Introduction
Chemists, with their innate interest in molecular-level understanding, have often approached the study of biological systems by using models. Models, or synthetic analogs of natural molecules, have spanned a wide variety of constructs ranging from small-molecule biomimetics, through proteins and DNA to which unnatural functionality has been added. In the past, the need for a model system was obvious: one could not easily study biological systems in molecular-level detail because they were too complex for the methods available at the time. Many would argue that this is no longer the case—the ability to study biomolecules directly has advanced greatly. Nonetheless, models can still provide valuable insight that is difficult to tease out by studying the natural systems directly. Indeed, in studying a large, complex biological system, assumptions must be made to simplify the investigation. However, studies of models often demonstrate that such simplifying assumptions within the more complex biological system are not valid, a finding that can be difficult to obtain directly.
Models allow for a focus on one particular aspect of a biological structure, complex or pathway, which makes it possible to separate small contributions to stability, reactivity and mechanism. In addition, there are still many biological examples for which structural information is limited and for which models can thus provide an advantage in regard to the level of structural detail available. Studies in models can then lead researchers to ask new questions in the biological system, which may lead to new findings that would have been overlooked otherwise.
Moreover, the sophistication of model systems has advanced considerably, such that initial studies in small-molecule models can subsequently be incorporated into the biological molecule or assembly. This allows for the investigation of increased complexity and demonstrates the connectivity between the two. Indeed, in many recent examples, the model is the biological molecule itself rather than some small-molecule mimic, which makes it possible to insert the model directly into a larger biological context. This provides a method to probe the biological target more deeply than can be done otherwise.
Advances in biomimetic chemistry
The concept of building small models for biological systems is a time-honored one in chemistry. Some of the most important and influential early studies were focused on understanding the physical mechanisms of enzymatic catalysis. One important question being asked was what physicochemical strategies enzymes used to greatly enhance reaction rates. Small-molecule models were widely studied to investigate such strategies for catalysis. An important example of this kind of approach was the study of entropic factors and of the ways in which enzymes could pay entropic costs at transition states by sacrificing free energy of substrate binding. A widely read review by Page and Jencks1 gives many examples of small-molecule model studies addressing this issue. A related small-molecule model approach for studying enzymatic catalysis took isolated chemical features of protein enzymes, such as functional amino acid side chains, enzyme cofactors and binding moieties, and assembled them into small-molecule models2. The study of these bio-inspired molecules was termed 'biomimetic chemistry'. Commonly, their catalytic properties were assessed with small-molecule models for larger biological substrates as well (Fig. 1). Such work yielded useful insights into the contributions of general acid-base catalysis, covalent catalysis, catalysis by strain induction and cofactor catalysis in enzyme mechanisms.
Figure 1: Model system for selective oxidation by cytochrome P450.
The enzyme model contained an oxidizing metal and multiple binding functionalities (cyclodextrins, shown as conical toroids). The model for a steroid substrate contained terminal, hydrophobic moieties that could be bound by the cyclodextrins, bringing the C6 
-proton in position for selective hydroxylation (product shown in red) by the tetrameric cyclodextrin complex21.
Model systems remain important to the understanding of biology, but the types of molecules being studied and the questions being asked are often more complex than they once were. This change has certainly been influenced by advances in the field of biology, but it may owe the most to the increasing sophistication and number of tools available to aid these studies. Examples of biological methods that 'model builders' have adopted include procedures for making molecules (PCR, mutagenesis, molecular evolution) and for tracking them (fluorescent proteins, fluorescence confocal microscopy, flow cytometry). Also critical are methods for analysis and construction that incorporate both chemistry and biology, such as genetically encoded tags, expressed protein chemical ligation and non-natural amino acid mutagenesis. On the chemical side, new synthetic organic methods have made it much easier to prepare model molecules as single enantiomers than once was possible. Also important for model building and tracking is the development of new reactions that functionalize large biomolecules and of reactions that can function 'orthogonally' inside cells.
Not only do chemists have increasingly powerful tools that allow for the assembly and study of more complex molecules, but they are also asking more biologically sophisticated questions. This, of course, follows largely from the rapid gain in understanding of how cells function and from the increase in knowledge about the molecular bases of diseases. Thus, as well as continuing to ask the old questions of how specific amino acids or nucleotides function, for example, researchers are now synthesizing and modifying whole enzymes and studying the models directly in the macromolecule context. They are moving beyond studies of solitary biomolecules and are addressing how those interact with other cellular species and tracking how their models function directly inside cellular systems. Two specific examples illustrating how chemical model systems contribute useful information about biological systems are given below.
