Engineered Carbonic Anhydrase VI-Mimic Enzyme Switched the Structure and Affinities of Inhibitors

1 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports engineered carbonic Anhydrase Vi-Mimic enzyme Switched the Structure and Affinities of Inhibitors Justina Kazokaitė1,6, Visvaldas Kairys2, Joana Smirnovienė1, Alexey Smirnov1, Elena Manakova3, Martti tolvanen4, Seppo Parkkila5 & Daumantas Matulis 1 Secretory human carbonic anhydrase VI (CA VI) has emerged as a potential drug target due to its role in pathological states, such as excess acidity-caused dental caries and injuries of gastric epithelium. Currently, there are no available CA VI-selective inhibitors or crystallographic structures of inhibitors bound to CA VI. The present study focuses on the site-directed CA II mutant mimicking the active site of CA VI for inhibitor screening. The interactions between CA VI-mimic and a series of benzenesulfonamides were evaluated by fluorescent thermal shift assay, stopped-flow CO2 hydration assay, isothermal titration calorimetry, and X-ray crystallography. Kinetic parameters showed that A65T, N67Q, F130Y, V134Q, L203T mutations did not influence catalytic properties of CA II, but inhibitor affinities resembled CA VI, exhibiting up to 0.16 nM intrinsic affinity for CA VI-mimic. Structurally, binding site of CA VI-mimic was found to be similar to CA VI. The ligand interactions with mutated side chains observed in three crystallographic structures allowed to rationalize observed variation of binding modes and experimental binding affinities to CA VI. This integrative set of kinetic, thermodynamic, and structural data revealed CA VI-mimic as a useful model to design CA VI-specific inhibitors which could be beneficial for novel therapeutic applications. Human carbonic anhydrases (CAs) are widespread enzymes known for over 80 years1. CAs regulate both intracellular and extracellular pH homeostasis through the catalysis of reversible carbon dioxide hydration to bicarbonate and proton. To date, there are twelve catalytically active human CAs, which display diverse sub-cellular localization, tissue-specific expression, and kinetic properties2,3. Among a broad spectrum of CA-linked research areas, clinical investigation is a major focus due to the implication of abnormal CA levels or their activities in diseases, such as glaucoma4, epilepsy5, obesity6, and cancer7. Therefore, many efforts have been dedicated over years to design CA isoform-selective compounds exhibiting sufficient affinity properties8. These derivatives would be prospective for the translation into the clinic because of therapeutic efficacy without inducing undesired side effects caused by inhibited vital off-target CAs. However, it is a challenging task because of the high structural homology among human CAs9. CA VI is the only secreted human CA isoform found in saliva10, serum11, milk12, respiratory airways13, and alimentary canal14. Several studies have indicated the immunological CA VI function15,16 and have presented associations of CA VI with bitter taste perception17,18 or protection of excess acidity-caused complications, including dental caries19,20 and injuries of esophageal or gastric epithelium21. The link of CA VI with certain cancers, such as that of salivary glands, has been speculated by gene comparison study22, which have shown close relation of CA VI with CA IX, a marker of tumors23. Thus, there is a demand for effective and selective CA VI inhibitors, which would be relevant to determine the exact physiological role of CA VI. For more than five decades, the most widely applied method in the search of CA isoform selective inhibitors has been the stopped-flow assay of the catalytic activity of CO2 hydration (SFA)24,25. However, SFA has several 1Department of Biothermodynamics and Drug Design, Institute of Biotechnology, Vilnius University, Saulėtekio 7, Vilnius, LT-10257, Lithuania. 2Department of Bioinformatics, Institute of Biotechnology, Vilnius University, Saulėtekio 7, Vilnius, LT-10257, Lithuania. 3Department of Protein-DNA Interactions, Institute of Biotechnology, Vilnius University, Saulėtekio 7, Vilnius, LT-10257, Lithuania. 4Department of Information Technology, University of Turku, FI-20520, Turku, Finland. 5Tampere University, Faculty of Medicine and Health Technology; Fimlab Ltd, Tampere University Hospital, Arvo Ylpön katu 34, FI-33520, Tampere, Finland. 6Present address: Division of Biochemistry, the Netherlands Cancer Institute, Amsterdam, the Netherlands. Correspondence and requests for materials should be addressed to J.K. (email: [email protected]) Received: 17 February 2019 Accepted: 15 August 2019 Published: xx xx xxxx open 2 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ limitations, such as the largely unknown CO2 concentration and unfeasibility to measure inhibition constant below several nM26. Therefore, biophysical techniques, such as the fluorescent thermal shift assay (FTSA) and isothermal titration calorimetry (ITC), are promising alternatives to screen CA-targeting derivatives. FTSA is a high-throughput method exhibiting minimized biomolecule consumption and low limitations for binding affinity, thereby both strong (picomolar) and weak (millimolar) compounds can be identified during the same experiment26–29. ITC allows the direct determination of stoichiometry and thermodynamic parameters, such as affinity, enthalpy, entropy, and heat-capacity, during a single or several titration experiments but it demands relatively