Metal ions are often bound to proteins in living organisms and play a crucial role in protein structure stabilization and enzyme catalysis. Metal ions are the smallest protein cofactors bearing individual properties such as electron-acceptor ability, positive charge, flexible coordination sphere, specific affinity and varying valence state. The distribution of metal ions in cells depends on their affinities to proteins and complexing agents such as phosphate buffer. The binding affinity is a critical parameter quantifying the driving force of such interactions, but it may be challenging to determine experimentally.

Many techniques are used to investigate metal–protein interactions, such as mass spectrometry, NMR-spectroscopy, spectrophotometric techniques, isothermal titration calorimetry and others, each having strengths and limitations [1]. Here we emphasize that the affinities of metal ions to proteins measured by fluorescence-based thermal shift assay (FTSA, TSA, also termed differential scanning fluorimetry (DSF) or ThermoFluor). This method often used to characterize the thermal stability of proteins and to obtain the enzyme–inhibitor (protein–ligand) affinity constants based on the ligand thermal stabilization effect on the protein. This technique can determine some of the widest ranges of protein–ligand dissociation constants (Kd), spanning from molar to picomolar [2,3]. We have recently introduced a web-based server to apply complex thermodynamic equations to user-supplied thermal shift data and think that it will expand the use of TSA for protein-ligand affinity measurements [4]. In our opinion, this technique yields the most accurate protein-ligand Kd values.

Carbonic anhydrase (CA) enzyme was discovered >80 years ago [5] and still is one of the most investigated model proteins. Human CA isozymes belong to α family and there are at least 7 families of structurally variable enzymes performing the same function. It is responsible for catalyzing the reversible conversion of carbon dioxide to bicarbonate. Twelve catalytically active CA isozymes are found in human body, variably expressed in all cell types [6]. The isozymes CAIX and CAXII are associated with cancer growth [7,8]. Human CAI and CAII are the main off-target isozymes as they are widespread in red blood cells and many other tissues. Drugs targeting CAIX and CAXII should bind weakly to CAI and CAII to avoid toxic side effects. All human catalytically active α-CA isozymes have a conical 15 Å diameter active site cleft. The active site is very similar among CA isozymes and therefore it is difficult to design compounds that bind only CAIX. The centrally located zinc ion is coordinated by three histidine residues His94, His96, and His119 (amino acids numerated according to CAII), and a water/hydroxide molecule.

Two cytosolic human CA isozymes CAI and CAII, and bovine CAII isozyme are often used as model proteins for investigating protein stability and folding, the role of metal ions, and protein-ligand interaction. The simplicity and stability of the protein structure determines their wide application. These enzymes are monomeric, single chain, medium molecular weight proteins bearing no disulfide bridges. They are relatively inexpensive and easy to make using bacterial protein expression systems. CAI and CAII are well characterized by biochemical and biophysical methods and the enzymatic mechanism of substrate hydrolysis has been studied extensively [9].

The zinc ion may be removed from the native CA resulting in the apo‑carbonic anhydrase (apoCA) protein that fully loses enzymatic activity. However, the catalytic activity is fully reconstituted in proportion to the saturation of added Zn2+ [10,11]. Other divalent metal ions may be introduced into the active site instead of zinc. The binding of divalent metals to the apoCA active site occurs at different rates and with varying affinities. The affinity of divalent metal ions for wild-type CAII follows the trend suggested by the Irving-Williams series in the order of Hg2+ > > Cu2+ > Zn2+ > Cd2+, Ni2+ > Co2+ > Mn2+ [11]. However, deviations from this affinity trend can occur for some variants of CAII such as H119Q CAII [12]. Reactivation of apoCA by Zn2+ ion occurs instantaneously, while the introduction of a Co2+ ion into the bovine apoCA takes several hours depending on Co2+ concentration [10]. The Co2+ binds much more slowly than Zn2+ as it has to change its octahedral geometry in the solvent to the tetrahedral coordination at the active site of the protein [13].

The picomolar binding affinity of zinc ion for apoCAII was determined by equilibrium dialysis (Kb = 1.3 × 1012 M−1) [14]. However, the results by the isothermal titration calorimetry were several orders of magnitude lower (the observed Kb ≈ 2 × 106 M−1, the intrinsic pH-independent Kb ≈ 2 × 109 M−1) [15] and did not match the data obtained by equilibrium dialysis. Zinc binding to apoCAII was favorably driven by enthalpy (ΔHint = −16 kcal/mol). Unfavorable entropy (-TΔSint = 3.5 kcal/mol) is more difficult to interpret but is thought to be associated with the conformational changes of the His64 residue and stabilization of the side-chain residues of several α-helical regions on the outer surface of the enzyme [15]. Due to different binding affinity results, there is a need to examine the binding of metal ions to the apoCA isozymes by other methods.

