Crystal structure correlations with the intrinsic thermodynamics of human carbonic anhydrase inhibitor binding

Protein preparation

Recombinant human carbonic anhydrases isoforms were expressed and purified by affinity chromatography as described previously (CA I in Čapkauskaitė et al. (2013), CA II in Čapkauskaitė et al. (2012), CA XIII in Sūdžius et al. (2010), and CA XII in Jogaitė et al. (2013).

Chemistry

The synthesis, chemical structural characterization, and the purity of 4-substituted-2,3,5,6-tetrafluorobenzenesulfonamides 1–5 and 10 has been described in Dudutienė et al. (2013), chlorinated benzenesulfonamides 6–9 in Čapkauskaitė et al. (2013) and 3,4-disubstituted-2,5,6-trifluorobenzenesulfonamides 11–16 in Dudutienė et al. (2014) and Dudutienė et al. (2015). The purity of all compounds was >95%, as verified by HPLC.

X-ray crystallography

Crystallization conditions of seven crystal structures are listed in Table 1. All crystallization experiments were carried out in Cryschem sitting drop plates (“Hampton Research” Cat. Nr. HR3-160) at 18–20 °C. Crystals of CA II and CA XII were soaked with 0.5 mM solution of the ligand prepared in the crystallization buffer. The quality of the electron density of the compounds is presented in Fig. 1. The crystal structures were solved by molecular replacement using MOLREP software (Vagin & Teplyakov, 1997). Initial models were the following: 3HLJ for CA II, 2NNO for CA XIII, and 1JD0 for CA XII. The 3D models of compounds were generated using AVOGADRO (Hanwell et al., 2012). The library files which contain complete chemical and geometric descriptions of compounds were created using LIBCHECK (Vagin et al., 2004). Model building and refinement were made using COOT (Emsley & Cowtan, 2004) and REFMAC (Murshudov, Vagin & Dodson, 1997) programs, respectively. The graphics were generated using MOLSCRIPT (Kraulis, 1991), BOBSCRIPT (Esnouf, 1999), RASTER3D (Merritt & Bacon, 1997) and PYMOL (Schrödinger, 2010) programs. Coordinates and structure factors have been deposited to the RCSB Protein Data Bank (PDB). PDB access codes are listed in Table 2.

Conformational changes

The root-mean-square deviations (RMSD) between CA-inhibitor complexes and apo proteins without ligand were calculated by Perl script for protein backbone atoms of chains A of crystal structures. The RMSD of the CA II complexes in comparison with apoCA II was 0.23 ± 0.04 Å (compound 7—0.30 Å, 3—0.22 Å, 4—0.29 Å, 10—0.18 Å, 8—0.19 Å, 12—0.17 Å, 13—0.24 Å, 9—0.25 Å, 2—0.22 Å, 14—0.21 Å, 6—0.26 Å); for CA I complexes vs apoCA I—0.17 ± 0.03 Å (compound 3—0.14 Å and 5—0.20 Å); for CA XII complexes vs apoCA XII—0.27 ± 0.10 Å (compound 10—0.26 Å, 11—0.52 Å, 1—0.18 Å, 15—0.32 Å, 16—0.18 Å and 12—0.17 Å) and for CA XIII complexes vs apoCA XIII—0.51 ± 0.04 Å (compound 9—0.54 Å, 10—0.46 Å and 11—0.53 Å). The differences between main chain atoms between different crystal structures of the same protein have been reported to be around 0.25–0.40 Å (Chothia & Lesk, 1986). RMSD values of peptide backbone between apo proteins and complexes with inhibitor were in range 0.17–0.51 Å and we conclude that no significant conformational changes occur upon ligand binding.

Calculation of molecular surface

The buried surface area (BSA) and accessible surface area (ASA) values were calculated using Voronota (v. 1.10.1544, Olechnovič & Venclovas, 2014) program. Hydrogen atoms were added to models by PERL script. The initial parameters for contacts calculation were set as follows: (1) rolling probe radius was default (1.4 Å); (2) values of Van-der-Waals radii were the following: Cl atom—1.75 Å, C—1.7 Å, F—1.47 Å, N—1.55 Å, O—1.52 Å, S—1.80 Å, Zn—1.39 Å, and H—1.09 Å.

