Water Networks Contribute to Enthalpy/Entropy Compensation in Protein–Ligand Binding

The mechanism (or mechanisms) of enthalpy-entropy ( H/S ) compensation in protein-ligand binding remains controversial, and there are still no predictive models (theoretical or experimental) in which hypotheses of ligand binding can be readily tested. Here we describe a particularly well-defined system of protein and ligands—human carbonic anhydrase (HCA) and a series of benzothiazole sulfonamide ligands with different patterns of fluorination—that we use to define enthalpy/entropy ( H/S ) compensation in this system thermodynamically and structurally. The binding affinities of these ligands (with the exception of one ligand, in which the deviation is understood) to HCA are, despite differences in fluorination pattern, indistinguishable; they nonetheless reflect significant and compensating changes in enthalpy and entropy of binding. Analysis reveals that differences in the structure and thermodynamic properties of water surrounding the bound ligands are an important contributor to the observed H/S compensation. These results support the hypothesis that the molecules of water filling the active site of a protein, and surrounding the ligand, are as important as the contact interactions between the protein and the ligand for biomolecular recognition, and determining the thermodynamics of binding.


Introduction
The hydrophobic effect-the tendency of nonpolar molecules or parts of molecules to aggregate in aqueous media-is central to biomolecular recognition.It now seems that there is no single "hydrophobic effect" [1][2][3][4] that adequately describes the partitioning of a small apolar ligand between both i) an aqueous phase and a non-polar organic phase (e.g., buffer and octanol), and ii) bulk aqueous buffer and the active site of a protein (i.e., biomolecular recognition).3][4][5][6][7][8][9][10] Clarifying the role of water in the hydrophobic effect in protein-ligand binding would be an important contribution to understanding the fundamental, mechanistic basis of molecular recognition.Resolving this mechanism would, however, still leave a (presumably) related phenomena unresolved: so-called, enthalpy-entropy compensation (H/S compensation).
H/S compensation is often encountered in the putative design of tight-binding, lowmolecular-weight ligands for a protein. 11,12Changes in the structure of the ligand often lead to opposite and compensating changes in the enthalpy and entropy of binding, but result in surprisingly small changes in the free energy of binding.][20] Human carbonic anhydrase II (HCA, EC 4.2.1.1)is an excellent model system with which to study H/S compensation because it allows us to answer the question: "Do changes in the structure of the networks of hydrogen bonded waters -that result upon ligand bindingdetermine H/S compensation?".Thiazole-based sulfonamide ligands bind to HCA with an enthalpy-dominated hydrophobic effect, 2,4 and are not compatible with an entropy-dominated hydrophobic effect proposed by Kauzmann and Tanford. 21,224][15] HCA is conformationally rigid and undergoes minimal (< 1 Å) conformational changes upon binding of most arylsulfonamide ligands 23 -and, more importantly for this study thiazole-based sulfonamide ligands 2,4 -and allows us to focus solely on rearrangements of solvent within the active site of the protein, and not on contributions caused by conformational changes in the protein.
In order to reduce artifacts in our ITC measurements that could lead to perceived H/S compensation we: i) measured the binding of a standard sulfonamide (methazolamide), whose concentration was determined accurately with NMR standards, to HCA and obtained the concentration of active protein; ii) prepared stock solutions of each ligand, and used these stocks solutions for each experiment to eliminate changes in the concentration of the ligand between experiments; iii) compared the binding stoichiometry of each ligand with the methazolamide standard to obtain an accurate concentration of each ligand; iv) accounted for the uncertainties associated with the nonlinear fits used to analyze the thermograms (∆H o bind less than ~10%), and compared the average value of (n = 7 runs) each ligand with a Student's t-test with a 95% confidence interval.
In this paper we wished to determine if selectively replacing the hydrogen atoms of the benzothiazole moiety with fluorine atoms would change the network of waters in the active site of HCA, and result in an H/S compensation similar to that observed between H 4 BTA and

