Discovery of uncompetitive inhibitors of SapM that compromise intracellular survival of Mycobacterium tuberculosis

SapM is a secreted virulence factor from Mycobacterium tuberculosis critical for pathogen survival and persistence inside the host. Its full potential as a target for tuberculosis treatment has not yet been exploited because of the lack of potent inhibitors available. By screening over 1500 small molecules, we have identified new potent and selective inhibitors of SapM with an uncompetitive mechanism of inhibition. The best inhibitors share a trihydroxy-benzene moiety essential for activity. Importantly, the inhibitors significantly reduce mycobacterial burden in infected human macrophages at 1 µM, and they are selective with respect to other mycobacterial and human phosphatases. The best inhibitor also reduces intracellular burden of Francisella tularensis, which secretes the virulence factor AcpA, a homologue of SapM, with the same mechanism of catalysis and inhibition. Our findings demonstrate that inhibition of SapM with small molecule inhibitors is efficient in reducing intracellular mycobacterial survival in host macrophages and confirm SapM as a potential therapeutic target. These initial compounds have favourable physico-chemical properties and provide a basis for exploration towards the development of new tuberculosis treatments. The efficacy of a SapM inhibitor in reducing Francisella tularensis intracellular burden suggests the potential for developing broad-spectrum antivirulence agents to treat microbial infections.

persistence. Furthermore, guinea pigs infected with the ∆sapM Mtb strain showed total absence of mycobacteria in lungs and spleen at 16 weeks post infection, demonstrating the critical role of SapM in Mtb pathogenesis 24 . Importantly, the lungs, liver and spleens of these animals showed only few tubercles, negligible histopathological tissue damage, and all the infected animals survived the infection 24 . Clearly, the role of SapM is important for Mtb persistence, but it also impacts on the host ability to fight the infection. Our hypothesis is that if we could block the action of SapM with a small molecule, we may be able to enhance infection clearance and help to improve TB treatments.
The sapM gene is unique in the mycobacterial genome and there is no human orthologue to SapM 22,25 , suggesting the potential to develop specific and selective inhibitors. Recently, we have reported that SapM can be inhibited using l-ascorbic acid (L-AC) and 2-phospho-l-ascorbic acid (2P-AC), and that inhibition reduced mycobacterial survival in human THP1 macrophages 22 . However, these are low potency inhibitors with IC 50 values > 200 µM.
Here we have screened 1503 compounds from commercial and in-house libraries and identified three new potent (IC 50 < 10 µM) and selective inhibitors of SapM that behave as uncompetitive inhibitors. These compounds significantly reduce up to 70% of Mtb burden in human THP-1 macrophages at a low concentration of 1 µM. Two of these inhibitors are characterised by a benzylidenemalononitrile scaffold and share a tryhydroxy-benzene moiety with the third inhibitor, a Gallic acid derivative. The most potent inhibitor, compound 1, also reduces the burden of Francisella tularensis in MH-S macrophages. F. tularensis is the causative agent of the lethal disease tularaemia 26 and secretes the virulence factor AcpA [27][28][29] , a homologue of SapM that shares the same mechanism of catalysis and inhibition 22 . Our results demonstrate the potential of exploiting antivirulence agents to impair bacterial intracellular survival for the treatment of TB, with possible applications to other microbial infections.

Results and discussion
To identify new inhibitors of SapM, we screened four different compound libraries covering a wide range of pharmacological active drugs (LOPAC-1280 and LOPAC-Pfizer), phosphatase inhibitors (Enzo-BML-2834), and in-house compounds from other drug discovery programmes. A total of 1491 compounds were initially screened for activity against SapM. The primary screening resulted in a total of 18 compounds that reduced SapM specific activity (SA) by ≥ 50% at 100 µM (Table 1, Fig. 1A). These compounds were re-purchased and re-tested to confirm their activity. Six of the re-tested compounds showed ≥ 50% inhibition of SapM (Table 2, Fig. 1B). False positives or negatives are expected as compounds tend to degrade or precipitate after long-term storage in DMSO 30 , but the number of hits identified in our screening is similar with previous reports using the same libraries 31,32 .
The distribution of compound activity on SapM is homogeneous across the libraries (Fig. 1A), indicating an unbiased screening. Of note, some compounds increased the SA of SapM, an effect reported for allosteric activators 33 . In total, six compounds (1-6), five from the LOPAC-1280 library and one from the Enzo-BML-2834, were considered for further analyses (Table 2). These compounds are reported to target different proteins: Tyrphostin 51 and HI-TOPK-32 are kinase inhibitors 34,35 , RWJ-60475 is a tyrosine phosphatase inhibitor 36 , Galloflavin is a lactate dehydrogenase inhibitor 37 , YM-26734 is a phospholipase A2 inhibitor 38 and (6R)-BH4, Sapropterin dihydrochloride, is a cofactor for nitric oxide synthesis 39 used in the clinic to treat pulmonary hypertension 40 .

