High Throughput Screen Identifies Small Molecule Inhibitors Specific for Mycobacterium tuberculosis Phosphoserine Phosphatase*

Background: Phosphoserine phosphatase (PSP) is an essential enzyme involved in l-serine biosynthesis. Results: High throughput screen was performed to identify specific PSP inhibitors with activity against intracellular bacteria. Conclusion: Validation of PSP as a drug target would lead to identification of scaffolds with a novel mechanism. Significance: This is the first report demonstrating selective inhibition of M. tuberculosis PSP enzyme. The emergence of drug-resistant strains of Mycobacterium tuberculosis makes identification and validation of newer drug targets a global priority. Phosphoserine phosphatase (PSP), a key essential metabolic enzyme involved in conversion of O-phospho-l-serine to l-serine, was characterized in this study. The M. tuberculosis genome harbors all enzymes involved in l-serine biosynthesis including two PSP homologs: Rv0505c (SerB1) and Rv3042c (SerB2). In the present study, we have biochemically characterized SerB2 enzyme and developed malachite green-based high throughput assay system to identify SerB2 inhibitors. We have identified 10 compounds that were structurally different from known PSP inhibitors, and few of these scaffolds were highly specific in their ability to inhibit SerB2 enzyme, were noncytotoxic against mammalian cell lines, and inhibited M. tuberculosis growth in vitro. Surface plasmon resonance experiments demonstrated the relative binding for these inhibitors. The two best hits identified in our screen, clorobiocin and rosaniline, were bactericidal in activity and killed intracellular bacteria in a dose-dependent manner. We have also identified amino acid residues critical for these SerB2-small molecule interactions. This is the first study where we validate that M. tuberculosis SerB2 is a druggable and suitable target to pursue for further high throughput assay system screening.

Tuberculosis (TB) 2 caused by Mycobacterium tuberculosis is an enormous health burden in developing countries. The World Health Organization currently estimates that ϳ1.8 billion people are latently infected with M. tuberculosis, there are 9 million new incident rates and ϳ2 million annual deaths arising from TB infections (1). The current DOTS (directly observed treatment short course) regimen involves drugs that inhibit either DNA replication or transcription or cell wall biosynthesis but is very complex and lengthy. Numerous factors such as poor patient compliance or treatment failure have led to emergence of various drug-resistant (multidrug-resistant, extensively drug-resistant, and totally drug-resistant) TB strains (2). The major challenge in the field of TB drug development is to identify drug targets that are vulnerable in vitro and to identify scaffolds (i) with a novel mechanism of action, (ii) that have the potential to shorten chemotherapy, (iii) that target drug-resistant and latent bacteria, and (iv) that are compatible with current TB and anti-retroviral therapy (3). In the past decade, substantial progress has been made in development of genetic tools to identify and biochemically characterize metabolic pathways that are essential for M. tuberculosis growth in vitro. These pathways are potential drug targets to identify scaffolds with a novel mechanism of action (4 -7). L-Serine biosynthetic pathway represents one such mechanism, and enzymes involved in L-serine biosynthesis have been predicted to be essential for M. tuberculosis growth in vitro (4 -7). In bacteria, there are two distinct pathways involved in L-serine biosynthesis (8,9). The first pathway involves serine hydroxy methyl transferase that catalyzes simultaneous reversible conversion of glycine and 5,10-methylenetetrahydrofolate to serine and 5,6,7,8-tetrahy-drofolate, respectively (10). In an alternative pathway, 3-phosphoglycerate dehydrogenase (PGDH) oxidizes 3-phosphoglycerate to 3-phosphohydroxy pyruvate in a NAD ϩ /NADH-dependent manner. Phosphoserine aminotransferase (PSAT), a PLP (pyridoxal-5Ј-phosphate)-dependent enzyme converts 3-phosphohydroxy pyruvate to O-phospho-L-serine that is subsequently dephosphorylated by phosphoserine phosphatase (PSP) into L-serine (11). Both 3-phosphoglycerate dehydrogenase and PSAT homologs in M. tuberculosis have been extensively biochemically characterized, and their crystal structures have also been determined (12)(13)(14). In a recent study, it has been shown that intracellular cyclic AMP regulates the levels of PSAT enzyme, and extracellular addition of L-serine restores the growth defect of M. tuberculosis crp mutant in vitro (15).
PSP enzymes belong to the haloacid dehalogenase (HAD) superfamily of enzymes that are known to regulate diverse cellular functions such as membrane transport, metabolism, signal transduction, and nucleic acid repair (16). The HAD family of enzymes are characterized by the presence of three specific motifs: motif I, DXDX(T/V); motif II, (S/T)XX; and motif III, KX 18 -30 (G/S)(D/S)XXX(D/N) (17,18). These enzymes have an absolute requirement for divalent metal ion, and the conserved Asp residue in motif I forms a phosphoenzyme intermediate with the substrate via nucleophilic attack (19,20). HAD family of enzymes possess the Rossmannoid ␣/␤ fold and an inserted "cap" that regulates substrate access to its active site (21)(22)(23). PSP enzymes have been extensively biochemically characterized in various microorganisms and have been demonstrated to facilitate entry of Porphyromonas gingivalis into host cells by modulating host cytoskeletal architecture, innate immune responses, and dephosphorylating colicin and NF-〉 (24 -26).
