Discovery of Mycobacterium tuberculosis Protein Tyrosine Phosphatase B (PtpB) Inhibitors from Natural Products

Protein tyrosine phosphatase B (PtpB) is one of the virulence factors secreted into the host cell by Mycobacterium tuberculosis. PtpB attenuates host immune defenses by interfering with signal transduction pathways in macrophages and, therefore, it is considered a promising target for the development of novel anti-tuberculosis drugs. Here we report the discovery of natural compound inhibitors of PtpB among an in house library of more than 800 natural substances by means of a multidisciplinary approach, mixing in silico screening with enzymatic and kinetics studies and MS assays. Six natural compounds proved to inhibit PtpB at low micromolar concentrations (< 30 µM) with Kuwanol E being the most potent with K i = 1.6 ± 0.1 µM. To the best of our knowledge, Kuwanol E is the most potent natural compound PtpB inhibitor reported so far, as well as it is the first non-peptidic PtpB inhibitor discovered from natural sources. Compounds herein identified may inspire the design of novel specific PtpB inhibitors.


Introduction
Tuberculosis (TB) kills nearly 2 million people annually. The World Health Organization (WHO) declared TB as a global health emergency, which highlights the importance of TB as a major threat to humans [1]. Drug resistance and patient noncompliance are two key factors that affect the success rate of conventional treatments against TB. Therefore, there is an urgent need to identify novel therapeutic targets for TB treatment as well as new drugs that could act on them.
In the last decade, exoenzymes protein tyrosine phosphatase A (PtpA) and B (PtpB) have emerged as promising therapeutic targets to discover new anti-TB agents [2][3][4][5]. These enzymes are secreted into the host cell by Mycobacterium tuberculosis (Mtb) and attenuate host immune defenses by interfering with the host signaling pathways [6,7]. Thereby, PtpA and PtpB inhibition by small molecules could impact Mtb survival in the host and open the way for the development of innovative therapeutic strategies. Particularly, the localization outside of the mycobacterial cell wall, which is difficult to penetrate, renders these enzymes attractive drug targets.
In previous works we have investigated the inhibitory activity of natural compounds analogues toward PtpA and PtpB from Mtb. In particular, we have first identified potent PtpA inhibitors (IC 50 = 8.4 -53.7 μM) by screening a series of naphthylchalcones against this enzyme [8]. Subsequently, we showed that these chalcones inhibit PtpA by means of a competitive and selective mechanism of action (K i ranging from 5 to 21 µM) as well as are endowed with a significant inhibitory activity towards Mtb growth in infected macrophages [9]. We have also demonstrated the inhibitory properties of synthetic sulfonyl-hydrazones against PtpB, identified as competitive inhibitors with K i values between 2.5 and 15 μM [10]. In our last work, a hundred synthetic chalcones have been investigated for their activities against PtpA and PtpB, and six presented competitive mechanism of action with K i values between 8 and 29 μM [11].
In light of recent advances in understanding the pathological involvement of these phosphatases in Mtb growth and proliferation [12], and following our research interest in modulating these enzymes, here we focused on the discovery and characterization of natural compounds as PtpB inhibitors.
In vivo studies performed with activated macrophages of guinea pigs have shown that gene inactivation of this enzyme provoked accelerated mycobacterial cell death after macrophage invasion [13]. More recently, Zhou and co-workers proposed that PtpB promotes mycobacterial survival in vitro by inhibiting extracellular signal-regulated kinase 1/2 (ERK1/2) and p38 pathways and increasing the phosphorylation of Akt, resulting in reduced production of interleukin-6 (IL-6) and decreased apoptotic activity, respectively [14]. Alber and coworkers have synthesized a strong, competitive and selective PtpB inhibitor, namely OMTS [(oxalylamino-methylene)thiophene sulfonamide] showing an IC 50 of 0.44 μM, and solved the three-dimensional structure of the PtpB-inhibitor complex by means of X-ray crystallography [15]. Other groups also have successfully identified inhibitors of PtpB: indole derivatives with selectivity indexes up to 100 [16], cyclic hexapeptides from cyanobacterium Tychonema sp. with IC 50 around 8.0 µM [17], an isoxazole with K i value of 0.22 μM [18], selective indolin-2on-3-spirothiazolidinones with IC 50 values of 35.5 to 1.2 µM [19] and, recently, benzofurans with sub-micromolar inhibitory activity [20]. Based on these evidences, PtpB has emerged as an important target for anti-TB pharmacological intervention and new inhibitors are in high demand [21].
The screening of natural compounds libraries is a consolidated strategy in drug discovery, which employs the criteria of biological prevalidation and relevance to nature [22]. Natural products have long been recognized as an important source of therapeutically effective agents [23], also because they embody rigid, non-flat three dimensional structures which may positively influence the probability of clinical success of a drug [24]. Indeed, natural products can offer unprecedented opportunities for finding novel hits or leads against a wide range of biological targets. In previous works [8,9,11], we were pioneers in testing libraries of chalcones to find PtpA and PtpB inhibitors, underlining that screening natural products libraries may fuel the discovery of bioactive molecules. Moreover, in silico screening is a widely appreciated and reliable tool for prioritizing small molecules for biological testing. Accordingly, in this work we screened in silico an in house library of natural compounds by means of a structure-based approach composed of molecular docking, rescoring and visual inspection to prioritize few natural compounds as possible PtpB inhibitors that were subsequently assayed in vitro. Results of inhibition studies, kinetic measurements and mass spectrometry (MS) assays allowed elucidate their mechanism of action.