Two case studies of model systems in chemical biology
Cation-
interactions. An excellent example of the insight that model systems can provide to biology comes from studies of the cation-
interaction (Fig. 2a). Specifically, this provides an example of how non-biological models can lead to new questions being asked within biological systems. Cation- interactions are now well established as an important contributor to protein structure and function, but such interactions first caught the interest of the molecular recognition community as a result of host-guest recognition studies in aqueous solution carried out by Dougherty and others in the 1990s3, 4. These experiments demonstrated that cations, in particular tetraalkylammonium groups, bind preferentially to an aromatic host molecule rather than being solvated by water. Experimental systems provided affinities for binding in an aromatic pocket as well as information on their selectivity relative to hydrophobic guests. Experimental and theoretical studies indicated that the interaction consists of a unique charge-quadrupole interaction between the positive charge on the ammonium group and the partial negative charge on the face of the aromatic ring that, much like a hydrogen bond, provides a directional electrostatic component to recognition. The potential importance of this interaction in biological systems was quickly recognized5. Statistical analyses of protein structures indicated that cationic side chains (arginine and lysine) were found in close proximity to aromatic residues much more often than would be predicted statistically6. Studies of these interactions in model peptide systems supported the importance of such interactions in stabilizing secondary structure7, 8, 9 and also provided critical information about their enthalpic and entropic driving force and their role in providing selectivity10.
Figure 2: Evolution of model systems for proteins and DNA.
(a) Dougherty's host guest system for studying cation- interactions led to Waters' 
-hairpin peptide for studying N-methylation of cation- interactions in proteins. These model systems provided insight into the importance of the interaction in biological systems, here typified by the HP1 chromodomain bound to trimethyllysine from the H3A histone protein as part of the intact nucleosome. (b) A nonpolar shape mimic of thymidine (dichlorotoluene deoxyriboside; dL), used to study electrostatic and steric effects in DNA replication, is placed into an intact DNA helix (shown here is a DNA structure containing a related compound). This modified DNA is used to gain insight into the function of a DNA polymerase (shown in complex with DNA) in vitro and in E. coli, in which dL was found to be replicated accurately.
Wadsworth Center, New York State Department of Health
Full size image (95 KB)Cation- interactions have now been found to contribute to numerous biological processes, including protein folding and protein-protein interactions, the binding of small molecules to proteins, and ion channel function. For example, using in vivo non-natural amino acid mutagenesis, Dougherty, Lester and co-workers were able to demonstrate that a cation- interaction was critical for the binding of acetylcholine to the nicotinic acetylcholine receptor11. This is an excellent example of a model system that consists of a whole protein studied in vivo, which is possible owing to the advent of non-natural amino acid mutagenesis as a new tool for probing mechanisms within natural systems.
Most recently, the study of the 'histone code', which has a key role in controlling gene expression, has led to another example of the significance of cation- interactions in biology. Lysine methylation of histone proteins has been shown to turn on a protein-protein interaction with proteins containing a chromodomain. The key interaction is the binding of the trimethyllysine into an aromatic pocket, suggesting contribution from a cation- interaction12. Studies in model peptides have demonstrated the effect of lysine methylation not only on the magnitude of a single cation- interaction, but also on enthalpic and entropic driving force. This provides information about the selectivity of the interaction13 that would be difficult to determine from the protein-protein complex directly. As well as providing insight into this critical protein-protein interaction, these studies demonstrate the general fact that the breakdown of biomolecular recognition into hydrophobic and hydrophilic interactions, which is commonly used to simplify the study of complex systems, lacks the sophistication to accurately describe the features that provide selectivity in biological systems.
Steric and hydrogen bonding effects in DNA replication. Not only are modern model systems often more complex than older ones, but they can take on greater biological significance if there is a way to smoothly integrate data from pure synthetic molecules in vitro to studies carried out in living cells. A series of recent studies that make use of model systems for nucleic acids is illustrative of such an approach (Fig. 2b). This line of study started with the design of small-molecule models for DNA bases. It has long been recognized that electrostatic effects, including Watson-Crick hydrogen bonding, are important to nucleic acid structure. To evaluate the contributions of such electrostatic effects to the properties of DNA, isosteric molecules lacking polar functionality have been made. For example, 2,4-difluorotoluene was used as a model for the polar base thymine, and 4-methylbenzimidazole was used as a mimic of adenine14. Some of these analogs were close enough to nucleobases in size and shape that they effectively allowed the separation of electrostatic effects from steric effects.
The small-molecule models were studied structurally and their noncovalent interactions were measured. They were small enough that high-level ab initio calculations of structure and electronics could readily be performed. In addition, they were then incorporated into synthetic oligonucleotides, which subsequently served as chemical models for natural DNA. These studies used biophysical methods such as measurement of free energies of helix stabilities, and structural methods such as two-dimensional NMR structure determination. The data showed that electrostatic effects such as hydrogen bonding were important for the selectivity of base pairing in DNA. Moreover, they helped to uncover the central role of base stacking in the stability of the double helix and the importance of van der Waals and solvation effects in this stacked interaction15.
The synthetic models for natural DNA were then carried on to a level of greater complexity, where interactions between DNA and enzymes were studied. The class of enzymes chosen first was the DNA polymerases responsible for the replication of bacterial chromosomes. In this respect, the data obtained with pure synthetic DNAs alone served as an important baseline: if properties with the enzyme were similar to those seen with the DNA alone, then one might ascribe much of the properties to the DNA itself. However, if different properties were seen, then it was necessary to invoke the influence of the polymerase.