Isothermal titration calorimetry

ITC experiments for the binding of 13–14 and 16 to CA II, CA XII and CA XIII were performed using VP-ITC instrument (Microcal, Inc., Northampton, MA, USA, now part of Malvern Instruments Ltd, UK) with 5–10 µM protein solution in the cell and 50–100 µM compound solution in the syringe. A typical experiment consisted of 25 injections (10 µl each) within 3–4 min intervals. Experiments were carried out at 37 °C in a 50 mM sodium phosphate or TRIS chloride buffer containing 100 mM NaCl at pH 7.0, with a final DMSO concentration of 1%, equal in the cell and syringe. Protein stock solutions were dialyzed against the buffers that were used to prepare the ligand solutions. ITC data were analyzed using MicroCal Origin software. All calorimetric titration data were presented after subtracting the background signal. The first point from the 5 µL injection in the integrated data graph was deleted. The binding constants, enthalpies and entropies of binding were estimated after fitting the data with the single binding site model.

Determination of intrinsic thermodynamics of binding

The observed energies of binding, including the Gibbs energies (observed affinities), enthalpies, and entropies of the protein-ligand pairs were determined by the fluorescent thermal shift assay (Pantoliano et al., 2001; Lo et al., 2004; Matulis et al., 2005; Cimmperman & Matulis, 2011) and isothermal titration calorimetry (Wiseman et al., 1989; Harding & Chowdhry, 2001; Broecker, Vargas & Keller, 2011). The measured observed thermodynamic parameters of binding depend on the assay conditions such as pH and buffer because the binding of sulfonamide to CA is linked to several protonation reactions, the protonation of hydroxide bound to the Zn(II) in the active site of CA, the deprotonation of sulfonamide group of the inhibitor, and the (de)protonation of the buffer (Baker & Murphy, 1996; Krishnamurthy et al., 2008; Baranauskienė & Matulis, 2012; Morkūnaitė et al., 2015). For the structure–activity relationship and correlation of thermodynamic parameters with the X-ray structures, the intrinsic thermodynamic parameters of sulfonamide anion binding to the Zn-bound water form of CA were calculated.

The intrinsic binding thermodynamics of compounds 1–12, and 15 were taken from references given in the Table 3. Isothermal titration calorimetry (ITC) was used to determine the enthalpy of 13–14 and 16 binding to CAII, CA XII, and CA XIII (the intrinsic affinities, determined by the fluorescent thermal shift assay, were taken from Dudutienė et al. (2015). We have shown previously (Smirnovienė, Smirnovas & Matulis, 2017) that the K_b values observed by FTSA and ITC methods were in good agreement and varied up to 3 times. However, the K_b determined by ITC method were not used in this study, because some compounds bound CAs with observed K_d < 10 nM and Wiseman c factor was too large for accurate K_b determination. To avoid discrepancies only the K_b determined by FTSA were used.

The intrinsic thermodynamic parameters for compounds 13–14 and 16 were calculated using Eqs. (1)–(4), where the pK_a of sulfonamide protonation was 7.88 and the enthalpy of sulfonamide deprotonation −24.3 kJ/mol.