Molecular modeling
Coordinates of the target receptor for structure-based molecular modeling were retrieved from the Protein Data Bank, under the accession code PDB ID: 2OZ5 [15]. This structure has been solved by X-ray crystallography at 2.00 Å resolution and represents the only PtpB structure in complex with a small molecular inhibitor which was available at the time of experiments. Coordinates of the protein-ligand complex were energy minimized with Amber11 in a box of explicit TIP3P water molecules (10 Å buffer), by using the ff99bsc0 force field for the protein and the GAFF force field for the OMTS ligand [39,40]. The MM-GBSA python script was used for calculating the delta energy of ligands binding to PtpB, by following a procedure already described [41].
Coordinates of the OMTS were then manually removed and such generated protein structure was used as target receptor during docking calculations with GOLD Program 4.1.2 [42]. The binding site was centered on the CD2 atom of Phe161 and included all PtpB atoms within 20 Å. The highest accuracy of the GOLD genetic algorithm (200%) was used for docking the unique library. The GoldScore function was used.
The GRID program was used for probing the potential energy of interaction of the OH2 probe atom within the catalytic site of PtpB [43,44]. The Grid center was placed in correspondence of Tyr125 and was of 17.04, 14.40, 16.50 Å (x, y and z axes). Grid maps were then visualized with Ligandscout 3.0 from Inte:ligand [45].

Measurement of PtpB inhibition (IC 50 )
The phosphatase assays were carried out similarly as previously described [8,11], in 96-well plates containing 8 µL of diluted compounds in DMSO (final concentration 4%), 20 mM imidazole pH 7.0, 160 µL of MilliQ water and 2 µL of recombinant PtpB (70 ng/µL, in Buffer D -20 mM Tris-HCl pH 8.0, 50 mM NaCl, 5 mM EDTA, 20% glycerol and 5 mM DTT). The mixture was maintained for 10 min at 37 °C, followed by addition of 20 mM p-nitrophenyl phosphate (pNPP), in order to start the reaction. The enzyme hydrolyzes the substrate (pNPP), releasing p-nitrophenol. The absorbance was measured with a UV-VIS spectrophotometer (TECAN Magellan Infinite M200) for 10 min at 37 °C (at 410 nm with readings every 1 min). Negative controls were performed in the absence of enzyme or compounds, and positive controls in the presence of enzyme and 4% DMSO. The fraction of residual activity was calculated as the difference in absorbance observed at 2 and 7 minutes of enzyme reaction, obtained by the average of two experiments carried out in triplicate. The IC 50 values were determined with increasing concentrations of inhibitor (100 nM − 100 μM) versus percentage of residual activity, which was calculated as the difference between the observed absorbance at 2 and 7 min of enzyme reaction, obtained by the average of three independent experiments carried out in triplicate. The experimental data were analyzed with GraphPad Prism 5.0 and the IC 50 values determined by linear regression. It is important to stress the fact that all compounds are soluble in the assay mixtures at the described experimental conditions.