large quantities of proteins and has limitations for assessing the binding affinity26,30,31. Importantly, two types of variables can be distinguished when binding reactions are carried out by FTSA or ITC: the observed parameters obtained from experimental setup and the intrinsic values calculated according to the corresponding observed data. Most studies on the development of CA inhibitors usually provide only observed binding parameters, which are dependent on experimental conditions and might be misleading. Both the CA and inhibitor exist in different protonation states in the solution compared with ones in the complex. Therefore, protonation-deprotonation reactions are required to initiate the binding of inhibitor to CA. Only intrinsic values subtract energetic contribution of binding-linked protonation events and thus are relevant for the rational drug design32–35. Due to the recent advances in the structural and in silico biology, production of target recombinant proteins, including CAs, in large quantities is of high demand for in vitro inhibitor screening of drug-candidates during preclinical research. The literature lists a number of host cells for expression of recombinant proteins. Among microorganisms, the enterobacterium Escherichia coli (E. coli) is selected frequently owing to numerous advantages, such as rapid growth, easy genetic manipulation, and relative cost effectiveness36,37. However, the stability of heterologous protein in E. coli can be influenced by the several factors, including mRNA instability, codon bias, protein aggregation, toxicity, and lack of post-translational modification38,39. Therefore, different, more efficient strategies to obtain functionally active recombinant proteins in high yield are required for screens of chemical compounds with the aim to identify hits in the initial stages of drug discovery. The goal of the present study was to design a CA II-based CA VI model protein, named as CA VI-mimic, for the search of CA-isoform selective inhibitors. As CA VI-mimic, mutant of CA II containing five point mutations, such as A65T, N67Q, F130Y, V134Q, L203T, was generated via site-directed mutagenesis. CA II was selected as a core for CA VI-mimic because purification yield of CA II from E. coli is ~10-fold higher than CA VI, CA II has highest catalytic efficiency among CAs, and CA II is confirmed as a stable CA protein for X-ray crystallography. Here enzymatic activity and inhibition of CA II, CAVI-mimic, and CA VI was determined by SFA. Biophysical studies on inhibitor binding to CA II, CA VI-mimic, and CA VI were carried out by ITC and FTSA. X-ray crystallography and computational modeling were used to compare positions of several inhibitors in the active sites of CA II, CA VI-mimic, and CA VI. Observed and intrinsic thermodynamics were in line with structural results which confirmed the relevance of CA VI-mimic as a CA VI model protein. The most tested benzenesulfonamides bound to CA VI-mimic in a manner corresponding to their interactions with CA VI but not CA II, thereby emphasizing suitability of the investigated CA II mutant mimicking CA VI for inhibitor screening. Results Enzymatic activity of CA VI-mimic correlates with CA II, but not CA VI. Studies on inhibitor selectivity towards diverse human CA isoforms are important to develop efficient compounds for the treatment of diseases caused by abnormal levels or activities of a particular CA isoform. Therefore, it is essential to evaluate inhibitor affinity to all human CAs, including CA VI. Since our previous study40 indicated a low yield of recombinant CA VI from E. coli, we generated CA II mutant as a CA VI model protein (CA VI-mimic) for inhibitor screening. Inhibitor affinities towards CA II and CA VI-mimic were expected to differ in the way imitating inhibitor binding to CA VI, but not CA II (Fig.1A). Thus, negligible differences between inhibitor affinities towards CA VI and CA VI-mimic were presumed. According to computational modeling, five point mutations A65T, N67Q, F130Y, V134Q, L203T were chosen (Figs1B–D and S1) and introduced into the active site of CA II. The catalytic activity of CA VI and CA VI-mimic to catalyze CO2 hydration reaction was measured by SFA (Figs2A and S3). Analysis of kinetic data showed that site-directed mutagenesis did not significantly affect either the catalytic activity or pKa of zinc-bound water molecule of CA II. Catalytic constants (kcat) of CA II and CA VI-mimic did not differ (kcat values were 6.0 × 105 s−1), whereas kcat for CA VI was lower than CA VI-mimic by 3-fold (1.9 × 105 s−1). In the pH range 5.9–7.0 Michaelis constants (KM) as well as kcat values of the carbon dioxide hydration reaction were comparable: 7.3 ± 2.9 mM for CA II, 8.8 ± 1.8 mM for CA VI-mimic and 9.9 ± 3.2 mM for CA VI. However, in the pH range 7.1–8.4 KM values of CA VI-mimic (6.8 ± 2.0 mM) were closer to CA II (4.7 ± 1.0 mM) than to CA VI (11.3 ± 0.7 mM). Interestingly, maximum catalytic activity of CA VI was observed at pH 7.0–8.0 and it decreased at pH above 8.0. The determined pKa value of zinc-bound water molecule of CA VI was 6.6 ± 0.2. The observed inhibition constants by SFA correlated with dissociation constants determined by FTSA. Typical SFA curves of CA II, CA VI-mimic and CA VI inhibition by compound 39 are shown in Fig.2B. Influence of buffer and pH for the observed binding affinity of ethoxzolamide to CA VI-mimic. Biophysical methods, such as ITC and FTSA, enable measurements of observed thermodynamics and