Structure-activity relationship for SapM inhibitors.
Among the six inhibitors, three compounds: Tyrphostin 51 (1), Galloflavin (2) and YM-26734 (3) completely abolished the enzymatic activity of SapM, and share the presence of a trihydroxy-benzene group (Table 2), also present in polyphenols reported as phosphatase inhibitors 41 . In order to gain structure-activity relationship (SAR) insight we searched for related compounds to 1 and 2 in the libraries tested and for commercially available derivatives. Compound 3 was not included in this search because of its complexity and poor drug-like properties according to the Lipinski's rule of five (Ro5) 42 .
Additionally, nine compounds containing hydroxyl-substituted benzene rings were identified from the primary screening (20-28 in Table S1), where 20 and 25 exhibited ~ 30% reduction in SA and the rest had poor or no activity (Table S1). A group of polyphenols found in the literature 41 (29)(30)(31)(32)(33), including catechins and Table 1. Summary of compounds screened from commercial and in-house libraries and structure-activity relationship (SAR) analyses.
However, the number of hydroxyl groups (-OH) in the benzyl ring of tyrphostins, has a substantial contribution to potency because removal of one or more -OH results in loss of inhibition (Table S1). Compounds 1, 13 and 14, with three -OH abolished enzyme activity completely, whereas 7, 15 and 16 with two -OH only inhibit by ~ 30-50% and 8, 9 and 12 with one -OH inhibit by less than 21%.
For other hydroxyl-substituted benzene compounds it is clear that the presence of -OH alone is not sufficient for full potency. Compound 2 with three -OH inhibits SapM activity completely at 100 µM, but 26, 28, 31-33 with more than three -OH show none or poor activity (Suppl . Table S1). Thus, the presence of at least two -OH, appears to be necessary for inhibition particularly for the benzylidenemalononitrile scaffold, but not sufficient for full potency.
Potency and selectivity of SapM inhibitors. IC 50 values of key compounds showing substantial inhibition on SapM in the primary and SAR screens were determined using dose-response curves (Fig. 2B). Compound 4 was not used as it precipitates at > 100 µM, preventing the ability to reach saturation. The more potent compounds are 1, 2, 13 and 14, with IC 50 values ranging from 6 to 14 µM (Fig. 2B, Table 3). These compounds are up to 40 times more potent than the 2P-AC and L-AC inhibitors that we have previously reported (IC 50 > 200 µM) 22 . Furthermore, all inhibitors show selectivity towards SapM over other secreted phosphatases in Mtb (Table 3): MptpB, a lipid phosphatase with a similar mode of action to SapM 18,44 and MptpA, a tyrosine phosphatase that regulates phagosome acidification [45][46][47] . Importantly, compounds 1, 2, and 13 are also selective over the human phosphatases PTP1B (phosphotyrosine specific) and VHR (dual specificity phosphatase) ( Table 3).

SapM inhibitors show an uncompetitive mechanism of inhibition.
Kinetics studies showed that the best four inhibitors, 1, 2, 13 and 14, behave as uncompetitive inhibitors, as indicated by the characteristic pattern of parallel lines in the Lineweaver-Burk plots (Fig. 2C), where increasing amounts of inhibitor causes a reduction of both the K m and V max values 48 . Uncompetitive compounds bind to the enzyme-substrate complex instead of the free enzyme 48 . The potency of uncompetitive compounds is then enhanced as the substrate concentration in the reaction increases 49,50 . This mechanism is similar to the one we observed for L-AC 22 , indicating a general mechanism of inhibition for this target.