The intrinsic binding constant K_{b_intr} is equal to the observed binding constant K_{b_obs} divided by the fractions of deprotonated inhibitor and protonated Zn-bound water form of CA: (1)${\displaystyle K_{\mathrm{b\_int}}=\frac{K_{\mathrm{b\_obs}}}{f_{{\mathrm{RSO}}_2{\mathrm{NH}}^{-}}\ast f_{{\mathrm{CAZnH}}_2\mathrm{O}}}.}$ The fractions of the deprotonated inhibitor and the Zn bound water form of CA may be calculated: (2)${\displaystyle f_{{\mathrm{RSO}}_2{\mathrm{NH}}^{-}}=\frac{10^{pH-p{K}_{\mathrm{a\_sulf}}}}{1+10^{pH-p{K}_{\mathrm{a\_sulf}}}}}$ (3)${\displaystyle f_{{\mathrm{CAZnH}}_2\mathrm{O}}=1-\frac{{10}^{pH-p{K}_{{\mathrm{a\_ZnH}}_2\mathrm{O}}}}{1+{10}^{pH-p{K}_{{\mathrm{a\_ZnH}}_2\mathrm{O}}}}.}$ The observed enthalpy (Δ_bH_obs) is the sum of all protonation events and the intrinsic binding reaction: (4)${\displaystyle {\Delta }_b{H}_{\mathrm{intr}}={\Delta }_b{H}_{\mathrm{obs}}-n_{\mathrm{sulf}}{\Delta }_{\mathrm{b\_proton\_sulf}}H-n_{\mathrm{CA}}{\Delta }_{\mathrm{b\_proton\_CA}}H+n_{\mathrm{buf}}{\Delta }_{\mathrm{b\_proton\_buf}}H}$ where Δ_bH_obs is the observed binding enthalpy, n_sulf = f_RSO2NH⁻ − 1 is the number of protons uptaken by the inhibitor, Δ_{b_proton_sulf} H is the enthalpy of inhibitor protonation, n_CA = 1 − f_CAZnH2O is the number of protons uptaken by the Zn-hydroxide, Δ_{b_proton_CA} H is the enthalpy of CA protonation, n_buf = n_sulf + n_CA, is the net sum of uptaken or released protons and Δ_{b_proton_buf} H is the enthalpy of buffer protonation.

Similar binders

The sulfonamide-bearing benzene ring was observed in the same orientation for all pairs of “similar binders” (S1-9, Fig. 3). Differences in the binding of ligands within these pairs were observed only in the regions where different substituents were present in the compounds of the pair. The intrinsic thermodynamic parameters of binding are listed in Table 3. Figure 5A shows the energies, including the Gibbs energy, enthalpy, and entropy of binding, for every analyzed pair of similar binders, as a function of the sum of accessible and buried surface areas of the hydrophobic substituents of inhibitors in each matched pair. The accessible and buried surface areas of substituent groups, marked green or magenta in Fig. 3, were calculated by VORONOTA algorithm (Olechnovič & Venclovas, 2014).

The comparison of intrinsic Gibbs energy of binding for each molecular pair in the group S1-9 (Fig. 5A) showed that there was a relatively small increase in affinity with an increase in the accessible and buried surface area. Additional hydrophobic surface seemed to not have a large impact on the binding affinities. However, the intrinsic enthalpies of binding in all pairs of similar binders were significantly less exothermic for compounds bearing larger hydrophobic substituents. Additional contacts between the protein and the ligand did not make the enthalpy more favorable. The entropies of binding mirrored the enthalpies—the enthalpy-entropy compensation was observed (Fig. 5A).

The compounds 2, 3, and 4 are para-substituted fluorinated benzenesulfonamides exhibiting nearly identical binding affinity for CA II (Figs. 3B and 3H, pairs S2 and S8). Fluorinated benzene rings of these compounds bind to the active site of CA II in almost identical mode and their hydrophobic para-substituents interact with the hydrophobic cavity formed by the amino acids Phe131, Val135, Leu198, Pro202, and Leu204. The properties of para-substituents in these compounds are different: propyl of compound 3 is small and flexible and can adopt an optimal conformation for interactions, while 2 and 4 have large and rigid adamantane and aromatic heterocyclic groups, respectively. The hydrophobic area of the para-substituents increases in 4 and 2 as compared to 3 and the entropic gain of the corresponding binding reaction increases too (Fig. 5A and Table 3). The identical binding mode of these compounds allows to propose that ΔH_interactions for these compounds can be similar in the binding to CA II. ΔH_{conformational} gain also can be similar due to (1) identical binding mode to CA II and (2) congeniality between these compounds. As mentioned previously, the pH and buffer influence to ΔH_exchange gain is eliminated because we analyze the intrinsic binding parameters. We hypothesize that the enthalpy-entropy differences are caused by the reorganization of the solvent near the hydrophobic surfaces of the compound, but not of compounds themselves. The changes of ΔH_desolvation and −TΔS_desolvation seem as first candidates for explaining the changes between binding thermodynamics of these compounds. The changes of desolvation entropy can be explained by increase of translational and rotational mobility (freedom) of water molecules which were more ordered near the hydrophobic surfaces of compounds before desolvation.