thereafter calculations of intrinsic affinities. The observed binding profiles are altered by linked reactions and therefore, only intrinsic binding parameters can be correlated with compound structures, thereby revealing structural reasons for protein-ligand binding affinity. Binding energetics are significantly affected by several protonation-deprotonation events which are necessary for the binding of sulfonamide derivative to CA. Only deprotonated sulfonamides can interact with the zinc cation in the active site of pronated CA, containing zinc-coordinated water molecule (protonated hydroxy group). In this study, observed and intrinsic affinities of inhibitor binding to CA VI-mimic were determined and compared 3 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ to their affinities towards CA II and CA VI. The obtained experimental data by FTSA on interactions between ethoxzolamide (EZA) and CA VI-mimic in buffers with different pH showed that pH remarkably influenced the observed binding Gibbs energy (ΔbGobs, Fig.3A). The dependence of ΔbGobs on pH has also been observed previously40,41 when EZA binding to CA II or CA VI was measured. The strongest interaction was determined near neutral pH and became weaker both in acidic and alkaline pH. Sulfonamide group usually has pKa in the range between 7 and 10, whereas CA isoforms have pKa around 7. Therefore, diminished EZA affinity in acidic solution was because the fraction of binding-ready deprotonated form of EZA decreased by 10-fold with every pH unit. Similarly, EZA affinity decreased in alkaline solution because the fraction of binding-ready CA with the Figure 1. (A) The mimic of CA VI was prepared from CA II by site-directedmutagenesis of amino acids that differ between two CA isoforms. The CA VI-mimic protein served as a model of compound binding to CA VI. (B) Active site of CA II (PDB ID: 3KS3). Dark red molecular surfaces mark the positions of point mutations introduced in the active site of CA II to resemble CA VI by making a multiple-residue mutant of CA II (CA VImimic). (C) Active site of CA VI (PDB ID: 3FE4). The light blue areas are buried molecular surfaces between interacting molecules in the homodimeric complex. Dark red molecular surfaces mark the equivalent positions between multiple-residue mutant of CA II (CA VI-mimic) and CA VI. The labels belong to CA VI (CA II numbering). (D) Superposed structures of the binding pockets of CA II (rose; PDB ID: 3M96), CA VI (blue; PDB ID: 3FE4), and CA VI-mimic (green; PDB ID: 6QL2). The mutated residues of CA II are colored dark red. The zinc ion in the active site of each CA isoform is shown as a grayish sphere in panels (B–D). 4 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ zinc-bound protonated hydroxide (water molecule) decreased. According to U-shaped curve as the global fit of experimental data extrapolated to 25 °C, intrinsic binding Gibbs energy (ΔbGintr) change upon EZA interaction with CA VI-mimic was determined to be −52.4 kJ/mol which was 7.8 kJ/mol greater than the highest experimentally observed value (−44.6 kJ/mol at pH 7.1). Difference of ΔbGintr (ΔΔbGintr) between EZA interaction with CA VI-mimic and CA VI were smaller (ΔΔbGintr = −1.5 kJ/mol) compared to that between CA VI-mimic and CA II (ΔΔbGintr = 6.1 kJ/mol). Observed standard enthalpy changes (ΔbHobs) upon EZA binding to CA VI-mimic formed an X-shaped curve which depended on pH and buffer (Fig.3B). The same tendency has been found previously40,41 when ΔbHobs of EZA binding to CA II or CA VI was analyzed. Results were obtained by ITC titration at 25 °C in two buffers exhibiting different protonation enthalpies: sodium phosphate (Pi) and TRIS. Upon EZA-CA VI-mimic titration, more than 20 kJ/mol difference in ΔbHobs was observed in same buffer at different pHs (in TRIS buffer: −81.6 kJ/ mol at pH 6.5, −57.2 kJ/mol at pH 8.5; in Pi buffer: −38.9 kJ/mol at pH 5.2, −84.9 kJ/mol at pH 8.4). To dissect protonation influence, intrinsic enthalpy (ΔbHintr) of EZA interaction with CA VI-mimic was globally fitted to be −62.0 kJ/mol. Difference of ΔbHintr (ΔΔbHintr) between EZA interaction with CA VI-mimic and CA VI were smaller (4.0 kJ/mol) compared to that between CA VI-mimic and CA II (11.0 kJ/mol). Thus, ΔbHintr were in line with ΔbGintr, confirming that CA II mutant was mimicking CA VI for EZA binding. Furthermore, analysis of Uand X-shaped curves obtained by FTSA and ITC, respectively, led to the characterization of two important parameters of CA VI-mimic: ionization constant (pKa) and enthalpy of protonation Figure 2. Catalytic activity, inhibition and binding profiles of CA II (red squares), CA VI-mimic (wine triangles) and CA VI (royal circles). (A) The plot of kcat dependence on pH by stopped-flow CO2 hydration assay (SFA). Solid lines were fit using single protonation model. (B) Inhibition of CAs by compound 39 using SFA. Data points were fit to the Morrison eq. (solid lines)74,75. The insets show raw activity curves of CA catalyzed reaction without added inhibitor (red, wine, royal lines), CA inhibited reaction with 313 nM added compound 39 (magenta, dark yellow, purple lines) and spontaneous CO2 hydration reaction (pink, orange, cyan lines) in the absence of CA. (C) Dosing curves of compound 39 binding to CAs by fluorescent thermal shift assay (FTSA). Data points show the ΔTm as a function of the total concentration of compound 39 added and the lines are simulated using fitting parameters when temperature is 37 °C, CA concentration is 10 µM, enthalpy