation state of the copper ion.

452

7 Controlling molecular motion

One approach involves the use of catenanes that exist in at least three different co-conformations. If one ring in a [2]catenane has three stations – A, B, and C, for example – the other ring makes a full circumrotation by either visiting the stations in the order A-B-C-A or A-C-B-A. Controlling this sequence thus allows controlling the directionality of the motion, which is not possible for an oscillating motion between only two stations. An example of such a catenane is 7.25 (Figure 7.20) [30]. The large ring in this compound contains four stations A, B, C, and D to which the tetralactam binds with different affinities. Binding to the fumaramides A and B is preferred, with the tetralactam binding less strongly to B than to A because the N-methyl groups in B destabilize the interactions for steric reasons. The least preferred station is the succinic amide ester C, and the amide D does not play a role for the behavior of 7.25. Station A is located next to a benzophenone unit, which allows selectively isomerizing this station into a maleamide derivative by irradiation at 350 nm. The isomerization of B is achieved by irradiating at 254 nm, while station C is not photoactive. If all double bonds of 7.25 are E-configured, the tetralactam is preferentially located at station A. Isomerizing this station to a maleamide unit forces the ring to move to B and when this subunit is also isomerized further to C. The thermal regeneration of the E-configured double bonds finally induces the ring to move back to A. The tetralactam in 7.25 thus moves in a defined sequence from A to B to C and back to A if the proper sequence of stimuli is applied. None of these steps occurs in a directional manner, however, because the next station can be reached in each case from either side.

O O D

O

O N H

NH O

N H H N

N H H N

(CH 2)4

A

O HN O

O

E/Z

A

D

B

(CH 2) 12

hv (350 nm)

C A

(CH 2)4

B

N

O O

O C

A

O NH

D

B

hv (254 nm)

B

O N

D

C

C

7.25 (CH 2) 12

Figure 7.20: Molecular structure of 7.25 and sequential switching of the position of the tetralactam in this [2]catenane by irradiating first at 350 nm, then at 254 nm, and finally by isomerizing the resulting Z,Z form of the catenane back into the original E,E form by thermal treatment.

453

7.3 Catenane-derived machines

A directional motion thus requires an additional component that blocks the motion into the wrong direction such as the second ring in the [3]catenane 7.26. In this compound, the two rings are preferentially located at the two fumaramide stations (A and B) (Figure 7.21) [30]. The irradiation at 350 nm induces the ring 1 to move counterclockwise (as drawn) to the succinic amide ester C because the movement in the other direction is prevented by ring 2. The isomerization of B now forces the other ring to move, and since there are no good stations left, it moves counterclockwise to the isolated amide D. The thermal isomerization of both double bonds back to the E-configurations causes ring 1 to move to station B and ring 2 to station A. At this point, both rings have swapped their positions with respect to the original state. Repeating the cycle once more results in another exchange, with both rings having made the full circumrotation of the large ring in a counterclockwise direction when reaching their original locations. If the larger ring in a catenane only contains two stations, it is normally impossible to tell which way the smaller ring takes when moving from one station to the other, unless the large ring contains blocking groups that prevent the smaller ring from moving in a certain direction as we have seen in rotaxane 7.14. This concept is realized in 7.27 (Figure 7.22) [31]. Starting from the co-conformation in which the ring resides at the fumaramide station, the tetralactam can move counterclockwise or clockwise to the succinamide after photochemical isomerization of the double bond.

1 1 O D

O

A O

NH O

O N H

B

N H

A H N

N H H N

(CH2)4

D

2

1

O HN

B C

(CH2)4

A

O C

O NH

O

7.26 (CH2) 12

N H HO N

B

O

N H H N

1 hv (254 nm)

hv (254 nm) hv (350 nm) 1

O 2

N

O

hv (350 nm) D

D (CH2) 12 2

O

C

B

O

O

E/Z

A 2

A

D O

N

A

B

2 D B

C 2

2 A

B

C

C

D 1

1

C

Figure 7.21: Molecular structure of 7.26 and counterclockwise circumrotation of the two tetralactams in this [3]catenane along the perimeter of the large ring by repeating the stimuli irradiation at 350 nm, irradiation at 254 nm, and thermal treatment twice.

454

7 Controlling molecular motion

(CH2)6

O

Ph O

Ph

1. Photochemical isomerization (E 2. Desilylation 3. Silylation

Ph

Z) O

Ph O

N H O

HN O

HN

O H N

O

O

NH

O

N H

O

O O H N

HN

NH

OO Si O

O

7.27

O HN

HN

O Si

Ph

N H O

HN

NH

NH

O NH

Ph

OO

O

O

O

(CH2)6

(CH2)6

(CH2)6 4. Thermal isomerization (Z 5. Detritylation 6. Tritylation

E)

Figure 7.22: Molecular structure of 7.27 and sequence that induces the clockwise (as drawn) circumrotation of the tetralactam along the perimeter of the large ring.

Both ways are, however, blocked by bulky protecting groups. The directionality of the movement is therefore controlled by which protecting group is cleaved after the isomerization. Cleaving the silyl ether (with tetrabutylammonium fluoride), for example, induces the clockwise motion of the tetralactam to the succinamide. The complete clockwise circumrotation along the large ring is completed by reinstalling the silyl group, thermally isomerizing the maleamide group to a fumaramide, and cleaving the trityl group (with boron trichloride). The process can start again after the reintroduction of this group. The corresponding counterclockwise circumrotation is achieved by reversing the sequence with which the protecting groups are cleaved.

7.4 Machines without mechanical bonds Although the examples in the previous chapter show that rings in catenanes can be made to perform directional movements, the modes of operation of these interlocked systems are complex, requiring a sequence of stimuli, which makes using them to perform work difficult. Not only molecules with mechanical bonds, but also structurally simpler compounds with moving parts whose spatial arrangement is

455

7.4 Machines without mechanical bonds

changed in a controlled manner, exhibit typical traits of machines. Examples are Kelly’s triptycene derivative 7.4, discussed in Section 7.1, that, however, cannot make a full unidirectional rotation. In addition, molecular walkers exist in which one subunit performs a directional movement along the track of another subunit by reversibly cleaving and re-connecting covalent bonds, a behavior resembling that of the proteins dynein and kinesin [32]. Even more impressive are the molecular motors developed by Ben L. Feringa in which two components perform a continuous unidirectional rotation relative to each other [33]. These motors are very efficient, but unlike interlocking molecules, they work according to the rules of molecular rather than supramolecular chemistry. Because of his important contributions to the development of molecular machines, Ben L. Feringa shared the Nobel Prize in chemistry in 2016 with Jean-Pierre Sauvage and J. Fraser Stoddart. Feringa’s motors are based on chiral sterically congested alkenes, an example of which is compound 7.28 (Figure 7.23a) [34]. This molecule contains a thioxanthene unit connected via a double bond to a group of three condensed six-membered rings, one of which contains an R-configured stereogenic center. Because of the bulky substituents, the double bond is forced out of planarity, and one substituent has to adopt a helical arrangement.

(a)

S

H 3 CO

H 3 CO S 7.28

(b)

S 7.29

7.30

S

S

hν (365 nm) 180° clockwise rotation

H 3CO S ( R )-(M,E )-7.28

Helix Δ (60° C) inversion

H 3CO S ( R )-(P,Z)-7.28

Δ (60° C)

Helix inversion

S

S

hν (365 nm) H 3 CO S ( R )-(P,E )-7.28

180° clockwise rotation

H 3 CO S ( R )-(M,Z)-7.28

Figure 7.23: Structures of the molecular motors 7.28, 7.29, and 7.30 (a) and mode of operation of 7.28 that involves four steps, two photochemical E/Z isomerizations and two thermal helix inversions (b). The respective states are shown schematically and in the form of the calculated structures.

456

7 Controlling molecular motion

To understand how this motor operates, it is helpful to consider the thioxanthene unit as the stator around which the other half of the molecule, the rotor, moves. The full rotation involves four steps – two of which are photochemical E/Zisomerizations driving the respective molecules into thermodynamically disfavored states, and two thermal relaxations of the structures thus generated. The complete cycle is shown schematically together with the corresponding calculated structures in Figure 7.23b. The cycle starts from the E-configured stereoisomer of 7.28, which prefers an M-helical arrangement of the rotor with the methyl group in a pseudoaxial position, pointing away from the aromatic ring of the thioxanthene unit. Irradiation at 365 nm switches this isomer into the Z-form. Because of steric reasons, the rotor can only move clockwise during this process (when viewed from the side of the double bond carrying the rotor) and has to additionally undergo a transition from the M-helix to the P-helix to prevent the subunits from clashing. This helix inversion involves the flipping of the six-membered ring in the rotor, thus bringing the methyl group into the unfavorable pseudoequatorial position. The molecule subsequently relaxes in a thermal process, keeping the configuration at the double bond intact, but squeezing the two stacked aromatic rings past each other. Another helix conversion thus occurs at the rotor, leading back to an M-configured arrangement with the methyl group in the pseudoaxial position. Repeating these steps initially affords again a thermodynamically disfavored intermediate with a P-configured rotor that relaxes into the thermodynamically stable structure from which the cycle started. The rate at which this motor operates depends crucially on the thermal steps that convert the thermodynamically disfavored into stable structures. If the corresponding conformational rearrangements are associated with large energy barriers, full rotation is slow. Since the magnitude of this energy barrier depends sensitively on structural aspects, appropriate structural changes allow varying the rate of rotation of these motors over several orders of magnitude. The current record holder in terms of speed is compound 7.29 that operates in the MHz regime (Figure 7.23a) [35]. The light-induced unidirectional rotation of the subunits in these alkenes allows driving the motions of systems that are significantly larger than the motors themselves, which is the typical hallmark of motors [36]. Compound 7.30 (Figure 7.23a) has been shown to induce the rotation of a 5 × 28 μm glass rod, for example [37]. The actual experiment involved using small amounts of 7.30 as a dopant in a cholesteric liquid crystalline phase in which the mesogens are either arranged in a P- or an Mhelical pattern. The actual orientation of the mesogens depends on the chirality of the dopant, in this case on the helicity of the rotor in 7.30. If the helicity changes, which is the case during each step of the rotation, the mesogens in the liquid crystalline phase have to follow. The corresponding movement is transmitted to the surface of the liquid crystalline phase, causing a glass rod placed on top of the film to rotate. The light-induced isomerization of 7.30 from the E- to the Z-configuration leads to a

7.4 Machines without mechanical bonds

457

clockwise rotation, for example, which comes to a halt after ca. 10 min when the reorientation of the cholesteric phase is complete. After turning off the light source, rotation starts again but in the opposite direction because the helicity of the rotor changes back to the initial form during the thermal relaxation. The motion of the glass rod therefore mirrors the changes in the helicity of the rotor during switching and not the full rotation of 7.30. Can a molecular motor power a car?

Another fascinating application of such molecular motors is their use to induce the gliding of molecules on surfaces, for example the nanocars developed in the group of James M. Tour [38]. An example of such a nanocar is 7.31 (Figure 7.24), featuring a Z-shaped oligo(phenylene ethynylene)-derived chassis with four fullerene wheels that rotate around the alkynyl axles. After deposition onto a gold surface, these molecules can be imaged by using scanning tunneling microscopy (STM) as four bright spots representing the fullerene moieties [39]. Because of the relatively strong adhesion force between the fullerene wheels and the underlying gold, 7.31 remains stationary on the surface up to 170 °C. When further increasing the temperature, 7.31 starts to move through a combination of translation and pivoting. The translation occurs perpendicular to the axis, demonstrating that it involves the

OC10H21

OC10H21

H21C10O

H21C10O

H21C10O

OC10H21 7.31

H21C10O

OC10H21

OC10H21

H21C10O

OC10H21

H21C10O

Figure 7.24: Structure of the nanocar 7.31 and series of STM images, showing the movement of 7.31 on a gold surface at ca. 200 °C. The orientation of the nanocar is determined by the separation of the fullerene wheels, thus allowing distinguishing whether the motion occurs perpendicular to the axis or not. The images also show that the translation of the nanocar on the surface is accompanied by a pivoting motion. Images adapted with permission from [39]. Copyright American Chemical Society, 2005.

458

7 Controlling molecular motion

rotation of the fullerenes around the axles. Pivoting is likely due to rotations of pairs of wheels in opposite directions. Several attempts were made to combine these nanocars with Feringa’s motors, with the goal of converting their normally uncontrolled thermal motion on the surface into a directional light-controlled one. An example is 7.32 that integrates the motor 7.29 into a chassis containing four p-carborane wheels (Figure 7.25) [40]. These wheels were chosen to avoid the quenching of the necessary photoisomerization by strongly absorbing fullerene moieties. It was expected that the motor would push the molecule forward every time it makes contact with the surface during the rotation. Although 7.32 could be successfully synthesized and deposited on a copper surface, a lateral movement upon light irradiation was not observed, maybe because one motor is insufficient to propel a car of this size. A more successful design involves incorporating 7.29 into a single axis with two adamanyl-derived wheels at the ends [41]. The corresponding nanoroadster 7.33 diffuses at temperatures above 150 K on a copper surface and becomes substantially faster upon irradiation, indicating that the built-in motor contributes to the motion when it starts to rotate. The Feringa group showed that the nanocar 7.34 with four motors as wheels also moves in a directional manner across a surface [42]. With the development of these nanocars, a competition was launched to find out which molecular design would lead to the fastest nanocar. The first race, in which six teams from all over the world competed for the championship, was organized in Toulouse in 2015. It involved a 100 nm long track on a gold surface that had two bends. Four STMs were used to follow the race. The car that set the speed

= B10C2H12

H13C6

C6H13

H 13 C 6

C 6 H 13

S

S

7.32

7.33

7.34

Figure 7.25: Molecular structures of nanocars 7.32, 7.33, and 7.34 with built-in motors.

Bibliography

459

record traveled with a maximum velocity of 95 nm per hour, meaning that it would have to drive more than 1,000 years to travel the distance of one meter. This and another car did, however, reach the finishing line while others simply returned after a few nanometers, broke down, or were lost on the surface. These achievements may seem not very remarkable for Formula 1 enthusiasts, but they nevertheless demonstrate how well matter can be manipulated at the molecular level today. It is thus fitting to return to physicist Richard Feynman at the end of this chapter by referring to his famous lecture entitled “There’s Plenty of Room at the Bottom” [43]. In this lecture, given in 1959 at the annual American Physical Society, Feynman speculated about many potential advantages the miniaturization of devices potentially has, also referring to tiny machines in this context and their potential uses. Since 1959, many of Feynman’s ideas that may have been purely speculative at his time have come to pass, and we have seen in this chapter that controlling motion at the molecular scale is no longer just a vision. The future thus awaits the next step, namely, bringing these developments to applications.

Bibliography [1]

Original image taken by Janice Haney Carr and published under the Public Domain license (CDC PHIL ID# 6937). [2] Videos on YouTube that give interesting insight into cellular processes can be found under the titles “Inner Life of a Cell” or “Your Body’s Molecular Machines”. [3] Kelly RT. Progress toward a rationally designed molecular motor. Acc. Chem. Res. 2001, 34, 514–22. [4] Kelly TR, Bowyer MC, Bhaskar KV, Bebbington D, Garcia A, Lang F, Kim MH, Jette MP. A molecular brake. J. Am. Chem. Soc. 1994, 116, 3657–8. [5] Kelly TR, Sestelo JP, Tellitu I. New molecular devices: in search of a molecular ratchet. J. Org. Chem. 1998, 63, 3655–65. [6] Feynman RP, Leighton RB, Sands M. The Feynman Lectures on Physics, Vol. 1, AddisonWesley: Reading, MA, 1963. [7] Kelly TR, Silva RA, De Silva H, Jasmin S, Zhao Y. A rationally designed prototype of a molecular motor. J. Am. Chem. Soc. 2000, 122, 6935–49. [8] Kay ER, Leigh DA, Zerbetto F. Synthetic molecular motors and mechanical machines. Angew. Chem. Int. Ed. 2006, 46, 72–191. [9] The Nobel Prize in Chemistry 2016 (Accessed April 30, 2020, https://www.nobelprize.org/ nobel_prizes/chemistry/laureates/2016/press.html). [10] Anelli PL, Ashton PR, Ballardini R, Balzani V, Delgado M, Gandolfi MT, Goodnow TT, Kaifer AE, Philp D. Molecular meccano. 1. [2]Rotaxanes and a [2]catenane made to order. J. Am. Chem. Soc. 1992, 114, 193–218. [11] Ashton PR, Brown CL, Chrystal EJT, Parry KP, Pietraszkiewicz M, Spencer N, Stoddart JF. Molecular trains: the self-assembly and dynamic properties of two new catenaries. Angew. Chem. Int. Ed. Engl. 1991, 30, 1042–5. [12] Anelli PL, Spencer N, Stoddart JF. A molecular shuttle. J. Am. Chem. Soc. 1991, 113, 5131–3.

460

7 Controlling molecular motion

[13] Bissell RA, Córdova E, Kaifer AE, Stoddart JF. A chemically and electrochemically switchable molecular shuttle. Nature 1994, 369, 133–7. [14] Armaroli N, Balzani V, Collin JP, Gaviña P, Sauvage JP, Ventura B. Rotaxanes incorporating two different coordinating units in their thread: synthesis and electrochemically and photochemically induced molecular motions. J. Am. Chem. Soc. 1999, 121, 4397–408. [15] Marlin DS, González Cabrera D, Leigh DA, Slawin AMZ. An allosterically regulated molecular shuttle. Angew. Chem. Int. Ed. 2006, 45, 1385–90. [16] Keaveney CM, Leigh DA. Shuttling through anion recognition. Angew. Chem. Int. Ed. 2004, 43, 1222–4. [17] Altieri A, Bottari G, Dehez F, Leigh DA, Wong JKY, Zerbetto F. Remarkable positional discrimination in bistable light- and heat-switchable hydrogen-bonded molecular shuttles. Angew. Chem. Int. Ed. 2003, 42, 2296–2300. [18] Chatterjee MN, Kay ER, Leigh DA. Beyond Switches: ratcheting a particle energetically uphill with a compartmentalized molecular machine. J. Am. Chem. Soc. 2006, 128, 4058–73. [19] Bruns CJ, Stoddart JF. Rotaxane-based molecular muscles. Acc. Chem. Res. 2014, 47, 2186–99 [20] Jiménez MC, Dietrich-Buchecker C, Sauvage JP. Towards synthetic molecular muscles: contraction and stretching of a linear rotaxane dimer. Angew. Chem. Int. Ed. 2000, 39, 3284–7. [21] Wu J, Leung KCF, Benítez D, Han JY, Cantrill SJ, Fang L, Stoddart JF. An acid-base-controllable [c2]daisy chain. Angew. Chem. Int. Ed. 2008, 47, 7470–4. [22] Badjic JD, Ronconi CM, Stoddart JF, Balzani V, Silvi S, Credi A. Operating molecular elevators. J. Am. Chem. Soc. 2006, 128, 1489–99. [23] Nguyen TD, Tseng HR, Celestre PC, Flood AH, Liu Y, Stoddart JF, Zink JI. A reversible molecular valve. Prof. Natl. Acad. Sci. U.S.A. 2005, 102, 10029–34. [24] Nguyen TD, Leung KCF, Liong M, Pentecost CD, Stoddart JD, Zink JI. Construction of a pHdriven supramolecular nanovalve. Org. Lett. 2006, 8, 3363–6. [25] Chen S, Wang Y, Nie T, Bao C, Wang C, Xu T, Lin Q, Qu DH, Gong X, Yang Y, Zhu L, Tian H. An artificial molecular shuttle operates in lipid bilayers for ion transport. J. Am. Chem. Soc. 2018, 140, 17992–8. [26] Cheng C, McGonigal PR, Schneebeli ST, Li H, Vermeulen NA, Ke C, Stoddart JF. An artificial molecular pump. Nat. Nanotechnol. 2015, 10, 547–53. [27] Lewandowski B, De Bo G, Ward JW, Papmeyer M, Kuschel S, Aldegunde MJ, Gramlich PME, Heckmann D, Goldup SM, D’Souza DM, Fernandes AE, Leigh DA. Sequence-specific peptide synthesis by an artificial small-molecule machine. Science 2013, 339, 189–193. [28] Livoreil A, Dietrich-Buchecker CO, Sauvage JP. Electrochemically triggered swinging of a [2]-catenate. J. Am. Chem. Soc. 1994, 116, 9399–9400. [29] Grunder S, McGrier PL, Whalley AC, Boyle MM, Stern C, Stoddart JF. A water-soluble pHtriggered molecular switch. J. Am. Chem. Soc. 2013, 135, 17691–4. [30] Leigh DA, Wong JKY, Dehez F, Zerbetto F. Unidirectional rotation in a mechanically interlocked molecular rotor. Nature 2003, 424, 174–9. [31] Hernández JV, Kay ER, Leigh DA. A reversible synthetic rotary molecular motor. Science 2004, 306, 1532–7. [32] von Delius M, Leigh DA. Walking molecules. Chem. Soc. Rev. 2011, 40, 3656–76. [33] Kassem S, van Leeuwen T, Lubbe AS, Wilson MR, Feringa BL, Leigh DA. Artificial molecular motors. Chem. Soc. Rev. 2017, 46, 2592–2621. [34] Koumura N, Geertsema EM, Meetsma A, Feringa BL. Light-driven molecular rotor: unidirectional rotation controlled by a single stereogenic center. J. Am. Chem. Soc. 2000, 122, 12005–6

Bibliography

461

[35] Klok M, Boyle N, Pryce MT, Meetsma A, Browne WR, Feringa BL. MHz Unidirectional rotation of molecular rotary motors. J. Am. Chem. Soc. 2008, 130, 10484–5. [36] García-López V, Liu D, Tour JM. Light-activated organic molecular motors and their applications. Chem. Rev. 2020, 120, 79–124. [37] Eelkema R, Pollard MM, Vicario J, Katsonis N, Serrano Ramon B, Bastiaansen CWM, Broer DJ, Feringa ML. Nanomotor rotates microscale objects. Nature 2006, 440, 163. [38] Vives G, Tour JM. Synthesis of single-molecule nanocars. Acc. Chem. Res. 2009, 42, 473–487. [39] Shirai Y, Osgood AJ, Zhao Y, Kelly KF, Tour JM. Directional control in thermally driven singlemolecule nanocars. Nano Lett. 2005, 5, 2330–4. [40] Chiang PT, Mielke J, Godoy J, Guerrero JM, Alemany LB, Villagómez CJ, Saywell A, Grill L, Tour JM. Toward a light-driven motorized nanocar: synthesis and initial imaging of single molecules. ACS Nano 2012, 6, 592–7. [41] Saywell A, Bakker A, Mielke J, Kumagai T, Wolf M, García-López V, Chiang PT, Tour JM, Grill. Light-induced translation of motorized molecules on a surface. ACS Nano 2016, 10, 10945–52. [42] Kudernac T, Ruangsupapichat N, Parschau M, Maciá B, Katsonis N, Harutyunyan SR, Ernst KH, Feringa BL. Electrically driven directional motion of a four-wheeled molecule on a metal surface. Nature 2011, 479, 208–11. [43] Feynman RP. There’s plenty of room at the bottom. Eng. Sci. 1960, 23, 22–36.

8 Mediating molecular transformations CONSPECTUS: The potential scope of a receptor goes far beyond its ability to merely bind a molecule. A receptor can also change the reactivity of the bound substrate, particularly if functional groups along the cavity approach the substrate in the complex and actively participate in the reaction. Such a receptor thus accelerates a chemical transformation and possibly also controls its outcome because of structural constraints imposed on the substrate in the complex. These processes can require the receptor to be present in stoichiometric amounts if it binds the product with similar affinity or even stronger than the starting material, but the receptor can also act catalytically, rendering its mode of action closely related to that of enzymes. In this chapter, we will see examples of all of these cases. In addition, a few catalysts from the field of asymmetric organocatalysis are discussed to show how important supramolecular concepts are in this context. At the end of the chapter, we learn how to design molecules that mediate their own formation.

8.1 Introduction Two molecules with complementary functional groups A and B have to meet in solution before they can react. The rate of reaction depends on a number of parameters such as the temperature, the concentrations of the reacting molecules, and the Gibbs free energy of activation ΔG‡. The latter contains contributions from the activation enthalpy ΔH ‡, which correlates with the extent to which bonds have to be broken and the reacting molecules distorted on the way to the transition state, and the activation entropy ΔS‡, which reflects the organization of the transition state and the degrees of freedom the reaction partners have to give up for the reaction to occur. (a)

(b)

EM

O COOMe O

A + B A

B

kinter

10 9 M

4

10 4 M

7

10 3 M

O

A B

O

OH kintra

3

–MeOH

COOH

O

COOH

A B

–H2O O

–

OSO 2 Me

O –

–MeSO3

Figure 8.1: Schematic schemes comparing an intermolecular and an intramolecular reaction between two functional groups A and B (a), and examples of effective molarities of intramolecular cyclization reactions (b).

https://doi.org/10.1515/9783110595611-008

464

8 Mediating molecular transformations

If A and B are part of the same molecule, they usually react much faster than when they are separated (Figure 8.1a). The corresponding rate enhancement is quantified by relating the rate constants of the intramolecular and the intermolecular reactions by using equation (8.1). EM = kintra =kinter

(8:1)

Since kintra is associated with a first-order and kinter with a second-order reaction, the two rate constants have different dimensions and the ratio kintra /kinter , hence, yields a concentration. In practical terms, equation (8.1) specifies the often unrealistically high concentration of a reaction partner that is required for the intermolecular reaction to have a pseudo-first-order rate constant identical to the rate constant of the intramolecular reaction (kintra = EM kinter ). In other words, the ratio kintra /kinter represents the kinetic equivalent of the effective molarity EM we encountered in its thermodynamic expression in Section 5.5.1 about self-assembly. EM was used in this context to assess the extent to which the intramolecular association of two binding partners is favored over the intermolecular one. In its kinetic form, the effective molarity describes how much faster an intramolecular reaction proceeds with respect to the corresponding intermolecular counterpart, with large EM values denoting particularly effective intramolecular pathways. To illustrate this aspect, effective molarities of several cyclization reactions are given in Figure 8.1b [1]. One reason for the high rates of many intramolecular reactions is the proximity of the reacting groups when they are part of the same molecule. This proximity makes it more likely that both groups meet, rendering the activation entropy of the intramolecular reaction to be less unfavorable than that of the corresponding intermolecular one. An alternative explanation is that the loss of the translational entropy associated with the intermolecular reaction is paid for in the intramolecular counterpart by positioning the reacting groups within the same molecule. Large rate enhancements thus result unless the intramolecular reaction is associated with the build-up of strain, which has an adverse effect on ΔH ‡. Similar effects are operative when the proximity of the reacting groups is enforced by organizing them within a noncovalently stabilized complex (Figure 8.2a). In this way, the activation entropy ΔS‡ also becomes more favorable in comparison to the reaction outside the cavity because the reacting molecules give up their translational degrees of freedom when forming the complex. Since the kinetics of an intermolecular reaction changes from a second-order reaction in the absence of the receptor to a pseudo-first-order reaction in the complex, the extent to which complex formation influences the reaction rate is again estimated on the basis of EM values [2]. An example is the cucurbit[6]uril (CB[6])-induced azide–alkyne cycloaddition between 8.1 and 8.2 (Figure 8.2b). This reaction proceeds via a ternary complex in which both reaction partners are included into the cucurbituril cavity with their ammonium groups binding to the carbonyl groups along the rims [3]. The azide and the alkyne group are thus arranged close to each other, allowing them to

465

8.1 Introduction

react. In the absence of CB[6], the reaction in HCOOH (88%)/H2O 1:1 (v/v) at 40 °C is slow and associated with a rate constant kinter of 1.16 × 10−6 M−1 s−1. If CB[6] is present, the pseudo-first-order rate constant kintra amounts to 0.019 s−1, giving a remarkably high EM of 1.6 × 104 M [2]. This value is still two orders of magnitude lower than the upper limit of 4 × 106 M predicted for the EM of an intramolecular reaction in which the reacting groups are optimally arranged so that no strain develops during the reaction and the entropy changes are limited to the actual bond-breaking and bond-making processes [4]. Nevertheless, it is much higher than many other receptor-mediated EM values, which span a range of seven orders of magnitude, from 10−3 to 104 M [2]. Lower values typically suggest that the reaction involves a substantial reduction of the number of rotatable bonds in the reaction partners, or that the two molecules spend a significant amount of time in a wrong mutual orientation within the cavity.

(a) A + B

A B

kinter

kintra

OO O N N N NN N

(b)

+ H3N

N3 8.1

NH 3 +

+ 8.2

A B

A B

OO

O

N N N N NN

N N NN N N N N N N NN O O O OO O CB[6] HCOOH (88%)/H2O 1:1 (v/v) 40 °C EM = 1.6

H3N +

N N N

+ NH 3

8.3

10 4 M

Figure 8.2: Comparison of the intermolecular reaction between two molecules with functional groups A and B and their pseudo-intramolecular reaction within a receptor cavity (a). The cucurbit [6]uril-mediated 1,3-dipolar cycloaddition reaction between azide 8.1 and alkyne 8.2 to afford the 1,4-disubstituted 1,2,3-triazole 8.3 in (b) is an example of a reaction that proceeds significantly faster in the presence of a receptor.

Another important aspect of receptor-mediated reactions is that they can proceed with an altered regio- or stereoselectivity in comparison to the same transformation in the absence of the receptor. In the reaction shown in Figure 8.2b, for example, the CB[6] induces the exclusive formation of the 1,4-disubstituted product 8.3, completely

466

8 Mediating molecular transformations

suppressing the formation of the 1,5-disubstituted isomer that is normally also formed. The structure of the product thus reflects the preferred mutual arrangement of the two substrates within the CB[6] cavity. Looking more closely at the first-order rate equation of a receptor-mediated reaction shows that the rate with which the complex is consumed not only depends on kintra , but also on the extent to which the starting materials are complexed, that is, on cC (Figure 8.3). For an effective substrate conversion, it is therefore not sufficient that complex formation affects the activation parameters of the reaction. The complex also needs to be present in substantial amounts. If it does not form at all, no rate enhancement can be expected, whereas a high stability of the intermediate complex is beneficial. A receptor therefore not only has to ensure the proper arrangement of the binding partners within the cavity, it also has to bind them efficiently to have an effect.

Ka = A Substrate S

cc

dcc

cS cR

dt

A Receptor R

= Kintra cc B

Complex C

Figure 8.3: Both steps of a receptor-induced substrate conversion. The extent to which the intermediate receptor–substrate complex is formed in the first step depends on its thermodynamic stability, that is, Ka , and the rate of the second step correlates with the concentration of the complex cC and the rate constant kintra .

