Dynamic Combinatorial Chemistry: The Unexpected Choice of Receptors by Guest Molecules

Survival of the fittest? Amplification of the best receptor from a dynamic combinatorial library does not always proceed according to survival of the “fittest”. However, it has recently resulted in the discovery of some unexpected receptors for various substrates, as illustrated by the example of a catenane receptor (see picture; green) for the neurotransmitter acetylcholine (red). Although our knowledge in the broad field of chemistry has increased enormously over the past decades, we are still unable to give precise predictions on the behavior of molecules in solution. Therefore, combinatorial approaches and high-throughput experimentation have become increasingly important by facilitating the rapid discovery of lead compounds and the optimization of processes in which trial-and-error is the dominant approach. In the mid-1990s, a fundamentally new concept in this area, dynamic combinatorial chemistry (DCC), was introduced by the groups of Sanders,1 Lehn,2 and others.3, 4 Whereas traditional combinatorial libraries are based on the preparation of kinetically stable compounds and subsequent screening for their function, this new concept takes advantage of reversible bonds, covalent or noncovalent, by generating a random (virtual) library whose members are in dynamic equilibrium. The formation of the library is under thermodynamic control, which gives rise to new properties such as adaptation; that is, the library readjusts its distribution by external stimuli. Upon addition of a target molecule, the mixture of potential receptors in the library re-equilibrates in favor of the best receptor(s) for the target. By using this strategy the preparation and screening of the receptor for a certain target is integrated in one step,4d as the receptor is chosen by the target. Libraries of receptors of variable complexity have been constructed, with the simplest comprised of a mixture of macrocycles of variable sizes based on identical building blocks (type A; Figure 1 a). More complex systems are composed of a mixture of different building blocks, thus forming receptors based on homo- and heteroaggregates (type B; Figure 1 b). These approaches have been successfully applied to obtain a variety of receptors to bind different classes of guests, such as Li+ and ammonium ions, and even a neurotransmitter.5–7 A dynamic combinatorial library (DCL) consisting of macrocycles of variable sizes based on identical building blocks (type A), and a DCL of similar-sized macrocycles based on different building blocks (type B). The elegance and power of the concept of amplification (and thus identification) of the best receptor was an issue of recent debate. Computer simulations8 and thermodynamic/statistical analysis, supported by experimental data,8a, 9 on dynamic combinatorial libraries (DCLs) of the types shown in Figure 1 indicated that certainly not in all situations does the best receptor need to be amplified by the addition of the target. In these complex systems, in which receptors are in dynamic equilibrium, the energetically most favorable situation of the entire system determines the outcome of an amplification experiment. In situations where the concentration of the receptor depends on the composition of the library, the best receptor is not likely to be amplified at the expense of a lower total concentration of receptor. This problem does not arise for conventional, kinetically stable receptors because the composition of the receptor remains the same during the entire experiment. The amplification of a receptor from a DCL can only happen because the receptors compete simultaneously for both the building blocks that they are built from and the target guest. As different receptors are commonly built from the same building blocks, an increase in the concentration of certain receptors causes a decrease in the concentration of others (Figure 1). Importantly, a shift in the equilibrium as a result of binding the guest leads in some scenarios to a decrease in the total concentration of relevant receptors. For example, amplification of the larger macrocycles of type A (An) leads to a decrease in the number of the smaller ones and, thus, reduces the total concentration of receptors available for guest binding. A comparable situation can occur for the type B receptors (Figure 1) if significant binding of the target guest requires the presence of a specific building block (e.g. building block A, as shown in entry 3 of Table 1). Amplification of the homoreceptors entirely built from this building block (A-A-A) occurs at the expense of the heteroreceptors that contain this building block (e.g. A-A-B, A-B-C, etc.) and thus leads to a decrease in the total concentration of receptors that make a significant contribution to the total free energy of the entire system. This phenomenon can easily lead to amplification of the weaker binding receptors at the cost of the better receptors, because the total free energy gained by guest binding by a few strong receptors can be less beneficial than a multitude of somewhat weaker interactions (see Table 1, entry 3). This effect is especially important when the relative free energies (ΔG0=−R T lnK) associated with binding of the target are of the same order of magnitude8–10 and when the concentration of the target is relatively high.9b, 10 Entry AAA AAB BBA BBB BBC BCC CCC AAC ACC ABC 1 Krel 1 1 1 1 1 1 1 1 1 1 crel [%] 3.7 11.1 11.1 3.7 11.1 11.1 3.7 11.1 11.1 22.2 ΔΔG0 [kcal mol−1] 0 0 0 0 0 0 0 0 0 0 2 Krel 1000 1 1 1 1 1 1 1 1 1 crel [%] 28.1 0.5 3.4 7.2 21.5 21.5 7.2 0.5 3.4 6.8 ΔΔG0 [kcal mol−1] −4.1 0 0 0 0 0 0 0 0 0 3 Krel 900 600 300 1 1 1 1 600 300 300 crel [%] 0.1 2.1 22.8 0.5 1.6 1.6 0.5 2.1 22.8 45.7 ΔΔG0 [kcal mol−1] −4.0 −3.8 −3.4 0 0 0 0 −3.8 −3.4 −3.4 To explain this phenomenon, the selection procedure in a DCC can be compared to the economic trade situation of retailers with respect to their customers. If sufficient customers (guests) are available, the most profitable situation is reached when as many shops (receptors) as possible are available to satisfy (bind) as many customers (guests) as possible, even though the profit per shop (binding energy per receptor) is not as high for all shops as that for the most profitable one. A turning point is reached at lower demands, and below a critical value the