of unfolding is 690 kJ/mol for CA II and CA VI-mimic, and 480 kJ/mol for CA VI, enthalpy of binding is −42 kJ/mol, heat capacity of binding is −0.8 J/ (molK) and the reference melting temperature is 56.8 °C for CA II, 63.1 °C for CA VI-mimic, and 47.6 °C for CA VI. The ΔTm shift is equal for CA VI-mimic and CA VI, but Kd’s differ due to different enthalpies of unfolding. The insets show CA unfolding curves at 0 and 200 µM inhibitor 39 concentrations. (D) Isothermal titration calorimetry (ITC) curves of EZA binding to CA VI-mimic in phosphate (Pi, pH 6.1 (▲) and 8.4 (●)) and TRIS buffer (pH 6.9 (■)) at 25 °C. Lines were fitted using single binding site model. The insets show raw data ITC curves at 10 µM CA VI-mimic concentration. Different observed enthalpies of binding illustrate the presence of binding-linked protonation reactions that must be accounted for the determination of intrinsic binding parameters. 5 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ (ΔpH) of the zinc-bound water molecule (Table1). The pKa of CA VI-mimic was determined to be 7.2 at 25 °C as the average of two pKa values evaluated independently by two techniques: 7.1 by FTSA and 7.3 by ITC. The pKas of CA II and CA VI-mimic matched each other within the error margin of 0.2 pH unit42, whereas pKas of CA VI and CA VI-mimic significantly differed by 1.0 pH (25 °C). Thus, target five point mutations of CA II, which were introduced to design CA VI-mimic, did not affect amino acids surrounding zinc in active sites of CA II at the level causing significant difference of pKas between CA II and CA VI-mimic. Moreover, ΔpH for CA VI-mimic was assessed to be −38.0 kJ/mol at 25 °C. The difference of ΔpH between CA VI and CA VI-mimic (6.0 kJ/mol) was 2-fold lower than difference of ΔpH between CA II and CA VI-mimic (12.0 kJ/mol). Therefore, experiments with one inhibitor EZA resulted in both pKa and ΔpH for CA VI-mimic, which are essential parameters to determine intrinsic energetics of any other inhibitor binding to CA VI-mimic. Hydrophobic substituents and fluorine substituents significantly affected intrinsic inhibitor binding affinity for CA VI-mimic. Here 43 benzenesulfonamide derivatives binding to CA VI-mimic was measured by FTSA and inhibition constants of several selected compounds were confirmed by SFA. Trifluoromethanesulfonamide (TFS), EZA, and methazolamide (MZM) were used as controls. Structures of tested compounds are shown in Fig.4, while dissociation constants (Kd) are listed in Tables2 and S1 (examples Figure 3. (A) Comparison of observed Gibbs energy changes (ΔbGobs) upon EZA binding to CA II (red), CA VI-mimic (dark red), and CA VI (blue) as a function of pH (25 °C). Experiments were performed by FTSA in universal buffer (50 mM sodium phosphate, 50 mM sodium acetate, and 25 mM sodium borate). The pKa for CA VI-mimic was determined to be 7.1. (B) The observed enthalpy changes (ΔbHobs) upon EZA binding to CA II (red), CA VI-mimic (dark red), and CA VI (blue) as a function of pH in two different buffers (sodium phosphate (Pi) and TRIS), which have different protonation enthalpies. Experiments were performed by isothermal titration calorimetry (ITC) at 25 °C. The dashed line shows the intrinsic binding enthalpy (ΔbHintr), which is independent of pH. The pKa for CA VI-mimic was determined to be 7.3. Thermodynamic binding parameters of EZA binding to CA II and CA VI have been previously published40,41. Red arrows indicate difference in ΔbGintr or ΔbHintr of EZA binding to CA VI-mimic compared to CA II or CA VI. Protein pKaΔpG, kJ/mol ΔpH, kJ/mol TΔpS, kJ/mol CA IIa7.1 −40.5 −26.0 14.5 CA VI-mimic 7.2 −41.1 −38.0 3.1 CA VIb6.2 −35.4 −32.0 3.4 Table 1. Thermodynamic parameters of protonation of zinc-bound hydroxide anion of studied CA isoforms as determined by FTSA and ITC at 25 °C. The uncertainty of the pKa values determined by FTSA and ITC is approximately 0.2 pH units, while for the change in Gibbs energies and enthalpies it is approximately 2 kJ/mol. aData taken from41; bData taken from40. 6 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ of raw and integrated data of inhibitor binding to CA VI-mimic by FTSA and ITC at different pHs are indicated in Fig.2C,D, respectively). According to observed thermodynamics, EZA was shown to be the strongest binder to CA VI-mimic with observed Kd (Kd_obs) of 17 nM. From a series of fluorinated benzenesulfonamides, compounds 22 and 26 bearing substituents at para position were characterized to be the most potent CA VI-mimic inhibitors bound with observed Kd (Kd_obs) in the range of 50–67 nM. The comparison between binding affinities of corresponding fluorinated and nonfluorinated compounds (6 vs 30, 15 vs 16, and 22 vs 23) showed that fluorination significantly increased observed binding affinity and diminished pKa of inhibitor sulfonamide amino group. For instance, Figure 4. Chemical structures of 1-46 compounds designed as CA inhibitors. Compounds 1-3 are standard inhibitors of CAs that we used here as control compounds (TFS, EZA, and MZM). 