Enzyme kinetics
To determine the mechanism of inhibition of the compounds, they were screened by the same methodology described before, however, varying concentrations of pNPP (at least seven concentrations ranging from 0.2 and 12.8 mM) for each concentration of compound (at least three concentrations ranging from 1 to 40 μM). The reaction rates were expressed as specific activity of the protein (μmol pNP.min −1 .mg −1 ) and the pNPP concentration in mM. The p-nitrophenol released (1/V) was quantified and analyzed by the Lineweaver−Burk plot
Chemical and physico-chemical features of all compounds were predicted with QikProp [47], to determine reasonable bioavailability as well as drug-like properties. In this respect, MW, LogP, polar surface area (PSA), number of rotatable bonds, hydrogen bond donors (HBD) and acceptors (HBA) were calculated and compared with those of the 95% of commercial drugs. Notably, 92% of compounds have features within the limits set by 95% of commercial drugs, thus emphasizing the drug-likeness of this unique library and its suitability for drug discovery purposes.
After the analysis of possible ionization and tautomeric states at physiological conditions, a sample of 1014 structures from this library was then chosen for docking studies.

Molecular docking
The ability of some docking programs to reproduce the X-ray determined binding conformation of OMTS within the active site of PtpB was preliminary checked (data not shown). The ligand OMTS was self-docked into the binding site of PtpB whose coordinates were retrieved from the X-ray structure (PDB ID: 2OZ5) [15] and the docked conformation corresponding to the lowest free energy (or highest score) provided by each program was selected as the most probable binding pose. While most programs and functions failed to dock correctly OMTS, the GoldScore function implemented in GOLD [42] provided satisfactory results. In particular, the best agreement between computational and X-ray structural data was obtained by self-docking OMTS toward the receptor structure which was previously energy minimized with Amber11 [39], and including conserved water molecules (RMSD between crystal and docking pose = 0.7441 Å, see also Figure S1). The position and number of water molecules retained within the PtpB catalytic site was established by a GRID analysis performed using the OH2 probe atom ( Figure S2). The parameter set used for self-docking OMTS was further used to dock the in house unique library within the catalytic site of PtpB. After docking and visual inspection, top 10% ranking compounds were selected for rescoring.

Rescoring with MM-GBSA and virtual hits selection
It is widely appreciated in computer-aided drug design that rescoring docking poses with a scoring function different to that used in generating docking poses could better describe the ligand binding energy toward a receptor [41,48]. In previous studies we evaluated the capability of the Molecular Mechanics Generalized Born Surface Area (MM-GBSA) method in rescoring docking poses generated with GOLD [41,48]. Similar results were discussed also by other research groups [49]. Here, we used the MM-GBSA method for rescoring docking poses of selected molecules and to predict their delta energy of binding (∆E).
Based on the calculated ∆E and the chemical diversity, fourteen compounds were deemed top priority and selected for biological investigations in vitro (docking score and rescoring energy of virtual hits are reported in Table S1).

