Novel thiazolidinone-containing compounds, without the well-known sulphonamide zinc-binding group acting as human carbonic anhydrase IX inhibitors

Abstract A small collection of 26 structurally novel thiazolidinone-containing compounds, without the well-known sulphonamide zinc-binding group, were synthesised and tested in enzyme inhibition assays against the tumour-associated hCA IX enzyme. Inhibition constants in the lower micromolar region (KI < 25 μM) have been measured for 17 of the 26 compounds. Even though the KI values are relatively weak, the fact that they do not contain a sulphonamide moiety suggests that these compounds do not interact with the active site zinc ion. Therefore, docking studies and molecular dynamics simulations have been performed to suggest binding poses for these structurally novel inhibitors.


Introduction
Carbonic anhydrases (CAs; EC 4.2.1.1) are a structurally diverse family of metallo-enzymes that catalyse the reversible hydration of carbon dioxide to bicarbonate and protons 1 . Due to this physiologically important reaction, CAs are important in controlling pH values and supplying bicarbonate ions for various processes. The human CA isoform IX (hCA IX), which belongs to the a-subfamily of CAs (aCAs), is expressed in the stomach and peritoneal lining. More interestingly, hCA IX is upregulated in many solid hypoxic tumours and it helps the tumour cell to function in an acidic environment [1][2][3][4][5] . This provides an advantage over healthy cells in the tumour microenvironment. As such, the development of selective hCA IX inhibitors may provide novel compounds for the cancer chemotherapy.
Many inhibitors of hCA IX has been synthesised and tested in the last decade, including substituted-phenylacetamido aromatic sulphonamides, open saccharin analogues, probenecid analogues, and isatin analogues [5][6][7][8][9][10][11] . Many CA inhibitors (including hCA IX inhibitors) bind to the active site zinc ion via a so called zinc-binding group (ZBG), which is a sulphonamide in most cases, and thereby block the reversible hydration of carbon dioxide 12 . One such example is compound SLC-0111, a sulphonamide-containing potent CA inhibitor, which is currently in clinical trials [13][14][15][16][17][18][19][20][21] . Other binding modes of inhibitors are also possible, including binding to allosteric sites and binding to the active site but without a direct interaction with the zinc ion 12 .
In this work, we synthesised and tested 26 structurally novel thiazolidinone-containing compounds, without the well-known sulphonamide ZBG, in enzyme inhibition assays against hCA IX. Inhibition constants were obtained in the lower micromolar region. Subsequently, docking studies and molecular dynamics simulations were performed to suggest binding poses for these compounds. until crystallisation. The crude product thus obtained was recrystallised from ethanol.

hCA IX enzyme inhibition assays
A stopped-flow instrument (SX.18 MV-R Applied Photophysics model) was used for assaying the CA-catalysed CO 2 hydration activity 35 . Inhibitor and enzyme were pre-incubated for 15 min for allowing the complete formation of the enzyme-inhibitor adduct. IC 50 values were obtained from dose response curves working at seven different concentrations of test compound (from 0.1 nM to 50 mM), by fitting the curves using PRISM (www.graphpad.com) and non-linear least squares methods, the obtained values representing the mean of at least three different determinations. The inhibition constants (K I ) were derived from the IC 50 values by using the Cheng-Prusoff equation, as follows: represents the CO 2 concentration at which the measurement was carried out, and Km the concentration of substrate at which the enzyme activity is at half maximal. All enzymes used were recombinant, produced in Escherichia coli as reported earlier 36 . The concentration of hCA IX used in the assay was 12.1 nM.

Preparation of compound series 3 and 4 for docking studies
The three-dimensional structures of all ligands, including all possible stereoisomers, were prepared (MOE software package, v2016.08, Chemical Computing Group, Inc, Montreal, Canada).
Afterwards, the ligands were energy minimised using the MMFF94x force field.

Preparation of hCA IX structure
The crystal structure of the catalytic domain of hCA IX in complex with acetazolamide (pdb: 3iai; 2.20 Å) was obtained from the RCSB protein databank. The structure was protonated using the Protonate3D tool 37 of the MOE software package and subsequently the obtained structure was energy-minimised using the AMBER12:EHT force field. The protein atoms of subunit A and the corresponding active site zinc ion were retained and all other atoms were omitted.

Docking protocol
The GOLD Suite software package (v5.6.1, CCDC, Cambridge, UK) and the ChemScore scoring function were used to dock the compounds into the hCA IX structures (25 dockings per ligand). The binding pocket was defined as all residues within 13 Å of a centroid corresponding to the location of the acetazolamide C2 atom. The best three docked poses were retained for each ligand.