7 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ Inhibitor Lab. name pKa_SA Kd_obs (nM) Kd_intr (nM) CA II CA VI-mimic CA VI CA II CA VI-mimic CA VI 1. TFS 6.02 20 33 14 8.0 15 1.2 2. EZA 7.82 1.3 (<5.0) 17 (<54) 33 (<54) 0.073 1.1 0.40 3. MZM 6.86 100 330 830 27 97 44 4. VD10-9 8.12 46 130 430 1.4 4.7 2.7 5. VD10-12 8.61 18 100 1100 0.19 1.2 2.4 6. VD10-13 8.14 11 130 200 0.32 4.2 1.2 7. VD10-14 8.84 91 330 1000 0.57 2.4 1.3 8. VD10-16 8.47 9.1 200 1200 0.13 3.3 3.7 9. VD10-18 7.80 3.4 140 200 0.20 9.8 2.5 10. VD10-35 7.28 17 130 67 2.5 22 2.1 11. VD10-39b 7.85 83 140 130 4.6 8.8 1.5 12. VD10-45 7.69 5.8 140 140 0.43 12 2.2 13. VD10-49 7.83 0.65 110 200 0.037 7.2 2.3 14. VD10-50 8.02 9.6 (<43) 330 (140) 1000 (630) 0.37 15 7.9 15. VD11-9 8.05 1.7 100 400 0.061 4.1 3.0 16. VD11-36 10.1 12 500 1100 0.0039 0.20 0.077 17. VD11-67 8.67 5900 100 000 100 000 55 1000 190 18. VD11-51 7.07 3.3 130 160 0.68 31 6.6 19. VD11-56 7.97 20 330 500 0.86 16 4.4 20. VD11-71 8.67 500 2500 3300 4.6 26 6.3 21. VD12-04 8.67 1800 3300 20 000 17 35 38 22. VD12-05 8.15 2.2 50 140 0.065 1.7 0.86 23. VD12-10 10.2 25 (<36) 500 (200) 830 (3900) 0.0070 0.16 0.048 24. VD12-17 8.67 1300 5000 11 000 12 52 21 25. VD11-4-2 8.01 56 100 67 2.2 4.5 0.54 26. VD11-10 7.22 1.2 67 140 0.21 13 4.8 27. VD11-16 7.87 35 1000 2000 1.9 59 22 28. VD11-17 7.87 50 140 330 2.6 8.5 3.6 29. VD11-28 7.87 6.7 200 110 0.35 12 1.2 30. VD11-31 9.96 140 1000 5000 0.070 0.55 0.50 31. VD11-39 7.87 33 2000 1000 1.8 120 11 32. VD11-61 8.53 3.3 140 330 0.042 2.1 0.87 33. E46 8.90 50 50 200 0.28 0.31 0.23 34. E11-6 8.70 5.6 330 1100 0.048 3.3 2.0 35. E11-11 9.40 1000 2000 13 000 1.8 4.0 4.5 36. E11-12 8.90 8.5 500 5000 0.047 3.1 5.7 37. E11-14 8.90 7.1 330 3300 0.039 2.1 3.8 38. E11-15 9.40 560 1300 5000 0.98 2.5 1.8 39. E11-18 8.90 28 (<54) 500 (300) 2500 (1600) 0.15 3.1 2.8 40. E11-37 9.60 3600 3300 6700 4.0 4.2 1.5 41. E11-36 8.30 8.3 500 3300 0.16 12 14 42. E11-52 8.70 2.9 330 1400 0.025 3.3 2.5 43. E11-53 9.60 1700 3300 10 000 1.9 4.2 2.3 44. E11-57 9.60 140 500 3300 0.16 0.63 0.75 45. E11-60 8.70 250 500 3300 2.2 4.9 5.9 46. E11-64 9.40 100 1100 17 000 0.16 2.2 5.5 Table 2. The Kd_obs and Kd_intr values (nM) for interactions between inhibitor and three CA proteins: CA II, CA VI-mimic, and CA VI. The observed inhibitor affinities for CA VI-mimic were obtained experimentally by FTSA (pH 7.0, 37 °C), whereas the intrinsic parameters were calculated from the corresponding observed data using pKa of 7.0 for CA VI-mimic at 37 °C as explained in the methods part. The standard error of Kd measurements is ±2-fold. The pKa values of applied sulfonamide amino group (pKa_SA) and inhibitor affinities towards CA II and CA VI have been already reported58,86. Dissociation constants Kds of selected compounds were confirmed by SFA (pH 7.5). Experiments were performed at 23 °C and observed Kds were extrapolated to 37 °C using van’t Hoff equation when enthalpy of binding is −42 kJ/mol. The values at 37 °C are given in parentheses. The determined Kds at 23 °C are given in TableS2. 8 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ Kd_obs for 30 and 6 binding to CA VI-mimic increased 8-fold upon fluorination (from 1000 nM to 130 nM), whereas Kd_obs for 23 and 22 increased affinity 10-fold (from 500 nM to 50 nM). Fluorines reduced pKa of sulfonamide group significantly: from 9.96 to 8.14 for inhibitors 30 and 6, respectively, and from 10.2 to 8.15 for compounds 23 and 22, respectively (Table2). Correspondingly, chlorine in most compounds also increased observed affinity and reduced pKa of inhibitor sulfonamide amino group. For example, upon chlorination Kd_obs for 40 and 41 interaction with CA VI-mimic increased 430-fold (from 3600 nM to 8.3 nM, respectively), while pKa values were lowered by 1.30 unit (from 9.60 to 8.30, respectively). To investigate structure-activity relationships, intrinsic Kd (Kd_intr) values for interactions between CA VI-mimic and investigated series of compounds were calculated. The largest differences between Kd_obs and Kd_intr values were determined for nonfluorinated benzenesulfonamides (16, 23, and 30), where the binding to CA VI-mimic differed 2500, 3200, and 1800-fold, respectively. Only five compounds (10, 18, 26, TFS, and MZM) exhibited lower than 10-fold difference between the Kd_obs and Kd_intr. According to intrinsic thermodynamics, the strongest binders were inhibitors 16, 23, 30, and 33 with Kd_intr in the range of 0.16–0.55 nM. Therefore, the strongest intrinsic interaction between inhibitor and CA VI-mimic was observed when inhibitor did not possess any fluorines in benzenesulfonamide scaffold and contained a hydrophobic substituent at para position, such as SCH2CH2CH3 (Kd_intr for inhibitor 23 was 0.16 nM) and SCH2CH2Ph (Kd_intr for inhibitor 16 was 0.20 nM). Exceptionally, inhibitor 33 was the strongest binder to CA VI-mimic with chlorine at ortho position and large hydrophilic group at meta position (Kd_intr was 0.31 nM). Replacement of the methyl group (inhibitor 23) by hydrophilic hydroxyl group (inhibitor 30) weakened intrinsic binding affinity more than 3-fold (from to 0.16 nM to 0.55 nM). Moving on to the structural analysis of fluorinated benzenesulfonamides, two inhibitors were determined to be the strongest binders to CA VI-mimic: compound 5 bearing 4-Morpholinyl group at para position (Kd_intr was 1.2 nM) and 22 with SCH2CH2CH3 group at para position (Kd_intr was 1.7 nM). In line with results obtained from nonfluorinated compounds, hydrophobic contacts between inhibitors and CA VI were identified to be significant because the exchange of methyl group (inhibitor 22) by hydroxyl group (inhibitor 6) or carboxyl group (inhibitor 19) weakened intrinsic interaction by 2 and 9-fold, respectively. Apparently, the number of methyl groups of substituents at para position had significant effect on intrinsic binding affinity. The inhibitor 32 with SCH2Ph bound to CA VI-mimic 2-fold stronger than inhibitor 15 with SCH2CH2Ph