REVIEWS Drug Discovery Today Volume 12, Numbers 13/14 July 2007 Protein-binding pockets are usually thought to be solvated. However, recent studies indicate that this may not necessarily be the case, leading to unexpected gains in ligand binding affinity. Water, water everywhere ��� except where it matters? Steve W. Homans Astbury Centre for Structural Molecular Biology, University of Leeds LS2 9JT, UK Biological processes depend on specific recognition between molecules with carefully tuned affinities. Because of the complexity of the problem, binding affinities cannot reliably be computed from molecular structures. Modern biophysical techniques can decompose the problem to determine the thermodynamic contributions from protein, cognate ligand and solvent. Such studies applied to a model protein with a hydrophobic binding pocket have resulted in some surprising findings. For example, binding is not driven by the favourable entropic loss of solvent water from the binding pocket, but rather by favourable dispersion interactions arising from suboptimal hydration of the protein-binding pocket. Under these circumstances, one can anticipate particularly dramatic gains in binding affinity using shape complementarity to optimise solute���solute dispersion interactions, since these will not be offset by opposing solute���solvent dispersion interactions. All biological processes depend critically on highly specific recog- nition between molecules with carefully tuned affinities. Despite the universal nature of these interactions, our understanding of their molecular basis is severely limited. For example, despite a number of successes, it is still extraordinarily difficult to exploit routinely high-resolution structural data for a given complex in order to design molecules that inhibit binding. In other words, it is not trivial to predict binding affinity from structure. Thus, with notable exceptions, ���structure-based��� ligand design has enjoyed limited success. In the face of the emergence or re-emergence of diseases such as age-related neurodegenerative disorders or the relentless progress of antibiotic resistant bacterial strains that may soon reach epidemic proportions, the ability to design novel ligands at will that inhibit biomolecular interactions remains one of the major challenges in contemporary science. Ourlimitedabilitytopredictaffinityfromstructureisdueinlarge part to the complexity of the problem, whereby competing thermo- dynamic processes all contribute to binding affinity. The standard free energy of binding, which determines the strength of the inter- action, not only is governed by structural terms (loosely, enthalpy) but also involves the dynamics (loosely, entropy) of the interacting partners (Box 1). Thus, the affinity of a protein for a given ligand depends not only on the spatial positions of the interacting atoms but also on their dynamics, that is, how their positions change with time.Tofurthercomplicatetheproblem,a fullunderstandingofthe binding process requires knowledge of not only the structure and dynamics of the protein and ligand but also solvent water, which can have a dramatic influence on binding (Box 2). The hydrophobic effect A typical case in point is the classical hydrophobic effect, whose driving force is widely accepted as arising from solvent reorganisa- tion [1]. Water molecules cannot form hydrogen bonds with non- polar solutes, and this results in a disruption of the favourable hydrogen-bonding network of bulk water. Water molecules that are in contact with the solute are bonded more strongly to their neighbours, which results in an ordering of water molecules around thesolute.Thenatureoftheresultingorderinghasbeendescribedin various ways over the years, such as ���clathrates���, ���icebergs��� and ���flickering clusters��� [1���3]. More recent theoretical analyses [4] sug- gestthatwatermoleculesmoveawayfromthesolutesurfacetoform an interface that bears a similarity to that between a vapour and a liquid,suchthatthehydrophobicsurfaceis���dewetted���.Thissolvent- ordering phenomenon is consistent with the data on the solvation thermodynamics of small hydrophobic molecules [5]. The standard entropy of hydration, that is, the change in the standard entropy Reviews INFORMATICS Corresponding author: Homans, S.W. (s.w.homans@leeds.ac.uk) 534 www.drugdiscoverytoday.com 1359-6446/06/$ - see front matter �� 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2007.05.004