These studies were asking the simple, biologically important question, How do polymerases choose correctly among the four nucleotides in solution when a new base pair is made? The isosteric nucleobase mimics made it possible to separate electrostatic effects (solvation and hydrogen bonding) from steric effects, because steric differences were minimized. Initial studies of DNA polymerases made use of classical biochemical measurements: analysis of Michaelis-Menten kinetics of nucleotide insertion reactions in vitro. Surprisingly, the results indicated that some nucleotide mimics could behave quantitatively like natural ones in their biochemical replication properties: replication fidelity with difluorotoluene, for example, was essentially the same as with thymine. The enzyme DNA polymerase I selected adenine as a partner for both, with high efficiency and fidelity16.
These results were surprising because it had been widely believed that both hydrogen bonding and Watson-Crick geometry were important for replication fidelity. The use of isosteric models allowed it to be demonstrated that hydrogen bonds could be largely or completely removed while biochemical selectivity was still fully retained. This model study led to a new hypothesis: namely, that steric effects might be the main governors of biological replication fidelity.
Moving beyond the biochemical approach, the model was then inserted directly into an active biological system. The replication systems of bacterial cells are, of course, much more complex than a single enzyme, and even the DNA itself is much more complex in its structure and interactions. Thus, to assess biological relevance it is very useful to be able to extrapolate from the simple chemical model all the way into a living cell. To accomplish this, nonpolar nucleoside isosteres were inserted into DNA oligomers through automated solid-phase synthesis. These were then inserted into an M13 bacteriophage genome, and the resulting phage, containing a single non-natural DNA base at a single site, were used to infect living Escherichia coli. Results showed17 that both mimics were replicated in the cells; recovery and analysis of the daughter phage after replication showed that the expected complementary nucleobases were encoded. For example, phage with one difluorotoluene gave rise to new phage with thymine replacing it, showing that the model molecule encoded the faithful insertion of adenine opposite difluorotoluene, aided by the intact replication machinery of the cell. Thus the model system functioned correctly, giving biological activity with a non-biological molecule.
This biological confirmation of function without hydrogen bonding led to the design of new model molecules, whose shape and size were varied systematically to directly test the steric hypothesis of DNA replication. Like the original molecules, these were studied as small molecules, evaluated in pure oligonucleotides as a baseline, and measured with pure DNA polymerases in vitro. Results confirmed that small (sub-angstrom) changes in shape and size could yield effects of several kilocalories at the transition state in regard to the choice of DNA bases during replication18. Some of these size-varied molecules were also studied in living bacterial cells, where it was shown that the bacterial replication machinery is exquisitely sensitive even to size changes as small as fractions of a bond length. The experiments have led to new hypotheses about how polymerases regulate size and shape of their substrates and to new ideas about how DNA genomes evolve during replication.
This work started with a simple chemical model for a natural DNA base. As with earlier biomimetic approaches, the molecules were studied as chemically isolated systems for their structure and noncovalent interactions. However, taking the more recent approach of using model systems, these molecular mimics were carried from chemistry to biophysics to biochemistry and biology without changes to the model. Each stage acted as a baseline for studies of the larger, more complex and dynamic system above it.
This small-molecule model yielded information that could not have been obtained from studies of the natural biomolecules alone. The initial molecular design offered a way to alter electrostatics without substantially changing sterics, thus making it possible to evaluate the importance of electrostatics independently at the level of small molecules, then in DNA, then in an enzyme active site, and finally in the cell. Similarly, the subsequent studies used molecules that systematically changed sterics with little or no change in electrostatics, thus allowing size and shape effects to be evaluated independently. Studies of natural nucleotides alone could not have yielded the same kind of control. An additional bonus from such studies is that they can inspire the creation of new classes of non-natural substrates that can function in biochemical systems, such as the recent designs of new base pairs for replication and transcription19, 20.
Potential limitations of model systems
Of course one must also recognize the limitations in any model. The primary limitation of model systems in general is that they simplify the problem to be studied in some way and so may not be representative of the natural system. These simplifications can be manifest both in models that cannot completely recapitulate the complexity of the original system and in models that veer slightly from the original system to increase experimental tractability. Thus, when considering applications to specific biological systems, models are never perfect. For example, in some cases, the solvation environment of cation- interactions, as described in the first case study, may be different in the protein of interest than in the model system, which may result in different interaction energies. Similarly, some DNA isosteres (particularly adenine analogs) in the second model system are imperfect structural mimics, and thus have more than one variable that differs from that of the natural structure. Second, and related, when models are found to be inactive in a biochemical context, this can be difficult to interpret, because there can be multiple reasons for poor activity in a many-variable system. For this reason, models yield much more useful information if they retain biological activity (when substituted into biochemical or biological pathways) than they do if activity is abrogated.
Large is the new small
The current trend toward increasing complexity in research at the chemistry-biology interface will continue for scientists who study model systems. There is little doubt that there will be a growing number of chemical and physical tools available for monitoring structure, chemistry, interactions and dynamics in biological systems. It is critical to the success of a modern model system that its findings be assessable within biological context, either through direct incorporation or through comparative studies. Model systems themselves will continue to develop, with a growing trend toward taking chemical model systems from small molecules to biomacromolecules and from the test tube into living cells and organisms.