situation is reversed and a higher profit (binding energy) per shop (receptor) becomes most important. Therefore, the target-induced amplification in a DCL depends on the relative concentration of the guest. This behavior was first simulated by Severin and co-workers8 and was recently experimentally confirmed by Otto, Sanders, and co-workers, as well as by Saur and Severin.9 Indeed, at high concentrations of guest competition among receptors that are based on the same building blocks dominates the selection procedure, whereas at low concentrations of guest the receptor with the largest affinity is selected by the guest according to the best fit (most profit) principle. A recent theoretical analysis by Otto, Sanders, and co-workers10 revealed that also in very large DCLs, which consist of many potential receptors, moderately good hosts that are already present in relatively high concentrations in the absence of the guest are likely to be amplified further upon addition of an excess of the guest. On the basis of these findings, the groups of Sanders and Severin have reported a general strategy to select the best receptor from a DCL: 1) It is recommended to use conditions under which the building blocks (not the assemblies) are the dominant species in solution. 2) A DCL should be designed such that all receptors contain a common building block, which should be present in substoichiometric amounts with respect to the other subunits. 3) A relatively low concentration of the target should be used during the amplification experiment (a 1:10 ratio with respect to the total concentration of building blocks is used as a rule of thumb, taking the detection limitations into account). These results show that the field of DCL is more complex than previously anticipated and amplification of the best receptor is not always obvious. On the other hand, one might argue that problems occur only when the differences between the receptors are small. The value of the DCC approach has been recently demonstrated by some interesting examples that reveal receptors with exceptionally strong affinities for the target. A common feature of these examples is that the hosts have unexpected structures, and it is very unlikely that these receptors would have been discovered by more traditional design and synthesis strategies.11, 7 For example, the DCL depicted in Figure 2 is based on the reversible formation of a hydrazone from the hydrazide and a latent aldehyde (acetal functionality) of the peptide building block pPFm (1), which is a pseudopeptide based on proline and phenylalanine residues. The library consists of mostly cyclic oligomers 2–5 (2–4 dominate). Most remarkably, addition of the neurotransmitter acetylcholine (6) causes a gradual increase in catenane 7, which comprises two interlocked three-membered rings 3 and is not detected in the absence of the target. Acetylcholine binds to a single diastereoisomer of 7 in one specific conformation from the many possibilities, with an exceptionally high binding constant (1.4×107 M−1). The affinity of the trimer 3 and tetramer 4 for acetylcholine proved to be much weaker (1.5×103 and 5.7×103 M−1, respectively). These results are remarkable because, first, the amplification of a catenane is highly unexpected. From rational design one would never have predicted that two interlocked rings would form the most suitable binding motif for the neurotransmitter, as there are no other catenane structures known that strongly bind cations.12 Second, the DCC method proves useful as a synthetic method to prepare a catenane receptor that would have been most difficult to prepare without the template.13 Unexpected amplification of a catenane receptor from a DCL of cyclic peptides by the addition of the neurotransmitter acetylcholine.7 TFA=trifluoroacetic acid; DMSO=dimethyl sulfoxide. A second remarkable example was recently reported in which tetramethylammonium salts were used as target guests during the amplification process (Figure 3). The initial DCL consisted of macrocycles based on 8 and a smaller disulfide formed by reversible SS bond formation, and surprisingly the large macrocycle 11 was discovered as a good receptor. On the basis of size match–mismatch arguments, one would expect to bind much less strongly to the macrocyle meso-11 than to the macrocyles rac-9 and rac-10 (Figure 3). Detailed studies revealed that meso-11 is by far the best host in the mixture (reported binding constants of 4×106 M−1 and 8×102 M−1 for meso-11 and rac-9, respectively). This unexpected result, which is unlikely to have been discovered by traditional strategies, was explained by the induced-fit character of the host; meso-11 encapsulates the ion by folding completely around it (Figure 4).11 Unexpected amplification of a large, size-mismatched cyclic host from a DCL by the addition of .11 CPK models of the complex formed from meso-11 and in a) extended and b) folded conformations. The host binds the guest through an induced-fit mechanism. Reprinted with permission from Reference 11 Although “selection of the fittest” from a DCL does not necessarily always provide the “fittest”, the importance of dynamic combinatorial chemistry is clearer than ever before. DCLs are theoretically well understood now and it is clear that they behave as complex systems. Examples of amplification of unexpected receptors with high target affinities superbly demonstrate the proof of principle. The next challenge is to apply the DCC approach to existing problems that are hard to solve by traditional strategies. In addition, the application of DCC in new research areas should be explored. In this regard, supramolecular catalysis in which the catalytic activity is based on the complexation of the transition state seems a particularly promising area. An example of the successful application of DCC in catalysis has already been reported.6b A DCL that is similar to the one depicted in Figure 3 was used to select a supramolecular catalyst for a Diels–Alder reaction by the addition of a transition-state analogue (in this case the product). The host that was amplified indeed accelerated (although moderately) the envisioned Diels–Alder reaction. The relevance of DCC in, for example, transition-metal catalysis is still open to investigation. However, in general, unexpected results, new synthetic methods, and new applications are likely to result from future studies.