Because binding precedes substrate transformation, all effects that influence complex concentration have a direct impact on the rate of the reaction. A temperature increase, for example, has a negative impact on the degree of complexation, although it may be beneficial for the actual reaction. It is therefore usually necessary to identify a temperature at which complex formation is still guaranteed and the reaction is fast enough. Additional components in the reaction mixture that interfere in binding also affect the rate of the reaction. A binder that competes with the substrate for the receptor cavity inhibits the reaction, for example. Investigating the effects of inhibitors on the reaction rate thus allows confirming whether the receptor is involved in substrate conversion or not. Receptors can potentially also prevent the bound substrate from reacting with its partner. In this case, the receptor is itself an inhibitor like in the reaction shown in Figure 5.67. The bimolecular fusion of two molecules like that shown in Figure 8.4a is only one type of reaction that can be mediated by a receptor [5]. For this reaction to work, a driving force must exist to simultaneously bind the two substrate molecules inside the receptor cavity. The proximity thus achieved then promotes coupling without the need for a further participation of functional groups in the actual bond forming reaction.

8.1 Introduction

467

The arrangement of the reaction partners inside the cavity could, however, cause the reaction to proceed with a characteristic regio- or stereoselectivity. A drawback is that a ternary complex is converted during this reaction into a, for entropic reasons, usually more stable binary complex. The product thus often likes to stay inside the cavity after the reaction, preventing the receptor from mediating further transformations. Because of this product inhibition, the receptor usually needs to be present in stoichiometric amounts to achieve full conversion. Only if the product is bound weaker than the transition state, turnover is achieved. (a)

(c) A B

A B

X

X

A B

A B

(b) X

X

A B

A B

(d)

X Y

A B Y

(e) Cat A

Cat B

A

B

Figure 8.4: Schemes illustrating the types of reactions mediated by a receptor. Besides the fusion of two substrate molecules (a), a receptor also induces the fission of a substrate into smaller fragments by either directly reacting with the included substrate (b), or by transferring a simultaneously bound reactive species onto the substrate (c). The third mode of action involves transforming the substrate into another compound with the receptor actively participating in the conversion (d) or not (e).

A second type of reaction that can be induced by a receptor is the fission of the substrate into smaller fragments. This reaction could involve a functional group in the receptor that actively participates in the reaction by accepting one fragment of the substrate while the other is released (Figure 8.4b). If the product resulting from the covalent modification of the receptor is stable under the reaction conditions, the reaction only goes to completion if stoichiometric amounts of the receptor are present. If, on the other hand, the receptor is regenerated by cleavage of the bond formed in the first step, it acts as a catalyst. A fission of the substrate is alternatively initiated by the transfer of a reactive species onto the substrate that is also complexed by the receptor (Figure 8.4c). In this case, a covalent modification of the receptor does not occur in the course of the conversion. Since the products usually interact weaker with the receptor than the substrate because they are smaller and therefore bound by fewer interactions, turnover is possible.

468

8 Mediating molecular transformations

Finally, receptors also induce the transformation of a single substrate molecule. Such a reaction can again involve a prosthetic group in the receptor that would mediate the same transformation also in the absence of the receptor (Figure 8.4d). The role of the receptor is to facilitate this reaction and to thus make it more effective. If complex formation causes the stabilization of the transition state of the reaction, the confinement of the substrate in the receptor cavity alone is sufficient to achieve a rate enhancement without the participation of further functional groups (Figure 8.4e). In both cases, the reaction inside the receptor cavity can proceed with a regio- or stereoselectivity that differs from that observed in the absence of the receptor. Catalytic transformations are possible, although the substrate and the product are often bound with similar affinity. Do synthetic enzymes exist?

The ways with which synthetic receptors mediate reactions are strikingly similar to the mode of action of enzymes. In both cases, the reaction takes place within the confinement of a binding pocket and potentially involves prosthetic groups that end up being arranged close to the substrate after complex formation. In this environment, the conversion of the substrate proceeds more rapidly than outside the cavity and often with a characteristic selectivity. The reaction comes to a halt in the presence of a competitive inhibitor that binds more effectively to the binding site than the substrate(s). Because of these analogies, work on catalytically active receptors often aims at the development of enzyme mimics. Only few of the many systems obtained in this context achieve the catalytic activity of enzymes, however [6]. Probably the most important reason for the often poorer performance of supramolecular catalysts is that enzymes are much better fine-tuned to stabilize transition states than can be achieved by rational design [5]. Synthetic systems therefore often suffer from product inhibition and/or an unfavorable arrangement of the substrate in the cavity. These problems already became apparent in the pioneering work of Cram on the development of enzyme mimics [7]. Cram aimed at mimicking the action mode of chymotrypsin, a protease responsible for the cleavage of peptide bonds. This reaction involves the incorporation of the substrate into a binding pocket where the peptide group is arranged close to a serine side chain (Figure 8.5). The hydroxy group in this side chain is part of a so-called chargerelay system that additionally comprises the imidazole unit of a histidine and the carboxylate group of an aspartate or glutamate residue. In this catalytic triad, the carboxylate group causes the partial deprotonation of the imidazole NH group, which increases the basicity of the other imidazole nitrogen atom to such an extent that it is able to deprotonate the serine OH group. The so-formed alkoxide initiates the cleavage of the peptide bond by reacting with the peptide carbonyl group. An acylated enzyme is subsequently formed that is finally hydrolyzed, leading back to the active form of the enzyme.

8.1 Introduction

469

O O

O Ser

Asp –

O H N

O Asp

N H

O H N

His

O

R2

N H

O H N

O

–

O Ser

Asp

O Ser

N H

R1

N H

R2

His

O R1

–

Asp

N H

O H N

His

– R1 O 2 R HN O Ser N H

–R1–NH2

His R2

O

O O Ser

Asp –

O H N

N His

+ H2O –R2-COOH Asp

O –

O H N

O Ser N H His

Figure 8.5: Reaction scheme illustrating the mechanism of the cleavage of a peptide bond mediated by the charge-relay system of the enzyme chymotrypsin.

Cram used Corey–Pauling–Koltun models to design the compound 8.4 (Figure 8.6a), which was expected to hydrolyze amino acid esters in a similar way as peptides are cleaved by chymotrypsin [7]. In 8.4, the macrocyclic spherand-type subunit should induce affinity for the ammonium group of the substrate by serving as a hydrogen bond acceptor (Section 4.1.3). The charge-relay system involves a benzylic OH group, which assumes the role of the serine OH group. This group is flanked by an imidazole residue that interacts with a carboxylate group. Due to the demanding synthesis of 8.4, work initially focused on investigating simpler compounds whose structural complexity was gradually increased. In this context, it was shown that the receptor scaffold 8.5 alone indeed interacts with ammonium groups, leading to stable complexes in CDCl3 (e.g., log Ka = 9.7 for (H3C)3CNH3+). 8.5 was then converted into 8.6, which contains the benzyl OH group as the nucleophile, whose ability to induce the transesterification of amino acid esters was investigated in buffered CDCl3 by using the p-nitrophenol esters of L-alanine or other amino acids as substrates (Figure 8.6b) [8]. p-Nitrophenol esters were chosen because their carbonyl group is activated for a nucleophilic attack by the electron-withdrawing nature of the phenol group and because the resulting p-nitrophenolate anion is colored so that the course of the reaction can be easily followed by UV–vis spectroscopy. The corresponding measurements showed that the acylation of the benzyl alcohol is first order in the ratio of the buffer components, indicating that the deprotonated OH group is indeed the active nucleophile. The acetylation is moreover ca. 1011 times faster than that of 3-phenylbenzyl alcohol, a model compound lacking the receptor unit. Finally, the presence of an excess of NaClO4 as a competitive

470

8 Mediating molecular transformations

inhibitor significantly slows down the reaction. These results demonstrate that the complexation of the substrate by 8.6 and the thus resulting proximity of the alkoxide to the ester group causes the expected pronounced rate enhancement. (a) Nucleophile

N

O H

Complexation site

O H

O

O H

N

N

HN

H N N

O

N O O

O

O

N

O O

N N

N

O O

N

O O

N N

N

N O O

N

O

N

O O

N N

N

N

O O

N

O

N

8.6

8.5

O O

N

N 8.7

OH

O H

O2N

N N

O

N

8.4

(b)

N

O

N

+

N

O N

O

N

O O

HO

O

NO 2

N O O

O

–

H3N +

H H N O +O H O O O N

N

N N 3-Phenylbenzyl alcohol

N

(c)

N

OH

O H

O2N

O H O N

–

HN

N

N N N

O O N

O

O O

+ N

O O H3N +

N

HN

NO 2

N

Fast

N N

H H N O +O H O O O N

N

O

O

N

N N

Slow

N

H H N O +O H O O O N

N N

N

Figure 8.6: Molecular structures of the chymotrypsin mimic 8.4 proposed by Cram, the structurally simplified analogs 8.5, 8.6, and 8.7, and the model compound 3-phenylbenzyl alcohol (a). The reactions of 8.6 and 8.7 with the p-nitrophenol ester of L-alanine are shown in (b) and (c), respectively.

Motivated by these results, the imidazole containing receptor 8.7 was then prepared in a 30-step synthesis [9]. The subsequent kinetic studies revealed that this compound is acylated ca. 105 times faster than a noncomplexing model compound,

8.2 Stoichiometric transformations

471

even in the absence of an externally added base, indicating that the imidazole residue in 8.7 indeed contributes to the reaction. Unexpectedly, however, the reaction does not occur at the benzylic OH group, but at a nitrogen atom of the imidazole residue (Figure 8.6c) [10]. Accordingly, a derivative of 8.7 with a protected OH group also exhibits activity. The originally expected O-acylated product of 8.7 is produced only in a subsequent slow step by the migration of the acyl group from the acylated nitrogen to the oxygen atom. Since the mode of action of 8.7 thus differs from that of chymotrypsin, the synthesis of 8.4 was not attempted. Although Cram’s work demonstrated that an efficient transformation of suitable substrates is possible with properly designed receptors, it also showed that the deliberate design of enzyme mimics is far from trivial. The main reason is that efficient catalytic properties require a finely tuned interplay of binding properties and structural aspects of the receptor, which are influenced by external parameters such as solvation. Most importantly, high catalytic activity requires the receptor to bind to the transition state of the reaction, which makes a deliberate structural design challenging. Nevertheless, a number of interesting supramolecular catalysts have been identified, some of which are presented in the following sections. We start with systems that suffer from product inhibition and that therefore act in a stoichiometric fashion. In this context, systems are of particular interest in which the confinement within the receptor cavity induces a reactivity of the bound substrate not observed in the absence of the receptor. We then look at catalytic systems, the different strategies used to design them, and their mode of action. These catalysts span a wide range of structures, from covalently and noncovalently assembled receptors to coordination cages, including switchable systems and asymmetric organocatalysts. In the last part of the chapter, self-replicating molecules are presented that catalyze their own formation. The use of noncovalent interactions to control the assembly or optimize the properties of conventional catalysts is not considered [11].

8.2 Stoichiometric transformations 8.2.1 Transformation by functional group participation Given that enzymes operate in water, receptors active in the same environment should make a good starting point for the development of enzyme mimics. Particularly attractive in this context are cyclodextrins (CDs) (Section 4.1.4), not only because of their ability to bind various hydrophobic substrate in water, but also because they can be structurally varied in a wide range and thus tailored to meet the requirements of the substrate and its reaction. In addition, CDs are chiral and therefore potentially able to induce enantioselective substrate conversions. Of the known CD-based enzyme mimics, several are not catalytically active, however, either because they are irreversibly modified during the transformation or because of product inhibition.

472

8 Mediating molecular transformations

Notably, even native CDs accelerate the rate of certain reactions, for example the cleavage of phenol esters. First reports about this reactivity were published in the early 1960s by the group of Friedrich Cramer, followed shortly thereafter by important contributions from Myron L. Bender, who showed that CDs produce marked effects on the cleavage of phenyl acetates [12]. 3-Nitrophenyl acetate, for example, is cleaved in water at pH 10.6 and 25 °C ca. 100 times faster when an excess of α-CD is present. An even more pronounced effect is observed for 3-tert-butylphenyl acetate where the the reaction is accelerated by a factor of 226. The reason for these rate enhacements is the partial deprotonation of 2-OH groups along the wider rim of the CD cavity at high pH. Nucleophilic alkoxide groups are thus generated that approach the bound substrate in the complex and induce ester cleavage by reacting with the ester carbonyl group as illustrated in Figure 8.7. The reaction leads to an acetylated CD and the mode of action therefore resembles that of chymotrypsin, with the secondary hydroxy group assuming the role of the serine side chain of the enzyme. Since the CD acetylation is irreversible, the overall process is not catalytic. – O

(a) O

O OH

O

O

O

O

–

O

O

O

NO2

pH 10.6 NO2

NO2

O2N

α-CD

O

–

(b) COOR

Fe

COOR 8.8

R=

NO2

8.9

Figure 8.7: Schematic illustration of the α-cyclodextrin-mediated cleavage of 3-nitrophenyl acetate in water at pH 10.6 (a), and structures of the ferrocene and adamantane-derived substrates 8.8 and 8.9 (b).

Indications that the initial complexation of the substrate does indeed have an effect on the reaction are the considerably smaller effects of β- and γ-CD on the reaction rate as well as the influence of the substrate’s substitution pattern on the rate enhancement. 2-Nitro- and 4-nitrophenyl acetate are only cleaved, respectively, ten and three times faster in the presence of α-CD than in the absence, for example, suggesting that these substrates are not well arranged inside the cavity to react with a deprotonated 2-OH group. Considering that the bond cleavage is mediated by a nucleophilic alkoxide, factors of 100–200 are nevertheless modest. Ronald Breslow attributed these small effects to the fact that 3-nitrophenyl acetate has to

8.2 Stoichiometric transformations

473

partially leave the cavity to form the tetrahedral intermediate, which is detrimental for a fast conversion [13]. Accordingly, the ferrocene derivative 8.8, designed to orient the ester group in close proximity of a secondary OH group when bound by β-CD, is cleaved 330,000 times faster if the CD is present. With some ferrocene derivatives, rate enhancements of up to 5,900,000 have even been observed [14]. Orienting the ester close to the alkoxide group influences the first step of the ester cleavage, namely the formation of the tetrahedral intermediate. Product formation in the next step not only depends on the propensity of the leaving group to depart, but also on how facile the formation of the resulting planar ester group in the environment of the CD cavity is. If the substrate is too rigid, the corresponding structural rearrangement is difficult, causing the second step of the reaction to become rate determining. More flexible substrates such as the adamantane derivative 8.9 do not have this disadvantage [15]. In this case, the tetrahedral intermediate forms less rapidly than that of 8.8 because of the rotation of the ester group around the ethinyl group. This inherent flexibility facilitates the second step, however, rendering it much less dependent on the type of leaving group than in the case of the ferrocene. Breslow therefore concluded that it is necessary to consider the structural changes associated with the entire reaction when developing enzyme mimics to avoid an otherwise fast step to become rate determining [13]. Although native CDs also mediate reactions other than transesterifications [13], CD-based enzyme mimics are more often based on substituted derivatives whose reactivity is controlled by the appended functional group. An example is the β-CD derivative 8.10 (Figure 8.8a) with a pyridoxalamine group linked to the 6-position of one glucose unit [16]. Pyridoxal phosphate is the active form of vitamin B6 and responsible as cofactor in many biochemical processes for transamination reactions as well as decarboxylations, deaminations, and racemizations. Specifically, pyridoxalamine reacts with α-ketocarboxylic acids to form imines that, after a hydrogen shift followed by hydrolysis, afford the corresponding amino acids and pyridoxal (Figure 8.8b). In a (a)

(b)

N N

β-CD

HN

R

OH

R

O

+

COOH

OH

HN

–H2O

NH 2

COOH

3-(1 H-Indol-3-yl)-2oxopropanoic acid

R = β-Cyclodextrin

N

N S

H2N

R

HO N

8.10

HN

OH

HN

N

+ H 2O

+

R

N COOH

OH O

NH 2 COOH

Tryptophan

Figure 8.8: Molecular structure of the pyridoxalamine-containing β-CD 8.10 (a) and mechanism of the transamination reaction that converts an α-ketocarboxylic acid into an amino acid (b).

474

8 Mediating molecular transformations

similar fashion, 3-(1H-indol-3-yl)-2-oxopropanoic acid is converted by 8.10 into tryptophan in a reaction that is 200 times faster than the same conversion performed in the presence of a pyridoxalamine derivative lacking the CD ring. This result and the fact that α-ketocarboxylic acids with substituents smaller than the indole group are converted less readily into the corresponding amino acids are strong indications that the inclusion of the substrate into the CD ring plays a role in the conversion. Despite the chirality of 8.10, the enantioselectivity of product formation is small. Moreover, there is no turnover because the pyridoxalamine is not regenerated. Analogous catalytically CD derivatives are discussed in Section 8.3.1. Proximity effects also operate in the self-folding cavitands that were developed by Julius Rebek Jr. if they contain functional groups directed toward the cavity interior (Section 4.1.9) [17]. An example is 8.11 (Figure 8.9) with a Kemp’s triacid incorporated into one of the walls. In the complexes of this cavitand with suitable amines, the methyl ester group ends up in close proximity to the nitrogen atom of the guest. The methyl group is thus efficiently transferred onto this nitrogen atom in a nucleophilic substitution reaction, affording the free carboxylate group and the corresponding ammonium ion [18]. The reaction is particularly rapid in the case of quinuclidine that is converted into the N-methylquiniclidinium ion with a half-life of less than 3 min at room temperature. Amines less well bound by 8.11 react slower, and those that are preferably bound with the nitrogen atom oriented toward the cavity bottom such as morpholine not at all. A related Kemp’s triacid lacking the cavitand residue does not react with the same amines even at 100 °C, demonstrating that the main reason for the efficiency of 8.11 is the proximity of the reacting groups in the complex as well as their mutual orientation that is ideal for the substitution reaction. An additional advantage is the absence of solvent molecules in the complex that shield the reacting functional groups. Product inhibition, caused by the high affinity of the cavitand to the positively charged product and the irreversible modification of the carboxy group, prevent catalysis also in this case. R O R O R H N HO N H OH H R N N H H O O O

O

O

O R

ON

O

R R 8.11 (R = C2H5)

R

N

R O R O R H N HO N H OH H R N O N H H O N O O N O O O O R

ON

O

R

R R

N

R O – R O R H N HO N H OH H H R N O N H O N + O O O

N

O

O

O R

ON

R

R R

Figure 8.9: Molecular structure of the functionalized self-folding cavitand 8.11 and its reaction with quinuclidine, affording the corresponding free acid and the N-methylquinuclidinium ion. The front panel of the cavitand is not shown for reasons of clarity.

475

8.2 Stoichiometric transformations

The structurally related cavitand 8.12 bearing a 2-pyridone unit mediates the aminolysis of a choline-derived activated carbonate (Figure 8.10) [19]. The reaction involves the incorporation of the quaternary ammonium group of the substrate into the cavity of the cavitand, which arranges the carbonate group close to the 2-pyridone moiety. This group stabilizes the tetrahedral intermediate of the aminolysis by hydrogen bonding, thus favoring product formation. Although the rate enhancement is small, amounting to a factor of ca. two if 10 mol% of 8.12 are present, this cavitand exhibits turnover. It is presented here and not in Section 8.3.1 because of the structural relationship with 8.11.

R O H OH N N

H N R

O N HN H

O

N

O

NO2

O

+ O

O R

R 8.12 (R = C2H5)

R R

+ H2 N OH

N

O

O

N H NH2

O

O

O R

O N NH

O O

R

R

R R

O

N H – O O

O

+ O

N H

+ N

+

NO2

NO2 O

N H

R = cavitand

Figure 8.10: Molecular structure of the functionalized 2-pyridone-containing cavitand 8.12 and mechanism of the aminolysis of a choline-derived 4-nitrophenylcarbonate, illustrating the stabilization of the tetrahedral intermediate by the 2-pyridone unit. The front panel of 8.12 is not shown for reasons of clarity.

8.2.2 Transformation by confinement Bimolecular reactions between two substrates often proceed much faster if they are performed within the confined cavity of an appropriate receptor also if the receptor does not contain functional groups to mediate the bond forming reaction. The reason is again the proximity of the bound molecules, which causes the activation entropy of the reaction to be less unfavorable than that of the analogous reaction outside the cavity. The receptor could also induce characteristic effects on the regio- and/or stereoselectivity of the transformation as a consequence of the mutual orientation of the guests inside the cavity. Finally, the shielding of the reactive intermediates by the receptor potentially prevents unwanted side reactions, causing a reaction inside a cavity to proceed more cleanly than outside. In all cases, the receptor simply serves as a reaction vessel without being actively involved in the transformation [20]. Especially useful for these purposes are capsules or coordination cages in which the substrates are fully encapsulated and thus tightly held

476

8 Mediating molecular transformations

together. Such receptors have the further advantage that they are often assembled from simple building blocks, not requiring elaborate syntheses (Section 5.1). Since the actual reaction between the complexed molecules have to proceed spontaneously without the need for additional reagents or catalysts for which there is usually no room inside the cavity, the approach works particularly well for thermal cycloadditions such as Diels–Alder reactions or 1,3-dipolar cycloadditions. In these bimolecular reactions, the ternary complex between the receptor and the two substrate molecules is converted into the typically more stable binary complex of the product. Product inhibition is therefore usually observed, preventing catalysis. An early example of such a process is the cycloaddition between benzoquinone and 1,3-cyclohexadiene inside the cavity of Rebek’s molecular softball (Section 5.3.3), which is assembled from two molecules of 8.13 (Figure 8.11a) [21]. Product formation in this case proceeds 200 times faster than outside the softball, leading to a pronounced EM of 2.4 M. The selectivity of the reaction, which involves the preferred formation of the endo product, is not altered by the receptor. This is different if the Diels–Alder reaction between N-cyclohexylmaleimide and 9-hydroxymethylanthracene is performed in the presence of Fujita’s coordination cage assembled from [Pd(en)(NO3)2] and the tripodal ligand 8.14 (Section 5.6.3) [22]. Anthracene normally reacts in this reaction with the center ring acting as the dienophile because the loss of the aromatic stabilization is in this way smaller than that of other reaction pathways. If the reaction partners are incorporated into the coordination cage, however, their arrangement does not allow the cycloaddition to proceed in this manner. Instead, a product is formed that contains the bridge in a peripheral ring (Figure 8.11b). The open and solvent-exposed cavity of the bowlshaped receptor derived from ligand 8.15, in contrast, allows the cycloaddition to proceed with the normal regioselectivity (Figure 8.11c) [22]. Since its large cavity opening also facilitates guest exchange, this receptor acts as a catalyst. It mediates complete conversion of the substrates within 5 h when 10 mol% are present, whereas 24 h are required if the receptor concentration is reduced to 1 mol%. The hydrogen-bonded dimer of the deep cavitand 8.16 (Section 5.3.3) simultaneously binds phenylacetylene and phenylazide and thus drives the 1,3-dipolar cycloaddition between these two molecules (Figure 8.11d) [23]. This reaction is 30,000-fold faster inside the cavity than outside and exclusively affords the 1,4-disubstituted 1,2,3-triazole, whereas a mixture of the 1,4- and 1,5-disubstituted product is formed in the absence of the capsule. A way to induce reactions of molecules included into capsules or cages that do not react spontaneously is irradiation. Such photochemical reactions can proceed in a concerted fashion like the thermal cycloaddition discussed earlier, but they can also involve stepwise processes via intermediates whose reactivity is characteristically influenced by encapsulation. When irradiating α-(n-alkyl) dibenzyl ketones entrapped inside the dimer of Gibb’s octa acid 8.17 (Section 5.2), for example, radical species are formed by the homolytic cleavage of the α-carbon bond. The subsequent reactions of

477

8.2 Stoichiometric transformations

(a) O HN Ar HN

OH

O

N

O

N N

Ar N OH

O

N

N N O

O

OH

NH Ar NH

Ar N

O

OH

O +

O

8.13 (Ar = 4-n-heptylphenyl)

O

H 25° C p-Xylene-d10

H

O

O

(b) N O N

N

N N

N

O

+

N

HO

N

O

8.14

O

80° C, D2O HO

(c) N O N N

N

N

(d)

O N O

N

NO

NO O

R

R

O

N

N

O

10 mol% 25° C, D2O HO

H O N

O H N N

O

O

HO

8.15 H N

H O N O

O

+

N

O

O N N O

R R 8.16 (R = C11H23 )

N O O

N3 +

25° C Mesitylene-d12

N N N

Figure 8.11: Diels–Alder reactions between benzoquinone and 1,4-cyclohexadiene within the molecular softball (a), between N-cyclohexylmaleimide and 9-hydroxymethylanthracene in the presence of the coordination cages formed from [Pd(en)(NO3)2] and the tripodal ligands 8.14 (b) and 8.15 (c), and 1,3-dipolar cycloaddition between phenylacetylene and phenylazide inside the cavity of the self-assembled capsule formed from 8.16 (d).

these radicals characteristically depend on their orientation within the cavity (Figure 8.12) [24]. The substrate with the methyl group (n = 1) fills the capsule with the two aromatic residues occupying the hemispheres and the alkyl group residing in the equatorial region of the capsule. In this case, the initially formed radicals preferentially react under decarbonylation followed by recombination (Norrish type I reaction). While the formation of homocoupled products is unavoidable in solution, the reaction

478

8 Mediating molecular transformations

O –

O

O

O

O O

– O

O

O

– O

O

O

O

O

O O

O

–

O

O

O

O

OO

– O

O

O

– O

O O

– O 8.17

O

O

–

C n H 2n+1

O

C n H 2n+1 h·ν

n = 1–8

C n H 2n+1

Norrish type I product

C n H 2n+1 +

O

+

Rearranged products

O

+

O

+

Cn–2H2(n–2)+1

Norrish type II products

+ further minor products

Figure 8.12: Photochemical reactions of α-(n-alkyl) dibenzyl ketones incorporated into the dimer of the octa acid 8.17. The methyl derivative predominantly affords the decarbonylated heterocoupled product (a). The amount of this product decreases in favor of rearranged (b) and Norrish type II products (c) as the chain length becomes longer. Rearranged products are not observed for the substrates with hexyl, heptyl, and octyl chains. The octyl derivative affords predominantly the Norrish type II product.

only leads to heterocoupled products when performed inside the cavity. The arrangement of substrates with longer alkyl groups (n = 2–5) inside the cavity is less well defined because their alkyl and aryl groups compete for the space in one capsule hemisphere. As a consequence, additional reaction pathways for the photochemically generated intermediates become available, causing a progressive shift of the product ratio from the Norrish type I products toward rearranged and Norrish type II products. Once the length of the alkyl group reaches six or more carbon atoms, no more rearranged products are formed. The strong preference of the octyl group to reside within one hemisphere finally leads to the preferred formation of the Norrish type II product in the case of this substrate. By contrast, the same Norrish type II product is produced in solution in much smaller amounts. Thus, characteristic modes of inclusion of these substrates within the capsule cause them to react via distinct pathways that differ from those outside the cavity. Other examples of photochemical reactions inside the capsule formed from 8.17 are the photoisomerization of (E)-4,4ʹ-dimethylstilbene and the photochemical dimerization of 4-methylstyrene [25]. (E)-4,4ʹ-Dimethylstilbene can be photochemically isomerized

479

8.2 Stoichiometric transformations

into the Z-isomer, leading to a photostationary state in the absence of the capsule with an E/Z ratio of ca. 1:3. When (E)-4,4ʹ-dimethylstilbene is incorporated into the capsule and irradiated, a much lower conversion into (Z)-4,4ʹ-dimethylstilbene is observed because this bulky isomer is a poor guest (Figure 8.13a). Consequently, the capsule also prevents the formation of the phenanthrene that is formed in solution from (Z)-4,4ʹ-dimethylstilbene in small amounts. In the case of 4-methylstyrene, two molecules dimerize upon irradiation, giving preferentially 1,2-disubstituted cyclobutanes. The same reaction performed inside the capsule leads to an almost 1:1 mixture of two other products, one being the 1,3-disubstituted cyclobutane, again illustrating that the arrangement of the guests inside the capsule controls product formation (Figure 8.13b). (a) h·𝜈

+

Without capsule With capsule

+

18% 85%

76% 15%

6% 0%

(b)

2

h·𝜈

Without capsule With capsule

+

83% 0%

+

10% 0%

+

0% 55%

7% 45%

Figure 8.13: Effect of the dimer of octa acid 8.17 on the photoisomerization of (E)-4,4ʹdimethylstilbene (a) and the photoinduced dimerization of 4-methylstyrene (b).

Effects on the outcome of photochemical [2 + 2]cycloadditions are also observed of substrates included into the coordination cage derived from 8.14 [26]. Two acenaphthylene molecules dimerize inside this cage upon irradiation, for example, yielding only the syn-isomer of the product because the mutual arrangement of the reaction partners does not allow the formation of the more extended anti-isomer (Figure 8.14a). Of the four stereoisomers that result from the photodimerization of 1methylacenaphthylene, only the syn-isomer with the 1,3-cis dimethylated cyclobutane ring is formed in the presence of the cage (Figure 8.14b). The water-soluble deep cavitand 8.18 (Figure 8.15a) shows that receptors with cavities open to the surrounding solvent can also serve as reaction vessels [27]. In the absence of suitable guests, 8.18 adopts the kite conformation in water that is prone to self-assemble, forming an unreceptive velcrand dimer as discussed in Section 5.2. Suitable substrates such as long chain alkanes shift the equilibrium toward the vase conformation of the monomeric velcrand whose deep cavity thus

480

8 Mediating molecular transformations

(a) h·𝜈 2

+

syn

anti

(b)

2

h·𝜈

1,3-cis-syn

+

+

+

1,2-cis-syn

1,3-trans-anti

1,2-trans-anti

Figure 8.14: Effect of the coordination cage derived from ligand 8.14 on the photochemical [2 + 2] cycloaddtion of acenaphthylene (a) and 1-methylacenaphthylene (b). The products depicted in gray are not formed in the presence of the coordination cage whose structure is shown in the inset.

becomes available for substrate binding. In the corresponding complexes, the alkanes adopt U-shaped conformations with the chain ends exposed to the solvent while atoms near the center of the chain are buried deep within the cavity. The chain ends of the guests are thus pushed together by complex formation as illustrated by the calculated structure of the 12-aminododecanoic acid complex of 8.18 (Figure 8.15b). The result is a good preorganization of α,ω-difunctionalized substrates to cyclize. Ring closure is induced in the case of 12-aminododecanoic acid by treating the complex with an appropriate coupling reagent, leading to a ca. threefold higher yield of the cyclic product than in the absence of the cavitand (Figure 8.15c) [28]. In a similar way, bis(lactams) are formed from α,ω-diamines and suitable activated diesters in an up to ten times higher yield when performing the reaction under the influence of 8.18 (Figure 8.15d) [29]. Since the product remains inside the cavity after cyclization, no turnover occurs. In many of the examples discussed in this chapter, the receptor serves to preorganize the bound molecules, thus decreasing the Gibbs free energy of activation of a specific reaction pathway. As a consequence, the reaction becomes faster, allowing it to outcompete potential alternative pathways. It is thus important to note that the effect exerted by the receptors on the reaction outcome is nothing more than a kinetic template effect. The only difference to reactions in which the template serves to facilitate a receptor synthesis is that the receptor itself acts as template in these reactions by facilitating the transformation taking place inside its cavity. The principles on which the template effects rely are otherwise the same as explained before (Section 5.1).

