Molecular Recognition of the Catalytic Zinc(II) Ion in MMP-13: Structure-Based Evolution of an Allosteric Inhibitor to Dual Binding Mode Inhibitors with Improved Lipophilic Ligand Efficiencies

Matrix metalloproteinases (MMPs) are a class of zinc dependent endopeptidases which play a crucial role in a multitude of severe diseases such as cancer and osteoarthritis. We employed MMP-13 as the target enzyme for the structure-based design and synthesis of inhibitors able to recognize the catalytic zinc ion in addition to an allosteric binding site in order to increase the affinity of the ligand. Guided by molecular modeling, we optimized an initial allosteric inhibitor by addition of linker fragments and weak zinc binders for recognition of the catalytic center. Furthermore we improved the lipophilic ligand efficiency (LLE) of the initial inhibitor by adding appropriate zinc binding fragments to lower the clogP values of the inhibitors, while maintaining their potency. All synthesized inhibitors showed elevated affinity compared to the initial hit, also most of the novel inhibitors displayed better LLE. Derivatives with carboxylic acids as the zinc binding fragments turned out to be the most potent inhibitors (compound 3 (ZHAWOC5077): IC50 = 134 nM) whereas acyl sulfonamides showed the best lipophilic ligand efficiencies (compound 18 (ZHAWOC5135): LLE = 2.91).

Initial success in developing inhibitors against MMP-13 was achieved over a decade ago by modifying the natural substrate close to the scissile amide bond. By implementing structural elements that cannot be hydrolyzed by the enzyme, peptidomimetics were obtained [13]. Potent inhibitors resulted by adding strong zinc binding groups such as hydroxamic acid to interact with the catalytic domain [14,15]. Those inhibitors lacked sufficient selectivity due to high structural similarity among the various MMPs. Because of toxicity and severe side effects arising from scarce selectivity no compound with strong zinc chelating groups such as hydroxamic acids has survived clinical trials so far [16,17]. In this article we present a structure-based design approach to modify the allosteric MMP-13 inhibitor 1 to increase its potency by improving the molecular recognition between the enzyme and its ligand. This could be achieved by the introduction of a weak zinc binding fragment in order to interact with the catalytic center. Furthermore we enhanced the lipophilic ligand efficiency of the novel compounds compared to the allosteric inhibitor.

Molecular Modeling
In the first step of the design process we had a close look on the binding mode of 1 in the allosteric binding pocket of MMP-13. We used the crystal structure PDB 2OW9 (Figure 2a) [24] to dock the molecule into the cavity defining a pharmacophore query to maintain the three hydrogen bonds to the backbone amino acids Thr224, Thr226 and Met232, formed by the oxygen atoms of the phthalimide scaffold ( Figure 2b). It turned out that a zinc binding group could be attached to the molecule, favourably in meta position, to interact with the catalytic zinc ion ( Figure 2c). As the distance that has to be overcome added up to more than 5 Å (Figure 2d) we decided to implement a linker consisting of four and more heavy atoms between the aromatic ring system and the chelating group.  At the meta position of the aromatic ring an oxygen was added for synthetic reasons, followed by aliphatic linker chains between 3-6 carbon atoms in length as well as a terminal hydroxy group. The alcohol head group was implemented as a generic zinc chelator of small size to evaluate the optimal linker length ( Figure 3). We decided to focus on chain lengths of 4 and 5 carbon atoms ( Figure 3, yellow and magenta) as they showed lower root mean square deviation (RMSD) values (1.16 yellow and 1.35 magenta) than the other two linkers (1.56 green and 2.57 cyan). This indicates better quality of the docking poses [28]. In addition, visual inspection of the docking poses indicated that chain lengths of 4 and 5 carbon atoms seemed to cover the distance more efficiently than the other two linkers. Subsequently we examined different head groups as potential zinc chelators. Therefore we modified structure 2 by implementing a carboxylic acid (3), an acyl sulfonamide (4) and an amide (5) to obtain a set of ligands for which the recognition of the catalytic zinc should increase their affinity to the protein and therefore enhance the inhibitors potency ( Figure 4) [29,30]. For the purpose of this evaluation we executed docking experiments applying the identical pharmacophore as for the previously docked structure 1 (Figure 5a-d). All of the four examined weak zinc binding head groups showed the postulated molecular recognition of the metal ion. The alcohol and the acid as well as the amide exhibit monodentate binding, whereas the acyl sulfonamide acts as a bidentate chelator. According to our modeling experiments, the carboxylic acid interacts as a monodentate ligand. This corresponds well with the reported binding mode of carboxylic acids in the active site of MMP-13 ( Figure 6, crystal structure PDB 1ZTQ [31]). The designed compounds 2-5 displayed better scoring values (in the range of´18) compared to 1 (´12.8). In consideration of the modeling results described above, we decided to synthesize a small library of this type of molecules.