Results and Discussion
Fluorination of the benzothiazole ligand changes its electronic structure, but leaves the size of the ligand and its binding geometry relatively unchanged.Replacing all of the C -H bonds of the benzo-extension of H 4 BTA with C -F bonds results in a decrease of the average dipole moment, but does not result in large changes in: i) the pKa of the ligand (a decrease of 0.3 units); ii) the solvent accessible surface area of the ligand in crystal structures of the HCA: ligand complex (an increase of 34 Å 2 ); iii) the binding geometry of the ligands (a shift of 0.7 Å in the binding pocket toward the hydrophobic wall). 2 We compared the binding geometry of six of the partially fluorinated ligands (4-F 1 BTA, 7- provide information about some of the so-called "fixed" water molecules, but these techniques cannot capture the location of the majority of the waters in the active site (before or after the ligand binds). 3,26,27In addition, a water molecule that diffracts simply implies that it is spatially immobilized but does not provide information about the energetics of the water.
To understand the thermodynamics of the network of waters in the active site of each HCAligand complex better we performed WaterMap calculations (see Figure 4), 9,28,29 which are explicit solvent calculations that determine a free energy, enthalpy, and entropy value for each molecule of water in the active site of the protein-ligand complex 2,4 using an approach called inhomogeneous solvation theory. 30,31These calculations support the H/S compensation we observe in the ITC measurements of HCA-ligand binding, and show: i) 4-F 1 BTA and 4,6-F 2 BTA have the most favorable entropy of binding, and the least favorable enthalpy of binding.
These effects result from subtle disruptions in the network of waters in the solvent-exposed region of the binding site, as seen in Figure 5 (the 4-F 1 BTA results look highly similar to 4,6-F 2 BTA and are therefore omitted for clarity).The fluorination pattern of these ligands apparently disrupts the water network in a way that results in less restricted water motion and therefore better entropy.ii) WaterMap calculates 5,6-F 2 BTA to have the least favorable entropy of binding, a result that is also consistent with the experiment.The 5-and 6-positions of the benzoextension are directed toward bulk solvent, and the difluoro substitution is predicted to restrict the mobility of a local region of waters around the ligand.

Conclusions and interpretations
The results obtained from the thermodynamic analysis, X-ray crystal structures, and surrounding the ligand and filling the active site of a protein are as important as the structure of the ligand and the surface of the active site.

F 4 BTA.
Selective replacement of the C-H bonds of the benzo-group with C-F bonds allow us to study the binding of a set of ligands that are similar in size, but have entirely different group dipole moments and therefore different interactions with the networks of water in the active site.

F 1 BTA, 5 , 6 -F 2
F 2 BTA, 6,7-F 2 BTA, 4,6-F 2 BTA, and 4,7-F 2 BTA) to HCA with previously solved structures of HCA complexed with H 4 BTA4 and F 4 BTA2 .Each of the high-resolution crystal structures (Figure2, resolutions ranging from 1.25-1.50Å) show that a fluorine atom in the 4position causes the ligands to: i) bind to HCA with the same geometry as F 4 BTA; ii) rotate 180 o around the molecular axis of the sulfonamide-benzothiazole bond when there is not a fluorine in the 7-position.We attribute these changes in the position of the partially fluorinated ligands to a repulsive interaction between the fluorine atom of the ligand and Thr 200 (Figure1C).Surprisingly, the position of the amino acids lining the active site of HCA are not affected by the position or orientation of the ligand: the average root-mean square deviation (RMSD) of the heavy atoms of the amino acids lining the active site of all of the partially-fluorinated ligand with H 4 BTA and F 4 BTA is 0.132 Å and 0.112 Å, respectively.These crystal structures suggest that the interactions between the benzothiazole ligands and the binding pocket of HCA are mediated by the molecules of water in the active site, and not through the traditional lock-and-key model of direct interactions between protein and ligand.The binding affinity of benzothiazole is relatively unaffected by fluorination, and in certain cases is the result of compensating values of enthalpy and entropy of binding.We measured the enthalpies of binding (∆H o bind ) and the association constants (K a ) for the series of the partially fluorinated ligands in Figure 1A with isothermal titration calorimetry (ITC), and estimated the free energies (∆G o bind ) and entropies (−T∆S o bind ) of binding.Figure 3 plots the pK a -corrected values of ∆J o bind for each ligand (where ∆J = ∆G, ∆H, or -T∆S); these values represent the binding of the sulfonamide anion to HCA (details in the SI). 4,25We classified the ligands into three categories: i) ligands in which ∆H o bind and -T∆S o bind are unchanged, and result in an unchanged binding affinity (7-F 1 BTA, 5,6-F 2 BTA, 4,7-F 2 BTA, and 5,6,7-F 3 BTA); ii) ligands in which -T∆S o bind is significantly different, and results in an increase in binding affinity (6,7-BTA); iii) ligands in which ∆H o bind and -T∆S o bind are significantly different, but compensate and result in an unchanged binding affinity (4-F 1 BTA and 4,6-F 2 BTA).A change is considered significant if the values of ∆J o bind are statistically distinguishable from H 4 BTA at p <0.05 (by Student's t-test).Molecular dynamics simulations support the role of water in H/S compensation.Our previous studies of HCA-arylsulfonamide ligand complexes support the hypothesis that the network of waters within the active site of HCA is an integral component in enthalpically driven hydrophobic effects.Structural and thermodynamic data of the partially fluorinated ligands binding to HCA suggest that the network of hydrogen-bonded water molecules in the active site of a protein-ligand complex could be responsible for changes in the ∆H o bind and -T∆S o bind , but there are currently no experiments to probe the thermodynamic characteristics of individual water molecules directly in the active site of a protein.X-ray and neutron-diffraction data