Enzymatic assays
The inhibitory activity of the fourteen selected natural compounds towards PtpB was evaluated using the previously described methods [8,11], with a slight modification (see Materials and Methods). Six compounds showed significant inhibition of PtpB with IC 50 < 30µM, two were moderated inhibitors showing IC 50 between 30 and 100 µM (Figure 1), and six not exhibited enzymatic inhibition at 100µM ( Table 1).
The best inhibitory effect of PtpB was achieved by KuwE (IC 50 = 1.9 ± 0.5 µM), a polyhydroxylated Diels-Alder type adduct metabolite isolated from Morus nigra cell cultures [26]. Despite the shortlisted compounds show a significant chemical diversity, the active molecules can be attributed to two distinct molecular scaffolds (Figure 1). Moreover, the comparison of chemical structure of active molecules with their inhibition of PtpB enzymatic activity allowed for delineating rough structureactivity relationships. For instance, comparing anthrone derivatives PirIII (IC 50 = 5.4 ± 0.6 µM) and ∆3 (IC 50 = 26.7 ± 0.6 µM), the substitution of two hydrogen atoms with two hydroxyl groups in only one of the prenyl moieties of ∆3, reduced the PtpB inhibitory activity. Based on docking simulations, prenyl groups of PirIII are nicely docked in proximity of a cluster of hydrophobic residues within the PtpB catalytic cavity (Phe98, Leu101, Phe161, Leu199, Ile203, Val231, Leu232) and their hydroxylation such as in ∆3 causes the loss of hydrophobic interactions as well as a different positioning of the tricyclic scaffold in proximity of catalytic residues. The hydrogenation of prenyl groups in the isosophoranone M2H (IC 50 = 69.4 ± 4.5 µM) reduced the inhibitory potency with respect to M2 (IC 50 = 19.8 ± 2.3 µM), probably due to the loss of some pi-pi or pication interactions with Arg59, Arg63 and aromatic residues of the PtpB binding site such as Phe133. Moreover, comparing compounds 6016 (IC 50 = 19.2 ± 6.7 µM) and Ac3 (IC 50 = 33.2 ± 4.9 µM), the acetylation of the four hydroxyl groups in Ac3 provides a steric hindrance that, by docking, determines a different orientation of Ac3 within the PtpB catalytic cavity, leading to a slight loss of inhibitory activity in vitro. By comparing dihydrochalcones 59-triMe and Ega1, in spite of the different position of the prenyl and geranyl groups, the methylation of three hydroxyl groups in 59-triMe (IC 50 >100 µM) reduced drastically the inhibitory activity with respect to Ega1 (IC 50 = 13.4 ± 2.6 µM) due to the loss of H-bond interactions with a conserved water molecule and the protein backbone within Phe161 and Ala162. In summary, we found that the methylation or acetylation of hydroxyl groups in dihydrochalcones and trachypones, respectively, as well as the hydrogenation of prenyl groups in isosophoranones reduced significantly the PtpB inhibitory activity, whereas the presence of non-hydroxylated prenyl groups seems to be very important for the activity of anthrones. Other mono or polyhydroxylated compounds were not active (Hesp, Caf, Cyn, Buf-I and α-Cub). Overall, we found a satisfactory correlation between rescoring energy and enzymatic inhibition data of active compounds(-logIC 50 ), with the only exception of ∆3 whose binding affinity was overestimated by both docking and rescoring. By removing the outlier ∆3, we found a R 2 correlation value of 0.23 by comparing the -logIC 50 values with docking scores and a R 2 of 0.73 by comparing rescoring energies ( Figure S5), thus reinforcing the robustness of our computational approach as well as that rescoring docking poses with the MM-GBSA may provide a higher correlation with experimental data than docking.
It is worth noting that the presence of a carboxyl group in OMTS seems to be essential for the interactions with PtpB catalytic site [15] but, contrarily to expectation, compounds Caf and Cyn endowed with the carboxyl group were not active in our assays.

Kinetics measurements
Most potent PtpB inhibitors were then selected to investigate their mechanism of action with respect to the PTPs substrate pnitrophenyl phosphate (pNPP). Kinetic analysis revealed that compounds KuwE, 6016 and Ac3 act as PtpB competitive inhibitors, with K i values between 1.6 and 17.1 µM, while PirIII, Ega1 and ∆3 are non-competitive inhibitors, with K i values between 6.6 and 14.5 µM (Table 2). Figure 2 shows the Lineweaver-Burk plots of the PtpB inhibitors.

Selectivity assays
Human protein tyrosine phosphatase 1B (PTP1B) plays a critical role in regulating glucose homeostasis and body weight by acting as a key negative regulator of insulin and leptin signaling pathway, respectively [50]. This enzyme has been shown to increase insulin sensitivity and obesity resistance [51]. The tyrosine phosphatase family shares a catalytic domain with the conserved invariant sequence HCX 5 R [52,53]. Especially, PtpB (H159C160X 5 R166) has a Phe161, a Lys164 and an Asp165 that are conserved and differ from human PTP1B (H214C215X 5 R221) ( Figure 3) and might be therefore exploited to design inhibitors with improved selectivity. The most potent PtpB inhibitors were evaluated against the human PTP1B (Table 3), in order to monitor their selectivity index (SI) as the ratio between IC 50 measured towards human PTP1B and PtpB from Mtb. Results showed that the most potent compounds, namely KuwE, PirIII and Ega1 are slightly selective for PtpB, with a SI of 5.1, 2.3 and 1.6, respectively. Other compounds exhibited a SI lower than 1.