SapM inhibitors reduce mycobacterial burden in vitro.
Compounds with IC 50 < 10 µM (Table 3) (Fig. 3A,B). This reduction is comparable to that observed when deleting the sapM gene 24 , and consistent with the critical role of SapM in intracellular mycobacterial survival 23,24 . A similar effect is observed both in vitro 12,13 and in vivo 17,18 , when inhibiting MptpB, another secreted virulence factor involved in phagosomal maturation.
There is no substantial difference between treatment with 1 or 13, with similar IC 50, suggesting that the change in the nitrile position in the benzylidenemalononitrile core does not influence efficacy in cells. Compound 2 is marginally better than 1 and 13 at 24 h but the differences are not statistically significant. The efficacy of these compounds is more pronounced at 24 h, with reductions of 50-70% on Mtb burden, compared to 30-50% at 72 h. This would agree with the proposed role of SapM in the dephosphorylation of PI(4,5)P 2 and PI3P during phagocytosis 22 , at the initial stages of the infection. Efficacy is also superior to the one we observed for MptpB   www.nature.com/scientificreports/ inhibitors, which showed similar reduction of Mtb burden at concentrations of 80 µM or higher, instead of 1 µM 12,18 . None of the inhibitors tested affected acellular growth over the course of 9 days compared to the DMSO control (Fig. 3C). This is in agreement with previous reports (and our own data in Suppl. Fig. S1) that deletion of sapM impairs mycobacterial survival in macrophages and animal models of infection, but does not reduce growth in culture medium 24 . Furthermore, the compounds had no significant effect on the intracellular burden of ∆sapM (Suppl. Fig. S1), supporting specificity.
Compounds 1, 2 and 13 exhibit favourable drug-like properties ( Table 4). The physico-chemical properties calculated with the SwissADME tool 51 indicate that 1, 2 and 13 are relatively small molecules (Mw < 300 Da), highly hydrophilic, with predicted good solubility and with lead-like properties according to the Ro5 criteria 42 . Thus this makes them suitable candidates for further development and future validation in vivo. Binding of 1, 2 and 13 to SapM was confirmed by microscale thermophoresis (MST) (Suppl. Figs. S2, S3).

Compound 1 reduces Francisella tularensis intracellular burden. One of the closest homologues
of SapM is the secreted phosphatase AcpA from Francisella tularensis, which acts as a virulence factor for that bacterium 27,29 . We have demonstrated that SapM and AcpA share the same mechanism of catalysis 22 . We have also shown that AcpA inhibitors 2-PAC and L-AC, which reduce F. tularensis burden in vitro 52 , also inhibit SapM activity, and that 2-PAC significantly reduces intracellular survival of Mtb 22 . We hypothesized that SapM inhibitors may therefore have efficacy in reducing intracellular burden of F. tularensis. For this, the most potent compound 1 was confirmed to inhibit AcpA (Suppl. Methods and Suppl. Fig. S4) and selected to evaluate its efficacy in MH-S cells infected with F. tularensis.
As for Mtb, compound 1 reduces F. tularensis intracellular burden in a dose dependent manner, with a significant reduction (55%, ***p < 0.001) at 40 μM at 24 h post infection (Fig. 4A). Compound 1 is not cytotoxic to MH-S cells (> 70% viability) (Fig. 4B), and does not affect acellular growth of F. tularensis (Fig. 4C). The efficacy of compound 1 in reducing Mtb and F. tularensis burden in infected macrophages suggests the potential of developing antivirulence agents with a broad-spectrum activity to treat microbial infections.

Conclusions
In this study we have identified new potent inhibitors of SapM (IC 50 < 10 µM) that have an uncompetitive mechanism of inhibition. The inhibitors are specific to SapM and selective over other Mtb secreted phosphatases and human phosphatases. The best inhibitors 1 and 13, contain a benzylidenemalononitrile core and share a trihydroxy-benzene group with compound 2, also found in other polyphenol phosphatase inhibitors. Importantly, the best inhibitors, show significant efficacy in reducing intracellular Mtb burden at 1 µM concentration, and recapitulate the behaviour of the ∆sapM strain in macrophage infections 24 (and this study). These results are consistent with the critical role of SapM in phagosomal maturation arrest to increase Mtb survival and pathogenesis, and with the reported reduction of mycobacterial burden both in vitro 12,13 and in vivo 17,18 , when inhibiting MptpB, another secreted virulence factor. Notably, we see also reduction in intracellular burden of F. tularensis when using compound 1, suggesting the potential of developing antivirulence agents with a broad-spectrum activity to treat different microbial infections. This is the first report of potent inhibitors of SapM and provides a basis for further development to exploit the potential of this target in the treatment of TB.