Despite the importance of PSP enzymes in L-serine biosynthesis, biochemical characterization of mycobacterial PSP homologs has not been reported so far. In the present study, we have biochemically characterized SerB2 enzyme in vitro and developed a high throughput screening (HTS) assay sys-tem to identify novel SerB2 specific inhibitors. These identified new scaffolds that were (i) structurally different from known PSP inhibitors, (ii) selective in their ability to inhibit SerB2 enzyme in comparison with human PSP (HPSP) enzyme, and (iii) inhibited M. tuberculosis growth in vitro in a dose-dependent manner.

EXPERIMENTAL PROCEDURES
Chemicals, Strains, and Growth Conditions-Most of the chemicals used in the present study unless mentioned were purchased from Sigma-Aldrich. Various strains and plasmids used in the study are shown in Table 1. Escherichia coli strains XL-1 Blue and BL-21 (DE3, plysS) were used for cloning and expression studies, respectively. M. tuberculosis H 37 Rv and Mycobacterium bovis BCG strains were used for growth inhibition and macrophage infection studies. Various E. coli and mycobacterial strains were cultured in LB and Middlebrook medium, respectively, as per manufacturer's standard protocols. The antibiotics were used in the following concentrations: ampicillin (50 g/ml), kanamycin (25 g/ml), tetracycline (10 g/ml), and chloramphenicol (34 g/ml).
Multiple Sequence Alignment and Phylogenetic Analysis of PSP Homologs-Protein sequences of PSP homologs from various organisms were retrieved from the National Center for Biotechnology Information protein database. Multiple sequence alignment analysis was performed using the Clustal Omega (version 1.2.0) alignment tool and edited using GeneDoc (27). The evolutionary history was inferred using the neighbor joining method (28,29).
Construction and Validation of SerB1 and SerB2 Homology Models-The best templates for homology modeling of SerB1 and SerB2 proteins were identified using Position-Specific Iterative Basic Local Alignment Search Tool (PSI-BLAST) analysis in the Protein Data Bank. The homology models for SerB1 and SerB2 were built using Discovery Studios 2.5 (Accelrys). The built models were further refined with repetitive loop modeling and energy minimization studies. The refined homology models were finally validated with PROCHECK, Verify_3D, and Errat programs (30 -32).
Expression and Purification of PSP Proteins-For expression studies, both serB1 and serB2 were PCR-amplified and cloned into either pET28b or pMALc2x or pGEX4T-1. Various active site point mutants of SerB2 enzyme were generated by two-step PCR using gene specific primers having the desired mutations. E. coli BL-21 (DE3, plysS) transformed with either wild type or mutant constructs were grown in LB medium at 37°C. Protein expression was induced at A 600 nm ϭ ϳ0.8 with the addition of isopropyl ␤-D-galactopyranoside (IPTG) at a final concentration of 1 mM. The cultures were grown overnight at 18°C, pelleted by centrifugation, and frozen at Ϫ20°C for later use. Purification of His 6 -tagged and MBP (maltose-binding protein)-tagged proteins was performed by affinity chromatography using nickel-nitrilotriacetic acid or amylose resin, respectively, as per the manufacturer's recommendations. Purified proteins were visualized on 10% SDS-PAGE by Coomassie Brilliant Blue staining. The purified fractions were dialyzed, concentrated using Amicon Ultra-15 centrifugal units (Millipore), and stored as aliquots in enzyme storage buffer (50 mM Tris, pH 7.4, 100 mM NaCl, and 10% glycerol). Protein concentration was estimated using Coomassie Plus protein assay reagent (Thermo Fisher).
Steady State Kinetics of SerB2 Enzyme-To determine kinetic parameters for SerB2 enzyme, assays were performed in 25 l of reaction volume (100 mM Tris, pH 7.4, 5 mM MgCl 2 , 5 mM D⌻⌻) using varying concentration of O-phospho-L-serine (20 -320 M) and 1 M His 6 -SerB2 enzyme at 37°C for 10 min. The formation of inorganic phosphate (P i ) in enzyme reaction was monitored by measuring absorbance at 630 nm using Quantichrome phosphate assay kit (Bioassay Systems) as per manufacturer's recommendations. The initial velocities in enzyme reactions (rate of P i release in M/min) were plotted against concentration of O-phospho-L-serine using nonlinear regression to the Michaelis-Menten equation using Prism 5 software (GraphPad Software, Inc., version 5.01). The apparent K m , V max , and k cat /K m for SerB2 enzyme was determined from the plotted area. The substrate specificity for SerB2 enzyme was determined by performing assays in the presence of varying concentration of either O-phospho-L-serine or O-phospho-L-threonine. To determine optimum conditions for SerB2 activity, assays were performed using buffers in the pH range 6.0 -9.0 in the presence of various cations and nonionic detergent (0.01% Triton X-100) using 100 M O-phospho-L-serine. For identification of catalytically important residues, enzyme assays were performed using 100 M of O-phospho-L-serine in the presence of either 1 M wild type or mutant SerB2 protein.