Predicted binding mode of KuwE
The possible binding mode of KuwE within the catalytic site of PtpB was investigated by molecular docking (Figure 4). The hydroxyl groups of KuwE perform H-bond interactions with Glu60, His94, Tyr125, Met206 and two conserved water molecules. Only the H-bond with His94 and water molecules are in common with the binding of OMTS revealed by X-ray crystallography, although there is a clear shape overlapping between these inhibitors within the PtpB active site. KuwE aromatic rings belonging to the dihydroxyphenyl-vinyl moiety are involved in hydrophobic interactions with a cluster of PtpB hydrophobic residues such as Ile203, Met206, Ile207 and Phe161 located at the entrance of the catalytic site, while the prenyl group shows a fine overlapping with a phenyl ring of OMTS and binds in a hydrophobic sub-pocket in close proximity of Leu83.
Notably, KuwE performs a hydrophobic/aromatic interaction with the side chain of the non-conserved Phe161 (Figure 4) with geometry resembling the parallel displaced π-stacking interaction, thus providing a possible structural explanation for KuwE selectivity observed by enzymatic assays. Such structural feature could be exploited for the rational design of selective PtpB inhibitors as well as to improve the selectivity of already known PtpB inhibitors.

Peptide mass fingerprint analysis
To monitor the capability of KuwE to protect the PtpB catalytic site from the proteolytic cleavage by trypsin, as well as to support the proposed mechanism of inhibition, the peptide mass fingerprint (PMF) of PtpB in presence or absence of inhibitor was determined by mass spectrometry (MS). The sequence coverage by MS analysis of the tryptic digest of PtpB was 78% and 81%, respectively, in the presence or absence of KuwE. By comparing the PtpB PMF in the absence ( Figure 5A) and presence ( Figure 5B) of KuwE, a significant difference in fragment composition was observed. In fact, when PtpB proteolysis was carried out in the absence of KuwE, two fragments (m/z 1953 and m/z 2224) were observed, which are not present in the mass spectrum recorded in the presence of 300 µM of KuwE. Since KuwE is not able to inhibit trypsin by itself, these results suggest that KuwE inhibits the formation of the cleavage products of 1953 Da and 2224 Da by interacting within the PtpB catalytic site and protecting it from the proteolytic cleavage by trypsin. The peptide with m/z 2224  corresponds to the tryptic fragment of the complete sequence of the catalytic site ((R145)VVTLLAAGRPVLTHCFAGKDR(T167)) ( Figure  6), while the fragment with m/z 1953 corresponds to a part of the catalytic site with sequence ((R145)VVTLLAAGRPVLTHCFAGK(D165)), which include the His159 and the catalytic Cys160 residues. Notably, these evidences are in agreement with docking results showing that KuwE interacts within the PtpB catalytic site in close proximity of Cys160, Phe161, Lys164 and Arg166 which are present in the fragment cleaved by trypsin in absence of the inhibitor.
Furthermore, we monitored also the effect of the PtpB competitive inhibitor 70 [11] used as positive control, towards the trypsin proteolytic cleavage of PtpB. Results showed that this molecule is also able at least to reduce significantly the intensity of peaks m/z 1953 and m/z 2224 in PMF spectra recorded in presence of PtpB and trypsin ( Figure S4).
In summary, analysis of PMF suggests that KuwE may protect PtpB cleavage by trypsin by interacting and shielding the catalytic site. Indeed, the tryptic fragments were not detected in the PMF of PtpB recorded in presence of the competitive inhibitor KuwE. These results also support the interaction model proposed by molecular modeling for KuwE, which is shown in Figure 4.

Conclusions
In this work we reported on the discovery of natural compounds as potent inhibitors of PtpB by in silico screening of an in house unique library and enzymatic assays. Especially, KuwE showed the most potent inhibition (K i = 1.6 ± 0.1 µM) and, to the best of our knowledge, KuwE is the most potent natural compound inhibitor of PtpB from Mtb discovered so far, as well as this is the first report on a non-peptidic natural compound inhibitor of PtpB. Kinetic studies shed light on the mechanism of PtpB inhibition, whereas peptide mass fingerprint analyses performed by MS showed that KuwE was able to protect the PtpB catalytic site by the proteolytic activity of trypsin, thus reinforcing that this inhibitor may interact within the PtpB catalytic site, in agreement with docking results. Despite the success in the identification of natural compounds as potent inhibitors of PtpB, a low selectivity towards the human PTP1B was found. In this respect, the rational optimization of KuwE should account for challenging organic synthesis strategies.
Natural products continue to represent a unique source of chemical diversity for the discovery of hit and lead molecules. PtpB natural product inhibitors discovered and characterized in this study may serve as profitable tool to investigate the biochemical functions of PtpB as well as starting point for further optimization aimed at the development of anti-TB medicinally active agents.