and 4-fold stronger than inhibitor 9 with SPh. Most often, introducing diverse substituents at meta position did not change intrinsic binding affinity significantly (6 vs 25, 26 vs 28, and 26 vs 29), except for 26 vs 31 bearing large hydrophobic group which weakened interaction 9-fold. The compound 17 bearing two large and highly hydrophobic substituents at ortho and para positions was the weakest binder not only according to the observed parameters (Kd_obs was 100 µM), but also intrinsic data (Kd_intr of 1000 nM). Thermodynamically CA VI-mimic binds benzenesulfonamides similarly to CA VI but differing from CA II. To evaluate if CA VI-mimic based on CA II is a suitable CA VI model protein for inhibitor screening, observed and intrinsic affinities represented by logarithmic Kd values of inhibitor binding to CA II, CA VI-mimic, and CA VI were compared by applying linear regression. A higher linear correlation was determined between observed affinities of inhibitor binding to CA VI and CA VI-mimic (R2 = 0.79) compared to the observed affinities of inhibitor interaction with CA II and CA VI (R2 = 0.61; Fig.5A). Analysis of the calculated intrinsic parameters were in line with experimentally measured observed data, emphasizing a stronger correlation of the intrinsic thermodynamics of inhibitor binding to CA VI and CA VI-mimic (R2 = 0.74) compared to that of CA II and CA VI-mimic (R2 = 0.56; Fig.5B). Furthermore, regression line slopes indicating the comparison of inhibitor binding to CA VI and CA VI-mimic (0.74 for observed affinity, 0.95 for intrinsic affinity) were larger than the corresponding slopes for CA II and CA VI (0.54 for observed affinity, 0.58 for intrinsic affinity), Figure 5. Comparison of logKd values representing observed (A) and intrinsic (B) inhibitor binding affinities towards CA VI-mimic and CA II (red squares) or CA VI-mimic and CA VI (blue squares). Straight line represents a model of equal affinity of inhibitor binding to pairwise proteins. Red and blue dashed lines show linear regression models for inhibitor binding to CA II and CA VI, respectively. R2 values and linear equations are indicated. Experiments were performed by FTSA (pH 7.0, 37 °C). 9 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ thereby indicating a lower difference between the inhibitor binding towards CA VI-mimic and CA VI compared to CA II. The influence of investigated CA II mutations on inhibitor binding thermodynamics was further analyzed by calculating the absolute error (AE) values from logarithmic observed or intrinsic Kds of inhibitor binding to CA VI-mimic, CA II, and CA VI. According to the observed thermodynamics, binding affinities of only 12 inhibitors out of 46 tested compounds towards CA VI-mimic was more similar to CA II than CA VI. For the intrinsic data, only 6 compounds were identified as CA VI-mimic binders with the affinity more alike CA II compared to CA VI. Moreover, mean absolute errors (MAEs) as the averages for each AE were also evaluated. MAEs of AEobs,CA II and AEintr,CA II were equal to 1.1, while MAEs of AEobs,CA VI and AEintr,CA VI were significantly smaller, 0.47 and 0.41, respectively. Therefore, CA VI-mimic designed via site-directed mutagenesis from CA II was characterized to be a proper model of CA VI for observed and intrinsic inhibitor binding reactions. Differences in inhibitor binding affinities are due different binding modes as determined by crystallographic analysis of CA II and CA VI-mimic. Despite numerous attempts, the crystal structures of recombinant CA VI complexes with sulfonamide-based inhibitors were not obtained by soaking. Even though CA VI crystals survived soaking procedure, crystals did not contain the clear electron densities of inhibitors. The co-crystallization of CA VI protein with several inhibitors failed, as we did not obtain any crystals suitable for X-ray diffraction experiment. Most likely, CA VI complexes with sulfonamide-based inhibitors cannot be crystallized using crystallization conditions that are effective for the unbound CA VI protein. To structurally investigate the binding of benzenesulfonamides with CA VI, we have engineered CA VI-mimic and applied in crystallographic studies. We have solved crystal structures of CA VI-mimic complexes with three inhibitors (Fig.S2): EZA (PDB ID: 6QL2), inhibitor 14 (PDB ID: 6QL1), and 25 (PDB ID: 6QL3). These complexes were compared with the corresponding complexes composed of CA II and same ligands (EZA (PDB IDs: 3CAJ (X-ray), 6BCC (neutron diffraction)), inhibitor 14 (PDB ID: 4HT0), and 25 (4PYY)). The space groups and unit cell parameters of CA VI-mimic crystals were similar to those of CA II (Table3). There was one unique protein-ligand complex in the asymmetric unit. CA VI-mimic binding pocket was found to be similar to CA VI according to crystallographic studies (Fig.1D) followed by thermodynamic analysis. For this reason, the insights into the compound binding mode to CA VI-mimic are likely to be valid for analyzing the ligand binding data to CA VI. The comparison of binding mode of inhibitor 14 in the active sites of CA VI-mimic and CA II is shown in Fig.6A. Inhibitor 14 in the active site of CA VI-mimic had two alternative binding modes characterized by different positions of the fluorinated ring: the ring was either lodged between Leu198 and Thr200 side chains (colored cyan in Fig.6A), or located in the hydrophilic part of active site (colored blue). On the other hand, in the active site of CA II we had only one position