481

8.3 Catalytic transformations

(a)

O

O

O NO

N N

N

N N

N

N

(c) H2N

O O

O

O R

O

O

1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide N-hydroxysulfosuccinimide D2O

HN

R

R R

8.18 R = Cl

(b)

O O

O

COOH

–

+ N

N

(d) H2N

O NH2

O

O N

O

O

O

HN

NH

O

O O

N

O

D2O

Figure 8.15: Molecular structure of the deep cavitand 8.18 (a), calculated structure of the 2aminododecanoic acid complex of 8.18 (b), and schematic illustration of the cyclization of 12aminododecanoic acid (c) and an α,ω-diamine (d) under the influence of 8.18. The side chains of the cavitand in the calculated structure are omitted for reasons of clarity.

8.3 Catalytic transformations 8.3.1 Transformation by functional group participation Functionalized receptors are able to promote the chemical conversion of the bound substrate, but for the receptor to act in a catalytic fashion, several conditions must be met. First, the structural integrity of the receptor must be guaranteed during the full catalytic cycle. If substrate conversion leads to structural changes, a way must exist to reverse these changes and thus regenerate the receptor in its catalytically active form. Second, the receptor should ideally bind more strongly to the transition state of the reaction than to either the substrate or the product to prevent product inhibition. Although this situation would most closely mimic the behavior of enzymes, it is usually very difficult to achieve by design as we saw when discussing Cram’s chymotrypsin mimic. Therefore, the receptor should at least bind stronger to the substrate than to the product(s), allowing new substrate molecules to replace the product(s) from the complex after the reaction is complete. The latter is the case, for example, in an early CD-based hydrolase mimic developed in the Breslow group [30]. The respective bis(cyclodextrin) 8.19 (Figure 8.16a),

482

8 Mediating molecular transformations

which contains two β-CD rings connected via a linker with a copper(II)-2,2ʹ-bipyridyl unit, has a significantly higher affinity for substrates with two appropriately spaced hydrophobic subunits than a single β-CD ring has for either of these subunits. In substrates with a central ester group such as 8.20, the cleavable group is moreover placed close to the metal center. This metal center serves to activate water molecules and thus induces the rapid hydroxide-mediated hydrolysis of the bound substrate (Figure 8.16b). The resulting ternary complex is less stable than the complex of intact 8.20, allowing further substrate molecules to displace the products. Since the copper complex is also not consumed during the reaction, the overall process is catalytic with a pronounced rate enhancement over the uncatalyzed reaction by a factor of 220,000. (a)

(b) S

S N

N S

Cu2+

S N

N

Cu2+ HO

8.19

O O O O

NO2

NO2

8.20

Figure 8.16: Molecular structures of the bis(cyclodextrin) 8.19 and its substrate 8.20 (a), and proposed mechanism of the ester cleavage mediated by the central copper(II) complex (b).

Other CD-derived catalysts are the tetrakis(cyclodextrin) derivative 8.21 (Figure 8.17a) with four CD moieties surrounding a porphyrin-manganese(III) complex and the bis(imidazole) 8.22 (Figure 8.17b). Both compounds exhibit turnover because the substrate and the product of the respective reactions are bound with comparable affinities. In the case of 8.21, the manganese(III) complex has to be activated first by treatment with iodosobenzene [31]. The resulting oxo complex then mediates the epoxidation of stilbene derivatives or the oxidation of a nonactivated C–H bond in dihydrostilbene [32]. Note that the substrates used in these reactions contain bulky end groups to ensure that binding to the CD moieties positions the reacting groups close to the metal center. The same strategy allows the regioselective oxidation of C–H bonds in steroid-derived substrates. Bis(imidazole) 8.22 facilitates the hydrolysis of a cyclic phosphate, with one imidazole unit acting as a general base to activate a water molecule and the protonated form of the other imidazole group transferring its proton onto the intermediate [33]. The reaction is most efficient if the two imidazole units are located in neighboring glucose units along the β-CD ring. In addition, the product with the phosphate group on the oxygen atom in meta position to the tert-butyl group is almost exclusively

483

8.3 Catalytic transformations

(a) β-CD R

R R

S

β-CD

β-CD

8.21 N N Mn3+ N N

S

8.21 PhIO R

S

R 8.21 PhIO R

O

R

R HO

R=

S

NO2

H N

COOH

COOH

β-CD (b)

H + N

N N

H O

– O

H + N – O O O P O

O

R=

O

N

H

N O O P O N

8.22 H2O

N

8.22

N N

HO OH – H N+ O O P O N

– OH O

β-CD

HO

O P

O

O HO

–

O P

O OH

+

–8.22

Figure 8.17: Molecular structure of the tetrakis(cyclodextrin) 8.21 along with reaction schemes showing the transformations that are catalyzed by this compound (a). In (b), the schematic structure and mode of action of the β-CD bis(imidazole) 8.22 is shown. The phosphate ester shaded in gray is practically not formed (300 nm) unilamellar vesicles. These vesicles retain in their interior the solution present during their preparation, which makes it possible to trap additional components such as salts or dye molecules inside the vesicles. The solution is subsequently subjected to dialysis or size exclusion chromatography to separate the nonencapsulated compounds, thus yielding vesicles with a concentration gradient along their membranes (Figure 9.4c). After incorporating a transporting system into the bilayer, which is usually achieved by adding a concentrated solution of a transporter in a water-miscible solvent such as DMSO, membrane transport is initiated and followed by monitoring the change of the analyte concentration inside or outside the vesicles. If the transporter causes the release of ions from the vesicle interior, the increase of the concentration of the respective ion in the surrounding solution is followed electrochemically, for example, by using ion-selective electrodes. Conversely, it is possible to entrap dyes inside the vesicles that respond to the presence of certain analytes by a change in fluorescence. Dyes exist that sensitively report the pH change of the solution inside the vesicle, for example, or the presence of certain ions. These experiments are usually performed by first recording the signal in the absence of the transporter. The addition of the transporter at a certain time then results in the progressive increase of the concentration of the analyte, which is recorded. Finally, a detergent is added to destroy the vesicles and obtain the maximum observable signal under the chosen conditions. The resulting curves provide quantitative information about the transport rates and, in turn, the transport activities of the investigated systems. Special strategies exist to also obtain information about the transport mechanism. Whether a transporter acts as a carrier or channel is often assessed, for example, by adding compounds to the lipid bilayer that reduce the fluidity. This addition should decrease the rate of ion transport if it involves a carrier mechanism, whereas the transport rate through a channel should not be affected. A counterion effect on the transport rate is often indicative of a symport mechanism.

518

9 Transporting molecules

9.2 Cation transport 9.2.1 Channels The transport of monovalent metal ions such as sodium or potassium across lipid membranes is crucial in biological systems for signal transmission or the conversation of the membrane potential. Membranes thus contain specialized proteins to mediate cation transport and to ensure that it proceeds selectively. Conversely, compounds that disturb the membrane potential are harmful to cells and often exhibit antibacterial activity. A protein responsible for maintaining the membrane potential is the potassium channel protein. The structure of this channel is known at atomic resolution, providing insight into the structural reasons for its transport properties and its 105-fold selectivity for potassium over sodium ions [2]. This channel protein is a tetramer of smaller subunits that surround a central pore through which the potassium ions pass. The pore is characterized by wider tunnels at both sides lined with negatively charged amino acid side chains that serve to increase the local cation concentration with respect to the surrounding environment. In these regions of the protein, the cations are still fully solvated. The narrow region of the pore between these tunnels, the so-called selectivity filter, ensures potassium selectivity. This pore has a radius that exactly matches the diameter of a potassium ion and is lined by carbonyl groups as illustrated by the corresponding section of the crystal structure shown in Figure 9.5a. Potassium ions thus pass the pore after having shed their solvation shell and are efficiently stabilized within the selectivity

(a)

(b)

Figure 9.5: Section of the crystal structure of the potassium channel protein from Streptomyces lividans, illustrating the environment in the filter region of the pore (a), and NMR structure of the head-to-head dimer of gramicidin A in sodium dodecyl sulfate micelles (b). Only two of the four chains surrounding the potassium ions in the selectivity filter are shown in (a). The protons in (b) are omitted for reasons of clarity and the carbon atoms of the protein backbones of the two gramicidin A subunits are shown in different shades of gray.

9.2 Cation transport

519

filter by ion–dipole interactions with the carbonyl groups. Slightly larger rubidium ions also pass. Desolvated sodium ions, on the other hand, are too small to efficiently interact with the carbonyl groups while solvated ones are too large. They are therefore unable to move through the channel, explaining its pronounced potassium selectivity. A channel protein that causes the collapse of the membrane potential is gramicidin A. This linear α-helical peptide is composed of fifteen alternating L- and D-amino acids. Upon self-assembly of two gramicidin A molecules in the interior of a lipid bilayer, a channel is formed through which up to 107 monocations pass in per second. The structure of the head-to-head dimer of gramicidin A is shown in Figure 9.5b to illustrate the tube-like structure [3]. Both structures provide crucial information about the concepts on which the design of an artificial cation channel could be based. Such channels either encompass a single compound that spans the whole width of a lipid bilayer or an assembly of several subunits. Furthermore, pore producing compounds alone are sufficient to achieve transport, with transport selectivity in this case depending mainly on the pore diameter. The additional presence of binding sites for cations within the pore, such as anionic residues or oxygen atoms with which the cation interacts by ion–dipole interactions, is beneficial. All of these concepts have been realized in synthetic systems [4]. Synthetic channels are formed from the tube-forming p-octiphenyl-peptide conjugates developed by Stefan Matile, which were discussed in Section 5.3.4. These compounds assemble into tetrameric tubes whose height is comparable to the hydrophobic core thickness of lipid bilayers. If hydrophobic amino acid side chains in the peptide residues end up being arranged on the outer surface of the tube, selfassembly occurs within lipid bilayers. The resulting structures resemble natural βbarrels, pore-forming proteins containing cyclically arranged β-sheets of interacting peptide strands [5]. The pore diameter of these rigid-rod β-barrels can be controlled by varying the length of the peptide residues attached to the p-octiphenyl scaffold, while the transport properties depend on the nature of the substituents arranged along the inner pore surface, with negatively charged residues producing a suitable environment for cation transport. The p-octiphenyl derivative 9.1a (Figure 9.6) arranges aspartate residues along the inner surface of a relatively large pore (diameter ca. 10 Å), for example, and therefore preferentially transports potassium over chloride ions [6]. According to bilayer conductance measurements, the lifetime of the corresponding channels is rather short (τ < 0.5 ms), indicating that these βbarrels are not very stable. Stability increases in the presence of magnesium ions, which likely crosslink the carboxylate groups along the pore interior, but once these cations are present, the cation/anion selectivity of the channel is almost lost. Counterintuitively, the rigid-rod β-barrel derived from 9.1b with internal arginine and histidine residues is also cation selective, although it should contain positively charged guanidinium groups along the inner surface. The reason is the simultaneous

520

9 Transporting molecules

Peptide

O O

N H

O

H N O

1

R

N H

O

H N O

R

NH2

N H

2

O

O R

H N

R O

HN

Peptide

O

O

O

O

9.1a R1 = R2 =

COOH

N H

3

O O

O

NH

Peptide 9.1b R 1 =

Peptide

NH

R2 =

NH2

R

N H

R O

N 9.2 R = NH

NH

Figure 9.6: Molecular structures of the channel-forming p-octiphenyl-peptide conjugates 9.1a,b and of the cyclopeptide 9.2. For a schematic illustration of the structure of the tetrameric assembly formed from 9.1a,b, see Figure 5.34.

presence of phosphate ions in the aqueous solution that interact with the guanidinium residues and render the pore interior overall negatively charged. This situation changes when protonating the phosphate and the imidazole groups. The excess of positive charges thus produced inside the pore leads to an inversion of ion selectivity. Other applications of this versatile family of β-barrel mimics exist, including the use for the transport of anions and larger (charged) organic molecules as well as for sensing and catalysis [5]. Ghadiri’s self-assembling cyclopeptides (Section 5.3.4) have also been used to mediate cation transport. The cyclic peptide 9.2 (Figure 9.6), for example, selfassembles within a lipid bilayer to form stacks of ca. six rings which act as discrete ion channels [7]. According to single channel conductance measurements, the rate of potassium transport amounts to 1.9 × 107 ions s–1, which is greater than that induced by gramicidin A under similar conditions. Arranging oxygen atoms along the inner surface of a synthetic ion channel often relies on the use of pore-forming crown ether derivatives, for example the α-helical peptide 9.3 or the p-octiphenyl derivative 9.4 (Figure 9.7) [4]. When these compounds are incorporated into lipid bilayers, their crown ether moieties stack, thus producing channels through which cations pass. An alternative successful design of unimolecular ion channels are the so-called hydraphiles, which were introduced by George W. Gokel [8]. These large rings, an example is 9.5, contain crown ether units, two of which face the membrane surfaces when incorporated into a lipid bilayer, forming the channel gates. The alkyl chains and the remaining crown ethers span the bilayer, thus generating a hydrophilic region within its interior that is crucial for ensuring good transport activity. Particularly good central residues are those that interact with water molecules

521

9.2 Cation transport

O

O

O

O O

O

O

O

O O

O O

O

N

O O O

O

O

O

O S O

S 3

Boc

H N

O

O

H N

N H

N H

O

O

H N

N H

O

O

H N

OMe

3

O

O O N

O

O

O 9.3

O

O

O

9.4

O

N N

NH3

N O

N

O

O

O

O

O

O

O

O

O

9.5 O

N

O

N N

N O

O

COOR

ROOC O

HO

O

ROOC O

OH 9.7

COOR

ROOC

O

COOR

O O

O

ROOC

COOR 9.6

R=

O O O

S O

O

O

O O

S

COOH

O

Figure 9.7: Molecular structures of the crown ether derived unimolecular ion channels 9.3, 9.4, 9.5, and 9.6. The tartaric acid derivative 9.7 is a structurally simpler analog of 9.6 with similar ion transport activity. Compound 9.5 is a hydraphile.

while those that strongly bind to ions reduce the transport activity. The propensity of oligoethylene chains to interact with cations also lowers transport activity, which is why hydraphiles with hydrocarbon chains are typically more active. Somewhat related to the design of 9.5 is the hexasubstituted crown ether 9.6 described by Thomas M. Fyles in which the crown ether ring serves as the central relay, while the six substituents ensure the arrangement of the channel within the bilayer [9]. This compound exhibits a transport activity comparable to that of gramicidin A. Although it is tempting to attribute this activity to the presence of the crown ether, the fact that the structurally simpler analog 9.7 has about the same level of activity indicates that the crown ether ring mainly serves to arrange the substituents in a proper fashion but has a limited functional role [4].

522

9 Transporting molecules

9.2.2 Carriers Natural macrocyclic products such as valinomycin 9.8, nonactin 9.9 and other members of the nactin family, or the acyclic monesin 9.10 (Figure 9.8a) transport cations across lipid membranes by using a carrier mechanism. They thus induce the collapse of the transmembrane electrochemical gradient, which is harmful to cells and renders these ionophores useful and potent antibiotics. Their mode of action involves the complexation of the cation inside a cleft or cavity where it engages in ion–dipole interactions with surrounding oxygen atoms. To illustrate such a structure, the potassium complex of valinomycin, which was already discussed in Section 4.1, is shown in Figure 9.8b. Complex formation in this case produces a hydrophilic layer around the ion, which is mainly characterized by a cyclic array of the isopropyl groups of the amino acid and the α-hydroxycarboxylic acid subunits. This layer allows the complex to partition into the hydrophobic environment of a lipid bilayer without requiring an

(a)

(b) O O

N H

O

O

O

HN

O NH

O O

O

O

O

HN

O

O

NH

O

O

O

H N

O

O 9.8 H O

H

H

O O

H

O

H

O O

O H

O O

H

O H

O

O

H

OH OH

O O

O

O

H

O

OH

H O

H

O 9.9

H

OH 9.10

Figure 9.8: Molecular structures of valinomycin 9.8, nonactin 9.9, and monesin 9.10 (a), and crystal structure of the potassium complex of 9.8 (b).

9.3 Anion transport

523

anion to compensate the positive charge. Valinomycin thus mediates the uniport of potassium ions along the concentration gradient and does so selectively over slightly smaller and therefore less strongly bound sodium ions. Other ionophores form similar cation complexes but are often less selective than 9.8. The structures of these ionophores serve as an inspiration to develop structurally related cation carriers. There exists, for example, a whole family of macrocyclic peptides and depsipeptides with structures and cation transport properties related to valinomycin [10]. The number of crown ether derivatives investigated with respect to their cation transport properties is even larger. This large body of work mainly involved the use of U-tube transport experiments, thus also providing information about the structural parameters influencing the extraction of metal ions from the aqueous into the organic phase and affording a correlation between transport efficiency and the properties of crown ether complexes [11]. Less is known about the ability of crown ethers to mediate cation transport across lipid membranes although the toxicity of crown ethers suggests that they might be able to disrupt the transmembrane potential in a similar way as valinomycin [12].

9.3 Anion transport 9.3.1 Channels A structurally diverse family of proteins is responsible for the selective transport of chloride anions across cell membranes. These so-called chloride ion channels (ClC) are involved in the stabilization of the membrane resting potential and mediate the flow of salt and water across epithelial barriers. According to crystal structures, these proteins form a homodimer within the membrane in which the two subunits are arranged in opposite orientations [13]. Each subunit has its own selectivity filter in which a desolvated chloride anion is stabilized by a combination of electrostatic interactions with the dipoles of α-helices converging toward the binding site and direct hydrogen bonding interactions with two backbone NH groups and the OH groups from a serine and a tyrosine residue. While anions find an ideal environment within this filter region, cations would experience repulsive interactions, explaining the pronounced ion selectivity of these channels. The pore allows the passage of chloride and bromide but is blocked by larger anions such as iodide or thiocyanate, which thus inhibit chloride transport. Synthetic systems that allow the incorporation of pores into lipid bilayers with an excess of positive charges along the inner surface can again be based on Matile’s rigid-rod β-barrels, which are discussed in the previous chapter [5]. In the context of anion transport, another transporter family introduced by Matile is also noteworthy because its mode of action relies on anion–π interactions, a type of noncovalent interactions not very often used in supramolecular chemistry. These transporters are

524

9 Transporting molecules

rigid rod-like molecules containing electron-deficient aromatic subunits with the right length to span lipid bilayers. The oligo(p-phenylene)-N,N-naphthalenediimide 9.11 is an example (Figure 9.9) [14, 15]. After incorporation into the bilayer, the aromatic subunits of this so-called π-slide produce regions within the hydrophobic part of the membrane characterized by a substantial positive electrostatic potential. These regions serve to guide the movement of the anion across the membrane, which likely proceeds via a hopping mechanism. Transport activity depends on the peripheral substituents, with 9.11a carrying one charged substituent showing the highest activity, while the activities are substantially lower of 9.11b with charges at both termini or 9.11c with two uncharged residues. This trend suggests that the orientation of the rods within the membrane is important for transport activity. Anion selectivity, on the other hand, is best for the uncharged 9.11c because it benefits from the proximity of rods within the membrane [16]. These systems usually transport chloride more efficiently than the other halides and possess selectivity for nitrate among oxoanions. R1

O O

O

O

O

O

O

N

N

N

N

N

N

O

O

O

O

O

O

HN NH

9.11a R1 = NHBoc, 9.11b R1 = R2 = NH3 9.11c

+

R2 = NH3

+

O

R2

R1 = R2 = NHBoc

Figure 9.9: Molecular structures of the π-slides 9.11a–c and electrostatic potential surface of a truncated analog of 9.11a to illustrate the positive electrostatic potentials produced by the naphthalenediimide subunits. The color coding covers a potential range from –75 to +75 kJ mol–1, with red and blue signifying values greater or equal to the absolute maximum in negative and positive potential, respectively (Boc = tert-butyloxycarbonyl).

Structurally more closely related to natural anion channels are the amphiphilic peptides developed in the Gokel group of which 9.12 is an example (Figure 9.10) [17]. The two octadecyl chains at the N-terminus help to anchor this compound within the membrane. The CH2OCH2 residue resembles the glyceryl residue of phospholipids, and the (Gly)3Pro(Gly)3 subunit mimics peptide sequences that are involved in the anion transport of natural chloride channels.

9.3 Anion transport

O H37C18

O

N C18H37

O

N H

H N O

O N H

H N

N O

O

O N H

H N

525

O O

9.12

O

Figure 9.10: Molecular structure of the channel-forming amphiphilic peptide 9.12 that promotes chloride transport.

9.12 dimerizes within the lipid bilayer to form pores with a diameter of 7–8 Å that promote chloride transport with a ca. 10-fold selectivity over potassium. Each structural element of this compound has a characteristic impact on transport activity. A derivative of 9.12 with a leucine instead of the proline subunit is significantly less active, for example, while a derivative with decyl instead of the octadecyl residues has a higher activity but is essentially unselective. The characterization of the behavior of a series of structural analogs of these transporters thus provided deep insight into principles governing anion transport [17].

9.3.2 Carriers Prominent natural anion carriers are the prodigiosins, pyrrole alkaloids produced by microorganisms such as Serratia marcescens [18]. These alkaloids are structurally rather diverse but the pyrrolylopyrromethene unit with a methoxy group in 4-position of the central ring is found in all derivatives (Figure 9.11). Prodigiosins possess a number of biological activities, some of which linked to their ability to act as antiporters for chloride and bicarbonate or as symporters for protons and chloride anions across liposomal membranes. The proton-coupled chloride transport is somewhat related to phase-transfer catalysis in that the protonation of the weakly basic nitrogen atom of the prodigiosin azafulvene subunit converts the whole molecule into a large lipophilic cation that associates with the chloride counterion and thus mediates its transport across the membrane. Chloride binding is mainly due to electrostatic interaction but additional stabilizing effects of hydrogen bonds and/or charge transfer interactions could also play a role.

R

1

R2

N NH

H N OCH3

Figure 9.11: General structure of prodigiosins. The different members of the prodigiosin family differ in the substituents R1 and R2.

526

9 Transporting molecules

The structural simplicity of prodigionsins and their cleft-like structure, featuring a converging arrangement of hydrogen bond donors, indicate that typical anion receptors with suitably placed binding sites should also be able to act as anion carriers. The number of synthetic compounds that mediate anion transport is indeed large as work of Philip A. Gale clearly shows [19]. The examples depicted in Figure 9.12 demonstrate that some transporters resemble natural carriers while others are structurally entirely unrelated, containing squaramide or urea groups for anion binding or the binding sites attached to a tripodal scaffold.

tBu NH

F3C

N

F3C

OCH3

S

O N H

O H

NH

HN

O

O 9.13

Bu

O

N H

N H

N H

F5C6

HN

NC

N

NH NH S F5C6

Bu

O

NH

HN

CF3 S HN C F 6 5

NH

CN HN

O O H

CF3

HN

NH O

HN N

9.14

Bu O

N H

O

Bu Bu O

O

N H

N H O

9.15

Bu

O 9.16

Figure 9.12: Examples of anion receptors that mediate anion transport, including the derivatives 9.13–9.16 of which 9.13 and 9.14 are preorganized for anion binding, rendering them carriers, whereas the lack of preorganization causes 9.15 and 9.16 to be inactive.

All of these compounds contain hydrogen bond donors to interact with the substrate, with the isophthalamide derivatives 9.13, 9.15, and 9.16 representing particularly instructive systems because they illustrate the importance of available hydrogen bond donors for anion transport [20]. The two hydroxy groups in 9.13 stabilize a conformation in which both NH groups are able to simultaneously interact with an anion. Despite its structural simplicity, this compound therefore efficiently mediates chloride/ nitrate and chloride/bicarbonate exchange across lipid bilayers. For similar reasons, pyridine-2,6-dicarboxamides such as 9.14 are also efficient chloride carriers [21]. The diamides 9.15 and 9.16, on the other hand, that are either not well preorganized (9.15) or fixed in a conformation in which the NH groups are unavailable for anion binding (9.16), are inactive. Anthony P. Davis introduced cholic acid-based anion receptors, so-called cholapods, as chloride transporters [22]. A member of this family containing three thiourea groups along the concave face of the steroidal scaffold is 9.17 (Figure 9.13). These

527

9.3 Anion transport

OC20H41

S

NH

OC8H17

O

HN HN

O

S S NHR NHR

S S

NH

HN

RHN

RHN

9.17 (R = 4-nitrophenyl)

S

NHR 9.18 (R = 3,5-bis(trifluoromethyl)phenyl)

RHN

HN

NH S

NH

S NHR

RNH 9.19 (R = 3,5-bis(trifluoromethyl)phenyl)

Figure 9.13: Molecular structures of cholapod 9.17 and the structurally simpler analogs 9.18 and 9.19 derived from trans-decaline and cyclohexane, respectively.

binding sites are too far away from each other to interact intramolecularly, but have the optimal distance to simultaneously interact with a chloride anion. As a consequence, cholapods possess very high chloride affinities in chloroform. According to the results of binding studies performed by using Cram’s picrate extraction method, the chloride complex of 9.17 has a log Ka of 11.3, for example. The lipophilicity of these receptors moreover renders their anion complexes soluble in the hydrophobic environment of lipid bilayers, allowing cholapods to act as efficient chloride/nitrate antiporters. Transport activity generally correlates with chloride affinity, but there seems to be an affinity limit beyond which no increase in the transport rate is possible and any further increase of affinity becomes unproductive [23]. The steroidal scaffold is not required for transport activity since anion receptors such as 9.18 and 9.19, based on trans-decalin or cyclohexane, respectively, are also very effective [22]. Another family of steroid-derived transporters are the molecular umbrellas developed by Steven L. Regen [24]. These compounds typically arrange several cholic acid subunits or other steroid-derived residues on a flexible scaffold. The disubstituted spermidine derivative 9.20 (Figure 9.14) illustrates this structure. Depending on the polarity of the environment, these compounds adopt different conformations. In polar media, the hydrophilic faces of the cholic acid subunits are exposed to the

O

O N H

N H

N O

R

HO

OH OH

HO OH

HO

9.20 Figure 9.14: Molecular structure of the molecular umbrella 9.20. The substituent R in this compound can be varied and allows tuning the properties.

528

9 Transporting molecules

solvent, while they converge in apolar solvents, producing a polar cavity that is shielded by the hydrophobic outer faces of the steroid hydrocarbon scaffolds. Accordingly, molecular umbrellas solubilize polar substrates such as nucleotides, oligonucleotides, and also hydrophilic peptides within the interior of lipid bilayers and transport these substrates from one side of the membrane to the other by passive diffusion.

9.4 Water transport Although water molecules are able to penetrate the hydrophobic interior of lipid bilayers directly, specialized proteins exist that are involved in regulating the osmotic pressure and allow water to rapidly move from one side of the membrane to the other [25]. These so-called aquaporins consist of bundles of transmembrane αhelices that span the membrane and surround a channel through which water molecules can penetrate but ions or other solutes cannot. The channel consists of larger conical cavities on both sides that are linked through a narrow ca. 20 Å long pore, giving it an hourglass-like shape. In the larger cavities, the water molecules are intermolecularly hydrogen-bonded as in bulk solution. The pore, on the other hand, is so narrow that the water molecules have to move through in single file. They enter the pore with the oxygen atom first but change the orientation as they travel further to enable hydrogen bonds to asparagine side chains within the pore region. This reorientation prevents the water molecules within the pore to form a continuous chain of hydrogen bonds, which, together with the presence of positively charged or polarized residues positioned along the pore surface, explains why aquaporins do not allow protons to pass. Examples of synthetic systems capable of transporting water across lipid bilayers are the ureido imidazole derivative 9.21 and the pillar[5]arene 9.22 (Figure 9.15). 9.21 crystallizes from water to form a layered structure, stabilized by intermolecular hydrogen bonds between the urea groups [26]. This structure contains channels surrounded by four imidazole units in which an infinite chain of water molecules resides. Each water molecule donates one hydrogen bond to its neighbor and one to a nitrogen atom of a surrounding imidazole residue, similar to the arrangement of water molecules within the pore of aquaporins (Figure 9.15a). 9.21 is believed to self-assemble into similar structures within the hydrophobic interior of lipid bilayers because liposomes dispersed in water and filled with aqueous sodium chloride rapidly start to swell in the presence of 9.21 as a result of water molecules traversing the channels from the outside to the inside. The pillar[5]arene 9.22 produces a similar effect [27]. This compound contains substituents on both sides of the pillar[5]arene ring with hydrazide subunits, which stabilize a tube-like structure by intramolecular hydrogen bonds that is long enough to

9.4 Water transport

(a)

(b)

R

R

R

529

RR

O O HN O NH O HN NH NH NH HN NH HN O O HNO O O

O

N

O N H 9.21

N H

N H O O HN O NH O HN NH NHNH HN NH HN O OHN O O O

O

O

O

O O

O

O O

O O

O

O O O HNO HN O NH NH NH NH NH O HN O HN HN O O O

O O O HN O HN O NH NH NH NH NH O HN O HN HN O O O

R

R

R

R

9.22 (R = (CH2)2COOMe)

R

Figure 9.15: Molecular structure of the self-assembling ureido imidazole 9.21 and section of its crystal structure to illustrate the arrangement of water molecules within the channel formed from stacked subunits (a), and structure of the pillar[5]arene 9.22 and crystal structure of a shorter analog of 9.22 with a chain of four water molecules within the cavity. Hydrogen atoms in the crystal structures except those at nitrogen atoms are omitted for reasons of clarity. The side chains of 9.21 are truncated in the crystal structure in (a). The water molecules are shown as space-filling models. The protons of the water molecules in the crystal structure in (b) were not located.

span a lipid bilayer. Such tubes host linear chains of water molecules in the solid state (Figure 9.15b), suggesting that they are able to allow the passage of water molecules after their incorporation into a lipid bilayer, which is indeed the case. Since the water molecules within the channels formed by 9.21 and 9.22 form a continuous chain of hydrogen bonds, both channels also allow the passage of protons that are passed on

530

9 Transporting molecules

from one water molecule to the next through the formation and simultaneous cleavage of OH bonds (Grotthuss mechanism). Developing channels that transport water but not protons is therefore still an unsolved challenge.