Figure 5 shows an additional hydration site localized above the 5 , 6 -F 2 6 -F 2 6 -F 2 BTA 6 -F 2 BTA
BTA ligand (denoted by a black arrow).WaterMap predicts an additional region of entropically unfavorable water to the side of the 5,6-F 2 BTA ligand (denoted by the dashed oval), further contributing to the unfavorable entropy of binding.iii) 6,7-F 2 BTA has a value for the entropy of binding between the values of 4,6-F 2 BTA and 5,6-F 2 BTA.As seen in Figure5, the region to the right of the ligand (see dashed oval) is similarly unfavorable to 5,BTA, but the region extending toward bulk solvent does not have the additional localized hydration site observed in 5,6-F 2 BTA (see hydration site indicated by the black arrow in 5,that is missing from 6,7-F 2 BTA)The improved free energy of binding of 6,7-F 2 BTA (compared to H 4 BTA) arises from desolvation of the ligand upon binding, and most likely is not from a rearrangement of the solvent in the active site of HCA.6,7-F 2 BTA is the only ligand that binds to HCA with a higher binding affinity than H 4 BTA; this difference in free energy of binding (∆G o bind, x-FyBTA −∆G o bind,H4BTA = ∆∆G o bind, Fluorination ) is 1.1 ± 0.3 kcal mol -1 more favorable (FigureS1).The increased binding affinity of 6,7-F 2 BTA over H 4 BTA cannot be attributed to differences in buried hydrophobic area upon ligand binding (FigureS1), but arises from a more favorable entropy of binding.There are two plausible explanations for this increased binding affinity: i) the reorganization of the waters in the active site of HCA upon binding of the ligand are responsible for the more favorable -T∆S o bind of 6,7-F 2 BTA; or ii) the desolvation of 6,7-F 2 BTA, in addition to the desolvation of the active site of HCA, influences the ∆G o bind .Entropy-driven binding is compatible with the mechanism of the hydrophobic effect proposed by Kauzmann and Tanford, but is the only thiazole-based ligand studied thus far that is not enthalpy-dominated.The contribution of the enthalpy (∆H o OW ) and entropy (-T∆S o OW ) of partitioning of H 4 BTA and 6,7-F 2 BTA to the free energy of partitioning (∆G o OW ) mirrors the contributions of ∆H o bind and -T∆S o bind in the binding of these ligands to HCA (Figure S1).WaterMap predicts 6,7-F 2 BTA to be roughly equientropic with H 4 BTA.WaterMap only accounts for first-order water correlation terms and only considers regions of high water density for the thermodynamic calculations and not for desolvation of the ligand.The most plausible explanation for this mirroring of trends of partitioning and binding is therefore that the desolvation of the ligand, and not just the desolvation of the active site of HCA, influences the ∆G o bind .The HS compensation observed in the binding of 4-F 1 BTA and 4,6-F 2 BTA arise from a reorganization of the solvent.The values of ∆H o bind and -T∆S o bind of 4-F 1 BTA and 4,6-F 2 BTA are significantly different than H 4 BTA (> 4 kcal mol -1 ), but compensate and result in unchanged binding affinities.Interestingly, these ligands are the only two that rotate within the active site (Figure 2), possibly to reduce the unfavorable interaction of the fluorine at the 4-position with the backbone carbonyl of Thr 200 if the pose did not undergo rotation.The conserved binding of these ligands to HCA, and the conserved structure of the side chains of the amino acids of HCA in the active site suggest that neither changes in structure nor interaction of the protein and ligand are plausible candidates for significant changes in the thermodynamics binding.The most plausible candidate for the source of these compensated changes in ∆H o bind and −T∆S o bind is therefore the networks of hydrogen bonds of waters in the active site, and surrounding the ligands in solution and in the protein-ligand complex.WaterMap predicts the same trends for the observed thermodynamic of binding; the changed networks of water of 4-F 1 BTA and 4,in the active site of HCA clearly indicate the importance of water in H/S compensation (Figure 5).

Figure 1 :
Figure 1: (a) Structures of the partially fluorinated ligands used in this study, and their

Figure 4 .
Figure 4. Diagram comparing of the pK a -corrected thermodynamics of ΔJ o bind results from ITC measurements with WaterMap calculations for ∆G o bind (a), ∆H bind (b), and -T∆S o bind (v).

4 BTA) and perfluorobenzothiazole sulfonamide (F 4 BTA) to HCA are indistinguishable, 2 and
dynamics simulations described in this work show that a series of ligands with different electronic structure bind to HCA with very similar values of ∆G o bind , but with very different (and compensating) values of ∆H o bind and −T∆S o bind .These results suggest that the "size" (a term we cannot presently disaggregate into surface area, molecular volume, or dipole moment) of the ligand, and thus the water that is displaced from, or perturbed in the active site of HCA, is primarily responsible for the ∆G o bind ; it also implies that changes in the structure of the networks of hydrogen bonded waters, that result upon ligand binding, determine the values of bind .This water-centric view of ligand binding-and H/S-compensationcannot be rationalized by the lock-and-key principle, and suggest that the molecules of water