when the molecule is transferred from the gas phase to water solution, is invariably negative. It therefore follows that the associa- tion of two hydrophobic molecules in aqueous solution with the burial of hydrophobic surface area will be characterised by a favour- able entropy of binding because of the expulsion (and increase in entropy) of solvent water molecules at the solvent interface. Entropy-driven binding has thus long been taken as a characteristic thermodynamic binding signature of hydrophobic associations. A second parameter that has been ascribed to hydrophobic association is a negative change in the heat capacity at constant pressure (DCp). Again, this is related to solvent organisation. Experimental data for the transfer of small hydrophobic molecules from non-aqueous to aqueous solution are usually accompanied by a positive DCp [6]. This can readily be understood in terms of the above model of the hydrophobic effect ��� the ordered solvent molecules surrounding the solute are able to ���soak up��� more thermal energy without a concomitant rise in temperature because they have a lower kinetic energy than bulk solvent. Conversely, the loss of ordered solvent water molecules from a binding inter- face will result in a negative DCp for the binding process. Enthalpy-driven hydrophobic effects? Despite these apparently logical and consistent arguments, there exist discrepancies. Over two decades ago, Ross and Subramanian noted that, of the relatively limited thermodynamic data for ���hydrophobic interactions��� that were available at the time, a sub- stantial number were characterised by a thermodynamic binding signature that is enthalpy driven [7]. In order to investigate these discrepancies further, a programme of work was undertaken in our own laboratory, where we endeavoured to decompose the thermo- dynamics of binding of a ligand���protein interaction into contri- butions from the ligand, protein and solvent. We reasoned that, because of the obvious complexity of the problem, it would be appropriate to work with a structurally well-characterised model system. Moreover, it was apparent that it would be easier to try to delineate the thermodynamic differences in binding between a series of distinct, but structurally related, ligands for a given protein, than to attempt to dissect the interaction process for a single ligand de novo. With these constraints in mind, the mouse major urinary protein (MUP) was selected for further study. MUP is an abundant protein found in male mouse urine that binds pheromones, where subtle recognition of a series of related com- pounds underlies its biological function. The crystal structure of MUP-I isolated from urine was solved by Bocskei �� and co-workers [8]. The protein has a typical lipocalin fold that consists of an eight-stranded b-barrel and a single a-helix, and the interior of the barrel forms a hydrophobic cavity. A number of small hydropho- bic molecules can bind within the cavity, and the protein is thus an ideal model system with which to study ���hydrophobic inter- actions��� of a series of related ligands. The chosen ligands were 2-methoxy-3-isobutylpyrazine (a natural pheromone) and various derivatives such as 2-methoxy-3-isopropylpyrazine (Figure 1). At the outset, we had no knowledge of the thermodynamics of binding of these ligands to MUP. Isothermal titration calorimetry Drug Discovery Today Volume 12, Numbers 13/14 July 2007 REVIEWS BOX 1 The essentials of binding thermodynamics for a ligand���protein association. The interaction of a ligand with a protein can be written in the form of a standard chemical equilibrium: P �� L , PL (1) The association constant Ka (or equivalently, the reciprocal of the dissociation constant 1/Kd) for this reaction is related to the standard free energy of binding as follows: DGb �� RT ln Ka (2) Note that the standard free energy of binding DGb must not be confused with the free energy of binding DGb, which by definition is zero at equilibrium. The standard free energy of binding is in turn comprised of the standard enthalpy of binding DHb and the standard entropy of binding DSb (multiplied by the absolute temperature T): DGb �� DHb TDSb (3) The enthalpy can be loosely defined as the static, structural component of the association, whereas the entropy is related to the dynamics of the interacting partners. BOX 2 The formulation of the interaction in terms of Equation 1 (Box 1) is in fact a gross oversimplification of the binding process, since it ignores solvent water. A complete formalism involves a thermodynamic cycle as shown, where we account for the fact that the ligand and protein are associated with H2O (Scheme 1). Here, DGb is the observed standard free energy of binding, DGi is the ���intrinsic��� solute���solute interaction in the absence of solvent, and the terms DGsb and DGsu are related to the solvation of the complex and the free species, respectively. Since G is a state function, the sum over the cycle is zero, and we can thus write: DGb �� DGi �� DGsb DGsu (4) We can equate DGsb with the solvation free energy of the complex DGsolvPL, whereas DGsu comprises the sum of the solvation free energies of the free protein and free ligand, DGsolvP �� DGsolvL. Thus, finally we can write: DGb �� DGi �� DGsolvPL DGsolvP DGsolvL (5) A similar equation can be written for the enthalpy and entropy of binding since these are also state functions. The thermodynamic decomposition approach involves the determination of each component on the right-hand side of Equation 5. SCHEME 1 Thermodynamic cycle for a ligand���protein interaction in solvent water. www.drugdiscoverytoday.com 535 Reviews INFORMATICS