Bibliography [1] [2]

[3] [4] [5] [6]

[7] [8]

[9] [10] [11] [12]

[13] [14] [15]

[16] [17] [18] [19]

Sessa G, Weissmann G. Phospholipid spherules (liposomes) as a model for biological membranes. J. Lipid Res. 1968, 9, 310–8. Doyle DA, Cabral JM, Pfuetzner RA, Kuo A, Gulbis JM, Cohen SL, Chait BT, MacKinnon R. The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science 1998, 280, 69–77. Townsley LE, Tucker WA, Sham S, Hinton JF. Structures of gramicidins A, B, and C incorporated into sodium dodecyl sulfate micelles. Biochemistry 2001, 40, 11676–86. Fyles TM. Synthetic ion channels in bilayer membranes. Chem. Soc. Rev. 2007, 36, 335–47. Sakai N, Mareda J, Matile S. Artificial β-barrels. Acc. Chem. Res. 2008, 41, 1354–65. Sakai N, Sordé N, Das G, Perrottet P, Gerard D, Matile S. Synthetic multifunctional pores: deletion and inversion of anion/cation selectivity using pM and pH. Org. Biomol. Chem. 2003, 1, 1226–31. Clark TD, Buehler LK, Ghadiri MR. Self-assembling cyclic β3-peptide nanotubes as artificial transmembrane ion channels. J. Am. Chem. Soc. 1998, 120, 651–6. Gokel GW, Daschbach MM. Coordination and transport of alkali metal cations through phospholipid bilayer membranes by hydraphile channels. Coord. Chem. Rev. 2008, 252, 886–902. Fyles TM, James TD, Kaye KC. Biomimetic ion transport: on the mechanism of ion transport by an artificial ion channel mimic. Can. J. Chem. 1990, 68, 976–8. Ovchinnikov YA, Ivanov VT. Conformational states and biological activity of cyclic peptides. Tetrahedron 1975, 31, 2177–209. Inoue Y, Gokel GW. Cation binding by macrocycles. Marcel Dekker: New York, 1990. Hendrixson RR, Mack MP, Palmer RA, Ottolenghi A, Ghirardelli RG. Oral toxicity of the cyclic polyethers – 12-crown-4, 15-crown-5, and 18-crown-6 – in mice. Toxicol. Appl. Pharm. 1978, 44, 263–8. Dutzler R, Campbell EB, Cadene M, Chait BT, MacKinnon R. X-ray structure of a ClC chloride channel at 3.0 Å reveals the molecular basis of anion selectivity. Nature 2002, 415, 287–4. Gorteau V, Bollot G, Mareda J, Perez-Velasco A, Matile S. Rigid oligonaphthalenediimide rods as transmembrane anion-π slides. J. Am. Chem. Soc. 2006, 128, 14788–9. Jentzsch AV, Hennig A, Mareda J, Matile S. Synthetic ion transporters that work with anion–π interactions, halogen bonds, and anion–macrodipole interactions. Acc. Chem. Res. 2013, 46, 2791–800. Gorteau V, Bollot G, Mareda J, Matile S. Rigid-rod anion-π slides for multiion hopping across lipid bilayers. Org. Biomol. Chem. 2007, 5, 3000–12. Yamnitz CR, Gokel GW. Synthetic, biologically active amphiphilic peptides. Chem. Biodiv. 2007, 4, 1395–412. Davis JT. Anion binding and transport by prodigiosin and its analogs. Top. Heterocycl. Chem. 2010, 24, 145–76. Busschaert N, Caltagirone C, Van Rossom W, Gale PA. Applications of supramolecular anion recognition. Chem. Rev. 2015, 115, 8038–155.

Bibliography

531

[20] Santacroce PV, Davis JT, Light ME, Gale PA, Iglesias-Sánchez JC, Prados P, Quesada R. Conformational control of transmembrane Cl– transport. J. Am. Chem. Soc. 2007, 129, 1886–7. [21] Yamnitz CR, Negin S, Carasel A, Winter RK, Gokel GW. Dianilides of dipicolinic acid function as synthetic chloride channels. Chem. Commun. 2010, 46, 2838–40. [22] Valkenier H, Davis AP. Making a match for valinomycin: steroidal scaffolds in the design of electroneutral, electrogenic anion carriers. Acc. Chem. Res. 2013, 46, 2898–909. [23] Edwards SJ, Valkenier H, Busschaert N, Gale PA, Davis AP. High-affinity anion binding by steroidal squaramide receptors. Angew. Chem. Int. Ed. 2015, 54, 4592–6. [24] Janout V, Regen SL. Bioconjugate-based molecular umbrellas. Bioconjugate Chem. 2009, 20, 183–92. [25] Agre P. Aquaporins water channels (Nobel lecture). Angew. Chem. Int. Ed. 2004, 43, 4278–90. [26] Le Duc Y, Michau M, Gilles A, Gence V, Legrand YM, van der Lee A, Tingry S. Barboiu M. imidazole-quartet water and proton dipolar channels. Angew. Chem. Int. Ed. 2011, 50, 11366–72. [27] Hu HB, Chen Z, Tang G, Hou JL, Li ZT. Single-molecular artificial transmembrane water channels. J. Am. Chem. Soc. 2012, 134, 8384–7.

10 Detecting molecules CONSPECTUS: One drop of an indicator is sufficient to determine whether a solution is acidic or basic. Supramolecular probes work in a similar way by changing their color or another easily measurable physical property when they sense the presence of a certain analyte. Sensing relies on interactions between the analyte and a receptor unit within the probe, which in turn affect the optical or redox properties of an also present reporter unit. Both units either reside within the same molecule or are located in different molecules that interact with each other in an analyte-controlled manner. Either way, the selectivity with which the analyte is detected correlates and is limited by the selectivity of the receptor unit. This limitation is overcome by combining several different receptors in the assay, each of which interacts with the substrate in a slightly different manner. A characteristic response pattern is thus generated, which allows detecting several analytes with a single assay or differentiating between structurally closely related analytes. This sensing strategy resembles the way with which we perceive smells or tastes, showing that supramolecular strategies allow imitating yet another fundamental principle of natural systems.

10.1 Introduction When two molecules come together to form a complex, the environment of protons or of chromophores in the binding partners changes, which in turn affects their absorption or emission properties. These changes can be followed by using the techniques discussed in Section 2.2 that allow characterizing receptor-substrate complexes with respect to structure and stability. Once the properties of a complex are known, a receptor can also serve as a probe to analyze if and how much of its substrate is present in a sample of unknown composition. If the addition of the receptor to a solution does not produce the signal that is normally associated with complex formation, for example, one can conclude that the substrate is absent (if complex formation is not prevented by competing analytes). A measurable effect, on the other hand, not only qualitatively confirms the presence of the substrate but also allows quantifying its concentration by considering known properties of the complex such as its stability. Such measurements are the basis of supramolecular analytical chemistry [1]. This branch of supramolecular chemistry deals with devising strategies to reliably and conveniently detect one or several different analytes with the help of appropriate receptors, ideally without the need to use elaborate equipment. Receptors that change their color upon complex formation, for example, signal the presence of the analyte like an indicator reports a change of pH. Quantitative measurements are in this case possible by using photometers, which are indeed widely used in supramolecular analytical chemistry to follow binding events. Of the other techniques discussed in Section 2.2, fluorescence spectroscopy and electrochemical methods are

https://doi.org/10.1515/9783110595611-010

534

10 Detecting molecules

relevant if the emission or redox properties of the binding partners are affected by complex formation, respectively. Attractive as supramolecular approaches certainly are for achieving the sensitive and selective detection of a given analyte, their role in analytical chemistry should not be overestimated. With the vast array of sophisticated analytical equipment available today, it is rare that a problem can be solved with a supramolecular approach but not with another existing technique. The simplicity and selectivity of supramolecular systems are still advantageous. In addition, it is possible to devise systems by using the concepts of supramolecular chemistry that operate in a manner similar to the way smells or tastes are perceived by natural receptors. Complex mixtures of compounds can thus be analyzed in a relatively simple way with an accuracy that is sometimes difficult to achieve by using other techniques. Supramolecular methods thus have a firm place in analytical chemistry as this chapter will show, but before we come to specific examples, it is important to introduce and define relevant terms. What is a sensor?

The above-mentioned receptor that changes its color when binding to the analyte suggests that a good term for such a receptor would be indicator. Although very appropriate, this term is, however, only rarely used, maybe to avoid confusion with pH indicators. Alternatively, the term probe is correct and well established. The most widely used term for receptors that allow the sensing of their substrate is sensor or more precisely chemosensor, but these terms are somewhat problematic and their use has therefore been criticized [2]. The reason is that a real sensor is more than just a device signaling an event. A sensor also transmits a signal, which then triggers a response. A motion detector that just registers someone moving in the vicinity is not very practical, for example, but serves a purpose when a light or an alarm is subsequently turned on. In combination with feedback loops, sensors can even be used for more sophisticated applications such as process control. Chemosensors do not have these abilities and therefore only accomplish one half of the tasks of a real sensor. The term chemosensor in supramolecular analytical chemistry for the entity that does the sensing is nevertheless so firmly established since many years that it is not helpful avoiding it [3]. After all, it is the context that tells whether the term refers to a supramolecular or a real-world sensor, just like it is normally not a problem that the word bat has different meanings in different contexts. The signal generated by a chemosensor is a direct consequence of the formation of a complex or, in other words, of reversible non-covalent interactions with the analyte. Like most other supramolecular systems, chemosensors accordingly operate under equilibrium conditions, causing their behavior to be subject to thermody-

10.1 Introduction

535

namic control (Figure 10.1a). They are therefore able to respond in real time to variations in the analyte concentration, which allows using them to track reactions or other processes. In this respect, they differ fundamentally from another class of molecular probes, so-called chemodosimeters [4]. A chemodosimeter is a compound whose irreversible kinetically controlled reaction with the analyte, which can involve the cleavage or the formation of a covalent bond, triggers a change of the optical properties (Figure 10.1b). Such a compound is therefore also a useful optical probe, but since the analyte is covalently bound or otherwise irreversibly changed when it is sensed, a chemodosimeter only reports the maximum concentration of the analyte it was exposed to and is unable to monitor changes in analyte concentration. Its working principle moreover does not involve a molecular recognition event, which is why chemodosimeters do not belong to the realm of supramolecular chemistry and are therefore not considered further here.

(a)

+

The signal intensity reflects the concentration of the complex in the equilibrium

+

The signal intensity reflects the number of analyte molecules consumed in the reaction

(b)

Figure 10.1: Schematic illustration of the working principles of a chemosensor (a) and a chemodosimeter (b).

The detection of the analyte can be based on a single chemosensor molecule or an ensemble of molecules. Single molecules contain both a receptor and a reporter unit and are usually designed to target one analyte selectively or a family of analytes such as amino acids or nucleotides. Sensing ensembles serve similar purposes but consist of mixtures of noncovalently interacting receptor and reporter molecules. The ways with which they generate the signals therefore differ fundamentally from that of chemosensors. An assay can moreover involve any number of components, with an increase in the complexity of the assays usually having a beneficial effect on the fidelity of analyte detection. Accordingly, more complex systems often allow analyses that are more difficult to achieve with a similar precision by using

536

10 Detecting molecules

other approaches. A variety of strategies thus exist to achieve the selective and sensitive detection of molecules with the help of chemosensors. These strategies are discussed separately in the following parts of this chapter.

10.2 Single analyte sensing 10.2.1 Direct optical sensing A probe that responds to the presence of an analyte by changing its optical properties needs to contain two components: a receptor unit that interacts with the analyte and a chromophore that acts as the reporter unit by signaling the binding event. The integration of both units into the same molecule usually requires the additional presence of a spacer, which leads to the receptor–spacer–reporter paradigm proposed by A. Prasanna de Silva for the development single-molecule chemosensors (Figure 10.2) [5].

Reporter

Spacer

Receptor

Figure 10.2: General structure of a chemosensor combining a receptor and a reporter unit within the same molecule, covalently linked through a suitable spacer.

Designing such optical chemosensors is very modular and can involve a wide range of different receptor and reporter units to realize the desired properties. One criterion to consider is the analyte selectivity that is controlled by the receptor unit. When designing a probe for cations, for example, this unit can be based on any type of available cation receptors, provided that it binds to the analyte in question. Additional aspects to consider are the stability of the respective complex, which determines the minimum concentration of the analyte that can be detected, and the medium in which the assay should be performed. Since the selective detection of a single analyte is usually targeted, the extent to which this analyte is preferred over competing substrates is also important. If selective binding cannot be guaranteed because the analyte is structurally too closely related to other components in the mixture, single-molecule chemosensors reach their limits. The reporter unit determines the spectroscopic technique required to follow the binding event and the sensitivity of the probe. If, for example, chromophores are

537

10.2 Single analyte sensing

used as reporter units that react to analyte binding by the shift of an absorption band in the visible range of the electromagnetic spectrum or by an increase or decrease of fluorescence, it is possible to conveniently detect the presence of the analyte with the naked eye. Spectral changes that are not immediately visible are followed by using UV–vis or fluorescence spectrometers, which also make possible quantitative measurements, with the latter technique usually being more sensitive. The spacer unit, finally, ensures the optimal mutual arrangement of the receptor and the reporter unit. If sensing requires a direct electronic coupling of both units, the spacer unit may be absent. Its presence is crucial if a very close proximity of the receptor and reporter unit has a negative impact on their respective functions. Directly linking both units potentially compromises analyte binding, for example, which needs to be avoided. Different photophysical processes can be responsible in the reporter unit for the development of the signal, with early chemosensors mostly relying on internal charge transfer (ICT), sometimes also called photoinduced charge transfer (PCT). The reporter unit of such systems usually features a push-pull π-electron system like in crown ether 10.1 (Figure 10.3), which represents one of the first crown ether-based chemosensors, described already in 1978 by Fritz Vögtle [6]. Photoexcitation facilitates the transfer of an electron between the conjugated subunits of such chromophores from one side to the other, leading to a charge distribution in the excited state that differs substantially from that in the ground state. The receptor-induced positioning of an ion close to one side of the chromophore stabilizes or destabilizes this charge distribution, depending on the ion’s charge, thus influencing the frequency at which the photoexcitation occurs. In the case of 10.1, for example, the complexation of a potassium ion causes a 20 nm hypsochromic shift of the absorption band, while sodium or rubidium ions have

NO2 O –

N

OOC

N

–

–

OOC COO N N O

O

COO

–

O

O N

O O

O

–

COO

F

O

O

N

10.2

O

O O

10.1

O

F

O

N O

O

O

O

10.3

O

Figure 10.3: Molecular structures for the three ICT chemosensors 10.1 (for K+), 10.2 (for Ca2+), and 10.3 (for Na+).

538

10 Detecting molecules

smaller effects, consistent with the generally lower affinity of 18-crown-6 derivatives for ions smaller and larger than potassium (Section 4.1). The effect of the cations on the excitation wavelength of the absorption band is due to their interaction with the lone pair of the crown ether nitrogen atom that prevents this lone pair from participating in the conjugation with the aromatic π-system. In a similar way, ions also affect the emission wavelength of fluorophores with ICT properties. In general, a cation interacting with the donor moiety of a push-pull system (in 10.1 the amino group) decreases the transition dipole moment, thus causing a shift of the ICT band to higher frequencies, while an anion interacting with the acceptor moiety causes the opposite effect. Other examples of chemosensors that rely on ICT are the chelating ligand 10.2 that allows Ca2+ sensing and the xanthene-derived crown ether 10.3 (Figure 10.3), which is commercially available under the name CoroNa™ Green and exhibits an increase in the green fluorescence intensity upon Na+ binding. Both of these chemosensors are useful for cellular imaging (Section 11.2.3) [1, 5]. Fluorescence chemosensors alternatively rely on photoinduced electron transfer (PET) for signal generation. These chemosensors contain a fluorescence chromophore and at least one functional group in close proximity that is able to donate an electron. Once the fluorophore is excited, this group transfers an electron into the now singly occupied HOMO of the ground state. As a consequence, the excited electron cannot return into this orbital by emitting a photon, and the molecule therefore has to choose nonradiative pathways to dissipate its excess energy (Figure 10.4a). Fluorescence is turned on by preventing the electron transfer that causes quenching. The exact behavior thus depends on whether quenching occurs in the uncomplexed or the complexed state. If the interactions with the analyte prevent quenching, the chemosensor goes from the nonfluorescent state to the fluorescent state upon analyte binding, rendering it a so-called turn-on chemosensor (Figure 10.4b). Conversely, if analyte binding enables PET, the chemosensor reacts to the presence of the analyte by turning off its fluorescence. Both types of chemosensors exist but a turn-on behavior is preferable. The reason is that it is easier to detect light against a dark background than light disappearing in a bright environment. Imagine an office building with all lights turned on behind the windows. If the light in one room is turned off it will take a while to find the respective window by just looking at the building from the outside. If, on the other hand, all windows are dark, the room in which the light is turned on is easily spotted. The important issue in the context of sensing is that turn-off chemosensors become darker as the analyte concentration rises. Reliably measuring the feeble light emitted from the sample when varying the analyte concentration against the background noise is therefore challenging. In the case of turn-on chemosensors, the signal increases with increasing analyte concentration so that background noise does not interfere in the measurement.

10.2 Single analyte sensing

539

(a) LUMO

LUMO

h

PET prevents fluorescence

HOMO

HOMO Donor Fluorophore in ground state

Donor Fluorophore in excited state

(b) PET

Reporter

Donor

no PET

Receptor

Chemosensor is "turned off"

Reporter

Donor

Receptor

Chemosensor is "turned on"

Figure 10.4: Schematic illustration of photoinduced electron-transfer (PET) (a), and working principles of a turn-on chemosensor (b). During PET, the donor transfers an electron into the depleted orbital produced upon photoexcitation, thus preventing the system to relax to the ground state by the emission of a photon.

Figure 10.5 shows examples of PET chemosensors. All of these systems have in common that they contain one or more benzylic amino groups attached to the anthracene fluorophore as donors. The chemosensors differ in the functional groups responsible for analyte recognition, with 10.4 containing a crown ether for metal ion binding, 10.5 additional ammonium groups to interact with a dihydrogenpyrophosphate anion, and 10.6 boronic acid groups to form esters with carbohydrates as we have seen in Section 4.2.4. All receptors are nonfluorescent in the absence of their substrate as a result of PET mediated by the free electron pairs of the benzylic amino groups (provided that the pH of the solution is sufficiently high to ensure that these amino groups are not protonated). Once the analytes are bound, the electron transfer is no longer possible because the amino groups participate in the stabilization of the respective complexes. In 10.4, the lone pair of the amino group coordinates to the bound metal ion, while the respective electron pairs in 10.5 accept protons from the substrate. In 10.6 the lone pairs of the nitrogen atoms coordinate and thus stabilize the tetrahedral form of the boronic esters formed with the sugar. All of these compounds therefore start to fluoresce upon analyte binding.

540

10 Detecting molecules

O

+ NH3

O

+ H 3N

H3N

NH3

O

+ NH + N H

O N

O

10.4

+ HN + N H

N

(HO)2B

N

10.5 O

10.6

O

N

O O

B(OH)2

N

O

COO − COO −

10.7

Figure 10.5: Molecular structures of the PET chemosensors 10.4 (for K+), 10.5 (for H2P2O72−), and 10.6 (for glucose), and of the lab on a molecule 10.7.

Appropriately designed PET chemosensors also respond to more than one analyte. The chemosensor 10.7, for example, contains a crown ether moiety to interact with Na+, a phenyliminodiacetate group as a chelating group for Zn2+, and a tertiary amino group that can accept a proton [7]. When none of these analytes is present, 10.7 is non-fluorescent because of PET. Each analyte alone or even any combination of two analytes does not turn on the fluorescence because in every case a group remains that quenches the anthracene chromophore. If, for example, Na+ binds to the crown ether and a proton to the amino group, there is still an electron pair left in the phenyliminodiacetate moiety that causes quenching. If Na+ and Zn2+ are bound simultaneously, the crown ether moiety is responsible for PET. The fluorescence of 10.7 is therefore only turned on if all three analytes are simultaneously present. This ability to simultaneously analyze three different inputs is the reason why 10.7 is termed “lab on a molecule.” Other such multifunctional chemosensors are know. Their behavior is often described by using Boolean algebra [8]. 10.7, for example, represents an AND logic gate. Conformational changes induced in a chemosensor upon analyte binding give rise to an optical signal if these changes affect the distance or orientation of appended chromophores. The calixarene derivative 10.8, for example, contains two pyrene units which are located in close proximity in the uncomplexed form of 10.8 (Figure 10.6) [9]. In this form, the chemosensor exhibits the typical pyrene excimer band at ca. 480 nm in the UV–vis spectrum, which derives from the interaction of a ground state and an excited state pyrene ring. Both rings are pushed apart when a sodium ion binds to the oxygen atoms along the narrower rim of 10.8 (Section 4.1.7), causing the excimer band to disappear. Related chemosensors exist that respond to analyte binding also by the appearance of the excimer band.

10.2 Single analyte sensing

O

O O

O

O

O

O

O O Et Et

541

O

O

O 10.8 Figure 10.6: Molecular structure of the calix[4]arene derivative 10.8 that responds to the presence of Na+ by the disappearance of the pyrene excimer band.

If a conformationally flexible chemosensor contains two different fluorophores, Förster resonance energy transfer (FRET) is possible. Sensing in this case involves selectively exciting one fluorophore, the donor, whose energy is then transferred nonradiatively to the other fluorophore, the acceptor, which emits a photon. The excitation of the donor thus leads to the fluorescence of the acceptor. This process requires the donor emission spectrum to overlap in a certain range with the acceptor absorption spectrum and it also depends on the relative orientation of the donor emission and the acceptor absorption dipole moments. Most importantly for sensing, the efficiency of energy transfer correlates with the distance r between donor and acceptor with a dependence of r − 6 . As a consequence, the intensity of the donor emission progressively decreases as the distance between the two groups becomes smaller, while the emission intensity of the acceptor increases. Although some applications of FRET for chemosensing purposes exist, this process is more frequently used to probe distances between different sites in biomolecules. Additional mechanisms causing an optical response of a chemosensor are the rigidification of the molecule upon substrate binding, which usually causes an increase of the fluorescence intensity because of the reduced number of nonradiative pathways that allow the excited state to dissipate its energy, and the effect of analyte binding on the absorption properties of metal-containing receptors [10, 11]. In addition, a number of cyclodextrin derivatives are known whose interaction with the analyte causes the expulsion of an appended chromophore from the cyclodextrin cavity. The change in environment of the chromophore from apolar, if it resides within the cavity, to polar, if it is exposed to the outside aqueous solution, causes a reduction of the fluorescence intensity, which thus signals the presence of the analyte [12].

542

10 Detecting molecules

10.2.2 Indirect optical sensing The second approach to achieving the optical sensing of a specific analyte does not involve probing the properties of the receptor itself but those of an additional component, which interacts with the receptor in the absence of the analyte but is replaced from the binding site when the analyte is present [13]. This strategy, termed indicator displacement assay (IDA) by Eric V. Anslyn, who introduced it into supramolecular analytical chemistry and demonstrated its usefulness through many examples [14], is illustrated in Figure 10.7.

+

Receptor–dye complex

Receptor–analyte complex

Free dye

Figure 10.7: General principle of an indicator displacement assay (IDA). Sensing involves measuring the changes of the optical properties of the dye when it is displaced by the analyte from the complex with the receptor.

Accordingly, the actual probe consists of a mixture of a receptor and a suitable reporter molecule. This reporter is usually a dye, often a conventional pH indicator whose optical properties change when binding to the receptor. The UV–vis spectrum in the absence of the analyte reflects the optical properties of the complex between receptor and dye and the extent to which it is formed. When the analyte is added, a competition arises between the analyte and the dye for the binding site of the receptor. Since an optical signal only results if the analyte wins this competition, a crucial prerequisite for the strategy to work is that the complex between the receptor and the analyte is more stable than that between the receptor and the dye. Under these conditions, the increase of the analyte concentration causes the progressive displacement of the dye from the receptor binding site, signaled by the associated spectral changes. The degree of displacement allows calculating the stability constant of the receptor–analyte complex, but the underlying math differs from that used for determining the stability of a 1:1 complex because three rather than two components are involved in the equilibrium. The exact formalism is explained in Appendix 12.2. Two examples from the Anslyn group should illustrate this concept. The first involves the use of receptor 10.9 (Figure 10.8) that arranges three 2-aminoimidazoline units around a 1,3,5-triethyl benzene ring [15]. The six substituents have an alternating

543

10.2 Single analyte sensing

up–down arrangement so that the three heterocyclic units end up to be positioned on the same side of the benzene ring (Section 4.1.12). Since 2-aminoimidazolines are protonated at physiological pH, a well-preorganized binding site for anionic substrates results. 10.9 strongly interacts with tricarboxylates such as citrate but also binds to dicarboxylates, phosphates, and the trianionic dye 5-carboxyfluorescein, albeit with a significantly lower affinity. In 25 vol% water/methanol (buffered to pH 7.4) the log Ka of the citrate complex amounts to 5.5, for example, while the 5-carboxyfluorescein complex has a log Ka of only 3.7. The binding of 5-carboxyfluorescein to 10.9 is associated with an increase of its absorbance and emission bands. The displacement of the dye molecule from the binding site by the more strongly bound citrate is thus associated with the opposite effect. By using calibration curves that describe the concentration dependence of these optical changes, the citrate content of unknown samples can be quantified. This method allows determining, for example, how much citrate various sports drinks contain [15].

HN H

HN

N

H

H N + H N N H

HN

N NH + H H N N +

N H

H

N + N N H

NH +

HN N

Cu2+

HN + NH N NH –

N

N 10.10

10.9 HO O O H –

NH +

O

N HO – O N +NH H H N N + N H

O

O

COO

– H

O

O

N H

–

NH +

– HN O N H HN O O HN + – P – + NH HN O N NH Cu2+ N N N

– COO 5-Carboxyfluorescein

Figure 10.8: Molecular structures of receptors 10.9 and 10.10, schematic structures of their complexes with, respectively, citrate and phosphate, and molecular structure of 5carboxyfluorescein.

The orthophosphate receptor 10.10 (Figure 10.8) features three 2-aminoimidazoline units arranged around a binding site that contains a copper(II) ion [16]. It also interacts with 5-carboxyfluorescein in 50 vol% water/methanol (buffered to pH 7.4), which leads to a color change of the dye from yellow to light orange. The subsequent

544

10 Detecting molecules

addition of phosphate restores the original color, with the associated changes in the UV–vis spectrum allowing in this case the determination of phosphate in saliva. IDAs are also useful to follow enzymatic conversions. Such so-called supramolecular tandem enzyme assays were introduced by Werner M. Nau and involve the use of receptor-dye complexes as sensing ensembles that respond differently to the presence of the product of an enzymatic reaction than to the substrate [17]. Product formation can thus be sensed by optical measurements and can even be followed in real time without having to label the compounds involved in the reaction. The example in Figure 10.9 illustrates the concept [18]. In this reaction, lysine is converted by a decarboxylase into the diammonium form of 1,5-diaminopentane. The receptor-dye ensemble consists of cucurbit[7]uril (CB[7]) and the fluorescent dye dapoxyl. This dye is only weakly fluorescent in the medium in which the assay is performed but becomes strongly fluorescent when included into the cucurbituril cavity. The corresponding complex has a higher stability than the complex between CB[7] and lysine but dissociates in the presence of diammonium ions, which are better guests than dapoxyl. The formation of the decarboxylation product therefore causes a reduction of the fluorescence. Following this reduction allows determining the rate of the reaction or the enzymatic activity. It should be mentioned that the concept of indicator displacement also allows optical sensing if the receptor and the indicator are covalently linked and not just noncovalently assembled. The tripodal chemosensor 10.11, for example, contains two thiourea groups as anion recognition moieties and a naphthyl carboxylate as an anionic chromophore (Figure 10.10) [19]. In the absence of the analyte, 10.11 prefers a conformation in which the anionic substituent interacts with the urea groups. These intramolecular interactions are no longer possible in the glyphosate complex, so that the conformational rearrangement induced by the analyte leads to a pronounced reduction of the fluorescence. The aspect of indicator displacement accordingly plays a role in the sensing mechanism, but since reporter and dye are covalently linked, the probe strictly behaves as a direct chemosensor, similar to the cyclodextrin derivatives mentioned at the end of the previous chapter.

10.2.3 Direct electrochemical sensing The use of electrochemical methods is well established in analytical chemistry. Nonredox-active inorganic cations or anions are reliably detected, for example, with the help of ion selective electrodes, of which many are commercially available [20]. These electrodes typically contain ionophores embedded in a polymeric membrane that are responsible for the selective recognition of the charged guests. Alternatively, substrate sensing can also be achieved by using receptors immobilized on the surface of gold electrodes. The underlying principles will not be treated here. Instead, we will focus on molecular chemosensors that allow the use of electrochemical methods to

545

10.2 Single analyte sensing

− SO3

OO

O N NNN NN NN NNN NN

N N

O

O O

OO O

− SO3

O O O N N NNN N N N N O

N NN N N OO

H3N

+

Lysine decarboxylase

+

+ O O H3N O O O N N N N NNN NNN N N N N

OO

COO

−

H3N +

NNN NN

N N

N N

OO O H O 3N + O

N NN N N OO

N

N +

O

N

CB[7]-dapoxyl complex highly fluorescent

L-Lysine

CB[7] complex of the diammonium ion of 1,5-diaminopentane

Dapoxyl weakly fluorescent

Figure 10.9: Example of a supramolecular tandem enzyme assays that involves monitoring the decarboxylation of lysine by using the cucurbit[7]uril-dapoxyl complex as sensing ensemble. This complex is stable in the presence of lysine but releases dapoxyl during the formation of the more strongly bound reaction product.

O O

NH

Ph − HN O S O HN Ph HN HN

10.11 Strongly fluorescent

2−

O3P H2N +

+ −

O2 C

S Glyphosate

− Ph O − HN COO S H2N + O O HN Ph O NH P O − HN O O HN S −

Weakly fluorescent

Figure 10.10: Sensing of glyphosate with the covalently assembled receptor–dye conjugate 10.11. Analyte binding in this case prevents the anionic chromophore from interacting with the thiourea moieties.

detect the binding event. The design of such probes is based on the receptorspacer-reporter paradigm mentioned earlier. In this case, the spacer connects a receptor unit, that controls analyte selectivity, with a redox-active group, whose response to analyte binding is detected electrochemically. Analyte binding affects the electrode potential at which the redox process occurs, which is detectable by cyclic voltammetry. In spite of the closely related structures of optical and redox chemosensors, there is a fundamental difference in the way the signal is transmitted between their subunits. In the case of optical systems, analyte binding is transmitted from the receptor to the reporter unit, the latter of which changes its optical properties. Whether the reporter is in the ground or excited state has a small impact on the binding event because the photophysical processes underlying sensing are normally faster than the

546

10 Detecting molecules

binding equilibrium. This is different in electrochemical chemosensors in which switching the redox state of the reporter unit during the measurement has a direct impact on the strength with which receptor and analyte interact. The correlation is best understood on the basis of the square scheme in Figure 10.11a.

(a)

(b)

+e Rox

ER°

Rred O

O O

Fe

+S

Cox

Kox

Kred

EC°

+S

O

NH

NH

NH

NH

Fe

O

10.12

10.13

Cred

+e Figure 10.11: Cycle of equilibria associated with the binding of a substrate S to the oxidized and reduced version of the receptor R to form the corresponding complexes C and with the interconversion of the oxidized and reduced receptor and complex species (a). K stands for the association constants of the complexes and E 0 for the formal potential of the electron transfer reactions. The molecular structures of the electrochemical chemosensors, 10.12 (for Na+) and 10.13 (for HPO42−/H2PO4−) are shown in (b).

