The Binding of Benzoarylsulfonamide Ligands to Human Carbonic Anhydrase is Insensitive to Formal Fluorination of the Ligand

suggests that binding depends on a fine balance of interactions between HCA, the ligand, and molecules of water filling the pocket and surrounding the ligand, and that a simple analysis of interactions between the protein and ligand (Figure 1E), or of the physiochemical properties of the ligands, is insufficient to understand (or more importantly, predict) the energetics of binding. Our previous study of H 4 BT and H 8 BT showed that changes in the shape of the ligand also resulted in indistinguishable values of ∆ G obind . [3]

The hydrophobic effect (or the aggregated effects that we call "the hydrophobic effect") underlies the binding of many ligands to proteins. It involves three molecular participants: the surface of the binding pocket of the protein, the surface of the ligand, and the networks of water molecules that fill the pocket and surround the ligand. The molecular-level mechanism of the hydrophobic effect in protein-ligand binding remains a subject of substantial controversy. [1,2] There are three primary questions of interest: i) "Do hydrophobic effects reflect the release of structured, and hence entropically unfavorable, water from hydrophobic surfaces when the ligand and surface of the binding pocket come into contact?"; ii) "Do hydrophobic effects represent the release of free-energetically unfavorable water from hydrogen-bonded networks in the binding pocket or displacement by the ligand, and the release of free-energetically unfavorable (although perhaps different) waters from the hydrophobic surface of the ligand?"; and iii) "How important in free energy are the contact interactions between the protein and the ligand in the binding pocket?".
In a previous examination of these questions, [3] we compared the binding of a series of heteroarylsulfonamide ligands, and their "benzo-extended" analogs (Scheme 1), to human carbonic anhydrase II (HCA; EC 4.2.1.1). In these binding events, the addition of a benzo group: i) increased the hydrophobic surface area (and the volume) of the ligand; ii) generated two new van der Waals contacts between the ligand and hydrophobic wall of HCA; but iii) did not result in a significant increase in the area of contact between the hydrophobic surfaces of the protein and ligand. The free energy of binding of the arylsulfonamide ligands increased with the additional surface area buried upon binding from the benzo-extension by −20 cal mol 1 Å 2 , [3] an amount expected for normal hydrophobic effects (−20 to -33 cal mol 1 Å 2 ) [4] . The heat capacity of binding (∆C p o ) of the benzo-extended ligands was more negative than the corresponding Scheme 1. Arylsulfonamide ligands. Hydrophobic surface area is added to the heterocyclic ligands by: a "benzo-extension", denoted with a H 4 ; a "fluorobenzo-extension", denoted with a F 4 ; or a "tetrahydrobenzo-extension", denoted with a H 8  arylsulfonamide [3] -a change commonly considered to be a sign of a hydrophobic interaction. [5] We drew two conclusions pertinent to protein-ligand interactions from the previous study: [3] i) the balance of enthalpy and entropy responsible for the differences in the partitioning of a ligand, and its benzo-extended analog, between octanol and buffer is not the same as that responsible for differences in the binding of these ligands to HCA; and ii) the increased binding affinity of the benzo-extended ligands to HCA results from an increased favorability in the enthalpy of binding, and not from an increased entropy of binding. The enthalpy-driven binding observed between these ligands and HCA is not compatible with the mechanism of the hydrophobic effect proposed by Kauzmann and Tanford (KT), [4,6] but is similar to those observed in other protein-ligand systems. [7][8][9] We-along with Berne, [10] Chandler, [11] Friesner, [12,13] Klebe, [14] Ladbury, [15,16] Lemieux, [17] Rossky, [18,19] and Toone [20,21] -attribute this type of enthalpy-dominated hydrophobic effect to the release of water from the binding pocket upon binding of the ligand. Similar thermodynamic signatures characterize binding of the benzo and the tetrahydrobenzo derivative of T (see Scheme 1). [3] The objective of this work was to determine if replacing the four C -H bonds of the benzo moiety with four C -F bonds (i.e., "fluorobenzo-extension") would change the hydrophobic interactions of these ligands with HCA. Fluorocarbons are commonly believed to be "more hydrophobic" than homologous hydrocarbons, [22,23] but typical measures of hydrophobicitywhen corrected for differences in surface area-are very similar, if not indistinguishable. [9,23,24] We measured the partitioning of the benzo-and fluorobenzo-extended ligands between buffer and octanol, and found the surface area-corrected hydrophobicity of the ligands increases (by ~ 1.1 cal mol 1 Å 2 ) upon fluorination (summarized in the Supporting Information (SI)).

Benzo-and fluorobenzo-extended ligands bind to HCA with similar geometry. Crystal
structures of HCA complexed with F 4 BTA, H 4 BTA, and H 8 BTA (Figure 1) show that the geometry of binding of these ligands is similar in orientation despite their differences in shape, volume, and surface. The geometries of binding of F 4 BT, H 4 BT, and H 8 BT are also conserved (see SI).
Careful inspection of the crystal structures of H 4 BTA and F 4 BTA reveals that fluorination of the ligand shifts its position within the binding pocket by 0.7 Å ( Figure 1D) while the positions of the side chains of the amino acids lining the binding pocket of HCA do not change. We attribute this shift of F 4 BTA to an increased number of unfavorable interactions between the ligand and the binding pocket ( Figure 1E); in particular, the Coulombic repulsion between the fluorine atom on the ligand and the carbonyl of Thr 200, [25] a 3.0 Å distance.