The active site of crystal structure CA II-3 contains more water molecules in comparison with CA II-2 and CA II-4 (Figs. 3B and 3H). However, it is difficult to make accurate analysis of the water molecules because the number of solvent molecules that are visible in crystal structure depend on the resolution. There are more water molecules in the crystal structure CA II–3 with the resolution of 1.06 Å (17 waters by Pymol searching around the ligand within 9 Å distance) as compared to CA II-4 (1.90 Å resolution, six waters) and CA II- 2 (1.90 Å resolution, nine waters). Furthermore, the active site of crystal structure CA II-3 possibly contains more water molecules than it is currently observed in comparison with structures CA II-2 and CA II-4 due to the only one distinct conformation of the flexible His64 (which is important for the rate limiting proton transfer event from the active site). The active sites of crystal structures CA II-2 and CA II-4 contain electron densities of two conformations of His64, one of which disturbs the quality of the electron densities of water molecules in the hydrophilic part of the active site.

Compounds 3 vs 5 in the matched pair S3 are para-substituted fluorinated benzenesulfonamides bound in identical position in the CA I active site (Fig. 3C, pair S3). The binding of 3 by CA I described in Dudutienė et al. (2015) with an unfavorable entropic contribution (Table 3). The extremely high binding affinity was explained by the specific interactions between 3 and CA I where the fluorobenzene ring is fixed between His200 and Leu198 which restricts the mobility of fluorobenzene ring of the ligand from the other side by the aromatic-aliphatic interaction. The minor enlargement of the hydrophobic area (cyclohexane in 5 vs. propyl in 3) resulted in significant changes in the enthalpy-entropy contribution (ΔΔG =  − 0.9 ± 1, ΔΔH = 14.7, −ΔTΔS =  − 15.6 kJ/mol). The same binding affinities in the pair indicate that compounds interact with the active site in a similar way (similar ΔH_interactions) and the enlargement of the hydrophobic surface is the only factor that influences the difference in thermodynamics.

The unfavorable entropic contribution of 3 binding to CA I seem as uncompensated unfavorable changes of conformational entropy (–TΔS_{conformational}) due to the reduction in mobility of binding partners upon formation of complex.

Compounds 13 and 14 are meta/para-substituted fluorinated benzenesulfonamides that differ by meta-substituents. Both compounds have para-phenyl groups that were found in both crystal structures with CA II “stacked” with the fluorinated benzene ring. In such conformation, the para-phenyl group in both ligands dislocates the water molecules from the hydrophilic part of the active site of CA II (Fig. 3E, S5). The binding of 13 with CA II is enthalpy-driven and the release of water molecules from the active site did not make the entropy more favorable (Table 3, Fig. 5A) due to increase in translational and rotational mobility of waters released from the active site. The additional hydrophobic area of meta-group for the compound 14 induces large changes in enthalpy-entropy terms (ΔΔG = 0.3 ±1, ΔΔH = 26.3, −ΔTΔS =  − 26 kJ/mol, Table 3). On the other hand, the overall binding affinity is the same for both compounds.

The structural differences of 10 vs 12 bound in CA II (Fig. 3G, S7) are located in the meta-position of the benzene ring. These compounds have the same position of the fluorinated benzenesulfonamide and of sulfonyl moiety of para-group. The enlargement of the hydrophobic part induces the increase of the binding affinity by −2.3 ± 1 kJ/mol (Table 3) within this pair. The changes of the entropic and enthalpic contributions of the pair S7 (ΔΔG =  − 2.3 ± 1, ΔΔH = 10.3, −ΔTΔS =  − 12.6 kJ/mol) are more than two times smaller than in the pair S5 (13 and 14 in CA II, Fig. 5A), despite somewhat larger variations in the surface between 10 and 12. The enthalpy-entropy compensation takes place in both cases.