The vertical equilibria in this scheme describe the binding of the substrate S to the receptor R, which is either in its reduced or its oxidized state (Rred and Rox) to form the corresponding versions of the complexes (Cred and Cox). The horizontal equilibria describe the oxidation and reduction of the receptor R and of the complex C. The cyclic arrangement of these equilibria implies that the Gibbs free energy change of the complete cycle is zero. Using the expressions ΔG0 = − RT lnKa for the binding equilibria and ΔG0 = − nF ΔE0 for the electron transfer processes therefore allows deriving equation (10.1), which shows that the shift in the electrode potential of a redox chemosensor EC0 − ER0 induced by analyte binding is a direct consequence of the different affinities with which the oxidized and the reduced version of the receptor bind to the analyte.
 Kred nF EC0 − ER0 = RT ln Kox

(10:1)

Although the exact mechanism with which the redox states of the reporter unit affect complex stability varies from chemosensor to chemosensor [21], the correlation is straightforward. If a positive charge is generated in the reporter unit upon oxidation, for example, this charge has a negative impact on the binding of a cation in close proximity but favors the binding of an anion. Electrochemical chemosensors for cations therefore usually signal the presence of the substrate by a positive

10.3 Multiple analyte sensing

547

(anodic) shift of the redox potential (EC0 − ER0 > 0, the oxidation becomes more difficult), while the opposite is observed for the binding of anions. The modular design of electrochemical chemosensors allows a similar large structural variability as with optical systems. The receptor units are mostly based on established recognition elements while a few reporter units have proven to be particularly useful as sensing units. These units are either based on metal complexes or organic redox-active moieties, with ferrocene likely being the most frequently used metal-containing reporter and tetrathiafulvalene the metal-free one. The chemosensors 10.12 and 10.13 (Figure 10.11b) rely on ferrocene as the redoxactive reporter group, for example. They allow the detection of sodium ions (10.12) [22] and of hydrogenphosphate/dihydrogenphosphate ions (10.13) [23]. The electrochemical responses observed in the measurements are consistent with the abovementioned trends in that 10.12 exhibits a shift to larger potentials on sodium binding while the presence of the phosphate ions produces a shift to a smaller potential in the case of 10.13. Many other substantially more sophisticated chemosensors are known, with major contributions in the context of electrochemical anion sensing coming from the group of Paul D. Beer [24].

10.3 Multiple analyte sensing Taste and smell are senses that allow us to detect molecules in our environment. Although differing in their exact biochemical mechanisms, the general working principles of these senses are similar. When we smell coffee, for example, our olfactory system has to recognize at least forty volatile compounds and analyze their ratio. This task can obviously not involve a single nerve cell specialized to recognize coffee. Instead, we have a large number, likely thousands of different sensory neurons, each of which reacts to the components of the coffee aroma by recognizing characteristic properties that relate to structure, solubility, diffusion rate etc. A single component therefore triggers responses in many nerve cells but typically to varying degrees. All signals combined result in a characteristic response pattern that is analyzed by the brain and associated with the aroma of coffee. To perceive taste, specialized receptors exist on our tongues that typically work in concert with the smell receptors to produce similar signal patterns. Smell and taste receptors therefore do not need to be highly selective to allow us to sense aromas. Selectivity instead arises from the combination of many receptors, each of which recognizes specific traits in molecules but does not necessarily have to be selective for only a single molecule. The principles underlying our ability to smell and taste inspired the use of mixtures of receptors for sensing purposes [25]. This differential sensing strategy has a number of advantages. First, it allows the detection of several analytes with a single assay. Second, even structurally closely related analytes, for which the development of a selective receptor would be challenging, can be differentiated. Sensing

548

10 Detecting molecules

generally benefits from combining several nonselective receptors, similar to how smell and taste receptors work because the characteristic fingerprints produced by cross-reactive receptors, that bind to the same analyte with different affinities, provide more information about the identity of the analyte than the yes/no responses of selective receptors. Setting up an assay involves the simultaneous use of several chemosensors in the same solution or of spatially separated individual chemosensors in an array as illustrated schematically in Figure 10.12. Both approaches are either based on direct or on indirect sensing, but the use of sensing ensembles is often easier because it just requires combining several receptors and indicators without additional synthetic efforts. When combining several chemosensors in a single solution, the UV–vis spectrum of the mixture reflects the number and relative concentrations of all of the absorbing components (Figure 10.12a). These components either encompass the individual chemosensors or, if the assay is based on sensing ensembles, the receptors, indicators, and the corresponding complexes. The addition of an analyte to the solution causes characteristic perturbations in this composite UV–vis spectrum, which result from the differential interactions of the analyte with the chemosensors, or the extents to which the analyte shifts the equilibria in which the sensing ensembles are involved. Independent of the actual sensing mechanism, the perturbations in the spectrum are specific for each analyte, and therefore allow distinguishing different and even structurally closely related compounds. Spatially separating different chemosensors or sensing ensembles generally involves the use of microtiter plates of which each well contains an individual probe (Figure 10.12b). The presence of an analyte produces a characteristic response in each well, the combination of which is analyzed. Since the resulting patterns are again analyte-specific, this method also allows differentiating multiple analytes. Since the amount of data generated by such measurements is significantly larger than in assays involving only a single chemosensor, special chemometric methods such as principal component analysis (PCA) or linear discriminant analysis (LDA) have to be used for data analysis. These methods find a correlation between the structure of the analyte and its effect on the recorded UV–vis spectrum over the whole range of wavelengths, or on the color pattern generated in the microtiter well, by reducing the dimensionality of the data to the fraction that carries the relevant information. While the exact mathematical formalism of both methods differs, the general strategy to establish the correlation between the experimental result and the analyte structure is the same. The first step always comprises training the respective mathematical method in how to distinguish the analytes. To this end, a series of measurements is performed using all relevant analytes. On the basis of this training set, two- or three-dimensional score plots are generated that graphically illustrate the structural similarities between the analytes. An example is shown in Figure 10.12c. Each point in such a plot represents a single measurement, with measurements involving identical analytes ideally affording closely spaced clusters of points. The farther the clusters of different analytes

549

10.3 Multiple analyte sensing

1 5

2

(c) Score plot Analyte 1

3

4

λ (nm)

Individual optical probes

Patterns developing in an array of probes upon the addition of different analytes

1

1

2

2

3

3

4

4

5

5

LD/PC2

(b)

Effects of different analytes on the UV–vis spectrum

Mixture of optical probes

Absorbance

(a)

Analyte 4 Analyte 3 Analyte 2

LD/PC1

Figure 10.12: Schematic illustration of the strategies that are used to achieve multiple analyte sensing. In (a), several optical probes are combined in the same solution and the characteristic effects of the analyte on the UV–vis spectrum of the mixture is analyzed. In (b), individual optical probes are spatially separated to give a chemosensor array. The pattern produced when adding the analyte to each probe is then used to obtain information about its nature. An exemplary score plot resulting from such assays is shown in (c).

are separated, the greater the accuracy with which they are differentiated. Once these plots are available, the actual measurements are performed and analyzed. The following examples illustrate successful applications of these strategies. In an early attempt of multi-analyte sensing, the Anslyn group showed that combining the two receptors 10.14 and 10.15 in a sensing ensemble with bromopyrogallol red and pyrocatechol violet allows the differentiation of malate and tartrate (Figure 10.13) [26]. Both receptors structurally derive from the citrate receptor 10.9 but contain one (10.14) or two (10.15) boronic acids instead of 2-aminoimidazoline units to impart affinity for diols. The binding properties of both receptors differ, with 10.14 having a similar affinity for tartrate and malate, while 10.15 binds tartrate more strongly. When mixing the receptors and the two dyes in methanol/water, 3:1 (v/v) and adding increasing concentrations of malate or tartrate while keeping the concentrations of the other components constant, changes in the UV–vis spectrum result that are used as training set for the pattern recognition analysis. Based on the results, the concentrations of malate and tartrate can be determined with high precision, even in samples in which both analytes are present simultaneously.

550

10 Detecting molecules

Receptors

Analytes R1 H + N

H 10.14

R2

N N

N H

N H

R3

10.15

HO – HO B

H + N

H N

N

N H HO – HO B + N H

10.16 Dyes Pyrocatechol violet

OH

H

H HO – HO B N

COOH

HOOC

N

Citrate HOOC

+

COOH HO COOH

H

Bromopyrogallol red

Alizarin complexone

O OH

–

OH Malate

HO – HO B

+

COOH

HOOC

+

HO – HO B + N H

O

OH

H

R2 R1

Tartrate

HO – HO B

H + N

H

N

R3

–

O3S

Br

OH

O3S

OH

O N Br

OH OH

HO

OH

OH O

COOH COOH

OH

OH

Figure 10.13: Molecular structures of the three receptors 10.14, 10.15, and 10.16, the dyes bromopyrogallol red, pyrocatechol violet, and alizarin complexone, and the analytes malate, tartrate, and citrate.

When also including the tris(boronic acid) 10.16 and alizarin complexone in the assay (Figure 10.13), malate, tartrate, and citrate can be sensed simultaneously [27]. Since these three acids are constituents of wines, the UV–vis spectroscopic analysis of the response of the assay to different wine samples followed by linear discriminant analysis allows the pattern-based discrimination of wine varietals. The array-based assay shown in Figure 10.14, which was described by the group of Pavel Anzenbacher Jr. [28], detects different phosphate species in water and blood serum. This assay relies on direct sensing by using the chemosensors 10.17a-f, each of which responds to the presence of phosphate anions by a characteristic color change. An array containing these chemosensors embedded in hydrophilic polyurethane matrices thus exhibits color patterns when exposed to the analytes that are used as training set for a PCA. The resulting score plot illustrates that it is possible to differentiate phosphate, pyrophosphate, adenosine monophosphate, and adenosine triphosphate with this assay. The Severin group demonstrated that effective assays can be established in a surprisingly simple fashion. Setting-up the corresponding assay just involves

10.3 Multiple analyte sensing

551

(a) 10.17a,d R = R HN S HN R HN

HN NH O

R

O HN HN

NH

HN

R

NH

S

NH

S

HN

10.17a–c

S

N

O 10.17b,e R = N O

O

O

10.17d–f

10.17f

10.17e

10.17d

10.17c

10.17a

(b)

10.17b

COO

(c) 4

Water

H2PO4− HP2O73− AMP ATP

Water Serum Serum + AMP Serum + ATP Serum + H2PO4−

2 PC2 (25.1%)

Serum

10.17c,f R =

0

Serum + HP2O73−

−2 −4 −6 −6

−4

0 2 −2 PC1 (50.5%)

4

6

8

Figure 10.14: Molecular structures of the receptors 10.17a-f (a), differential responses of the polymer embedded chemosensors in an array in which the spots of the six receptors in each row are treated with the same analyte (b), and score plot illustrating the differentiation of the analytes after principal component analysis (c). Image and plot adapted with permission from [28]. Copyright Wiley-VCH, 2007.

dissolving three commercial dyes, namely, arsenazo I, methylcalcein blue, and glycine cresol red (Figure 10.15a) together with CuCl2 and NiCl2 in aqueous buffer at pH 8.4 [29]. Under these conditions, the five components afford a dynamic library of interconverting metal complexes. For the assay to work, it is neither important to know how many complexes are present in the mixture nor how they exactly look like. The only condition is that the interconversion of the complexes is fast and fully reversible. This approach thus represents an application of dynamic combinatorial chemistry (Section 5.7) for sensing purposes. Sensing relies on the spectral changes caused by analytes that shift the equilibrium state of the dynamic library. These optical changes are structure-sensitive as demonstrated by the fact that each of the six dipeptides Val–Phe, Gly–Ala, His–Ala, Ala–His, Phe–Pro, and Pro–Gly, produces a characteristic UV–vis spectrum when added to the library. The most pronounced changes are caused by His–Ala, likely because the imidazole unit located at the N-terminus of this dipeptide competes especially well with the dyes for binding to the metal ions. The spectral changes

552

10 Detecting molecules

(a)

(b)

OH

Arsenazo I

H2O3As

90

N

70

OH

NaO3S

SO3Na

% Cu

N

Gly-His-Gly vs . Gly-Gly-His

50 30

Methylcalcein blue

10 HO

O

O N

COOH 0.1 0.2 0.3 Total metal concentration (mM)

O N H

COOH

Glycine HO3S cresol red

90 N H

COOH 70 % Cu

OH

CuCl2, NiCl2 H2O, buffer

His-Gly-Gly vs. Gly-Gly-His

50 30 10

Dynamic library of metal-dye complexes 0.1 0.2 0.3 Total metal concentration (mM)

Figure 10.15: Components of a dynamic combinatorial chemosensor library for the differentiation of dipeptides and tripeptides (a), and effect of the composition of the mixture on the fidelity with which the tripeptides, Gly-Gly-His, Gly-His-Gly, and His-Gly-Gly are differentiated (b). In these graphs, the sum of the Cu2+ and Ni2+ concentrations is plotted vs. the percentage Cu2+ in the mixture. A deep red color indicates optimal tripeptide differentiation. Accordingly, the best differentiation is in both cases achieved at the highest total metal concentration but at different ratios of the metals. Plots adapted with permission from [30]. Copyright American Chemical Society, 2006.

induced by structurally closely related dipeptides such as Gly–Ala, Val–Phe, Ala–Phe, Phe–Ala, and D-Phe–Ala are sufficiently characteristic to allow differentiation by using a linear discriminant analysis for the data treatment. An advantage of this strategy is that it can easily be adapted to a specific analytical problem by varying the concentrations and ratios of the components [30]. When using the tripeptides Gly-Gly-His, Gly-His-Gly, and His-Gly-Gly as analytes, for example, a Cu2+/Ni2+ ratio of 1:3 is optimal for the differentiation of Gly-His-Gly and Gly-Gly-His, while a library containing exclusively Cu2+ provides the best discrimination between His-Gly-Gly and Gly-Gly-His (Figure 10.15b). There are certainly alternate ways to analyze mixtures of these tripeptides. Liquid chromatography coupled with mass spectrometry could work, for example, if the tripeptides, which are structural isomers, give

Bibliography

553

rise to characteristic fragmentation patterns. The above examples nevertheless demonstrate that supramolecular analytical chemistry offers potent solutions even for challenging analytical problems.

Bibliography [1] [2] [3] [4]

[5] [6] [7]

[8] [9] [10]

[11] [12] [13] [14] [15] [16] [17] [18] [19]

Anslyn EV. Supramolecular analytical chemistry. J. Org. Chem. 2007, 72, 687–99. Wolfbeis OS. Probes, sensors, and labels: why is real progress slow? Angew. Chem. Int. Ed. 2013, 52, 9864–5. Czarnik AW. Chemical communication in water using fluorescent chemosensors. Acc. Chem. Res. 1994, 27, 302–8. Kaur K, Saini R, Kumar A, Luxami V, Kaur N, Singh P, Kumar S. Chemodosimeters: an approach for detection and estimation of biologically and medically relevant metal ions, anions and thiols. Coord. Chem. Rev. 2012, 256, 1992–2028. de Silva AP, McCaughan B, McKinney BOF, Querol M. Newer optical-based molecular devices from older coordination chemistry. Dalton Trans. 2003, 1902–13. Dix JP, Vögtle F. Ion-selective crown ether dyes. Angew. Chem. Int. Ed. Engl. 1978, 17, 859. Magri DC, Brown GJ, McClean GD, de Silva AP. Communicating chemical congregation: a molecular AND logic gate with three chemical inputs as a “lab-on-a-molecule” prototype. J. Am. Chem. Soc. 2006, 128, 4950–1. Erbas-Cakmak S, Kolemen S, Sedgwick AC, Gunnlaugsson T, James TD, Yoon J, Akkaya EU. Molecular logic gates: the past, present and future. Chem. Soc. Rev. 2018, 47, 2228–48. Jin T, Ichikawa K, Koyama T. A fluorescent calix[4]arene as an intramolecular excimer-forming Na+ sensor in nonaqueous solution. J. Chem. Soc., Chem. Commun. 1992, 499–501. de Silva AP, Gunaratne HQN, Gunnlaugsson T, Huxley AJM, McCoy CP, Rademacher JT, Rice TE. Signaling recognition events with fluorescent sensors and switches. Chem. Rev. 1997, 97, 1515–66. You L, Zha D, Anslyn EV. Recent advances in supramolecular analytical chemistry using optical sensing. Chem. Rev. 2015, 115, 7840–92. Ogoshi T, Harada A. Chemical sensors based on cyclodextrin derivatives. Sensors 2008, 8, 4961–82. Wu J, Kwon B, Liu W, Anslyn EV, Wang P, Kim JS. Chromogenic/fluorogenic ensemble chemosensing systems. Chem. Rev. 2015, 115, 7893–943. Wiskur SL, Ait-Haddou H, Lavigne JL, Anslyn EV. Teaching old indicators new tricks. Acc. Chem. Res. 2001, 34, 963–72. Metzger A, Anslyn EV. A chemosensor for citrate in beverages. Angew. Chem. Int. Ed. 1998, 37, 649–52. Tobey SL, Anslyn EV. Determination of inorganic phosphate in serum and saliva using a synthetic receptor. Org. Lett. 2003, 5, 2029–31. Dsouza RN, Hennig A, Nau WM. Supramolecular tandem enzyme assays. Chem. Eur. J. 2012, 18, 3444–59. Hennig A, Bakirci H, Nau WM. Label-free continuous enzyme assays with macrocyclefluorescent dye complexes. Nat. Meth. 2007, 4, 629–32. Minami T, Liu Y, Akdeniz A, Koutnik P, Esipenko NA, Nishiyabu R, Kubo Y, Anzenbacher Jr. P. Intramolecular indicator displacement assay for anions: supramolecular sensor for glyphosate. J. Am. Chem. Soc. 2014, 136, 11396–401.

554

10 Detecting molecules

[20] Privett BJ, Shin JH, Schoenfisch MH. Electrochemical sensors. Anal. Chem. 2008, 80, 4499–517. [21] Beer PD, Gale PA, Chen GZ. Electrochemical molecular recognition: pathways between complexation and signalling. J. Chem. Soc., Dalton Trans. 1999, 1897–910. [22] Tetsuo S. Electrochemically switched cation binding in pentaoxa[13]ferrocenophane. Chem. Lett. 1986, 15, 275–6. [23] Beer PD, Cadman J, Lloris JM, Martínez-Máñez R, Padilla ME, Pardo T, Smith DK, Soto J. Selective electrochemical recognition of sulfate over phosphate and phosphate over sulfate using polyaza ferrocene macrocyclic receptors in aqueous solution. J. Chem. Soc., Dalton Trans. 1999, 127–34. [24] Hein R, Beer PD, Davis JJ. Electrochemical anion sensing: supramolecular approaches. Chem. Rev. 2020, 120, 1888–935. [25] Lavigne JJ, Anslyn EV. Sensing a paradigm shift in the field of molecular recognition: from selective to differential receptors. Angew. Chem. Int. Ed. 2001, 40, 3118–30. [26] Wiskur SL, Floriano PN, Anslyn EV, McDevitt JT. A multicomponent sensing ensemble in solution: differentiation between structurally similar analytes. Angew. Chem. Int. Ed. 2003, 42, 2070–2. [27] Gallagher LT, Heo JS, Lopez MA, Ray, BM, Xiao J, Umali AP, Zhang A, Dharmarajan S, Heymann H, Anslyn EV. Pattern-based discrimination of organic acids and red wine varietals by arrays of synthetic receptors. Supramol. Chem. 2012, 24, 143–8. [28] Zyryanov GV, Palacios MA, Anzenbacher Jr. P. Rational design of a fluorescence-turn-on sensor array for phosphates in blood serum. Angew. Chem. Int. Ed. 2007, 46, 7849–52. [29] Buryak A, Severin K. Dynamic combinatorial libraries of dye complexes as sensors. Angew. Chem. Int. Ed. 2005, 44, 7935–8. [30] Buryak A, Severin K. Easy to optimize: dynamic combinatorial libraries of metal-dye complexes as flexible sensors for tripeptides. J. Comb. Chem. 2006, 8, 540–3.

11 Applying supramolecular systems CONSPECTUS: We have now seen many facets of supramolecular chemistry but mainly concentrated on the fundamental aspects. At this point you may wonder if supramolecular chemistry also has practical uses. This chapter is devoted to these aspects. Examples from different fields will show that some of the receptors or concepts presented in the previous chapters have made their way into applications and have commercial value. Others have the potential to be applied soon. It could therefore be that you have unknowingly come across a supramolecular system already.

11.1 Introduction What is all of this good for?

After the serendipitous discovery of crown ethers, much work in the area of supramolecular chemistry was devoted to understanding noncovalent interactions and to developing new receptors. Another important driving force was devising systems that help to understand and imitate biological processes. Much of the work carried out in this context was curiosity-driven, without any specific application in mind. Yet, with supramolecular chemistry progressively developing into a mature science, the question naturally arose at some point whether applications are conceivable. For a while, the lack of applications in supramolecular chemistry was indeed criticized and supramolecular systems were believed to be interesting but mostly useless, but this criticism missed the mark. In a field as diverse and multidisciplinary as supramolecular chemistry, there is not one application or one field of applications. Rather, supramolecular systems can play a role in many areas and it is precisely this aspect this chapter aims to convey. The first applications of supramolecular systems developed early. Crown ethers, for example, have long been used as alternatives to phase-transfer catalysts to promote chemical reactions (Section 11.5). This application is so well established that the fact that it is actually based on supramolecular concepts is often overlooked. The receptors most widely used in applications are cyclodextrins, of which many tons are produced every year to meet the demand. This success story was, however, not without obstacles. When early toxicological studies suggested that the consumption of cyclodextrins could be problematic, the interest in these compounds initially diminished. Only after it was clearly established that cyclodextrins are toxicologically safe, their uses in a wider range of applications was explored. Today, cyclodextrins and their derivatives are employed as drugs, constituents of drug formulations, food additives, etc. Several other

https://doi.org/10.1515/9783110595611-011

556

11 Applying supramolecular systems

supramolecular concepts or systems also have a practical value or might find uses soon. In the following chapters, we take a brief look at these applications to illustrate that supramolecular chemistry can make, and partly already makes, an impact even in our daily lives.

11.2 Applications in medicine 11.2.1 Drugs Any supramolecular receptor that binds under physiological conditions to a species involved in a biological process potentially influences the outcome. More often than not, the effect is harmful and the respective receptor therefore toxic. Crown ethers, for example, transport metal ions across lipid bilayers and thereby disrupt the membrane potential. If a receptor produces a positive biological effect, however, it could also make a drug. An example is the γ-cyclodextrin derivative 11.1 (Figure 11.1), which is known as sugammadex and sold under the trade name Bridion™. This compound reverses the effects of rocuronium and vecuronium, drugs used in general anesthesia to induce neuromuscular blockade and muscle relaxation. Both anesthetics have relatively long durations of action, which is problematic if the intubation of the patient fails. In addition, they cause side effects such as muscle weakness after surgery and difficulty in breathing. In these cases, sugammadex helps to speed up the recovery of the patient’s muscle function by acting as a sequestration agent that binds to the drug molecules and thus prevents them from interacting with their biological target. The efficiency of complex formations is partly due to the perfect fit of the large hydrophobic steroidal systems of rocuronium and vecuronium in the cavity of the eight-membered γ-cyclodextrin ring. In addition, complex formation positions the quaternary ammonium groups of the drug molecules close to the sulfonate groups along the sugammadex cavity, which causes a further stabilization of the complex. Rocuronium is, for example, bound in water with a log Ka of 7.3 [1], which is one of the largest known stability constants for a γ-cyclodextrin complex. This example demonstrates that the interactions between a supramolecular receptor and a suitable target is sufficient to induce a biological effect, provided that binding is strong and selective under physiological conditions. The approval of sugammadex as a drug motivated research into whether other receptors could also serve as sequestration agents. Among the compounds tested in this context [2], particularly promising results were obtained for the acyclic cucurbiturils developed by Lyle Isaacs (Section 4.1.12). An example is 11.2 (Figure 11.1) [3], which binds to rocuronium approximately two orders of magnitude stronger than sugammadex (log Ka = 9.5) and, like sugammadex, induces the reversal of the neuromuscular block in anesthetized rats [4]. Acyclic cucurbiturils also effectively bind to a number of

557

11.2 Applications in medicine

O O NaO 3 S

+ N

H

N H S

O

NaO 3 S

S

O

O O O H H

+

O

H

N H

HO O HO

O OH

NaO 3 S

SO 3 Na S

HO

S

SO 3 Na

H O

O

H O O

S

– O O P O – O SO 3 Na

O

H

H

O

H HO O O

N

Rocuronium

O OH O OH

O O

O OH

Vecuronium

S

O HO

S

H

H O

HO

O O OH H

O

O

NaO 3 S

O

S

O

11.1 SO 3 Na

P O O –

–

11.3 NaO 3 S

O

O

O

N

N

N

N

O

N

N

N

N

H NaO 3 S

O

O

O

N

N

N

N

H H

O

11.2

O

N

N

N

N

SO 3 Na

H

O

HN O

O

SO 3 Na

Methamphetamine

Figure 11.1: Molecular structures of sugammadex 11.1, rocuronium, and vecuronium, of the acyclic cucurbituril 11.2 and methamphetamine, and of the molecular clip 11.3.

drugs of abuse and exhibit promising activity for the treatment of intoxications with amphetamines such as methamphetamine [5]. Since these cucurbituril derivatives have also not shown any signs of toxicity in a number of studies, they have the potential to develop into the next generation of pharmaceutically useful sequestration agents. Other supramolecular receptors that exhibit biological activities are calixarenes (Section 4.1.7). Positively charged calixarenes act as gene transfection agents [6], for example, while sulfonatocalix[4]arenes and certain other negatively charged calixarene derivatives possess antiviral, antibacterial, and antifungal activity [7]. Sulfonatocalix[4]arenes also allow the detection of posttranslational modifications in histone proteins and disrupt certain protein recognition events [8]. The mode of action often relies on the binding of the sulfonatocalix[4]arene ring to positively

558

11 Applying supramolecular systems

charged biomolecules or residues in biomolecules such as the methylated amino groups in lysine side chains formed during posttranslational histone modification. The ability to recognize lysine residues on protein surfaces also underlies the biological activity of the molecular clips developed by Klärner and Schrader (Section 4.1.12), which have a number of interesting properties [9]. These clips, an example of which is 11.3 (Figure 11.1), inhibit the oligomerization and aggregation of proteins involved in diseases such as Alzheimer’s and Parkinson’s disease. The development of Alzheimer’s disease is, for example, accompanied by the appearance of Aβ, a peptide with 36–46 amino acid residues that derives from the amyloid precursor protein and produces the amyloid plaques found in the brains of Alzheimer’s patients. Plaque formation involves the aggregation of misfolded forms of the Aβ peptide, leading to oligomers that are toxic to nerve cells. Aggregation then continues in an autocatalytic fashion because the initially formed Aβ oligomers induce misfolding of further peptides, eventually affording plaques. The clip 11.3 not only prevents this process by interacting with lysine and arginine residues in the Aβ peptide, but even dissolves existing fibrils. Importantly, 11.3 is active in vivo where it exhibits no toxicity, rendering it a promising drug candidate for the treatment of a disease that is likely to affect more and more people in the future. Remarkably, 11.3 also shows promising activity to treat viral infections such as HIV, with the mode of action again relying on the inhibition of the protein-protein aggregation processes responsible for the pathological effects. Supramolecular drugs are also useful to treat diseases caused by malfunctioning ion channels. The so far incurable genetic disorder cystic fibrosis (CF), which affects about one out of every 3,000 newborns, is such a disease. Its cause is a mutation in the gene for the CF transmembrane conductance regulator protein, which is an anion channel in epithelial cell membranes that controls the flux of chloride and bicarbonate anions. This dysregulation affects the transport of water across the membrane, leading to the formation of sticky mucus in organs containing epithelial cells, for example the lung. A potential strategy to treat this disorder involves the administration of a carrier that helps to restore anion transport [10]. Much work in this context is inspired by the anion transport properties of prodigiosins (Section 9.3.2). A number of synthetic anion carriers such as those developed in the group of Philip A. Gale or the cholapods developed by Anthony P. Davis indeed exhibit promising properties, mediating anion transport in epithelial cells [11]. Importantly, some of the most potent carriers possess almost no toxicity, showing that anion transport, by itself, does not necessarily trigger cell death.

11.2.2 Drug formulations The effectiveness of a drug depends primarily on how well it interferes in the biological process that causes the disease, but there are additional important parameters. One is bioavailability, that is, the fraction of the administered drug that reaches the

11.2 Applications in medicine

559

systemic circulation. If this fraction is low, which is the case, for example, when orally administering a drug with a low water solubility, the biological effect is usually weak, even if the drug molecule is intrinsically very active. Another problem arises if a drug is too unstable to ensure a sufficiently long shelf life. These aspects can be addressed by mixing the active drug with additional components that have no biological activity themselves, but serve to improve the drug’s performance. This process of drug formulation, which affords the tablets, capsules, liquids, or cremes sold in pharmacies, ensures that drugs are safely and easily administrated with the proper dosage, are stable, and induce the desired effects. A widely used strategy to improve the water solubility of the active component in a drug formulation involves the use of cyclodextrins that form water-soluble inclusion complexes with the respective drug molecule (Section 4.1.4) [12]. This approach benefits from the lack of toxicity if cyclodextrins are ingested orally or administered on the skin. It is also rather versatile, not only because three different cyclodextrins exist to adapt the binding properties of the receptor to structural parameters of the drug, but also because the availability of various methylated, hydroxypropylated, or sulfobutylated cyclodextrin derivatives allows a further tuning of the properties of the complexes. Besides increasing the solubility and, in turn, bioavailability, the complexation of a drug with a cyclodextrin has other favorable effects. It normally also improves the chemical stability, for example. In addition, the slow release of the active molecules from the complex prolongs the biological effect. In this context, cross-linked cyclodextrin-based materials can serve as useful drug depots. Finally, cyclodextrins enhance the transport of certain drugs across biological membranes, although the exact mechanism of this process is not yet fully understood. Thus, cyclodextrins represent versatile components in drug formulations, which is why they are found in many commercial products today, ranging from tablets and capsules to ointments and drop solutions. A sulfobutylated β-cyclodextrin derivative is marketed under the name Captisol®, for example, and is currently used in nine FDAapproved drugs. It is worth noting in this context that there are alternatives for cyclodextrins in drug formulations. In vivo studies demonstrated, for example, that cyclic and acyclic cucurbiturils also increase the solubility of bioactive molecules [13, 14].