of fluorinated ring – between Leu198 and Thr200. It looks like the replacement of Phe130 in CA II with tyrosine in CA VI-mimic enabled additional position of fluorinated ring of ligand due to a steric collision between the fluorine atom of fluorinated ring and the oxygen atom of Tyr130 side chain (the close contact found in the structure was 2.5 Å). The alternative position of the fluorinated ring of compound 14 in the active site of CA VI-mimic probably was available only due to spatial fluctuations of Tyr130 side chain. Also, due to a significantly larger size and the hydrophilicity of the side chain of Gln134 in CA VI-mimic compared to Val134 in CA II, the hydrophobic dimethylpyrimidine tail of inhibitor 14 was repelled in CA VI-mimic (see para-group of the cyan ligand, Fig.6A). Therefore, the change of size and the hydrophobicity/hydrophilicity of the residues 130 and 134 upon mutation could be rationalized as the main causes for the relatively significant difference in the binding affinities: inhibitor 14 bound to CA II 40-fold better than to CA VI-mimic (Kd_intr values were 0.37 nM and 15 nM for CA II and CA VI-mimic, respectively; Table2). The model compound EZA was bound similarly in active sites of CA II and CA VI-mimic (Fig.6B). Some discrepancy was present only in the positions of highly flexible ethoxy moiety. The aliphatic-aromatic interactions between the methyl group of Leu198 and the first ring of EZA was present in both cases. The larger side chain of Tyr130 slightly changed the position of EZA aromatic ring in CA VI-mimic as compared with CA II due to steric conflicts. It is important also to note the role of the residue 134 interacting with the hydrophobic tail of EZA, similarly to the observation for inhibitor 14 above. The hydrophobic Val134 sidechain in CA II was mutated into hydrophilic Gln134 in CA VI, leading to the worsening of the interaction. Thus, the mutations of residues 130 and 134 were the likely reasons for 15-fold stronger binding of EZA to CA II, as determined by intrinsic thermodynamics (Kd_intr values were 0.073 nM and 1.1 nM for CA II and CA VI-mimic, respectively; Table2). The intrinsic binding parameters of inhibitor 25 towards CA II and CA VI-mimic were comparable (2.2 nM vs 4.5 nM, respectively; Table2). In contrast, the binding modes of the compound found in crystal structures were different in these active sites. In CA II, inhibitor 25 had two alternative conformations: (1) the fluorinated ring located between Leu198 and Thr200, whereas the cyclooctyl ring replaced the side chain of Phe130 (Fig.6C, pink ligand); (2) the fluorinated ring positioned in the hydrophobic part of active site, while the cyclooctyl ring – in the hydrophilic part (Fig.6D, pink ligand). In CA VI-mimic, compound 25 had one well-defined conformation (Fig.6C,D, blue) in which the cyclooctyl ring replaced the Tyr130 side chain, whereas the fluorinated ring occupied the hydrophilic part of active site. We can explain the presence of the single conformation of inhibitor 25 bound to CA VI-mimic. It seems that the mutation in position 67 (asparagine to glutamine) allows the position of fluorinated ring in hydrophilic part of active site when the para-substituent of ligand does not have sterical collision with side chain of residue 67 (compare side chain conformations of asparagine (CA II) and glutamine (CA VI-mimic), Fig.6C,D). The same position of glutamine is found in the complexes of CA VI-mimic with inhibitor 14 and EZA ligands which means that compound 25 does not influence the position of side chain of residue 67 in CA VI-mimic. The replacement of asparagine (CA II) to glutamine (CA VI-mimic) creates the additional free space in the active site and allows for another binding mode. 16 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ 52. Pinard, M. A., Boone, C. D., Rife, B. D., Supuran, C. T. & McKenna, R. Structural study of interaction between brinzolamide and dorzolamide inhibition of human carbonic anhydrases. Bioorg Med Chem 21, 7210–7215 (2013). 53. Dudutienė, V. et al. Discovery and characterization of novel selective inhibitors of carbonic anhydrase IX. J. Med. Chem. 57, 9435–9446 (2014). 54. Krebs, J. F. & Fierke, C. A. Determinants of catalytic activity and stability of carbonic anhydrase II as revealed by random mutagenesis. J. Biol. Chem. 268, 948–954 (1993). 55. Behravan, G., Jonasson, P., Jonsson, B. H. & Lindskog, S. Structural and functional differences between carbonic anhydrase isoenzymes I and II as studied by site-directed mutagenesis. Eur. J. Biochem. 198, 589–592 (1991). 56. Sjöblom, B., Polentarutti, M. & Djinovic-Carugo, K. Structural study of X-ray induced activation of carbonic anhydrase. Proc. Natl. Acad. Sci. USA 106, 10609–10613 (2009). 57. Krishnamurthy, V. M. et al. Carbonic anhydrase as a model for biophysical and physical-organic studies of proteins and proteinligand binding. Chem. Rev. 108, 946–1051 (2008). 58. Linkuvienė, V. et al. Thermodynamic, kinetic, and structural parameterization of human carbonic anhydrase interactions toward enhanced inhibitor design. Quarterly Reviews of Biophysics 51 (2018). 59. Kovale vsky, A. et al. ‘To Be or Not to Be’ Protonated: Atomic Details of Human Carbonic Anhydrase-Clinical Drug Complexes by Neutron Crystallography and Simulation. Structure 26, 383–390.e3 (2018). 60. Koruza, K. et al. Using neutron crystallography to elucidate the basis of selective inhibition of carbonic anhydrase by saccharin and a derivative. Journal of Structural Biology, https://doi.org/10.1016/j.jsb.2018.12.009 (2019). 