Molecular dynamics simulations
All molecular dynamics simulations were performed using the NAMD software package (v2.12, Theoretical and Computational Biophysics group, NIH Center for Macromolecular Modeling and Bioinformatics, The Beckman Institute, University of Illinois at Urbana-Champaign) 38 . The select docked poses (ligand-enzyme complexes) were first placed into the centre of a box with periodic boundary conditions (minimal distance of 10 Å between protein and boundary). Afterwards, both water molecules (Tip3) and counter ions (NaCl) were added to generate a solvated and neutral system. After a steepest-descent energy minimisation (AMBER12:EHT), the system was first heated from 0 to 300 K during 100 ps followed by an 100 ps equilibration simulation (position restraints on all protein and ligand heavy atoms). Finally, the system was simulated for 1 ns at constant temperature (300 K, Langevin, default values) and pressure (1 bar, Nos e-Hoover Langevin, default values), without any position restrains. The only restraints applied were distance restraints to keep the zinc ion in the correct orientation towards His94, His96, and His119 (distance restraints between Zn and N atom of histidine: 1.8 Å; default settings). The timestep was set to 0.002 fs and all bonds were constrained using the ShakeH algorithm. In the mass spectra of 3c and 4g (M-H) À peaks were observed which confirmed their molecular weights. Further spectral details have already been presented in the "Materials and methods" section.

hCA IX enzyme inhibition studies
A small collection of 26 thiazolidinone compounds have been tested in enzyme inhibition assays against hCA IX and 17 compounds were identified with an inhibition constant (K I ) lower than 25 lM (Table 1). A subset of five compounds show K I values lower than 10 lM (3d: 1.4 lM; 3f: 1.2 lM; 3o: 1.1 lM; 4c: 2.4 lM; 4e: 7.7 lM). The measured K I values for these ligands are at least 44fold lower compared to the K I value of the potent hCA inhibitor acetazolamide (25 nM). However, it should be noted that these compounds do not contain a sulphonamide moiety that functions as a ZBG, which is present in many inhibitors of hCAs.
It is difficult to draw conclusive structure-activity relationships from these data since the K I values are relatively close to each other.
Compounds 3n and 4j show K I values of 22.5 and 21.3 lM, respectively. These two compounds have nitro groups on the para positions of their phenyl moieties. This may indicate the presence of a direct interaction between the nitro oxygen atoms and the zinc ion of the hCA IX active site. The other compounds may form interactions with the zinc ion via their hydroxyl or carbonyl groups.
Docking studies and molecular dynamics simulations have been performed to suggest possible binding poses for these compounds.
Docking studies into the active site of hCA IX All 26 thiazolidinone-containing molecules were docked into the active site of hCA IX, with and without a zinc-bound water molecule.
Docking studies without a zinc-bound water molecule suggests that only compounds with para-nitro substituents (compounds 3n and 4j; pose 1) may interact directly with the active site zinc ion ( Figure 1). The carbonyl group of the thiazolidinone ring forms a hydrogen bond with the side chain of Gln92, while the other carbonyl group forms a hydrogen bond to the side chain of Trp5. The unsubstituted phenyl group forms hydrophobic interactions with the side chain of Pro202. An additional hydrogen bond is formed between the hydroxyl group (R isomer) with the backbone of Ser3. In the S isomer, this hydroxyl group may interact with the backbone carbonyl group of Pro201 or point to the solvent. The methyl group at the R2 position (R isomer) may form a hydrophobic interaction with the side chain of Val131.
The docking studies with a zinc-bound water molecule in the active site suggest docked poses in which the ligand interacts directly with the water molecule (pose 2; Figure 2). In pose 2, the unsubstituted phenyl moiety forms a hydrophobic interaction with the side chain of His94 ( Figure 2). The carbonyl group forms a hydrogen bond to the side chain of Gln92 and the substituted phenyl ring is located in a hydrophobic pocket formed by Leu91, Val121, Val131, Leu135, Leu141, and Leu198. The ligands sulphur atom may form interactions with the backbone carbonyl of Pro201. The thiazolidinone carbonyl group is solvent exposed. The hydroxyl group of the ligand in the R isomer (turquoise, Figure 2) forms a direct hydrogen bond with the zinc-bound water molecule, while this is not observed for the hydroxyl group in the S isomer (purple, Figure 2). However, the distance and orientation of this hydroxyl group are such that it might be possible after slight conformational changes in this area.
Docking pose 2, with the hydroxyl group in either S or R orientation, was observed for all compounds 3 and 4, as the hydrophobic cavity formed by Leu91, Val121, Val131, Leu135, Leu141, and Leu198 seems to be large enough to accept the substituted phenyl group of the ligand.