11.2.3 Imaging Imaging methods are indispensable in modern medicine for the noninvasive diagnose of diseases. The use of X-rays is a classic method but there are many other options available such as ultrasound imaging (US), computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), single-photon emission CT (SPECT), and optical imaging (OI). Some of these methods require the simultaneous

560

11 Applying supramolecular systems

administration of a molecular agent that produces the measured signal or that aids the imaging procedure. In PET, for example, positron-emitting radioligands are introduced into the body whose binding to a protein or participation in a metabolic process is traced. Optical imaging builds on probes in which a suitable chromophore is linked to a binding unit that serves to recognize the biological target. In MRI, paramagnetic ions or molecules are used as contrast agents to reduce the relaxation times of the hydrogen nuclei of water molecules located in close proximity, thereby enhancing the signal intensity and improving the image resolution. The structural design of these agents and the tuning of their properties can rely on supramolecular principles as the following few examples illustrate. Although MRI contrast agents are mostly based on gadolinium(III) complexes with chelate ligands such as diethylenetriaminepentaacetic acid (DTPA) or 1,4,7,10tetraazacyclododecane-N,N,N,N-tetraacetic acid (DOTA) (Figure 11.2a), paramagnetic organic agents have advantages such as low cytotoxicity and good biodegradability. Organic radicals often produce a weaker contrast (only one unpaired electron instead of the seven of Gd3+), however, and suffer from low stability. The latter issue can be addressed by complexing them with a supramolecular receptor. The 2,2,6,6tetramethylpiperidinyloxyl (TEMPO) derivative 11.4 (Figure 11.2a) forms a complex with CB[8] (Section 4.1.11) that has a log Ka of 6.2, for example. CB[8] thus serves as a supramolecular protecting group, preventing the aminoxyl radical from decomposing [15]. By attaching multiple copies of the CB[8] complex of 11.4 to the outer surface of a tobacco mosaic virus, it is indeed possible to obtain a contrast agent that is much more stable than an analogous system with unprotected aminoxyl radicals and has a contrast strength close to that of the gadolinium(III)-DOTA complex. Optical imaging in living organisms requires the light used to excite the probe and the emitted light to penetrate the tissue without a substantial loss of intensity caused by interactions with biomolecules or the heating of water. The optimal wavelength region, the so-called phototherapeutic window, is between 650 and 1,000 nm, in the deep red and near-infrared (NIR) region. In terms of their absorption and emission properties, the intensely colored squaraines of the general structure shown in Figure 11.2b are, in principle, well suited for the development of optical probes, but have the disadvantage that they tend to aggregate in aqueous media and react with nucleophiles. A strategy introduced by the group of Bradley D. Smith to overcome these drawbacks involves threading squaraines through a tetralactam ring [16]. The corresponding rotaxanes have a substantially improved solubility and stability, which allow using them, after conjugating with suitable target units, for imaging purposes. The rotaxane 11.5, whose zinc–dipicolylamine complexes selectively bind to the surfaces of bacterial cells, allows the staining and visualization of bacteria, for example, even of bacteria within living organisms such as mice [17]. Figure 11.2c shows images of E. coli cells stained with 11.5,

561

11.2 Applications in medicine

(a)

O O O O

O

O

O

O

N N

Gd 3+ O

O

N Gd 3+ N

N

O

O

N

O

O

+ NH 2

N3

N

N O

O

O

O

11.4

O 3+

Gd3+-DOTA

Gd -DTPA

(b)

O

–

O

R

R N

2+

R

O

N

NH HN – O

N

R

R N

–O

N

2+

R

– O NH HN N O O

Squaraine

(c)

11.5 N

O

30 s

2 mm

410 s

720 s

R=

O

N H

N Zn 2+ N

O

N Zn 2+ N N

1,085 s

1,285 s

1,685 s

Figure 11.2: Structures of the Gd3+ complexes of DTPA and DOTA and of the TEMPO derivative 11.4 (a), general structure of a squaraine dye and molecular structure of the probe 11.5 (b), and images showing the binary fission of E. coli cells stained with 11.5. After starting the experiment, the cells were imaged using fluorescence microscopy at the times specified in the images. Image adapted with permission from [17]. Copyright Wiley-VCH, 2007.

which illustrate that binary fission occurs within the 30 min of the measurement, indicating that the probe is not harmful during this time. The field of medicinal imaging is broad and there are numerous other probes that rely on principles of supramolecular chemistry. The calcium and sodium probes discussed in Section 10.2.1 are two further examples, which also illustrate that chromophores can play a role that absorb or emit outside the NIR region if imaging does not involve whole animals but smaller organisms or cell cultures.

562

11 Applying supramolecular systems

11.3 Applications in separation processes 11.3.1 Chromatography Chromatographic separations mostly rely on physisorption processes between substances dissolved in the mobile phase and selectors dissolved in or attached to the stationary phase. The underlying interactions are noncovalent in nature and subject to thermodynamic control, rendering them closely related to the interactions that stabilize receptor–substrate complexes in supramolecular chemistry. As a consequence, supramolecular receptors make useful selectors in chromatographic separations, with the properties of the respective columns typically correlating with the binding properties of the immobilized hosts. An early example reported by Donald J. Cram involves the use of a polymeric stationary phase containing the chiral crown ether 11.6 as the selector (Figure 11.3) [18]. This phase allows the enantiomeric resolution of the ammonium salts of various α-amino acids and α-amino acid esters. In all cases, the L-enantiomer elutes prior to the D-enantiomer, showing that the immobilized (R,R)-configured crown ether interacts more strongly with the D-configured amino acids, which is consistent with the results of solution studies (Section 4.1.1). Me O O

O

O

O

( R,R )- 11.6

O Polystyrene

O

Me

Figure 11.3: Molecular structure of the immobilized chiral crown ether (R,R)-11.6 that allows separating the enantiomers of α-amino acids and α-amino acid esters.

Many other crown ether-based selectors were developed and commerzialized, mostly for enantiomeric separations – respective columns are available from different suppliers. Immobilized calixarenes can also serve as selectors, but the most widely used ones are cyclodextrins, which has several reasons. First, cyclodextrins are available in different ring sizes and therefore allow separating a broad spectrum of compounds, especially hydrophobic organic compounds with residues that can be incorporated into the cyclodextrin cavity. Second, cyclodextrins can be easily immobilized and structurally modified to tune their selector properties and adapt them to a specific application. Native cyclodextrins or those with polar functional groups are typically used in liquid chromatography and capillary electrophoresis, for example, while nonpolar derivatives such as alkylated cyclodextrins are more suitable for gas chromatography. Finally, cyclodextrins are chiral and therefore able to separate enantiomers, which is probably

11.3 Applications in separation processes

563

the most important application of cyclodextrin-based columns. Most suppliers of chromatography equipment have such columns in their portfolio.

11.3.2 Extraction Hydrophobic organic compounds brought into the environment as herbicides or pesticides or other chemicals such as polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls, dioxins, or benzypyrenes are hazardous and often persistent pollutants. They are also good guests for cyclodextrins, however, which is why the use of cyclodextrins to extract these pollutants from water or soil represents an attractive clean-up strategy. Soil polluted with phenanthrene and pyrene, for example, can be decontaminated by extraction with an aqueous solution of permethylated β-cyclodextrin [19]. After reextracting the hydrophobic pollutants from the aqueous phase, the cyclodextrin solution can be reused. Cyclodextrins and cyclodextrin derivatives such as hydroxypropyl-β-cyclodextrin are also useful to extract dibenzodioxins and polychlorinated dibenzofurans from water [20], or PAHs from the oil of oil spill sites [21]. The ability of cyclodextrins to operate in these complex environments clearly demonstrates their usefulness for environmental remediation where the lack of toxicity is a further advantage. Cation or anion receptors are useful for the removal of environmentally relevant charged species. Examples of cations that are of concern are toxic transition metals or the radioactive uranyl cation UO22+. Environmentally problematic inorganic anions range from nitrate and phosphate, which are responsible for the eutrophication of water bodies, over toxic cyanide and arsenate anions, to the radioactive pertechnetate. The extraction of these species is challenging because they occur in different matrices: sometimes they are dissolved in fresh or in salt water, and sometimes they are adsorbed on solids. A few selected examples should illustrate possible strategies. Although not environmentally relevant in itself, sulfate anions pose problems in the long-term disposal of radioactive waste, of which large amounts are currently stored in tanks. It has been proposed to transform this waste through vitrification into glass logs that can subsequently be deposited in a geological repository [22]. Sulfate ions present in the waste are not well soluble in borosilicate glass, however, and therefore reduce the long-term stability of such logs. For vitrification to be useful, it is thus necessary to selectively extract small amounts of sulfate from a matrix that is rich in nitrate anions, which is challenging because nitrate anions are much easier to transfer from a polar into a nonpolar environment than doubly charged and strongly hydrated sulfate anions. One solution to this problem was reported by the Sessler group [23]. They showed that exposing a solution of the nitrate salt of the doubly protonated form of the lipophilic cyclo[8]pyrrole 11.7 (Figure 11.4) in toluene (containing additional trioctylammonium nitrate) to 0.02 M aqueous Na2SO4 and varying concentrations of NaNO3 causes the macrocycle to selectively pick up sulfate from the aqueous phase

564

11 Applying supramolecular systems

R R

R

NH NO + H − HN N3 R

HN

HN N

R R 11.7 (R = C11H23)

2−

[(H17C8)3NH ]2SO4

R NH

R

R

N+ HN

SO42− HN NH + H HN N

NH

R R

R

+

HN

NH

R

NH

R

R

R

R

HN

NH

R

N NH

R

NH NO3 HN

R

R

R

R

−

R 2 [(H17C8)3NH ]NO3− +

Toluene Water Na+ NO3−

Na+

Na+

NO3− Na+

NO3−

Na+

SO42− Na+

NO3−

Figure 11.4: Molecular structure of the cyclo[8]pyrrole 11.7 and schematic representation of the sulfate extraction mediated by 11.7.

and release the nitrate counterions. Although the performance is not sufficient for practical use, this system shows that high sulfate-over-nitrate selectivity under the conditions of solvent extraction is achievable. We will see a different approach to solving the same problem in the following section. The tetrahedral pertechnetate anion TcO4– is radioactive with an additional danger arising from the fact that it migrates in the surface layers of the earth’s crust so that it easily enters the food chain. One pertechnetate isotope is used in diagnostic imaging, whereas those produced during the nuclear fuel cycle have very long halflives. Strategies to remove pertechnetate from water include solvent extraction and ion exchange but these methods suffer from low selectivity. Strategies relying on supramolecular receptors potentially permit selective removal, but although pertechnetate may look similar to sulfate, it is actually larger and much less strongly coordinating, rendering the development of suitable receptors difficult [24]. Several receptor classes nevertheless proved to be promising. Examples are the CTV derivative 11.8 (Figure 11.5) with which TcO4– can be removed almost completely by a single extraction from a 3 mM aqueous solution into nitromethane even in the presence of competing anions [25], the sapphyrin derivative 11.9 whose monoprotonated form binds in 2.5 vol% methanol/water to pertechnetate with a log Ka of 3.6 [26], and the polyazacryptand 11.10 [27]. The hexaprotonated form of this receptor binds in water to pertechnetate with a log Ka of 5.5, two orders of magnitude more strongly than to nitrate. The only anion that is bound with a similar affinity is perrhenate ReO4–, which structurally closely resembles the pertechnetate anion. The extraction of the positively charged radioactive uranyl cation UO22+ is achieved with macrocyclic ligands containing pyrrole moieties [22] and also with a tripodal Kemp’s triacid-derived receptor with three cyclically arranged carboxy groups [28]. The binding and subsequent extraction of transition metal cations

11.3 Applications in separation processes

HO

OH OH

3+

3 PF6− Fe

HO

N

N

O Fe

OMe MeO O

O

Fe O

O OMe

N H N

N H N

H N

11.8

565

HN

NH

HN

NH

HN

NH

N

N

11.9

11.10

Figure 11.5: Molecular structures of the pertechnetate binders 11.8, 11.9, and 11.10.

relies mostly on coordinative interactions mediated by appropriate chelating ligands rather than supramolecular receptors.

11.3.3 Precipitation Another strategy to remove sulfate from radioactive waste prior to vitrification involves the selective precipitation of the anion. This process should ideally be reversible, allowing recycling of the precipitation agent. That this strategy is indeed feasible was demonstrated by Radu Custelcean, who developed several selective sulfate precipitation agents. One is the tripodal tris(urea) 11.11 (Figure 11.6a) that binds to sulfate by forming a complex in which the anion is encapsulated between two interdigitating subunits of 11.11 and held tight by twelve hydrogen bonds [29]. To illustrate this arrangement, the crystal structure of the sulfate complex of 11.11 is shown in Figure 11.6b. This complex crystallizes from an aqueous solution containing 5 M NaNO3, 1.25 M NaOH, and 0.044 M Na2SO4, which closely resembles the radioactive waste mixture in terms of ionic strength, alkalinity, and sulfate concentration [30]. In the precipitate formed, the sodium cations interconnect individual capsules by binding to the nitrogen atoms in the pyridine units. The yield amounts to 90%, meaning that only 10% of the originally present sulfate anions remain in solution after one round of precipitation. The thus formed precipitate decomposes upon dissolution in fresh water, leading to the crystallization of uncomplexed 11.11, which can thus be recycled. A second strategy is similar but involves the use of the bis(guanidinium) salt 11.12, which is obtained in one step from the condensation of aminoguanidinium chloride and terephthalaldehyde. While the chloride salt of 11.12 is well soluble in water, the corresponding sulfate salt has a solubility product pKSP of 9.6, only slightly smaller than that of BaSO4 (pKSP = 10.0) [31]. Adding 11.12 to an aqueous solution of Na2SO4

566

11 Applying supramolecular systems

(a)

(b)

N NH O NH N HN

O

HN

HN HN

O N

N

11.11

(c) OHC

CHO +

2 H2N

NH H2N

N H

N

H N

H2N + H2N

N H

N N

−

2 Cl

H N

NH2 11.12

+ NH2

NH2

Cl− + NH2 Fitration followed by resuspension in aqueous HCl

Na2SO4 H2N + H2N

Solution

N H

N N SO42−

H N

NH2

+ NH2

Precipitate

Crystalline H N NH2 N NH

Filtration followed by resuspension in water

NaOH Crystalline H2N H2N

+ N H

N N SO42−

H N

+

NH2 NH2

Figure 11.6: Molecular structures of the sulfate binders 11.11 (a), crystal structure of the complex between two molecules of 11.11 and a sulfate anion (b), and scheme showing the synthesis of 11.12 and the cycle of sulfate precipitation and recovery of the precipitation agent. Hydrogen atoms except those on NH groups are not shown in the crystal structure for reasons of clarity.

thus leads to the formation of a precipitate that can be filtered off. The precipitation agent is recycled by decomposing the sulfate salt with NaOH, filtration, and dissolving the salt-free form in aqueus HCl as shown in Figure 11.6c. The process is selective since only the sulfate salt of 11.12 precipitates from an aqueous solution containing a mixture of chloride (0.1 M), nitrate (0.07 M), and sulfate (0.034 M) salts.

11.4 Applications in materials chemistry

567

11.4 Applications in materials chemistry 11.4.1 Self-assembled polymers Polymer materials are ubiquitous today and are used for many different purposes, ranging from packaging and construction to sophisticated organic electronic devices. Although the most important aspect that controls the properties of polymeric materials and, in turn, their field of application is the structure of the monomers, noncovalent interactions within polymer chains or between chains also play a role. In polyamides, for example, hydrogen bonds exist between the NH and C=O groups that stabilize the mutual orientation of individual chains and thus influence properties such as the melting behavior and flexibility. The presence of noncovalent interactions alone, however, does not qualify a polymer to be supramolecular. Supramolecular polymers are instead defined as “polymeric arrays of monomeric units that are brought together by reversible and highly directional secondary interactions, resulting in polymeric properties in dilute and concentrated solutions, as well as in the bulk” [32]. Noncovalent interactions thus serve in supramolecular polymers to hold the monomers together, which is achieved in a number of different ways. The most frequently used interactions are likely hydrogen bonds, requiring the presence of suitable acceptor and donor patterns in the monomers. A popular motif, introduced by Egbert (Bert) W. Meijer and Rint P. Sijbesma, is based on the 2-ureido-4[1H]-pyrimidinone (UPy) unit 11.13 (Figure 11.7a) [33]. This building block is conformationally stabilized by an intramolecular hydrogen bond and features a self-complementary DDAA pattern that mediates the formation of stable dimers in apolar solvents. The dimerization constant log Ka in chloroform amounts to 7.3, for example, and in toluene even to 8.0. These interactions enable monomers containing two UPy units to polymerize as schematically shown in Figure 11.7b [34]. This polymerization occurs spontaneously when dissolving a small amount of the ditopic monomer 11.14 (Figure 11.7b) in chloroform, producing a very viscous solution. The solution can be processed to afford an elastic bendable material that resembles a conventional polymer but has unique properties normally not seen for most polymeric materials. It remains responsive to external stimuli, for example, even after the polymerization is complete, which is due to the reversibility of the interactions between the monomer units. The addition of a monofunctional UPy derivative such as 11.13 to the viscous chloroform solution of 11.14 causes a shift of the polymerization equilibrium from long chains to shorter ones, for example, with the concomitant reduction of the viscosity. The polymer also depolymerizes upon heating so that it is possible at any time to transfer the material back into a processable state with a low viscosity. Cooling

568

11 Applying supramolecular systems

(a)

(c) O O

2

N H

H N H

N N

H

N H

N H

O

N

N

O N

11.13

R

R HN

NH HN

O O

NHHN

HN N

O

H N

H N

H

O PrO PrO

O

R O O

R NH NH

PrO

O HN

(b)

C 13 H 27 O

N

H N

N O H C 13 H 27

H N

O N H

H N H

11.15

N

(R = N

C 7 H 15 )

O HN O

11.14

O PrO PrO

HN O O

HN R

NHHN NH HN R

PrO

O O R

NH NH R

Figure 11.7: Dimerization equilibrium of the monotopic UPy derivative 11.13 (a) and structure of the ditopic monomer 11.14, which polymerizes as schematically shown in (b). The structure of the ditopic calix[4]arene-derived monomer 11.15 whose self-assembly leads to polycaps is shown in (c).

leads back to the polymer. The reversibility of the interactions between the monomers also ensures that damages such as scratches or cracks are eliminated simply by heating, which is a behavior known as self-healing [35]. The material properties can moreover be adapted to various applications by structurally varying the linking unit between the UPy moieties. Since self-assembly even occurs in water when shielding the interacting UPy units from the aqueous medium, biocompatible materials are also available for medical applications [36]. Supramolecular polymers based on the UPy binding motif are commercialized by a company named SupraPolix [37]. A variety of other recognition motifs give rise to supramolecular polymers, of which the calix[4]arene tetraureas were already discussed in Section 5.3.3. Covalently linking two such calixarenes through functional groups along their narrower openings gives rise to monomers that orient the wide openings containing the urea moieties in a diverging fashion [38]. These monomers, an example of which is 11.15 (Figure 11.7c), thus form chains of hydrogen-bonded capsules called “polycaps,” which are able to host guest molecules at every linking unit. In addition to hydrogen bonding, metal coordination is another frequently used strategy of preparing supramolecular polymers [39]. Covalently assembled polymers that are able to reversibly exchange their building

11.4 Applications in materials chemistry

569

blocks are called dynamers. They are structurally robust, but have many properties of supramolecular polymers [40].

11.4.2 Polymer networks A somewhat complementary approach of producing responsive materials involves tuning the material properties of conventional polymers by introducing supramolecular interactions. Poly(acrylamide) and poly(acrylic acid) are soluble in water, for example, affording free flowing viscous solutions, but can be transformed into hydrogels by cross-linking. The mechanical stability of the resulting materials is typically demonstrated by using the tube inversion test, which involves inverting a vial containing a polymer solution. While a polymer in the sol state flows freely, a gel remains at the bottom when the vial is turned upside down. Figure 11.8 illustrates this behavior.

Cross-linking

Figure 11.8: Schematic representations of the structures of a noncross-linked (left) and crosslinked polymer (right) and method to distinguish both states with a tube inversion experiment. While a solution of the noncross-linked polymer flows freely, the cross-linked version affords a gel that remains at the bottom of the vial when the vial is turned upside down.

Cross-links in polymers are typically covalent in nature but can also be realized by introducing receptor and substrate moieties along the polymer chain that engage in complex formation. The latter strategy has the advantage that the reversibility of the noncovalent interactions gives rise to stimuli-responsive and self-healing materials. An example is the acrylamide-derived hydrogel described by Akira Harada that is obtained by copolymerizing acrylamide with small amounts of two additional monomers, one containing an adamantyl group and the other a β-cyclodextrin (β-CD) residue (Figure 11.9) [41]. Since adamantane is a good guest for β-CD (Section 4.1.4), the interaction of both compounds produces cross-links in the polymer and thus induces typical gel properties [42]. The presence of uncomplexed binding partners in this gel moreover leads to self-healing properties. When cutting the gel, for example, the separated fragments contain uncomplexed adamantyl and cyclodextrin residues along their surfaces. Pressing the pieces back together causes them to re-adhere, affording a gel that has almost the same mechanical stability as before. The possibility to control the interactions of certain receptor-substrate combinations also allows the preparation of materials that can be deliberately switched

570

11 Applying supramolecular systems

r

r

x

H2N

y

HN

O

O

HN

z

HN

O

O

O

NH

β-CD

Figure 11.9: Chemical structure of the copolymer containing adamantyl and β-CD residues and schematic representation of the corresponding hydrogel.

between the sol and the gel state. Mixing poly(acrylic acid) with β-CD residues as the host polymer with a guest polymer containing ferrocene subunits affords a hydrogel, for example, in which the cross-links arise from the incorporation of the ferrocene residues into the cyclodextrin cavity (Figure 11.10) [43]. The electrochemical or chemical oxidation of the ferrocene to ferrocenium units causes the complexes to dissociate and the polymer state to change from gel to sol. The reduction leads back to the gel.

r HO

NH

O

Gel

x

y

HN

O

O Fe

NH

O

β -CD

Reduction Oxidation

r x

HO

O

O

y

HN

+

Sol Fe

O

O N H

NH O

Fe

HN

Figure 11.10: Chemical structures of the host copolymer with β-CD residues, the guest copolymer with ferrocene units, and schematic representation of the gel-to-sol transition upon oxidation and reduction of the ferrocene moieties.

11.4 Applications in materials chemistry

571

In a similar way, a hydrogel containing azobenzene and β-CD residues can be reversibly switched between the sol and the gel state by light [44]. Harada demonstrated that similar strategies allow controlling the assembly of macroscopic polymeric objects consisting of an acrylamide-derived hydrogel stabilized by the covalent cross-linker N,N’–methylenebis(acrylamide) if this gel additionally contains the residues of a third monomer (Figure 11.11) [45]. Two hydrogels, of which one type is functionalized with β-CD and the other with adamantyl residues, stick together, for example, when meeting in a petri dish that contains a small amount of water because the surface-exposed residues bind to each other (Figure 11.11a) [46]. When using four different hydrogels, one containing tert-butyl, one n-butyl, one α-, and one β-CD residues, self-sorting takes place (Figure 11.11b), with pairwise interactions only occurring between the tert-butyl and β-CD-containing hydrogels and between those containing n-butyl and α-CD-residues. This

r

r

x

H2N

R =

y

O

R

z

O

HN

O

HN

O

(a)

r N H

z

Agitation R =

N H

β -CD R =

R = N H

N H

(b) Agitation

R =

α -CD

N

R =

N H

N H

(c) hv (365 nm)

E N

R =

N

N H

N

Agitation

Z

Figure 11.11: Schematic structures of the hydrogels containing cyclodextrins as hosts and different types of guest units and their self-assembling behavior. The interaction of the hydrogels containing βCD and adamantyl groups in an agitated petri dish is shown in (a). Four different hydrogels with α-CD, β-CD, n-butyl, and tert-butyl residues self-sort as shown in (b), and (c) illustrates that the interactions of hydrogels containing α-CD, β-CD, and azobenzene units is controlled by light.

572

11 Applying supramolecular systems

behavior is due to the fact that α-CD binds only to n-butyl groups but not to bulky tert-butyl groups, while β-CD interacts with both alkyl residues but forms a more stable complex with tert-butyl groups. This complex is thus preferentially formed, leaving the hydrogel containing α-CD to interact with the gel containing the n-butyl residues. Self-assembly can again be controlled by external stimuli. When mixing hydrogels of which one contains α-CD and the other β-CD residues with a third gel containing Econfigured azobenzene units, for example, the azobenzene-containing gel binds preferentially to the gel containing α-CD residues (Figure 11.11c) [47]. When switching the azobenzenes with light into the Z-state, the gel containing the azobenzene units dissociates from the α-CD-containing gel and binds to the β-CD-containing gel. Also crown ethers, pillararenes, or cucurbiturils are useful recognition elements to control polymer properties [42]. An example of the use of cucurbiturils to achieve the adhesion of two macroscopic objects is the molecular velcro described by Kimoon Kim, which consists of two silicon surfaces, one modified with CB[7] and one with ferrocene units (Figure 11.12) [48]. These surfaces so strongly interact with each other under water that a 1 × 1 cm2 contact area suffices to support a 2 kg weight. This system thus demonstrates that many weak interactions lead to high binding strengths when working cooperatively. Since the stability of the complex between CB[7] and ferrocene is mediated to a large extent by “high-energy water” (Section 4.1.11), the adhesion in the above system is lost in the dry state. In water, it can be controlled by appropriate

N NN O O

NN

N O N O NN O NN

O NN

H2O

O

Si

NN

NN O NN O O NN N

1 cm

Si

O O

O O NN NN N O

2 kg

S Fe NH

HN

=

=

Figure 11.12: Schematic illustration of the assembly and molecular components of the molecular velcro, which consists of two 1 × 1 cm2 large silicon wafers, one containing CB[7] units and the other ferrocene residues on the surface. Both surfaces so strongly interact in water that they suspend a 2 kg weight.

11.4 Applications in materials chemistry

573

stimuli. An agent that oxidizes the ferrocene to more weakly binding ferrocenium units causes the surfaces to separate, for example. A somewhat different approach for incorporating cross-links into polymers through supramolecular chemistry was developed by Kohzo Ito [49]. Polymers are used in this case that contain a small number of macrocyclic receptors threaded onto the chains. An example are polyrotaxanes consisting of polyethylene glycol chains onto which α-CD rings are threaded (Figure 11.13). Covalently linking some but not all of the rings leads to a material in which polymer chains are held together by figure-of-eight-shaped ditopic receptor molecules. The network is thus mechanically stabilized but the chains nevertheless move relatively freely at the junctions. These so-called slide-ring gels have unusual mechanical properties. They relieve tension caused by mechanical deformations, for example, by allowing the polymer chains to slide past each other at the junctions, which is not possible in a conventional cross-linked polymer. This effect leads to scratch-proof properties that render these materials useful coatings for automobiles, cell phones, mobile computers etc. [50].

Cross-linker

Figure 11.13: Schematic illustration of the formation and structure of a slide-ring gel.

11.4.3 Self-assembled gels Gels are also formed from low molecular weight compounds that self-assemble in solution to form fibrillar structures. The resulting network of fibers traps large amounts of the surrounding solvent and prevents the bulk flow of the material. Depending on whether gel formation takes place in organic solvents or in water, organogels or hydrogels are formed. A typical scanning electron microscope (SEM) image of such a gel network is shown in Figure 11.14. Gel formation starts with the assembly of individual building blocks, the so-called low molecular-weight gelators, to initially afford one-dimensional fibrils that further aggregate into fibers and bundles. This assembly is driven by typical noncovalent interactions such as hydrogen bonding. In water, the hydrophobic effect also plays an

574

11 Applying supramolecular systems

NH

O

NH H N

H N

N

O HN

0.5 μm

O

HN

Figure 11.14: Typical SEM image of a gel formed from a low molecular weight gelator, in this case the depicted tris(urea). Image adapted with permission from [51]. Copyright The Royal Chemical Society, 2006.

important role. The exact nature of the interactions varies with the structure of the gelators, of which many are based on ureas, peptides, sugars, steroids, or nucleobases. The advantage of such self-assembled gels compared to gels made from crosslinked polymers is their stimuli-responsiveness and the ease with which their properties can be tuned by changing the structure and ratio of the building blocks. Many gelators are moreover biocompatible so that gels can easily be tailored for applications in medicine or cosmetics [52].

11.5 Applications in catalysis Although impressive progress has been made in the development of supramolecular catalysts as we have seen in Chapter 8, practical applications for such sophisticated systems are not yet in sight. Beneficial effects on the rate of a reaction not only arise, however, if a receptor brings the reaction partners together or stabilizes the transition state of a reaction, but also if it modulates the solubility or reactivity of a reagent or ensures mass transport. Even commercial receptors are useful for such purposes. Crown ethers, for example, represent efficient phasetransfer catalysts. In phase-transfer catalysis, the catalyst promotes the reaction by facilitating the transfer of an anionic reagent or reaction partner dissolved in an aqueous phase into the organic phase where it reacts with the substrate. Conventional phasetransfer catalysts are salts of quaternary ammonium ions that exchange their counterion at the liquid–liquid interface, thus mediating the dissolution of the anionic

575

11.5 Applications in catalysis

reagent in the organic phase. A similar effect can be achieved with crown ethers that interact with the positively charged counterion of the reagent by forming a complex. When this complex migrates into the organic phase, the preservation of charge neutrality requires the anion to follow, thus allowing it to participate in the reaction. The catalytic cycle of a crown ether-mediated phase transfer is illustrated in Figure 11.15a for a nucleophilic substitution reaction.

(a)

R –OSO2CH3

(b)

R –X CH3SO3−

X– O O

O

O

O

K+ O

O

O O

O K+

O

N O

O

O

O

O

O

O

O

N O

O O

O

N

N

Organic phase Kryptofix® 221

Kryptofix® 222

Aqueous phase KOSO2CH3

KX

Figure 11.15: General illustration of the course of a nucleophilic substitution under phase-transfer conditions mediated by a crown ether (a), and example of cryptands commercialized under the brand name Kryptofix® (b).

Crown ethers are efficient catalysts for a number of reactions, including substitutions, eliminations, or oxidations [53], and chiral crown ethers also mediate enantioselective reactions [54]. These compounds have the further advantage that, in contrast to quaternary ammonium salts, they also promote the dissolution of solid salts in organic solvents so that phase-transfer proceeds directly from the solid to the organic phase. 18-Crown-6, for example, dissolves potassium permanganate in benzene, affording a purple solution that represents an efficient oxidation reagent. In a similar manner, potassium cyanide can be dissolved in dichloromethane or potassium acetate in acetonitrile. Anions are more reactive in organic solvents because they lack the shell of water molecules that surround them in water. The extent to which they are really “naked” when dissolved in an organic solvent depends on how strongly they interact with the counterion. The rather exposed arrangement of a cation in the cavity of a crown ether allows it to ion-pair with the anion, which reduces the anion’s reactivity. As a consequence, cryptands that fully surround the cation are often more effective phase-transfer catalysts [55]. Many cryptand-based phase-transfer catalysts are commercial and sold under the brand name Kryptofix®. Examples are shown in Figure 11.15b.

576

11 Applying supramolecular systems

The ability of supramolecular receptors to promote phase-transfer is also attractive for industrial processes. An example is the hydroformylation of higher alkenes under biphasic conditions. In this process, the substrates and products reside in the organic phase, while the catalyst is dissolved in water so that it can be separated and recycled (Figure 11.16a) [56]. To ensure mass transfer, a phasetransfer agent must also be present that brings the substrate from the organic phase to the catalyst and returns the product into the organic phase, explaining why this strategy is associated with the term inverse phase-transfer catalysis. Several supramolecular receptors are useful for this purpose, with methylated cyclodextrins and calixarenes being particularly effective. If the course of the reaction is affected by the mode of binding, these receptors can also influence the selectivity of the conversion, that is, the ratio with which linear and branched products are formed. Suitable water-soluble cyclodextrins and calixarene derivatives with appended phosphine groups can also serve as ligands for the rhodium (I) complexes that mediate these reactions.

(a)

The branched aldehyde CHO

(b)

is typically also formed CHO Organic phase

[Rh]

Aqueous phase

Ph P Ph

CHO

Water-soluble macrocyclic receptor

Ph Ph P

Rh(I)-catalyst with solubilizing ligands

N H N

O H O

11.16

CO/H 2 Figure 11.16: Schematic illustration of biphasic hydroformylations (a), and structure of the selfassembled rhodium(I) complex 11.16 (b).