61. Pilka, E. S., Kochan, G., Oppermann, U. & Yue, W. W. Crystal structure of the secretory isozyme of mammalian carbonic anhydrases CA VI: implications for biological assembly and inhibitor development. Biochem. Biophys. Res. Commun. 419, 485–489 (2012). 62. Patrikainen, M. S. et al. Identification and characterization of a novel zebrafish (Danio rerio) pentraxin-carbonic anhydrase. PeerJ 5, e4128 (2017). 63. Linkuvienė, V., Krainer, G., Chen, W.-Y. & Matulis, D. Isothermal titration calorimetry for drug design: Precision of the enthalpy and binding constant measurements and comparison of the instruments. Anal. Biochem. 515, 61–64 (2016). 64. Paketurytė, V., Linkuvienė, V., Krainer, G., Chen, W.-Y. & Matulis, D. Repeatability, precision, and accuracy of the enthalpies and Gibbs energies of a protein–ligand binding reaction measured by isothermal titration calorimetry. European Biophysics Journal, https://doi.org/10.1007/s00249-018-1341-z (2018). 65. D udutienė, V. et al. Functionalization of fluorinated benzenesulfonamides and their inhibitory properties toward carbonic anhydrases. ChemMedChem 10, 662–687 (2015). 66. Zubrienė, A. et al. Benzenesulfonamides with benzimidazole moieties as inhibitors of carbonic anhydrases I, II, VII, XII and XIII. J Enzyme Inhib Med Chem 29, 124–131 (2014). 67. Dudutienė, V. et al. 4-Substituted-2,3,5,6-tetrafluorobenzenesulfonamides as inhibitors of carbonic anhydrases I, II, VII, XII, and XIII. Bioorg. Med. Chem. 21, 2093–2106 (2013). 68. Č apkauskaitė, E. et al. Benzenesulfonamides with pyrimidine moiety as inhibitors of human carbonic anhydrases I, II, VI, VII, XII, and XIII. Bioorganic & Medicinal Chemistry 21, 6937–6947 (2013). 69. Pettersen, E. F. et al. UCSF Chimera–a visualization system for exploratory research and analysis. J Comput Chem 25, 1605–1612 (2004). 70. Pei, J., Kim, B.-H. & Grishin, N. V. PROMALS3D: a tool for multiple protein sequence and structure alignments. Nucleic Acids Res 36, 2295–2300 (2008). 71. Beitz, E. TeXshade: shading and labeling of multiple sequence alignments using LaTeX2e. Bioinformatics 16, 135–139 (2000). 72. Cimmperman, P. et al. A quantitative model of thermal stabilization and destabilization of proteins by ligands. Biophys. J. 95, 3222–3231 (2008). 73. Carrigan, P. E., Ballar, P. & Tuzmen, S. Site-directed mutagenesis. Methods Mol. Biol. 700, 107–124 (2011). 74. Morrison, J. F. Kinetics of the reversible inhibition of enzyme-catalysed reactions by tight-binding inhibitors. Biochim. Biophys. Acta 185, 269–286 (1969). 75. Williams, J. W. & Morrison, J. F. The kinetics of reversible tight-binding inhibition. Meth. Enzymol. 63, 437–467 (1979). 76. B aranauskienė, L. et al. Inhibition and binding studies of carbonic anhydrase isozymes I, II and IX with benzimidazo[1,2-c][1,2,3] thiadiazole-7-sulphonamides. J Enzyme Inhib Med Chem 25, 863–870 (2010). 77. Kazlauskas, E. et al. Thermodynamics of Aryl-Dihydroxyphenyl-Thiadiazole Binding to Human Hsp90. PLoS One 7, e36899 (2012). 78. Goldberg, R. N., Kishore, N. & Lennen, R. M. Thermodynamic Quantities for the Ionization Reactions of Buffers. Journal of Physical and Chemical Reference Data 31 (2002). 79. Kabsch, W. XDS. Acta Cryst D 66, 125–132 (2010). 80. Vagin, A. & Teplyakov, A. MOLREP: an Automated Program for Molecular Replacement. J Appl Cryst, J Appl Crystallogr 30, 1022–1025 (1997). 81. Hanwell, M. D. et al. Avogadro: An advanced semantic chemical editor, visualization, and analysis platform. J Cheminform 4, 17 (2012). 82. Vagin, A. A. et al. REFMAC5 dictionary: organization of prior chemical knowledge and guidelines for its use. Acta Crystallogr D Biol Crystallogr 60, 2184–95 (2004). 83. Winn, M. D. et al. Overview of the CCP4 suite and current developments. Acta Crystallogr D Biol Crystallogr 67, 235–242 (2011). 84. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). 85. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255 (1997). 86. Čapkauskaitė, E. et al. Combinatorial Design of Isoform-Selective N-Alkylated Benzimidazole-Based Inhibitors of Carbonic Anhydrases. ChemistrySelect 2, 5360–5371 (2017). Acknowledgements Authors thank Gleb Bourenkov for the help with data collection at P14 EMBL beamline at PETRA III ring of the DESY synchrotron. Theaccess to thebeamline was supported by iNEXT (grant number 653706) funded by the Horizon 2020 program of the European Commission. The study was also supported by the grant S-MIP-17-87 from the Research Council of Lithuania. Author contributions J.K., J.S., V.K., A.S., M.T., S.P. and D.M. participated in the conception or design of the study; J.K. produced recombinant proteins, carried out biophysical assays and analyzed thermodynamic parameters; V.K. performed computational modelling; J.S. carried out enzymatic activity and inhibition measurements; A.S. and E.M. were responsible for X-ray crystallographic analysis; J.K. wrote the first version of manuscript; J.K., V.K., J.S., A.S., E.M., M.T., S.P. and D.M. contributed to manuscript drafting and approved the final version of manuscript. 17 Scientific RepoRtS | (2019) 9:12710 | https://doi.org/10.1038/s41598-019-49094-0 www.nature.com/scientificreports www.nature.com/scientificreports/ Additional Information Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-019-49094-0. Competing Interests: D.M. declares that he has patents and patent applications pending on CA inhibitors. Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. © The Author(s) 2019