Chemistry
Driven by the positive results of our computational design approach towards molecular recognition of the catalytic zinc ion, we synthesized compounds 2-5 according to the synthetic route summarized in Figure 7. The benzylated aminophthalimide derivative 6 was prepared according to a previously described procedure [27]. The acid 9 was obtained by saponification of 8 which was prepared by the nucleophilic substitution of 7 on benzyl-n-bromoalkylether. An amide coupling between the alkylated aminophthalimide 6 and the acid 9 led to compound 10 which could be hydrogenated to obtain the alcohol 2. By subsequent oxidation of the alcohol 2 the acid 3 was formed and finally converted to the compounds 4 and 5 by amide bond formation.

Biological Assays
Consecutively the synthesized compounds were tested for their inhibition activity against MMP-13 in in vitro assays. The values are averaged over triplicate determinations (Table 1). The carboxylic acids 3 and 16 turned out to be the most potent inhibitors within this series (IC 50 : 134 nM (3), 280 nM (16)). The acyl sulfonamides 4 and 18 show some of the best lipophilic ligand efficiencies (LLE: 2.53 (4), 2.91 (18)).

Discussion
We modified the allosteric inhibitor 1 of MMP-13 to increase its potency against the target enzyme and to elevate the LLE of the initial inhibitor. Increased potency could be achieved by the addition of a zinc binding fragment to the inhibitor. The carboxylic acid turned out to be the most potent option for the molecular recognition of the catalytic zinc(II) ion. The new inhibitors combine allosteric interaction with the recognition of the catalytic center to yield dual binding mode inhibitors. Compound 3 showed the highest affinity with an IC 50 value of 134 nM. Increased LLE values were achieved by the implementation of acyl sulfonamide fragments to 1 which lowered the clogP value of the respective inhibitor. By comparing the acyl sulfonamide 18 with its carboxylic acid counterpart 15 one observes that the LLE improves from 2.16 to 2.91, which is in the range of lead compounds, while the IC 50 values are comparable. Therefore acyl sulfonamides are good bioisosteres for substituting carboxylic acids as zinc binding functional groups if an increase of the LLE of MMP inhibitors is intended.
The results of the series with carboxylic acids as the zinc binding group clearly show that the selection of excessively short linkers leads to a loss in potency. Also linkers with six heavy atoms between the allosteric fragment and the zinc-recognizing acid cause diminished inhibition, leading to an optimal linker length of five heavy atoms. These findings harmonize with our modeling results as we observed better geometrical matches for linkers with medium lengths.
It should be mentioned that the critical observation of the generated docking poses with respect to binding angles and torsions was absolutely mandatory for the successful outcome of this study. The corresponding scoring values can only be compared for sets of molecules with comparable binding modes, but even then it is safer to examine the poses visually instead of using the scoring values for prioritizing compounds for synthesis.

General
All NMR spectra were recorded on a Bruker AVANCE III HD 500 One Bay spectrometer with a magnetic field of 11.75 T. For 1 H NMR spectra a frequency of 500 MHz resulted. Chemical shifts are reported in ppm from tetramethylsilane as internal standard. Data are reported as follows: chemical shift, integration, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, quint. = quintet, br. = broad, m = multiplet), coupling constants (Hz). For 13 C NMR spectra a frequency of 125 MHz resulted. Chemical shifts are reported in ppm from tetramethylsilane as internal standard. The multiplicities of the signals were determined by DEPT Measurements. High-resolution mass spectrometry was performed on an Agilent Technologies 6530 Q-TOF. NMR spectra, HRMS spectra and IC 50 curves can be found in the supplementary materials.

In Vitro Assays
IC 50 values were determined at Reaction Biology Corporation, Malvern, PA, USA in triplicates using 10 concentrations starting at 10 µM with 3 fold dilution. The substrate used for the determinations was the (5-FAM/QXLTM) FRET peptide. The buffer consisted of 50 mM HEPES at pH 7.5 with 10 mM CaCl 2 and 0.01% Brij-35. 0.1 mg/mL BSA was added before use. As a control inhibitor GM6001 was used.

Conclusions
In conclusion, we could employ molecular recognition for evolving an allosteric inhibitor through the addition of properly arranged weak zinc binding fragments to obtain an inhibitor with dual binding mode. In addition, the lipophilic ligand efficiency could be improved to lead-like level.
For the definite confirmation of the binding mode, a co-crystal structure of the complex is required. This work is currently ongoing in our lab.