Another modular concept of ligand development involves the use of noncovalent interactions to structurally stabilize chelating ligands. 11.16 (Figure 11.16b) is an example of a complex in which hydrogen bonds between the two heterocyclic rings help to bring the two phosphine residues together, facilitating their simultaneous interaction with the metal [57].

11.6 Applications in molecular electronics

577

11.6 Applications in molecular electronics Molecular electronics deals with the use of molecules to develop and assemble electronic devices. The molecular components used in this interdisciplinary research field often comprise compounds with extended π-systems linked to metals or assembled into films [58]. Assembling the devices involves a combination of covalent and noncovalent strategies so that drawing a line between molecular and supramolecular chemistry is difficult. We therefore look at only one example here that demonstrates the use of rotaxanes for the construction of an electronic memory device. The respective device is based on the idea of storing the binary information encoded in the “0” and “1” states of a bit in a bistable [2]rotaxane. To achieve a high information storage density, the respective rotaxanes must be arranged in an array that deliberately allows addressing and switching only some of them from one state to the other. The individual states must persist even after the trigger used for switching is turned off and there must be a way to retrieve the stored information. The feasibility of this concept was demonstrated by using the tetrathiafulvalene-containing rotaxane 11.17 developed in the Stoddart group (Figure 11.17) [59], which has a similar structure as the molecular valve discussed in Section 7.2. Sandwiching this rotaxane between 16 nm wide rows of orthogonally oriented Si and Ti nanowires gives rise to a crossbar architecture in which ca. 100 rotaxanes are present at each junction with their hydrophilic end groups preferentially oriented toward the Si wires. The switching of the rotaxanes at the junctions is achieved by applying high positive or negative voltage pulses that cause the oxidation and reduction of the tetrathiafulvalene groups. In the reduced state of the rotaxane, the blue box surrounds the tetrathiafulvalene unit as

16 nm wide Si nanowire

16 nm wide Ti nanowire O

O

O O

Si O

O

O

O O

O

O

O

O

O

O

O

O

O

3

O

S

S

S

S

+

O N

+N

O O

N

+N

O

O

Ti

+

11.17

Figure 11.17: Molecular structure of the bistable [2]rotaxane 11.17 and schematic illustration of the memory device assembled from it. In the respective crossbar architecture, rows of orthogonally arranged Si and Ti nanowires sandwich ca. 100 molecules of 11.17 at each junction.

578

11 Applying supramolecular systems

shown in Figure 11.17, while it moves to the naphthalene unit upon oxidation. Probing the respective states involves the use of a nonperturbing lower voltage. Although the first prototype of this device suffered from defects that reduced the reliability of the readout, a remarkably high storage density of 1011 bits per cm2 could be achieved, indicating that these molecular devices have the potential to be used for high-density data storage. Bistable rotaxanes have also proved to be promising building blocks for molecular electronics in other experimental setups [60].

11.7 Applications in consumer products 11.7.1 Textiles We now come to applications where supramolecular chemistry reaches our daily lives. Because it is essential that the compounds used in this context are nontoxic, available in large quantities, and inexpensive, only a few receptor types are relevant, of which the most important ones are again cyclodextrins. In the textile industry, the intensity and the distribution of the color on the fabric is improved by introducing the dye molecules in the form of their cyclodextrin complexes. Supramolecular receptors also allow reducing the residual color of the wastewater produced in dying processes. In this context, cucurbiturils are particularly useful because they form insoluble complexes with many dyes. Solid cucurbiturils allow the decoloration of the wastewater when used as a stationary phase in a column. The adsorbed dye molecules are selectively destroyed with ozone, while cucurbiturils are stable under these conditions. Cyclodextrins are furthermore used to impart special properties to finished textiles. Fixing cyclodextrins onto fibers leads to materials, for example, that adsorb unpleasant odors and thus reduce body odor development. Textile materials with cyclodextrins also serve as filters to remove organic compounds from the air, thus leading to protective cloths. Finally, complexes of cyclodextrins with drugs or perfumes immobilized on a fiber act as depots. If the thus treated cloths are not worn, the complexes are stable and the bound substances do not evaporate. When worn, the water on the skin and the slightly elevated temperature cause the complexes to dissociate and the active ingredients to be released. Underwear is on the market in Japan, for example, that contains the cyclodextrin complex of γ-linolenic acid to suppress the development of atopic dermatitis.

11.7.2 Food The only supramolecular receptors that are currently approved for use in foodstuff are cyclodextrins. Cyclodextrins serve in the food industry as emulsifiers because they

Bibliography

579

stabilize emulsions by interacting with their hydrophobic components. This property is useful to produce and ensure the long-term stability of sauces, cremes, and deserts. Cyclodextrins are also used to protect flavors and other active food ingredients against oxygen, heat, or light. The complexation of volatile flavors by cyclodextrins improves the shelf life of instant meals and reduces unpleasant tastes and odors. Many products that contain cyclodextrins are on the market, especially in Japan. Examples are powdered green tea or powdered flavors, spices, and herbs. In these products, cyclodextrins not only facilitate processing and act as protecting agents, they also influence the actual taste. Complexed flavors are often released slowly, which, for example, prolongs the taste of chewing gum.

11.7.3 Household Several other products are available in supermarkets that contain cyclodextrins. Examples are cosmetics that can contain cyclodextrins to solubilize, stabilize, and suppress the volatility of the fragrances. In addition, containers and wrapping materials are sometimes coated with cyclodextrins to bind preservatives or compounds with antibacterial properties. A widely known cyclodextrin-based supermarket product is manufactured by Procter & Gamble under the name Febreze™. It is sold as an odor eliminator and air freshener that is applied by spraying the solution onto fabrics or furniture. The active ingredient is hydroxypropyl β-cyclodextrin that masks the volatile odor molecules by complexation, thus reducing evaporation and the associated smells. The easiest way to conduct an experiment in supramolecular chemistry therefore involves spraying Febreze™ on your sofa. Is this the end?

Yes, this ends our journey through the field of supramolecular chemistry. We started with the basics and then discussed many different specific topics that characterize the field. One thing that has hopefully become clear is that supramolecular chemistry is a modern, diverse, and multidisciplinary field of research that is driven by the curiosity and creativity of the many contributing scientists. Some of their names, but by no means all of them, have been mentioned. If this book could spark your interest to join the community, your name may one day be among them.

Bibliography [1]

Cameron KS, Clark JK, Cooper A, Fielding L, Palin R, Rutherford SJ, Zhang MQ. Modified γ-cyclodextrins and their rocuronium complexes. Org. Lett. 2002, 4, 3403–6.

580

[2]

[3] [4]

[5]

[6]

[7] [8] [9] [10] [11]

[12] [13]

[14]

[15]

[16] [17]

[18]

11 Applying supramolecular systems

Shurpik DN, Mostovaya OA, Sevastyanov DA, Lenina OA, Sapunova AS, Voloshina AD, Petrov KA, Kovyazina IV, Cragg PJ, Stoikov II. Supramolecular neuromuscular blocker inhibition by a pillar[5]arene through aqueous inclusion of rocuronium bromide. Org. Biomol. Chem. 2019, 17, 9951–9. Ganapati S, Isaacs L. Acyclic cucurbit[n]uril-type receptors: preparation, molecular recognition properties and biological applications. Isr. J. Chem. 2018, 58, 250–63. Haerter F, Simons JCP, Foerster U, Moreno-Duarte I, Diaz-Gil D, Ganapati S, EikermannHaerter K, Ayata C, Zhang B, Blobner M, Isaacs L, Eikermann M. Comparative effectiveness of calabadion and sugammadex to reverse non-depolarizing neuromuscular-blocking agents. Anesthesiology 2015, 123, 1337–49. Ganapati S, Grabitz SD, Murkli S, Scheffenbichler F, Rudolph MI, Zavalij PY, Eikermann M, Isaacs L. Molecular containers bind drugs of abuse in vitro and reverse the hyperlocomotive effect of methamphetamine in rats. ChemBioChem 2017, 18, 1583–8. Bagnacani V, Franceschi V, Fantuzzi L, Casnati A, Donofrio G, Sansone F, Ungaro R. Lower rim guanidinocalix[4]arenes: macrocyclic nonviral vectors for cell transfection. Bioconjugate Chem. 2012, 23, 993–1002. Perret F, Lazar AN, Coleman AW. Biochemistry of the para-sulfonato-calix[n]arenes. Chem. Commun. 2006, 2425–38. Hof F. Host-guest chemistry that directly targets lysine methylation: synthetic host molecules as alternatives to bio-reagents. Chem. Commun., 2016, 52, 10093–108. Schrader T, Bitan G, Klärner FG. Molecular tweezers for lysine and arginine – powerful inhibitors of pathologic protein aggregation. Chem. Commun. 2016, 52, 11318–34. Busschaert N, Gale PA, Small-molecule lipid-bilayer anion transporters for biological applications. Angew. Chem. Int. Ed. 2013, 52, 1374–82. Li H, Valkenier H, Thorne AG, Dias CM, Cooper Ja, Kieffer M, Busschaert N, Gale PA, Sheppard DN, Davis AP. Anion carriers as potential treatments for cystic fibrosis: transport in cystic fibrosis cells, and additivity to channel-targeting drugs. Chem. Sci., 2019,10, 9663–72. Muankaew C, Loftsson T. Cyclodextrin-based formulations: a non-invasive platform for targeted drug delivery. Basic Clin. Pharmacol. Toxicol. 2018, 122, 46–55. Zhao Y, Pourgholami MH, Morris DL, Collins G, Day AI. Enhanced cytotoxicity of benzimidazole carbamate derivatives and solubilisation by encapsulation in cucurbit[n]uril. Org. Biomol. Chem. 2010, 8, 3328–37. Hettiarachchi G, Samanta SK, Falcinelli S, Zhang B, Moncelet D, Isaacs L, Briken V. Acyclic cucurbit[n]uril-type molecular container enables systemic delivery of effective doses of albendazole for treatment of SK-OV-3 xenograft tumors. Mol. Pharmaceutics 2016, 13, 809–18. Lee H, Shahrivarkevishahi A, Lumata JL, Luzuriaga MA, Hagge LM, Benjamin CE, Brohlin OR, Parish CR, Firouzi HR, Nielsen SO, Lumata LL, Gassensmith JJ. Supramolecular and biomacromolecular enhancement of metal-free magnetic resonance imaging contrast agents. Chem. Sci. 2020, 11, 2045–50. Gassensmith, JJ, Baumes JM, Smith BD. Discovery and early development of squaraine rotaxanes. Chem. Commun. 2009, 6329–38. Johnson JR, Fu N, Arunkumar E, Leevy WM, Gammon ST, Piwnica-Worms D, Smith BD. Squaraine rotaxanes: superior substitutes for Cy-5 in molecular probes for near-infrared fluorescence cell imaging. Angew. Chem. Int. Ed. 2007, 46, 5528–31. Sogah GDY, Cram DJ. Host-guest complexation. 14. Host covalently bound to polystyrene resin for chromatographic resolution of enantiomers of amino acid and ester salts. J. Am. Chem. Soc. 1979, 101, 3035–42.

Bibliography

581

[19] Petitgirard A, Djehiche M, Persello J, Fievet P, Fatin-Rouge N. PAH contaminated soil remediation by reusing an aqueous solution of cyclodextrins. Chemosphere 2009, 75, 714–8. [20] Cathum SJ, Dumouchel A, Punt M, Brown CE. Sorption/Desorption of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzo furans (PCDDs/PCDFs) in the presence of cyclodextrins. J. Soil Sediment Contamin. 2007, 16, 15–27. [21] Serio N, Levine M. Solvent effects in the extraction and detection of polycyclic aromatic hydrocarbons from complex oils in complex environments. J. Inclusion Phenom. Macrocyclic Chem. 2016, 84, 61–70. [22] Rambo BM, Sessler JL. Oligopyrrole macrocycles: receptors and chemosensors for potentially hazardous materials. Chem. Eur. J. 2011, 17, 4946–59. [23] Eller LR, Stȩpień M, Fowler CJ, Lee JT, Sessler JL, Moyer BA. Octamethyl-octaundecylcyclo [8]pyrrole: a promising sulfate anion extractant. J. Am. Chem. Soc. 2007, 129, 11020–1. [24] Katayev EA, Kolesnikov GV, Sessler JL. Molecular recognition of pertechnetate and perrhenate. Chem. Soc. Rev. 2009, 38, 1572–86. [25] Gawenis JA, Holman KT, Atwood JL, Jurisson SS. Extraction of pertechnetate and perrhenate from water with deep-cavity [CpFe(arene)]+-derivatized cyclotriveratrylenes. Inorg. Chem. 2002, 41, 6028–31. [26] Gorden AEV, Davis J, Sessler JL., Král V, Keogh DW, Schroeder NL. Monoprotonated sapphyrin-pertechnetate anion interactions in aqueous media. Supramol. Chem. 2004, 16, 91–100. [27] Alberto R, Bergamaschi G, Braband H, Fox T, Amendola V. 99TcO4−: selective recognition and trapping in aqueous solution. Angew. Chem. Int. Ed. 2012, 51, 9772–6. [28] Sather AC, Berryman OB, Rebek Jr. J. Selective recognition and extraction of the uranyl ion. J. Am. Chem. Soc. 2010, 132, 13572–4. [29] Custelcean, R. Urea-functionalized crystalline capsules for recognition and separation of tetrahedral oxoanions. Chem. Commun. 2013, 49, 2173–82. [30] Rajbanshi A, Moyer BA, Custelcean R. Sulfate separation from aqueous alkaline solutions by selective crystallization of alkali metal coordination capsules. Cryst. Growth Des. 2011, 11, 2702–6. [31] Custelcean R, Williams NJ, Seipp CA, Ivanov AS, Bryantsev VS. Aqueous sulfate separation by sequestration of [(SO4)2(H2O)4]4− clusters within highly insoluble imine-linked bisguanidinium crystals. Chem. Eur. J. 2016, 22, 1997–2003. [32] Brunsveld L, Folmer BJB, Meijer EW, Sijbesma RP. Supramolecular polymers. Chem. Rev. 2001, 101, 4071–98. [33] Sijbesma RP, Beijer FH, Brunsveld L, Folmer BJB, Hirschberg JHKK, Lange RFM, Lowe JKL, Meijer EW. Reversible polymers formed from self-complementary monomers using quadruple hydrogen bonding. Science 1997, 278, 1601–4. [34] Sijbesma RP, Meijer EW. Quadruple hydrogen bonded systems. Chem. Commun. 2003, 5–16. [35] Hart, LR, Harries JL, Greenland BW, Colquhoun HM, Hayes W. Healable supramolecular polymers. Polym. Chem. 2013, 4, 4860–70. [36] Goor OJGM, Hendrikse SIS, Dankers PYW, Meijer EW. From supramolecular polymers to multicomponent biomaterials. Chem. Soc. Rev. 2017, 46, 6621–37. [37] Bosman AW, Sijbesma RP, Meijer EW. Supramolecular polymers at work. Mater. Today 2004, 7, 34–9. [38] Castellano RK, Rudkevich DM, Rebek Jr. J. Polycaps: reversibly formed polymeric capsules. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 7132–37. [39] Whittell, GR, Hager, MD, Schubert US, Manners I. Functional soft materials from metallopolymers and metallosupramolecular polymers. Nat. Mater. 2001, 10, 176–88.

582

11 Applying supramolecular systems

[40] Roy N, Bruchmann B, Lehn JM. DYNAMERS: dynamic polymers as self-healing materials. Chem. Soc. Rev. 2015, 44, 3786–807. [41] Kakuta T, Takashima Y, Nakahata M, Otsubo M, Yamaguchi H, Harada A. Preorganized hydrogel: self-healing properties of supramolecular hydrogels formed by polymerization of host-guest-monomers that contain cyclodextrins and hydrophobic guest groups. Adv. Mater. 2013, 25, 2849–53. [42] Sinawang G, Osaki M, Takashima Y, Yamaguchi H, Harada A. Supramolecular self-healing materials from non-covalent cross-linking host–guest interactions. Chem. Commun. 2020, 56, 4381–95. [43] Nakahata M, Takashima Y, Yamaguchi H, Harada A. Redox-responsive self-healing materials formed from host–guest polymers. Nat. Commun. 2011, 2, 511. [44] Takashima Y, Nakayama T, Miyauchi M, Kawaguchi Y, Yamaguchi H, Harada A. Complex formation and gelation between copolymers containing pendant azobenzene groups and cyclodextrin polymers. Chem. Lett. 2004, 3, 890–1. [45] Harada A, Takashima Y, Nakahata M. Supramolecular polymeric materials via cyclodextringuest interactions. Acc. Chem. Res. 2014, 47, 2128−40. [46] Harada, A, Kobayashi, R, Takashima Y, Hashidzume A, Yamaguchi H. Macroscopic selfassembly through molecular recognition. Nat. Chem. 2011, 3, 34–7. [47] Yamaguchi H, Kobayashi Y, Kobayashi R, Takashima Y, Hashidzume A, Harada A. Photoswitchable gel assembly based on molecular recognition. Nat. Commun. 2012, 3, 603. [48] Ahn Y, Jang Y, Selvapalam N, Yun G, Kim K. Supramolecular velcro for reversible underwater adhesion. Angew. Chem. Int. Ed. 2013, 52, 3140–4. [49] Okumura Y, Ito K. The polyrotaxane gel: a topological gel by figure-of-eight cross-links. Adv. Mater. 2001, 13, 485–7. [50] Ito K. Novel entropic elasticity of polymeric materials: why is slide-ring gel so soft? Polym. J. 2012, 44, 38–41. [51] Stanley CE, Clarke N, Anderson KM, Elder JA, Lenthall JT, Steed JW. Anion binding inhibition of the formation of a helical organogel. Chem. Commun. 2006, 3199–201. [52] Hirst AR, Escuder B, Miravet JF, Smith DK. High-tech applications of self-assembling supramolecular nanostructured gel-phase materials: from regenerative medicine to electronic devices. Angew. Chem. Int. Ed. 2008, 47, 8002–18. [53] Gokel GW, Durst HD. Principles and synthetic applications in crown ether chemistry. Synthesis 1976, 168–84. [54] Oliveira MT, Lee JW. Asymmetric cation-binding catalysis. ChemCatChem 2017, 9, 377–84. [55] Landini D, Maia A, Montanari F, Pirisi FM. Crown ethers as phase-transfer catalysts. A comparison of anionic activation in aqueous–organic two-phase systems and in low polarity anhydrous solutions by perhydrodibenzo-18-crown-6, lipophilic quaternary salts, and cryptands. J. Chem. Soc., Perkin Trans. 2 1980, 46–51. [56] Obrecht L, Kamer PCJ, Laan W. Alternative approaches for the aqueous-organic biphasic hydroformylation of higher alkenes. Catal. Sci. Technol. 2013, 3, 541–51. [57] Breit B, Seiche W. Hydrogen Bonding as a construction element for bidentate donor ligands in homogeneous catalysis: regioselective hydroformylation of terminal alkenes. J. Am. Chem. Soc. 2003, 125, 6608–9. [58] Mathew PT, Fang F. Advances in molecular electronics: a brief review. Engineering 2018, 4, 760–1. [59] Green JE, Choi JW, Boukai A, Bunimovich Y, Johnston-Halperin E, DeIonno E, Luo Y, Sheriff BA, Xu K, Shin YS, Tseng HR, Stoddart JF, Heath JR. A 160-kilobit molecular electronic memory patterned at 1011 bits per square centimetre. Nature 2007, 445, 414–7. [60] Yu H, Luo Y, Beverly K, Stoddart JF, Tseng HR, Heath JR. The molecule-electrode interface in single-molecule transistors. Angew. Chem. Int. Ed. 2003, 42, 5706–11.

12 Appendices 12.1 Concentrations in a 1:2 binding equilibrium How do I do the math?

The fundamental equations to describe the formation of a 1:2 receptor–substrate complex were introduced in Section 2.1 but a strategy to solve these equations was not discussed. The approach used to generate the graphs in Figure 2.3 is presented here [1], but alternative methods exist [2]. We start from the laws of mass action in equations (12.1a,b) that describe the stepwise formation of the 1:2 complex. R + S Ð C11 Ka11 =

C11 + S Ð C12

c11 C cR cS

Ka12 =

c12 C 11 c cS C

ð12:1a,bÞ

The mass balances for these equilibria are specified in equations (12.2a,b). 12 cR = c0R − c11 C − cC

(12:2a)

12 cS = c0S − c11 C − 2 cC

(12:2b)

To reduce the number of unknowns, we first seek an expression for c11 C starting from equations (12.1a,b) and (12.2a), c0R − c11 C −

c11 C − Ka12 c11 C cS = 0 Ka11 cS

(12:3)

An expression for cS results by combining (12.1b) and (12.2b). 12 11 cS = c0S − c11 C − 2 Ka cC cS

(12:4)

The rearrangement of equation (12.4) affords equation (12.5). cS =

c0S − c11 C 1 + 2 Ka12 c11 C

(12:5)

When combining this equation with equation (12.3), equation (12.6) is obtained. c0R − c11 C −

12 11 c0S − c11 c11 C C 1 + 2 Ka cC − Ka12 c11 =0 C 0 11 11 Ka cS − cC 1 + 2 Ka12 c11 C

https://doi.org/10.1515/9783110595611-012

(12:6)

584

12 Appendices

Rearranging equation (12.6) affords the cubic equation (12.7).     
11 2 4 Ka12
11 3 2 12 12 0 cC + 1 − 2 Ka cR + 11 cC Ka 1 − 11 Ka Ka  
 1 0 0 − c0R + c0S + 11 + Ka12 c0S c0S − 2c0R c11 C + cR cS = 0 Ka

(12:7)

by using Newton’s iterative method. To this end, A solution for c11 C can be found

11 
11  ′ equation (12.7) is defined as f c11 C . The derivative of f cC , f cC , thus has the following expression.     
 4 Ka12
11 2 2 12 12 0 c c11 1 − + 2 1 − 2 K c + = 3 K f ′ c11 C a C a R C Ka11 Ka11 (12:8)  
 1 0 0 12 0 0 0 − cR + cS + 11 + Ka cS cS − 2 cR = 0 Ka We now calculate c11 C by using equation (12.9). 11 c11 C ðnewÞ = cC ðoldÞ −


 f c11 ðoldÞ
C11  f ′ c ðoldÞ

(12:9)

C

0 The first iteration requires a guess for c11 C ðoldÞ and using cR =2 is often a good start. The calculation is repeated by using the resulting value c11 C ðnewÞ as the new starting value c11 C ðoldÞ for the next iteration until convergence, which is typically reached 12 after five to ten steps. With the c11 C thus obtained, cC is calculated by using equation (12.11), which is obtained by combining (12.1b) and (12.2b) to give equation (12.10) followed by rearrangement.
 12 11 0 11 12 (12:10) c12 C = Ka cC cS − cC − 2 cC
 0 11 Ka12 c11 C cS − cC (12:11) c12 C = 1 + 2 Ka12 c11 C

Analogous expressions can be derived if complex formation involves the binding of two receptor molecules to one substrate. The simplest way to approach this situation involves replacing c0S for c0R and vice versa in the above equations.

12.2 Concentrations in an indicator displacement assay In an indicator displacement assay (IDA), the substrate S (analyte) is added to the solution of a receptor R and an indicator I. Both the substrate and the indicator bind to the same binding site of the receptor so that the two complexes CS and CI, of which CS needs to have a higher stability than CI, coexist in the equilibrium. If only

12.2 Concentrations in an indicator displacement assay

585

1:1 complexes are involved, the overall equilibrium can be broken down into two equilibria that describe the formation of the substrate complex CS and the indicator complex CI with the associated laws of mass action (12.12a,b). CI + S Ð CS + I R + S Ð CS KaS

=

cC

R + I Ð CI KaI =

S

cR cS

cC

I

cR cI

ð12:12a,bÞ

The concentrations cR , cS , cI , cCS , and cCI correlate with the total concentrations c0R , c0S , and c0I according to the mass balances (12.13a–c). cR = c0R − cCS − cCI

(12:13a)

cS = c0S − cCS

(12:13b)

cI = c0I − cCI

(12:13c)

We use these equations to derive an expression in which only one unknown concentration is left. If we chose cR as the unknown, we need to define all other unknown concentrations in terms of cR . To do so, we combine the above equations and thus initially arrive at the expressions (12.14a–c). cCS =

KaS cR 0 c 1 + KaS cR S

(12:14a)

c CI =

KaI cR 0 c 1 + KaI cR I

(12:14b)

c0I 1 + KaI cR

(12:14c)

cI =

Substituting these equations into (12.13a) leads to (12.15). cR = c0R −

KaS cR 0 KaI cR 0 c − c S 1 + KaS cR 1 + KaI cR I

Rearranging this equation gives the cubic equation (12.16). 
 KaS KaI ðcR Þ3 + KaS + KaI + KaS KaI c0S + c0I − c0R ðcR Þ2
 + ½1 + KaS c0S + KaI c0I − KaS + KaI c0R cR − c0R = 0

(12:15)

(12:16)

A solution for this equation is again found by using Newton’s method as explained in the previous section. Once cR is known, the other concentrations are calculated by using the equations (12.14a–c). The total concentrations of the host and the indicator are normally kept constant in an IDA, while the concentration of the substrate (analyte) is increased. The

586

12 Appendices

change of the solution absorbance Aobs during the titration is recorded and then plotted vs. the substrate concentration c0S . This measurement is performed at a wavelength at which neither the receptor nor the substrate absorb so that Aobs only reflects the contributions of the free indicator and of its complex according to the expression (12.17) (length of the cuvette = 1 cm). Aobs = AI + AcC = εI cI + εCI cCI I

(12:17)

Substituting the above derived equations (12.14c) and (12.14b) into (12.17) affords equation (12.18) that allows fitting the experimental results to obtain KaS . The parameters εI , εCI , and KaI should be determined prior to the titration to reduce the number of variables and thus increase the reliability of fitting. Aobs =


 c0I εI + εCI KaI cR 1 + KaI cR

(12:18)

Bibliography [1]

[2]

Hargrove, AE, Zhong Z, Sessler JL, Anslyn EV. Algorithms for the determination of binding constants and enantiomeric excess in complex host: guest equilibria using optical measurements. New J. Chem. 2010, 34, 348–54. Thordarson, P. Determining association constants from titration experiments in supramolecular chemistry. Chem. Soc. Rev. 2011, 40, 1305–23.

Index acceptor 39, 63, 297, 299, 402, 448, 538, 541 – halogen bond 54 – hydrogen bond 46–47, 49–53, 73–74, 77, 81–83, 85, 134, 142–143, 167, 218, 232–233, 260, 266–267, 270–271, 291, 405, 440, 469, 567 acetylcholine 58, 222, 287, 351, 380 actin 442 activation barrier 15, 24, 145, 178, 183 activation enthalpy 463 activation entropy 104, 255–256, 463–464, 475 active metal template synthesis 380, 395, 417, 449 adrenaline 222 alkaloid 330, 525 allosteric effect 98 allosterism 5, 93 Alzheimer’s disease 213, 248, 558 amplification factor 348, 358 anion–π interactions 60, 62, 169, 224–225, 227, 523 Anslyn, Eric V. 542, 549 antibiotic 102, 522 antiport 515–516, 525, 527 Anzenbacher Jr., Pavel 550 Aoyama, Yasohiro 173, 289 aquaporin 528 aromatic interactions 38, 138, 207–210, 213, 227, 313, 487 – displaced 63 – edge-to-face 62, 70, 72, 139–140, 142, 232 – T-shaped 62 association constant 8, 12, 31, 82, 90, 109, 184 – microscopic 94–95 autocatalysis 357, 499, 501–502, 504–505, 508, 558 avidity 90 Bacillus macerans 125, 128 Baekeland, Leo 153 Bakelite 154 Ballester, Pablo 168 BC50 parameter 22 Beer, Paul D. 412, 416, 418, 547 Beer–Lambert law 28 Behrend, Rolf 193, 196

https://doi.org/10.1515/9783110595611-013

Behrend’s polymer 194 Bender, Myron L. 472 Benesi–Hildebrand method 21 o-benzyne 185 Berryman, Orion B. 296 β-barrel 519, 523 Beuerle, Florian 340 Biedermann, Frank 200 bilayer 77, 248, 262, 291, 293, 556 – phospholipid 445, 513, 516–517, 524 binding isotherm 11, 17, 19–21, 29, 35 – fluorescence 30 – ITC 33–34 – NMR 27 – UV–vis 28 biological membrane 513, 559 biosensor 152 bis(rosette) 260, 274–277 blue box 141, 232, 399–403, 423, 435–436, 438, 447, 577 Böhmer, Volker 286, 409, 411 Boolean algebra 540 Borromean rings 388–389 Breslow, Ronald 473, 481 Bridion™ 556 Brownian motion 429, 434 Bruns, Carson J. 424 buckminsterfullerene 325, 370 Busch, Daryl H. 6, 105, 255 butterfly effect 358 c parameter 35 calixarene 124, 138, 153, 165, 169–171, 174, 186, 188, 190, 219, 221–222, 231, 274, 314–315, 330, 423, 484, 540, 557, 562, 568, 576 – 1,2-alternate conformation 156, 171 – 1,3-alternate conformation 156, 171 – cone conformation 155, 157, 170 – partial cone conformation 156 – pinched cone conformation 156, 171 – pleated loop conformation 158 – sulfonato- 164, 222, 232, 557 – tetraurea 286, 409 calixpyrrole 165, 225, 230, 563 – 1,3-alternate conformation 166 – strapped 168

588

Index

capsid 247–248, 278 capsule 169, 173, 180, 264, 278, 295, 297, 315, 409, 411, 476, 493, 496, 559, 565, 568 carbohydrate recognition 233, 539 carbonic anhydrase 348 carceplex 181 carcerand 172, 174, 181, 278–279, 281, 337 carrier 514–515, 517, 522, 525, 558 cascade complex 121, 229 catalysis 5, 62, 208, 295, 474, 476, 489, 495, 520, 574 catalytic triad 468 catenand 383, 386–387 catenane 142, 352, 370–374, 376, 378, 380, 383, 388, 392, 394–395, 398–399, 403–404, 416, 418, 420, 434–435, 451, 454, 487 – synthesis 383, 400–401, 413 – clipping 400–401 – entwining 383, 413 – threading 383, 413 catenate 383–384, 387–388 cation-π interactions 41, 57, 60–61, 62, 71, 112, 138, 141, 149, 151, 161, 166, 172, 190, 221–222, 230–231, 424, 496 cavitand 169, 174, 176, 181, 199, 315, 337 – deep 41, 174–176, 178–180, 221–222, 232, 282, 297, 476, 479 – octa acid 180, 222, 263–264, 476 – self-folding 176, 474–475 CGTase 128–129 channel 514, 516–520, 523–524, 528, 558 chaotropic effect 78, 131 charge density 43, 50, 57–58, 108, 113, 167, 231 charge-relay system 468–469 charge-transfer band 39, 202 charge-transfer interactions 39, 55, 63, 138, 142, 190, 232, 398, 400, 402, 412, 424, 435 chelate effect 88–91, 182, 220 chemodosimeter 535 chemometric methods 548 – linear discriminant analysis 548, 550, 552 – principal component analysis 548, 550 chemosensor 534–538, 540–541, 544, 548, 550 – array 548, 550 – electrochemical 546–547 – fluorescence 538