Molecular dynamics simulations
Molecular dynamics simulations have been performed to investigate whether the suggested docked poses 1 and 2 would be stable during molecular dynamics simulations. To this end, the ligand-enzyme complex was first placed into a box with periodic boundary conditions. Afterwards, both water molecules and counter ions (NaCl) were added to generate a solvated and neutral system. After a steepest-descent energy minimisation (AMBER12:EHT), the system was first heated from 0 to 300 K during 100 ps followed by a 100-ps equilibration simulation (position restraints on   all protein and ligand heavy atoms). Finally, the system was simulated for 1 ns at constant temperature (300 K) and pressure (1 bar), without any position restrains. The only restraints applied were dihedral and distance restraints to keep the zinc ion in the correct orientation towards His94, His96, and His119.
MD simulation of hCA IX-4j complex (pose 1) The simulation of compound 4j (with hydroxyl in R isomer) in the active site of hCA IX (docked pose 1; purple; Figure 1) indicates that the interaction between the oxygen atom of the nitro group and the zinc ion is not stable. The direct interaction is lost and a water molecule comes in between ( Figure 3). The newly obtained pose of compound 4j is remarkably similar to its original docked pose (0 ps), with the exception of the bridging water molecule and the absence of an interaction with Gln92 ( Figure 3). Hydrogen bonds are being formed with the side chain of Trp5 (84.3% of 1 ns simulation) and Thr199 (81.8% of 1 ns simulation) and the backbone of Ser3 (65.4% of 1 ns simulation). A similar result has been observed for the molecular dynamics simulation of compound 4j with the hydroxyl group in the S isomer (turquoise; Figure 1). Again the nitro group forms interactions with a zinc-bound water molecule and the backbone of Thr199 (89.2% of 1 ns simulation) or Thr200 (9.2% of 1 ns simulation). In addition, hydrogen bonds are formed with the backbone of Ser3 (12.2% of 1 ns simulation) and the side chain of Trp5 (86.1% of 1 ns simulation).
MD simulation of hCA IX-3d complex (pose 2) The simulation of compound 3d (R) reveals that the hydrogen bond between the hydroxyl group of the ligand (R isomer) and the zinc-bound water molecule is not stable and this bond is lost very early in the simulation (Figure 4). New hydrogen bonds are formed between the ligand and the backbone of Thr199 (69.3% of 1 ns simulation) or the side chain of Thr200 (45.0% of 1 ns simulation). An additional hydrogen bond is formed between the ligand and the side chain of Gln92 (16.0% of 1 ns simulation). One of the hydrogen atoms of the zinc-bound water molecule projects to the centroid of the unsubstituted phenyl moiety of the ligand. In addition, this unsubstituted phenyl group forms hydrophobic interactions with His94. The substituted phenyl group is located in the hydrophobic pocket formed by the side chains of Leu91, Val121, Val131, and Leu141. The two carbonyl groups and the sulphur atom form hydrogen bonds with other water molecules that have entered the active site.
The start conformation for the simulation of compound 3d (hydroxyl in S isomer; Figure 2, purple) indicates that no hydrogen bond between the ligand and the zinc-bound water molecule is formed. The hydrogen bond with the side chain of Gln92 seems to be stable (97.7% of 1 ns simulation), whereas the interaction between the ligand sulphur and the backbone carbonyl of Pro201 is less frequently observed (4.6% of 1 ns simulation). Occasionally, the ligand forms a hydrogen bond with the side chain of Thr200 (15.7% of 1 ns simulation). The substituted phenyl group of the ligand is located in the hydrophobic pocket lined by Leu91, Val121, Val131, and Leu141 during the whole simulation.

Conclusions
Thiazolidinone-containing compounds, without the well-known sulphonamide ZBG, inhibit the tumour-associated hCA IX enzyme with K I values in the lower micromolar range. Docking studies in combination with molecular dynamics simulations suggested several binding poses for these compounds, in which the ligand form no direct interactions with the active site zinc ion. These structurally novel compounds, without the well-known sulphonamide ZBG, may be interesting candidates as novel classes as CAIs.

Disclosure statement
The authors report no conflict of interest.