Interestingly, the Cu2+ ion binds to two sites on the apoCAII protein. The first copper binding site is in the active site and the second is near the N-terminus involving His-64 and His-4 residues [16,17]. The N-terminal Cu2+ site does not show any catalytic activity. It was demonstrated by circular dichroism and denaturation experiment that copper binding near the N-terminus results in a net unfolding of 6–8 residues. This is surprising as the copper binding constant for this site is high (Kb = 3.6 × 108 M−1) [17] and metal ions usually stabilize proteins by binding with such high affinity.

The catalytic activity of CA isozymes substituted by different divalent metals varies significantly. The Co2+ ion-containing CAII activity is about 50% of the native protein activity containing Zn2+ ion. The Ni2+ ion-replaced CAII showed low, up to 5%, activity level. However, the authors attribute this to possible contamination with the Zn2+ ions. Carbonic anhydrase isozymes that contained Cu2+, Cd2+, Pb2+, Hg2+, Be2+, and Fe3+ ions in the active site exhibited no activity. These results were almost indistinguishable between human and bovine CAII isozymes [10,11,18].

The structure of the apoCAII is almost identical to the native CAII protein. The position of the atoms did not change after the removal of the zinc ion [[19], [20], [21]]. In addition, the removal of the zinc ion from CAII did not affect the topological folding of the protein nor the network of water molecules in the active site. However, zinc removal affected general electrostatics and stability of the protein and the melting point of apoCAII lowered by 8 °C. The thermal mobility of atoms did not increase at the active apoCAII site but was on the surface of the enzyme. The orientation of proton shuttle residue His64 in the X-ray crystal structures of native CAII was nearly equally occupying the inward and outward positions, whereas in the apoCAII the His64 appeared predominantly in the outward orientation [20]. Interestingly, the 1.8-Å resolution neutron structure of apoCAII showed the His64 residue predominantly in the inward conformation [22].

The crystal structures solved for metal-substituted CAII showed that the metal coordination geometry in the active site may vary. For example, the copper ion in the CAII active site forms a trigonal bipyramidal geometry, allowing an additional bond with water molecule [16,23,24]. Metal coordination geometry and long-range electrostatic effects on restructuring water network are crucial for the catalytic activity of the enzyme. Metals that acquire non-tetrahedral geometry in the active site, such as octahedral (Ni2+) and trigonal bipyramidal (Cu2+), are responsible for the complete inactivity of the enzyme [24,25].

Primary sulfonamide compounds have been used as CA inhibitors for decades. Sulfonamide amino group forms a coordination bond with the Zn(II). Studies with bovine carbonic anhydrase have shown that several sulfonamides bound native CA up to 6 × 103 times stronger than the apoCA [25]. Furthermore, human apoCAII bound sulfonamide inhibitors up to 5 × 104-fold weaker than the native CAII [26]. Thus, the coordination bond with the zinc contributes significant part of the inhibitor–protein binding energy.

Metal-substituted CAs whose metal ions resemble the native CA geometry bind sulfonamides stronger than apoCA. For this reason, sulfonamides bind quite well to cobalt-substituted CA as it acquires tetrahedral coordination geometry like native CA. However, Co-CA has only 50% activity of native CA because after the coordination of the bicarbonate ion, the coordination geometry changes to octahedral, which reduces product release [24,25,27]. Other CAs with the metal ion acquiring octahedral or pyramidal geometries bind sulfonamides with reduced affinity.

Despite the catalytic activity of two metal-substituted CA isozymes, namely, CAI and CAII, was thoroughly explored, the binding of sulfonamide compounds was not extensively studied. Furthermore, there are no data on the effect of metal ion substitution in tumor-associated CAIX activity and the binding properties with sulfonamide compounds.

In this study, we aimed to investigate the affinities of the interaction of metal ions with three recombinant human apoCAs, namely, apoCAI, apoCAII and apoCAIX by applying fluorescent thermal shift assay. We substituted apoCAs with different divalent metal ions and determined the affinities of sulfonamide inhibitors. To the best of our knowledge, this paper is the first study comparing the affinities of inhibitors on the three metal substituted CA isozymes I, II and IX. Our study has shown that apoCAs bind the same metal ion with different affinities and the order of affinities did not fully follow the trend suggested by the Irving-Williams series.

However, most importantly, we demonstrated that the metal-substituted CA isozymes exhibited similar pH profiles of sulfonamide inhibitor binding to native CA. Compound affinities exhibited a maxima at near-neutral pH. The profiles allowed determination of the intrinsic affinities that are independent of pH or buffer. Only the intrinsic affinities should be directly correlated with the structure, not the affinities that are obtained experimentally [[28], [29], [30]]. We assembled the manually curated Protein-Ligand Binding Database (PLBD, https://plbd.org) that currently contains 597 compound intrinsic binding data [31]. Metal substitution demonstrated that the phenomenon is not limited to Zn, but applicable to numerous other metals. Finally, affinities to metal-free apoCAI, apoCAII and apoCAIX enabled dissection of the energy of the coordination bond for each isozyme. This dissection provided energetic contribution of the remaining chemical part of the inhibitor thus helping CAIX-selective drug design.