– optical 536, 545 – redox 545–546 – sensing ensemble 535, 544, 548–549 – turn-off 538 – turn-on 538 chiral memory 276 chloride ion channel 523 cholapod 526, 558 chromophore 23, 28, 533, 536–538, 540–541, 544, 560–561 – fluorescence 30 chymotrypsin 468–469, 471–472, 481 circular dichroism spectroscopy 217, 276, 414 clipping 382, 400–401, 403, 419–420 coalescence 24 co-conformation 386, 406, 435, 437, 439–440, 451, 453 cohesiveness 140 Collet, André 145 complementarity 85–86, 87, 118, 120, 150, 286, 499 complex stoichiometry 12, 17, 33–35 conformational reorganization 39, 57, 86–87, 96, 108, 113, 117, 178, 184, 210, 218, 223, 318, 373 constrictive binding 16, 184 contrast agent 560 Cooper, Andrew I. 338 cooperativity 5, 52, 67, 89–91, 93–95, 137, 270, 572 – allosteric 94–95 – chelate 91, 93, 251, 267 – intermolecular 97 – negative heterotopic 97 – negative 94, 96, 303 – polarization-enhanced 51 – positive 94, 96, 229–230, 303 coordination cage 302, 314, 321, 325, 471, 476, 479, 493, 496 copper(I)-catalyzed azide–alkyne cycloaddition 135, 191, 398, 449 Corey–Pauling–Koltun model 128, 469, 489 Cornforth, John 154 coronand 102, 112 coronate 102 CPK model 128, 469 Cram, Donald J. 5, 16, 29, 87, 109–110, 122, 124–125, 169, 174, 180–182, 184–185, 223, 337, 468–469, 471, 481, 527, 562

Index

Cramer, Friedrich 125, 472 Crick, Francis H. C. 6 cross-coupling reaction 163, 191 crown ether 4–5, 29, 44, 57, 89, 97, 101, 114–119, 124–125, 127, 152, 160, 162, 197, 205, 220–223, 228–229, 255–256, 330, 339, 378, 380, 391, 399–401, 403–404, 423, 435, 444–446, 487, 491, 520, 523, 537–540, 555–556, 572, 574–575 – chiral 109, 562 18-crown-6 103–104, 107–109, 117–118, 198, 405, 445, 538, 575 cryptand 4, 112, 114, 124, 138, 162, 220–223, 355–356, 575 – polyazacryptand 120–121, 225, 229, 335, 564 cryptate 114 cryptophane 146–147, 182, 221–222, 232, 278, 281 crystal engineering 55 cucurbituril 190, 194, 203, 214, 219, 221–222, 232, 330, 423–424, 464, 544, 572, 578 – acyclic 214, 222, 556, 559 cumulative protonation constant 30–31 Custelcean, Radu 565 cyclam 112 cyclen 112 cyclobutadiene 185 cyclodextrin 6, 125, 140, 176, 180, 182, 190, 194, 199, 221–222, 232, 420, 423, 471, 541, 544, 555, 562–563, 576, 578 – α-cyclodextrin 125, 127, 187, 424, 572 – β-cyclodextrin 125, 127, 419, 473, 481–482, 569–570, 572, 579 – γ-cyclodextrin 125, 127, 310, 556 cyclodextrin glycosyltransferase 128 cyclopeptide 152, 291, 293–294, 520 cyclophane 60, 137, 144, 153–154, 160, 165, 169–170, 186, 190, 193, 195, 199, 203, 214, 231–232, 353, 399, 413, 423–424, 484 cyclotriveratrylene 138, 180, 186, 228, 314, 330, 338, 423 cystic fibrosis 558 cystic fibrosis transmembrane conductance regulator 558 cytosol 77 Davis, Anthony P. 233, 526, 558 Day, Anthony 194, 197 de Silva, A. Prasanna 536

589

dehydration 130, 179, 200 desolvation 58, 76, 78, 81, 107–109, 130–131, 143, 184, 335 dibenzo-18-crown-6 101–103 dibenzo-24-crown-8 391, 404, 445 Diederich, François 140, 297, 484 diphenylmethane 139, 142, 144, 423 dipole moment 43, 45, 48–49, 59, 61, 65, 73, 83, 126, 131, 160, 176, 294, 538, 541 directionality 43, 51, 55, 227, 266, 295–296, 298, 408, 452, 454, 492 dispersion interactions 47, 64–65, 67, 72, 81, 89, 131, 149, 151, 232, 248, 487 dissociation constant 8, 22 donor 39, 63, 299–302, 306, 309, 314, 318, 351, 389, 398, 402, 437, 448, 538–539, 541 – halogen bond 54–55, 295 – hydrogen bond 46–47, 49–53, 73, 77, 81–83, 85, 105, 120, 133, 143, 166, 168, 213, 216, 218, 224–226, 232–233, 260, 266–267, 270–271, 289, 291, 405, 408, 413, 526, 567 dopamine 152, 202, 222 double-mutant cycle 69–71, 231 Dougherty, Dennis A. 141 dynamer 569 dynamic combinatorial chemistry 342, 358, 551 dynamic combinatorial resolution 351 dynamic covalent chemistry 326, 329–330, 340–342, 344, 360 dynamic library 342, 344, 346–349, 351, 353, 355, 358–359, 503, 551 dynein 455 effective molarity 92, 251, 253–254, 464, 476 electrostatic potential 38, 41, 43, 54–56, 59–60, 62–63, 81, 126, 190, 195, 211, 232, 524 enthalpy 13–14, 27, 32, 34–35, 39, 76–79, 87, 89, 118, 150–151, 200–202, 253–255 – activation 463 – hydration 43 enthalpy–entropy compensation 15, 151 entropy 13–14, 27, 33, 76–79, 87–88, 104, 108, 118, 133, 140, 151, 180, 184, 199, 253–255, 281, 315, 322, 324, 464–465 – activation 104, 255–256, 463–464, 475 – hydration 133 enzyme 2, 5–6, 86, 93, 128, 248, 348–349, 468, 471–472, 481, 485, 495, 498

590

Index

– carbonic anhydrase 348 – chymotrypsin 468–469, 471–472, 481 – coenzyme 173 – lipase 351 – mimic 125, 468, 471, 473, 489, 498 – pyruvate oxidase 484 equilibrium 7–8, 10–12, 15–17, 22, 24, 26, 34, 74–75, 81, 83, 87, 91–95, 157, 170, 231, 234, 249, 251, 255, 257–259, 266, 277–278, 311, 317, 322–324, 327–328, 332, 341–342, 345, 357, 402, 440, 479, 494, 505–506, 534, 542, 551, 567, 584 – binding 11, 74, 81, 177, 546 – complexation 177 – conformational 71, 156, 158, 160–161, 171, 176, 178, 187–188, 437 – thermodynamic 16, 246, 303, 305, 329, 334, 351, 421 equilibrium constant 71, 235, 251, 253, 322, 347 eutrophication 224, 563 excimer 540 extraction coefficient 29 Feringa, Bernard L. 398, 435, 455, 458, 491 Feynman, Richard 431–432, 459 Fischer, Emil 6, 86 Flood, Amar 227 fluorescence 30, 517, 533, 537–538, 540–541, 544 fluorescence spectroscopy 30, 533 foldamer 232–233, 290, 296, 415 – definition 215 Förster resonance energy transfer 541 French, Dexter 125 Freudenberg, Karl Johann 125 Fujita, Makoto 309, 476 Fyles, Thomas M. 521 Gale, Philip A. 526, 558 gauche effect 108 gelator 573–574 genetic code 2 Ghadiri, M. Reza 292, 506, 508, 520 Gibb, Bruce C. 180, 263, 476 Gibbs free energy 12–13, 16, 67, 75, 77, 90, 140, 262, 546 – activation 103, 105, 171, 178, 256, 380, 463, 480 Gibbs–Helmholtz equation 13, 133, 176

Gibson Harry W. 394 glycoluril 193–197, 202–203, 213, 279–281, 286 Gokel, George W. 520, 524 gramicidin A 519–521 Granja, Juan R. 293 Grotthuss mechanism 530 Gutsche, C. David 154 halogen bond 54, 227, 295, 297, 418 – acceptor 54 – comparison with hydrogen bond 55 – definition 54 – donor 54–55, 295 – geometry 55 – receptor 56 – σ-hole 54–55, 54–55 halogen bonding 224–225, 227, 419–420, 487 Hamilton, Andrew D. 143 Hamilton receptor 144, 232 Harada, Akira 569, 571 Heisenberg's uncertainty principle 24 helicate 297, 302, 306, 321, 323, 393 – chiral 303–304 – circular 304, 388, 393 – double 303, 305, 307, 309, 381, 392 – meso 304 – quadruple 309 – triple 297, 302, 304–305, 307 hemicarceplex 183–185 hemicarcerand 16, 172, 174, 178, 182–185, 218, 278, 337 hemicryptophane 147, 152–153, 182, 228, 278 hemicucurbituril 203 hemispherand 124, 152, 182, 221–222 heteroditopic receptor 152 heteroleptic complex 309, 312, 314, 384 heterotopic system 96 high dilution conditions 2, 116, 138, 149, 392 high-energy water 79, 130–131, 199–200, 572 Hill coefficient 96 Hill plot 95–96 Högberg, A. G. Sverker 169 homoleptic complex 312, 314, 384 homotopic system 97 Hoogsteen base pairing 208, 210, 501 Hopf link 388 host–guest chemistry 5 Huc, Ivan 218

Index

Hüning, Siegfried 141 Hunter, Christopher A. 80–82, 85, 142, 390, 406, 416 hydraphile 520 hydration 43–44, 46, 50, 66, 74, 77–78, 151, 180, 198–199, 224, 264 hydrogel 569–573 hydrogen bond 6, 54–55, 67, 73–74, 77–79, 83, 106, 109, 116, 119–120, 124, 126, 130–131, 133, 142, 144, 152, 155, 157–158, 164, 168, 176, 178, 188, 201, 207, 216, 222–223, 225, 227–228, 231, 266–267, 270–271, 274, 276, 279–282, 285–286, 289–290, 292, 295, 297, 334, 405, 409, 412, 416, 440, 476, 501, 504, 525, 528, 565, 567, 576 – acceptor 46–47, 49–53, 73–74, 77, 81–83, 85, 134, 142–143, 167, 218, 232–233, 260, 266–267, 270–271, 291, 405, 440, 469, 567 – bifurcated 52, 286, 295 – comparison with halogen bond 55 – cooperative 51 – correlation with donor acidity 48 – C–H donors 49 – definition 46 – donor 46–47, 49–53, 73, 77, 81–83, 85, 105, 120, 133, 143, 166, 168, 213, 216, 218, 224–226, 232–233, 260, 266–267, 270–271, 289, 291, 405, 408, 413, 526, 567 – geometry 51 – IR spectroscopy 47 – NMR spectroscopy 47 – pattern 53, 143, 208, 252, 266–267, 269, 288, 294–295, 334, 567 – polarization-enhanced 51 – resonance-assisted 50 – secondary 53, 267 – solvent effect 74 – strength 47 – water 77, 200–201 hydrogen bond parameter 81 hydrogen bonding 45–47, 49–51, 55, 68, 70, 73, 76, 82–83, 89, 113, 126, 138, 153, 158, 170, 176, 178, 208, 213, 221–222, 224–225, 230, 232–233, 252, 255, 260, 267, 269–270, 287–288, 294, 424, 440, 475, 487, 568, 573 hydrogen bonding interactions 38, 82, 113, 142, 166, 173, 179, 209, 286, 391, 404, 407, 411–412, 419, 440, 523

591

hydrophobic effect 78, 81, 85, 130, 138, 180, 213, 248, 255, 262–263, 265, 270, 317–318, 334, 402–403, 421, 423–424, 573 – nonclassical 78 hydrophobic interactions 78 indicator 3, 533–534, 542, 548, 584–585 indicator displacement assay 542, 544, 584–585 induced fit 5, 6, 86, 87, 166, 212, 215 information storage 5, 577 interlocked molecule 192, 329, 370, 373–374, 376–377, 380, 383, 386–387, 393, 395, 398–399, 405, 411–412, 416, 418, 423–424, 434–435, 437, 439 internal charge transfer 537 intrinsic median binding concentration 22 ion transport 62, 294, 517 – anion 520, 523–526, 558 – cation 518–520, 523, 556 ionophore 102, 215, 522–523, 544 IR spectroscopy 46 Isaacs, Lyle 213, 260, 556 isothermal titration calorimetry 27, 32 Ito, Kohzo 573 James, Tony D. 235 Janus molecule 270 jelly doughnut 281 Job, Paul 18 Job plot 19 Kamlet–Taft parameters 73, 81 katapinand 223 Kelly, T. Ross 431–432, 455 Kemp’s triacid 206, 208, 474, 487, 501, 564 Kim, Kimoon 194, 197, 572 kinesin 430, 455 kinetic template 105, 256, 480, 499 Klärner, Frank-Gerrit 211, 213, 558 knot 373, 376, 381, 390, 392–393, 403–404 – figure eight 421, 423 – pentafoil 394 – trefoil 370–372, 378, 381, 388, 390, 392, 414, 421 Koga, Kenji 139 Koshland Jr., Daniel E. 6, 86 Kryptofix® 114, 575

592

Index

lab on a molecule 540 lariat ether 57, 61, 112, 118, 220 – bibracchial 112 law of mass action 8, 10, 17, 26, 92, 95, 500, 583, 585 Le Chatelier’s principle 257 Lee, Chang-Hee 168 Lehn, Jean-Marie 2, 4–5, 114–115, 119, 121, 223, 270, 302, 307, 343, 348 Leigh, David A. 388, 390, 395, 440, 491 Lennard–Jones potential 64 lifetime 17, 23–24, 519 linear discriminant analysis 548, 550, 552 link 392–393, 403 – Borromean rings 388–389 – Hopf link 388 – Solomon link 388, 393, 421, 423 lipase 351 lipid membrane 445, 518, 522–523 liposome 513, 528 liquid membrane 110, 515 lock-and-key 6, 86 London dispersion interactions 65 Lüttringhaus, Arthur 376, 383 macrobicyclic effect 89, 118 macrocyclic effect 89, 117, 205, 208 macrocyclization 103–104, 117, 144, 215, 251–252, 330, 332–334, 352, 374, 378, 382, 399–400, 402, 405, 408, 414 magnetic anisotropy 140, 144 magnetic resonance imaging 152, 559 mass balance 8, 10, 17, 25, 28, 30, 583, 585 Mastalerz, Michael 338 Matile, Stefan 290, 519, 523 mechanical bond 373, 376, 386, 433–434, 454 mechanically interlocked molecule 373–374, 376–377, 380, 383, 386–387, 395, 398–399, 411–412, 416, 418, 423–424, 434–435, 437, 439 median binding concentration 22 Meijer, Egbert W. 567 membrane 77, 102, 220, 248, 262, 291, 293, 430, 513, 523, 528 – biological 513, 559 – lipid 445, 518, 522–523 – liquid 110, 515 – phospholipid 445, 513, 516–517, 524

– polymer 544 membrane potential 515, 518–519, 523, 556 membrane transport 102, 291, 513, 515, 517 – active 515 – antiport 515–516, 525, 527 – bilayer conductance 516 – carrier 514–515, 517, 522, 525, 558 – channel 514, 516–520, 523–524, 528, 558 – symport 515–517, 525 – U-tube 515 – uniport 515, 523 memory device 435, 577 metacyclophane 154, 170, 187 method of continuous variations 18 micelle 79, 255, 263, 265, 513 Mock, William L. 194, 196, 198 molar absorptivity coefficient 23, 28–30 molecular balance 71–72 molecular box 309 molecular cable car 445 molecular cage 147, 152, 180–181, 232–233, 278, 281 molecular capsule 169, 264, 278 molecular cleft 206–207, 232 molecular clip 206, 211, 213, 558 molecular daisy chain 442 molecular glue 202 molecular jelly doughnut 281 molecular machine 114, 192, 374, 398, 424, 429–430, 433–435, 451, 455 molecular motion 5, 424 molecular motor 433, 435, 451, 455, 457, 491 molecular muscle 442, 444 molecular pincers 206 molecular pump 447 molecular ratchet 431 molecular shuttle 437 molecular softball 280, 476 molecular switch 438–439 molecular tennis ball 279–280 molecular torsion balance 71 molecular train 437 molecular tweezers 206, 208, 211, 213, 221–222, 232 molecular umbrella 527 molecular velcro 572 monesin 522 multivalency 5, 52, 89–90, 137 myosin 430, 442

Index

nanocar 433, 457–458 nanoparticle 137, 445 nanoroadster 458 nanoswitch 489, 491 Nau, Werner M. 200, 544 Newton’s iterative method 584–585 Nitschke, Jonathan R. 322 NMR spectroscopy 27, 46, 71, 144, 168, 185, 188, 227, 264, 270, 283, 289, 298, 386, 392, 431, 436 – dynamic 156, 176 – spectrum 23–26, 47, 139–140, 142, 144, 151, 153, 156, 161–162, 173, 177–178, 181, 199, 212, 261, 275, 279, 283, 286, 431 – timescale 23, 25–26, 71, 156, 173, 176–178, 218, 276–277, 283, 286 Nobel Prize 5, 101, 114, 383, 398, 435, 455 Nolte, Roeland J. M. 213 nonactin 522 nonlinear regression 21, 27, 29–30 noradrenaline 222 octa acid 180, 222, 263–264, 476 Ogoshi, Tomoki 186, 190 olefin metathesis 329 – ring-closing 388, 390, 392, 410–411, 413 olympiadane 401 optical imaging 559–560 organocatalysis 485, 491 organocatalyst 471, 487 organogel 573 orthocyclophane 144, 187 Otto, Sijbren 334, 356 out-of-equilibrium 249, 357 packing coefficient 281 paracyclophane 138–139, 186 paraquat 64, 141–142, 192, 202, 423, 444–445, 447 Park, Chung Ho 223 Parkinson’s disease 213, 248, 558 passive metal template synthesis 395, 397 Pauling, Linus C. 6 Pauli repulsion 65 Pedersen, Charles J. 4–5, 101–102, 106, 114, 125, 223 peripheral crowding 272, 274 phase-transfer catalysis 525, 574

593

– inverse 576 phase-transfer catalyst 487, 574–575 phenol–formaldehyde resin 153 Philp, Douglas 503 phospholipid 445, 513, 516–517, 524 photoinduced charge transfer 537 photoinduced electron transfer 538 picrate extraction method 29, 123, 527 pillararene 138, 186, 221–222, 232, 330, 423, 572 π–π interactions 61, 72 π–π stacking 63 π-slide 524 podand 89, 112, 117, 205, 216 polyazacryptand 120–121, 225, 229, 335, 564 polycaps 568 polymer 137, 153–154, 193–194, 202, 544, 567–570, 572–574 polyrotaxane 573 porous liquid 338 positron emission tomography 559 postfunctionalization 202, 246, 329 potassium channel protein 518 predisposition 174, 254, 266, 269, 290, 302, 310, 321, 332, 340, 357, 390, 415 preorganization 5, 87, 89, 96, 108, 118, 122, 124, 133, 148, 154, 162, 168, 175–176, 182, 200, 205–208, 210, 215–216, 221, 223, 254–255, 266–267, 273–274, 376, 380, 395–396, 404, 411, 418–419, 480, 495, 498, 526, 543 pretzelane 409 principal component analysis 548, 550 principle of maximum site occupancy 254, 307, 327, 384 principle of microscopic reversibility 431 Pringsheim, Hans 6, 125 probe 533–536, 542, 544, 548, 560–561 product inhibition 467–468, 471, 474, 476, 481, 493, 495, 499–501, 505, 508 proximity effect 474 pseudo-high dilution conditions 116 pseudorotation 171 pseudorotaxane 373–374, 376, 380, 383, 391, 399, 403–404, 407–408, 416, 423, 445 pyrogallolarene 172, 181, 183, 289 pyrrole 22, 165, 169, 225, 230, 269, 525, 564 pyruvate oxidase 484

594

Index

quadrupole moment 50, 57 radioligand 560 rate equation 16, 466 Raymond, Kenneth N. 321, 494 Rebek Jr., Julius 173, 176, 206, 279, 281, 286, 289, 474, 476, 501–502 Rebek’s 55% rule 281 receptor 7 – acyclic cucurbituril 214, 222, 556, 559 – bibracchial lariat ether 112 – calixarene 124, 138, 153, 169–171, 174, 186, 188, 190, 219, 221–222, 231–232, 274, 286, 314–315, 330, 409, 423, 484, 540, 557, 562, 568, 576 – calixpyrrole 165, 225, 230 – carcerand 172, 174, 181, 278–279, 281, 337 – cavitand 41, 169, 174–176, 178–181, 199, 221–222, 232, 282, 297, 315, 337, 474–476, 479 – coronand 102, 112 – crown ether 4–5, 29, 44, 57, 89, 97, 101, 114–119, 124–125, 127, 152, 160, 162, 197, 205, 220–223, 228–229, 255–256, 330, 339, 378, 380, 391, 399–401, 403–404, 423, 435, 444–446, 487, 491, 520, 523, 537–540, 555–556, 562, 572, 574–575 – cryptand 4, 112, 114, 124, 138, 162, 220–223, 355–356, 575 – cryptophane 146–147, 182, 221–222, 232, 278, 281 – cucurbituril 190, 194, 203, 219, 221–222, 232, 330, 423–424, 464, 544, 572, 578 – cyclodextrin 6, 125, 140, 176, 180, 182, 187, 190, 194, 199, 221–222, 232, 310, 419–420, 423–424, 471, 473, 481–482, 541, 544, 555–556, 559, 562–563, 569–570, 572, 576, 578–579 – cyclophane 60, 137, 144, 153–154, 160, 165, 169–170, 186, 190, 193, 195, 199, 203, 214, 231–232, 353, 399, 413, 423–424, 484 – cyclotriveratrylene 138, 180, 186, 228, 314, 330, 338, 423 – foldamer 215, 232–233, 290, 296, 415 – hemicarcerand 16, 172, 174, 178, 182–185, 218, 278, 337 – hemicryptophane 147, 152–153, 182, 228, 278 – hemicucurbituril 203 – hemispherand 124, 152, 182, 221–222

– katapinand 223 – lariat ether 57, 61, 112, 118, 220 – pillararene 138, 186, 221–222, 232, 330, 423, 572 – podand 89, 112, 117, 205, 216 – polyazacryptand 120–121, 225, 229, 335, 564 – polyazamacrocycle 30, 42, 67 – pyrogallolarene 172, 181, 183, 289 – resorcinarene 138, 169, 186, 189–190, 263, 283, 288–289, 314, 330, 337, 423, 496 – spherand 122, 152, 182, 221, 223, 469 Regen, Steven L. 527 Reinhoudt, David N. 274, 278 relative permittivity 40, 66, 73–74, 76 replication 5, 359, 374, 498–499 – self-replication 357, 499–500, 502, 504–508 resorcinarene 138, 169, 186, 189–190, 263, 283, 288–289, 314, 330, 337, 423, 496 – boat conformation 171 – crown conformation 170 – diamond conformation 171 – rccc isomer 170 – rctc isomer 170 – rctt isomer 170 – rtct isomer 171 – saddle conformation 171 responsiveness 246, 323, 567, 569, 574 ring-closing metathesis 388, 390, 392, 410–411, 413 rocuronium 556 rosette 267, 271–274 – bis(rosette) 260, 274–277 rotaxane 142, 370, 373, 376, 380, 382, 390, 394–395, 397–399, 403–404, 416–420, 424, 430, 434–440, 442, 445–448, 451, 453, 491, 560, 577–578 – synthesis 382–383, 394, 403, 417, 419–420, 424, 447 – clipping 382, 403, 419–420 – shrinking 382 – slipping 382, 403, 447 – stoppering 383, 394, 403, 424 – swelling 382 – trapping 383, 417 Saenger, Wolfram 125 salt bridge 47, 67–68, 207, 221–222, 483, 487 – strength in water 68 Sanders, Jeremy K. M. 343, 351, 402, 420

Index

Sauvage, Jean-Pierre 114, 383–384, 392–393, 398, 435, 439, 442, 444, 455 Schardinger dextrins 125, 128 Schardinger, Franz 125 Schiff base 332 Schill, Gottfried 376, 383 Schmidtchen, Franz P. 120 Schmittel, Michael 312 Schneider, Hans-Jörg 171 Schrader, Thomas 213, 558 self-assembly 14, 161, 180, 192, 246, 389, 464, 489, 498, 508, 519, 568, 572 – definition 246 – metal-directed 255, 258, 309, 387 – subcomponent 322–323, 325, 394 self-complementarity 499 self-folding 176, 179, 262, 474 self-healing 568–569 self-healing material 569 self-organization 249 self-replication 357, 499–500, 502, 504–506, 508 self-sorting 260–262, 305–307, 314, 335, 571 – completive 260, 305, 307 – definition 260 – incomplete 260 – integrative 260 – narcissistic 260, 262, 305, 307 – social 260, 309, 314 sensing – differential 547 – ensemble 535, 544, 548–549 sensor 435, 534 Sessler, Jonathan L. 165, 563 Severin, Kay 550 shape selectivity 37 Shinkai, Seiji 235 Shionoya, Mitsuhiko 265 shrinking 382 signal transduction 2 Sijbesma, Rint P. 567 Simmons, Howard E. 223 Sindelar, Vladimir 203 size selectivity 37 slide-ring gel 573 slipping 403, 447 Smith, Bradley D. 229, 560 social isomerism 283 softball 280, 476 Solomon link 388, 393, 421, 423

595

solvation 13–14, 39–40, 43, 58, 66, 74–75, 77, 80–81, 83–84, 89, 107, 118, 133, 140, 176, 227, 234, 471, 518 solvent effect 13–14, 55, 58, 63, 73–74, 82–83, 130, 160, 164, 202, 227, 253 solvent reorganization 75–76, 78–80 solvophobic effect 78 solvophobic interactions 72 speciation diagram 32 spectroscopy – fluorescence 30, 533 – NMR 23–26, 27, 46–47, 71, 139–140, 142, 144, 151, 153, 156, 161–162, 168, 173, 177–178, 181, 185, 188, 199, 212, 227, 261, 264, 270, 275, 279, 283, 286, 289, 298, 386, 392, 431, 436 – UV–vis 23, 28–29, 39, 63, 101, 202, 402, 469, 540, 542, 544, 548–551 spherand 122, 152, 182, 221, 223, 469 squaraine 560 squaramide 526 square root law 500, 508 stability constant 8–9, 10, 11, 16–17, 21, 25, 27, 29–30, 32, 35, 92, 95, 149, 151–152, 167, 197, 199, 251, 310, 500, 533, 542, 556 standard state 12–13 star of David catenane 388, 394 steady state 16 steric gearing 205 steroid 173, 264, 482, 526–527, 556, 574 stimuli-responsive material 567, 569, 574 stimuli-responsive system 323 Stoddart, J. Fraser 141, 386, 389, 398–399, 402–403, 424, 435, 444, 455, 577 stoichiometry number 34 stoppering 383, 394, 403, 424 subcomponent self-assembly 322–323, 325, 394 substrate affinity 5, 98, 133–134, 141, 174, 193, 205, 208, 215, 222 substrate selectivity 2, 6, 86, 191, 205 substrate sensing 544 sugammadex 556 supramolecular chemistry – definition 3 supramolecular tandem enzyme assay 544 surfactant 513 swelling 382 symport 515–517, 525 systems chemistry 360, 502, 506

596

Index

template 105, 138, 181–182, 190, 197, 249, 255–257, 266, 281, 286, 317, 328–329, 342–349, 351–353, 355, 358–359, 376, 380–383, 394, 399–400, 409, 498, 500, 507 – anion 412–413, 416 – covalent 376, 383 – kinetic 105, 256, 480, 499 – metal 392, 394–395, 397–398, 417, 449 – thermodynamic 105, 257–258 template approach 148 template effect 6, 105, 116, 160, 190, 197, 259, 304, 310, 330, 342, 344–345, 355–356, 413, 480, 501 – definition 105 – thermodynamic 257 temple receptor 233 tennis ball 279–280 thermodynamic template 105, 257–258 thermogram 33 thiourea 137, 225, 487, 526, 544 Tian, He 445 Tiefenbacher, Konrad 497 timescale – human 17, 118 – NMR 23, 25–26, 71, 156, 173, 176–178, 218, 276–277, 283, 286 – UV–vis 28 titration 17, 21 – competitive 27, 35 – fluorescence 30 – indicator displacement assay 543, 586 – ITC 28, 32–34 – NMR 22, 26–27, 29 – potentiometry 30 – UV–vis 22, 28–30 tobacco mosaic virus 247, 560 topology 369, 376, 380, 383, 388, 390, 392, 401, 409–410, 414, 421 Tour, James M. 457 transmembrane electrochemical gradient 522 transport 2, 5, 102, 110, 291, 293, 519, 574 – active 515 – activity 446, 520, 524–525, 527 – anion 520, 523–526, 558 – cation 445, 518–520, 523, 556 – chloride 523, 525–526 – efficiency 523

– ion 62, 294 – membrane 513, 515, 517, 559 – mitochondrial potassium 102 – rate 446, 516–517, 527 – selectivity 519 – water 528, 530 trapping 383, 417 trianglimine 333 tube inversion test 569 unilamellar vesicle 517 uniport 515, 523 urea 22, 124–125, 193, 195, 203, 225, 286–287, 295, 409, 513, 526, 528, 544, 565, 568, 574 UV–vis spectroscopy 23, 28–29, 39, 63, 101, 202, 402, 469, 540, 542, 544, 548–551 – timescale 28 valinomycin 102, 106, 215, 522–523 van der Waals interactions 64, 66, 68, 78, 182, 286 van der Waals volume 281 van't Hoff plot 27 vecuronium 556 velcrand 175–177, 263 – kite 175–176, 263, 479 – vase 175–176, 297, 479 velcraplex 176 vesicle 79, 255, 263, 445, 513 – unilamellar 517 Villiers, Antoine 6, 125 virus 247–248 – tobacco mosaic 247, 560 vitamin 173 vitamin B6 473 vitrification 563, 565 Vögtle, Fritz 406, 414–416, 537 von Baeyer, Adolf 153, 165, 193 von Delius, Max 114, 355 von Kiedrowski, Günter 500–501 VSEPR theory 51 Walba, David M. 378 Warmuth, Ralf 185, 338 Wasserman, Edel 374–376 water structure 78 water – cluster 51, 200–201, 232

Index

– molecule 43–44, 46, 50, 77–79, 84, 87, 120, 126, 130, 133, 179, 199–202, 218, 224, 232, 262, 288–289, 297, 322, 482, 494, 520, 528–529, 560, 575 Watson, James D. 6 Watson–Crick base pairing 208, 210, 252, 501–502 Werner, Alfred 6 wheat germ agglutinin 233

Whitesides, George M. 272, 274 Whitlock, Howard W. 208, 211 Williamson ether synthesis 103, 162, 256, 384, 392 Wiseman parameter 35 Zimmerman, Steven C. 208, 213 Zinke, Alois 154

597