The pair of structures S1, where compounds 10 and 11 (Fig. 3A), that differ by the presence of the bulky cyclooctyl group in the meta-position (Fig. 2) are bound to CA XIII, is an interesting exception between reviewed pairs: the binding affinity is notably increased with the increase of the hydrophobic surface of the meta-group (Fig. 5A). Inhibitor 11 binds to CA XIII 30 times stronger than 10 (K_d 0.29 nM for 11 and 8.9 nM for 10, Table 3). The cyclooctyl group was found in the crystal structure bound to the hydrophilic part of the CA XIII active site displacing water molecules observed in the structure with compound 10 (Fig. 3A, S1). The entropic and enthalpic contributions are large for this pair (ΔΔG =  − 8.8 ± 1, ΔΔH = 39.0, −ΔTΔS =  − 47.9 kJ/mol). The compound 11 is entropy-driven ligand and large entropic gain seems to be obtained from the changes of desolvation entropy on the hydrophobic surface of ligand and maybe by the release of water molecules from the active site.

Compounds 9 and 8 are para-substituted 2-chlorobenzenesulfonamides, that contain pyrimidine and dimethylpyrimidine groups in the para-position, respectively (Fig. 2). We compared the crystal structures of this pair of inhibitors bound to CA II (Fig. 3I, pair S9) and CA XIII (Fig. 3F, pair S6). The increase of hydrophobic surface between 9 and 8 is small, it is an addition of two methyl groups to the para-pyrimidine. Therefore, for both pairs, the favorable impact to binding affinities is rather small and occurs due to negligible favorable changes of entropies vs changes of enthalpies (Table 3 and Fig. 5A). In CA II (pair S9): ΔΔG =  − 0.8 ± 1, ΔΔH = 6.5, −ΔTΔS =  − 7.3 kJ/mol and in CA XIII (pair S6): ΔΔG =  − 1.4 ± 1, ΔΔH = 3.2, −ΔTΔS =  − 4.6 kJ/mol. In CA XIII (pair S6), methylpyrimidine of compound 8 coincides with the pyrimidine ring of 9, whereas in CA II (pair S9), methylpyrimidine of 8 interacts with the other hydrophobic part of the active site than pyrimidine ring of 9 (Figs. 3F and 3I). Methylpyrimidine is larger than pyrimidine and could not occupy the same position as pyrimidine in both CA isoforms. The more spacious hydrophobic part of the CA II active site allowed for the reorientation of the methylpyrimidine along the hydrophobic surface whereas in CA XIII the methylpyrimidine is simply dislocated outwards.

Compounds 7 and 6 are homologous to compounds 9 and 8 respectively (Fig. 2), except that the benzene ring does not contain chlorine. This pair of ligands is represented by the crystal structures of the complexes with CA II (Fig. 3D, pair S4). The binding thermodynamics in S4 exhibits the distinct decrease of the binding affinity towards CA II: ΔΔG = 3.1 ± 1, ΔΔH = 13.6, and −ΔTΔS =  − 10.5 kJ/mol due to the uncompensated enthalpic loss and the entropic gain (Fig. 5A and Table 3).

Dissimilar binders

The thermodynamic parameters of the dissimilar inhibitor (Figs. 4A–4D) binding to CA XII are shown in Fig. 5B as a function of the surface of substituent at the benzenesulfonamide group. For the dissimilar binders, there is a significant increase of the intrinsic binding affinity between compounds in matched pairs (e.g., 318 times for the pair D1) when the accessible and buried surface areas increase. The intrinsic enthalpies of binding in all pairs of dissimilar binders were significantly more exothermic for compounds with larger hydrophobic substituents, contrary to the binding thermodynamics of similar binders (pairs S1–S9): pair D1 (ΔΔG =  − 14.7 ± 1, ΔΔH =  − 16.2, −ΔTΔS = 1.5 kJ/mol), pair D2 (ΔΔG =  − 8.7 ± 1, ΔΔH =  − 11.5, −ΔTΔS = 2.8 kJ/mol), pair D3 (ΔΔG =  − 10.5 ± 1, ΔΔH =  − 5.2, −ΔTΔS =  − 5.3 kJ/mol), and pair D4 (ΔΔG =  − 11.5 ± 1, ΔΔH =  − 22.4,−ΔTΔS = 10.9 kJ/mol). Additional contacts between the protein and the ligand made the enthalpy more favorable for this group. The entropy was significantly favorable for the pair D3, significantly unfavorable for D4, and the changes were negligible for D1 and D2. The compounds that are more hydrophobic in these pairs (15, 16, 12, and 11) differ by the meta-substituent from the reference compound 10 and have the same location of the fluorinated benzene ring in the CA XII active site. The meta-substituents are located in the hydrophobic environment and sterically dislocated the remaining part of the compound is bound in the hydrophilic concave of CA XII active site (Figs. 4A–4D). The favorable changes of the binding enthalpies of the dissimilar binders could be explained by the favorable location of the hydrophilic para-group which is sterically dislocate into the hydrophilic part of the active site where the formation of additional hydrogen bonds is more likely. The favorable ΔH_interactions may be the reason for the improved binding affinities.