Materials and methods
Production of recombinant proteins. SapM was expressed in Escherichia coli C41 (DE3) by auto-induction at 20 °C for 24 h as previously described 22 To determine the type of inhibition different inhibitor concentrations were selected to yield around 30% and 75% inhibition for each concentration of pNPP. This ensures sufficient signal to obtain accurate data while allowing a significant inhibition effect 48 . Velocity (V) was plotted as a function of pNPP concentration and fit in a Lineweaver-Burk plot (double-reciprocal) using GraphPad Prism 8.41. All assays were done in 96-well microplates and performed in triplicate in at least two independent experiments.

Bacterial and cell culture. THP-1 monocyte cell lines (ATCC) were cultured in Roswell Park Memorial
Institute-1640 medium (RPMI) (R8758-Sigma-Aldrich) containing l-glutamine supplemented with 10% heat inactivated fetal bovine serum (FBS, Invitrogen) at 37 °C in 5% CO 2 . Mtb strain H37Rv was grown in Middlebrook 7H9 broth (BD Diagnostics) or on Middlebrook 7H10 agar, both supplemented with 0.05% Tween 80, 0.2% glycerol and 10% OADC (Oleic Albumin Dextrose Catalase from Becton Dickinson Microbiology Systems) at 37 °C in 5% CO 2 . Mtb cultures were prepared using 1 ml of mid-log phase Mtb stock into 20 ml of fresh media and incubated static for 6 days prior to being used in infection or acellular assays. Cytotoxicity assays. A colorimetric assay using the tetrazolium dye 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was performed as described previously 54   Francisella tularensis infection assay. Francisella tularensis live vaccine strain (LVS), belonging to subspecies holarctica was used as described 55 . Briefly, to achieve the inoculum F. tularensis from a blood cysteine glucose agar (BCGA) plate grown for 48 h at 37 °C was resuspended in complete L-15 media (10% FCS, 5 mM l-Glutamine) (Life Technologies) to a 600 nm spectrophotometer optical density (OD) reading of 0.20 which corresponded to ~ 1 × 10 9 CFU/ml. MOI of 100 was achieved through serial dilutions and inoculum determined by plate count on BCGA. MH-S cells (ATCC, CRL-2019), an SV40-transformed alveolar macrophage cell line, were incubated for 2 h with the F. tularensis MOI 100 (bacteria:macrophage) at 37 °C to allow for cellular uptake of the bacteria. This was in the presence or absence of compound 1 (1 and 40 μM) and DMSO control. F. tularensis was removed and cells were washed with warm PBS, followed by 30 min incubation with 10 μg/ml gentamicin (Sigma-Aldrich) to kill any extracellular bacteria and then replaced with complete L-15 with 2 μg/ml gentamicin and compound 1 (1 or 40 μM) or DMSO control for 24 h. At 24 h cells were lysed with 4 °C water to determine bacterial burden. Statistical significance was evaluated by one-way ANOVA followed by a multiple comparison analyses of variance by Dunnett's test (GraphPad Prism 8.41 for Windows). Differences were considered significant at the 95% level of confidence. All experiments with F. tularensis were carried out in a biosafety level 3 containment facility. Acellular growth of Mycobacterium tuberculosis and Francisella tularensis. Mtb (4.4 × 10 7 CFUs) was cultured in 25 ml of Middlebrook 7H9 and compounds 1, 2 or 13 at 40 μM were added on day 0. Cultures were grown static over 9 days at 37 °C in 5% CO 2 . Controls were DMSO only and isoniazid at 0.14 µg/ml. F. tularensis (1 × 10 8 CFUs) was cultured in complete L-15 media (10% FCS) at 37 °C for 24 h in the presence or absence of compound 1 at 40 μM. Bacterial growth was monitored by OD at 600 nm. Experiments were performed in triplicate on at least two separate studies. Statistical significance was evaluated by two-way ANOVA followed by multiple comparison analyses of variance by Bonferroni test (GraphPad Prism 8.41 for Windows). Differences were considered significant at the 95% level of confidence.