The atomic composition of the benzo-extension does not affect binding affinity.
We measured the enthalpies of binding (∆H o bind ) and the association constants (K a ) for the series of ligands in Scheme 1 using isothermal titration calorimetry (ITC), and estimated the free energies (∆G o bind ) and entropies (−T∆S o bind ) of binding. To account for differences in the pK a of each ligand, we corrected the measured thermodynamic parameters to represent the binding of the sulfonamide anion to HCA (details in the SI). [26] Remarkably, values of ∆G o bind of the benzo-and fluorobenzo-extended compounds are indistinguishable at a 90% confidence level ( Figure 2A). Values of ∆G o bind , combined with the overall conserved binding geometry of each set of benzo-and fluorobenzo-extended ligands suggests that binding depends on a fine balance of interactions between HCA, the ligand, and molecules of water filling the pocket and surrounding the ligand, and that a simple analysis of interactions between the protein and ligand ( Figure 1E), or of the physiochemical properties of the ligands, is insufficient to understand (or more importantly, predict) the energetics of binding. Our previous study of H 4 BT and H 8 BT showed that changes in the shape of the ligand also resulted in indistinguishable values of ∆G o bind . [3] The increased binding affinity of TA (or T) upon benzo-and fluorobenzo-extension is an enthalpy-dominated hydrophobic interaction, and not the "classical hydrophobic" effect described by KT. We showed previously that the interactions between the benzo-extended ligands and HCA do not result from a "non-classical hydrophobic effect", [27] because the binding of H 8 BT is also enthalpy-dominated. [3] The partitioning of H 4 BTA and F 4 BTA from buffer into octanol ( Figure 2B) is, however, an entropy-dominated hydrophobic effect, and in agreement with the KT model.  are corrected for the ionization of the sulfonamide group in the buffer phase. [3] composition of the benzo group, and almost indistinguishable for three types of functional groups that have very different molecular properties.
While the ∆G o bind is unchanged upon fluorination, we observe significant and compensating changes in ∆H o bind and -T∆S o bind (Figure 2A). To elucidate potential sources for the differences in the ∆H o bind and -T∆S o bind we calculated the molecular mechanics implicit solvent binding energy (with Prime MM-GBSA calculations [28] ) of H 4 BTA and F 4 BTA, and decomposed these values protein-ligand binding. [29] We assume the role of the conformational entropy in the HCA-ligand complexes is minimal because of the rigidity of the binding pocket of HCA and the conserved binding geometry of each ligand; this rigidity stems from the fact that there is only a single, rotatable bond (the Ar-SO 2 NHbond) in the molecule.
The Prime MM-GBSA calculations predict a more favorable total free energy of binding of HCA with F 4 BTA over H 4 BTA (by < 3.0 kcal mol -1 ). The zero-temperature model used in MM-GBSA calculations, however, tend to overestimate the predicted magnitude of G o bind (by ~3fold [30]  crystallographically resolvable) waters in the binding pocket of HCA from six to ten. We summarize the number of fixed molecules of water for the benzo-and fluorobenzo-extended BTA and BMP ligands in Table 1. This result mirrors our previous study, [3] which showed that negative, and is compatible with our hypothesis that the interaction between the ligand and the protein is hydrophobic in nature. [5] The difference in the heat capacity of  non-polar surface area of the two ligands (−19 cal mol −1 K −1 ). [5] We attribute this discrepancy between the measured and predicted values of ∆C o p to the additional "fixed" waters observed in the crystal of F 4 BTA bound to HCA ( Figure 1D). The value estimated by Connelly for the ordering of a single water (−9 cal mol −1 K −1 ) [31] suggests that three additional molecules of water are fixed in the binding pocket of HCA when the hydrogen atoms of H 4 BTA are replaced with fluorine atoms, and is consistent with the four additional waters observed in the crystal structure.

Increases in binding affinity of ligands correlates with the number of waters released from the binding pocket of HCA, and not with the atomic composition or structure of the ligand.
The calorimetry and X-ray crystallography data for the binding of benzo-and fluorobenzoextended arylsulfonamide ligands to HCA reinforce our previous conclusion: [3]

Expression and Purification of Human Carbonic Anhydrase
We chose human carbonic anhydrase II (HCA, E.C. 4.2.1.1) as a model protein to study the hydrophobic effect in protein-ligand binding, because HCA: i) is well-characterized structurally, [1] and has a binding pocket comprised of a distinct "hydrophobic wall" and a distinct "hydrophilic wall"; [1] ii) is structurally rigid, and retains its secondary and tertiary structure upon binding of a ligand; [1,2] iii) can be expressed in E. coli in the quantities necessary for calorimetry experiments and X-ray crystallography; iv) crystallizes readily, and the conditions for growing crystals, reproducibly, are known. [3] We expressed HCA in BL21(DE3)pLysS competent cells (Promega, Madison, WI), transformed with a pACA plasmid containing the HCA gene, [4] [5] according to the procedures published by Fierke and coworkers. [5] After expression, the cultures of E. coli were pelleted (20 min, 10,000 x g, 4 o C), flash frozen under liquid nitrogen, and stored at -80 o C until needed.
We lysed the E. coli with BPER protein extraction buffer (Thermo Scientific) augmented