It is interesting to compare the compounds 12 (pair D4) and 16 (pair D3) that differ by only one bridged methyl group in the meta-substituent. The difference in the binding energy between these compounds (12 vs 16) is significant (ΔΔG = 1.0 ± 1, ΔΔH = 17.2, −ΔTΔS =  − 16.2 kJ/mol). Maybe this could be explained by the possibility of only one position of compound 12 para-group when bound in CA XII, where two hydrogen bonds are formed with the residues Asn64 and Pro200. The para-group of 16 was found in the alternate conformation, but in the compound 12 bound to CA XII both substituent groups were found in a single conformation. Another explanation can be the same as for similar binders, where the increase of the hydrophobic surface induces similar behavior of binding thermodynamics. Compound 12 and 16 bind to CA XII in identical mode.

The cases of perfect geometry fit in the interactions between protein and ligand

An extremely high affinity of the compound 1 to CA II and CA XII may be an example of a perfect geometrical fit of the ligand to the CA active site. The compound 1 has a relatively simple structure and its binding is enthalpy-driven (Fig. 4F, Table 3) exhibiting very high intrinsic affinity to both analyzed CA isoforms, 7.4 pM to CA II and 67 pM to CA XII. As seen in both crystal structures shown in Fig. 4F, the ligand is likely to bind so strongly due to the tight aromatic-aliphatic interaction between the benzene ring and a conservative leucine side chain (Leu198 in CA II and Leu197 in CA XII). An important difference between CA II and CA XII is the substitution of Phe131 in CA II with the alanine side chain in CA XII. A loss of the exothermic enthalpy was observed for the binding of 1 to CA XII as compared to CA II due to the possible absence of additional interaction of para-substituent of 1 with Phe131 in CA II.

Compounds of the pair S4 (Fig. 4E) exhibit an identical orientation of the benzenesulfonamide ring as compound 1 (Fig. 4F) bound to CA II and CA XII. The K_d values of 7 and 6 bound to CA II are 25 pM and 80 pM, respectively, and such values are similar to the binding of compound 1 to CA II and CA XII. Moreover, binding of these compounds are enthalpy-driven. The aromatic-aliphatic interaction looks like a possible reason of the favorable enthalpic gain (7: ΔH =  − 62.0 kJ/mol and 6: ΔH =  − 48.4 kJ/mol) of which a major part is due to ΔH_interactions.

Moreover, analysis of these pairs (1 in CA II and CA XII and matched pair S4) highlights that there are contacts between the para-substituents and the protein. In the crystal structures of 1 in CA II and CA XII, the loss of additional surface (Ala instead of Phe) changes the location of the propyl group which is now directed perpendicularly to the plane of aromatic-aliphatic interaction between Leu198 and benzene ring bound to CA XII (Fig. 4F). The propyl group of the compound 1 bound to CA II (Fig. 4F) is located in the hydrophobic cavity (formed by Phe131 and Pro202) and possibly increases the affinity of the aromatic-aliphatic interaction due to same direction of interactions between (1) propyl group and hydrophobic cavity and (2) benzene ring of compound and leucine side chain.

The matched pair S3 (Fig. 3C) exhibits similar behavior. Atoms of His200 side chain dislocate the fluorinated benzenesulfonamide ring of both compounds 5 and 3 and possibly increases the affinity of the aromatic-aliphatic interaction with Leu198. Compound 5 (of pair S3) binds to CA I with K_d = 1.5pM (Table 3) as previously confirmed both by the thermal shift assay and displacement ITC (Zubrienė et al., 2015).