General Synthetic Procedure for Fluorobenzo-extended Ligands.
If not stated otherwise, the fluorobenzo-extended ligands were synthesized following a general synthetic procedure (GP) developed by Chern et al. [6] The starting material (typically 10 mmol, 1 equiv.) was dissolved in dry tetrahydrofuran (THF) under an argon atmosphere, cooled to 78° C, and reacted with a 1.6 and allowed to warm to ca. 0° C before the intermediate aryl lithium species was quenched with dry gaseous sulfur dioxide. Sulfur dioxide was blown over the surface of the reaction solution for approximately five minutes to exchange the atmosphere in the flask, and the flask re-sealed with a balloon filled with gaseous sulfur dioxide. The reaction mixture was stirred for an additional hour, and the reaction mixture was allowed to warm to room temperature. Argon was then bubbled through the reaction suspension to remove the excess of sulfur dioxide, and the resulting slurry was concentrated in vacuo to yield the lithium benzoaryl-2-sulfinate as a pasty solid.
The crude lithium sulfinates were re-dissolved in aqueous sodium acetate, hydroxylamine-Osulfonic acid (2.5 equiv.) was added at 0° C, and the mixture stirred overnight at room temperature. The precipitate was isolated by filtration, washed with cold water, and the fluorobenzo-extended ligands recrystallized.

Physiochemical Characterization of the Heteroarylsulfonamide Ligands
We summarize the physico-chemical data collected for each sulfonamide ligand in Table S1, and discuss the procedures in detail below.
Determination of pK a . The pK a of T, TA, H 4 BT, H 8 BT, and H 4 BTA ( Figure S1) were determined previously. [2] We used the same procedure to determine the pK a of the fluorobenzo- Partitioning experiments. We measured the equilibrium constant of partitioning, between sodium phosphate buffer (10 mM, pH = 7.6) and octanol, for each of the benzo-and fluorobenzoextended ligands using the shake-flask method described previously. [2] Solution calorimetry. We followed the procedure reported previously, [2] and measured the heat of dissolution for solid samples of each ligand (5-10 mg) with a TAMIII calorimeter (TA Instruments). n.d. 6.9 (6.9) 6.9 --------a) NMR experiments were performed in DMSO-d 6 at 25° C. b) Calculated via logP = logD + log(1 + 10 pKa-pH ). c) Determined previously by Snyder et al. [2] d) Values estimated according to the literature. [1] Figure S1. NMR-based approximation of values of ΔH°i on for the fluorobenzo-extended ligands.
The solid black circles represent an independently measured value of ΔH°i on for the fluorobenzoextended ligands. The hollow circles represent an independently measured value of ΔH°i on from compounds for which ΔH°i on had been independently determined, and reported [2] : Iimidazole- pK a -correction of observed ITC data: We corrected the K a of each arylsulfonamide ligandto reflect the binding of the arylsulfonamide anion (ArSO 2 NH -) to the zinc-bound water form of HCA (HCA-Zn II -OH 2 + )-using a previously reported method, [15] which is explained in detail by Snyder et al. [2] Figure S2 shows ΔG°b ind , ΔH°b ind , and TΔS°b ind as function of SASA for each ligand.

Decomposition of the Free Energy of Binding Calculations
We calculated the binding energies of H 4 BTA and F 4 BTA using the MM-GBSA method [16] (as implemented in Prime [17,18] ). We prepared the initial crystal structure coordinates for the complex and then appropriately summed to calculate each contribution to the total binding energy. The anionic charge on the ligand was computed using a fit to the electrostatic potential derived from quantum mechanics at the RHF/6-31G* level of theory, [19] which has been shown to perform well with continuum solvation models. Energy terms are shown in Table S4.

Biostructural analyses
Protein Crystallization: Monoclinic crystals of HCA were prepared with the hanging drop diffusion method published by McKenna and coworkers, [3] and the crystals of HCA were left undisturbed (at 4 o C) until needed for soaking experiments.

Ligand Soaking Experiments:
We soaked the crystals of HCA in saturated solutions of the benzo-and fluorobenzo-extended ligands using the procedure described previously. [2]  Laboratory. [20] Reflections were indexed and integrated using HKL2000, and scaled using SCALEPACK. [21] Solution of Crystal Structures: Diffraction data were analyzed using the CCP4i suite of crystallography software [22] using previously published procedures. Table S3 summarizes the crystallographic details for each protein-ligand structure. Figure S3 shows the images of the binding pocket of HCA II occupied by the various arylsulfonamide ligands used in this study.
Only protein residues within a 5 Å distance from the ligands are shown. Figure S4 shows instructive overlays of crystal structures.