Far Ultraviolet Circular Dichroism Studies-CD spectra of various proteins were recorded on a JASCO-J-815 spectropolarimeter using a 1-mm-path length cuvette with a scan rate of 10 nm/min and averaged over five scans. The raw CD data were converted into molar ellipticity (⌽ M ) as follows, where ⌽ obs is the observed ellipticity (in degrees), d is the path length (in centimeters), and C is the protein concentration (molar). The ⌽ M of MBP was calculated and subtracted from the ⌽ M of MBP-SerB2 fusion protein to obtain molar ellipticity of free SerB2. ⌽ M was converted to mean residue ellipticity (⌽ MRE ) as follows, where n is the total number of amino acids in the protein.
High Throughput Screen to Identify PSP Inhibitors-Inorganic phosphate release was adapted for a high throughput screen to identify novel PSP inhibitors. This end point assay was performed in a final volume of 50 l and validated using varying concentrations (10 M to 100 mM) of known PSP inhibitors such as DL-AP 3 (DL-2-amino-3-phosphonopropionic acid) and sodium orthovandate in the presence of 1 M His 6 -SerB2 (33,34). The contents of small molecule library of the National Cancer Institute Developmental Therapeutic Program comprising 2300 structurally diverse compounds were transferred to 96-well plate (Nunc) at a final concentration of 100 M in assay buffer (containing 1 M His 6 -SerB2) for preliminary screening. All reaction plates included proper controls such as buffer only, no substrate, and DL-AP 3 control. The enzyme-scaffold mix was incubated at room temperature for 10 min, and the reaction was initiated by the addition of 200 M O-phospho-L-serine. After incubation for a further 10 min, the enzyme activity was measured as described above, and the percentage of inhibition was calculated. Half-maximal inhibitory (IC 50 ) concentration determination experiment was performed in the presence of increasing concentration of compounds in triplicates. IC 50 values were determined by nonlinear regression analysis using Prism 5 software.
Determination of Specificity of Primary Hits-The hits identified in our HTS were also evaluated for their ability to inhibit mycobacterial Ser/Thr phosphatase (PstP, Rv0018c), alkaline phosphatase (Bangalore Genei, Bangalore, India) and HPSP enzyme (Calbiochem) in vitro. PstP and alkaline phosphatase inhibition studies were performed in the presence of 100 M of phoshopeptide (K-R-pT-I-R-R; Millipore) and 200 M p-nitrophenyl phosphate, respectively, as per standard protocols. HPSP inhibition studies were performed in conditions similar to those standardized for SerB2 protein.
Surface Plasmon Resonance Studies-Surface plasmon resonance (SPR) experiments were performed at 25°C using BIACORE T200 (GE Healthcare). SerB2 was diluted to a concentration of 400 g/ml in 10 mM sodium acetate buffer, pH 4.0. The protein was immobilized on CM5 sensor chip by the use of amine coupling chemistry as per standard protocols. The free surface was blocked with 1 M ethanolamine-HCl (pH 8.5) and washed with 50 mM NaOH to remove free SerB2. The immobilization levels ranged from 8,000 to 10,000 resonance units (RU). For binding studies drugs at concentration of 1 mM in running buffer (20 mM Tris, pH 7.4, 200 mM NaCl, 0.005% Tween 20, 2% DMSO) were injected at a flow rate of 30 l/min for 2 min over the immobilized protein or a reference surface without protein. The surfaces were than washed with the running buffer and regenerated twice using 10 mM glycine, pH 2.5.
Cytotoxicity of Primary Hits-THP-1 (human acute monocytic leukemia) or Vero (green monkey kidney epithelial cell line) were cultured in Roswell Park Memorial Institute (RPMI) medium containing 10% FBS. These cell lines were procured from the National Cell Repository of the National Centre for Cell Science (Pune, India). THP-1 monocytes were differentiated into macrophages by the addition of 30 nM phorbol 12-myristate 13-acetate. For cytotoxicity assays, cells were diluted to a density of 5 ϫ 10 5 /ml in fresh medium, and 100 l was aliquoted into 96-well flat bottom plates (Nunc). After 24 h of incubation at 37°C, cells were overlaid with RPMI medium containing drugs (0.625-50 M) along with DMSO control in triplicate wells. After 96 h postincubation, cellular viability was assayed by measuring lactate dehydrogenase activity in culture supernatants using Cytotox96 nonradioactive cytotoxicity kit as per manufacturer's recommendations (Promega Corporation, Madison, WI). These assays were performed three independent times, and concentration causing 50% cytotoxicity (TC 50 ) was calculated.
Mycobacterial Growth Inhibition Assays-Minimum inhibitory concentration (MIC 99 ) against M. tuberculosis H 37 Rv and M. bovis BCG for these primary hits was determined using standard procedures. Bacterial cells were incubated at 37°C in the presence of varying concentration of drugs for 14 days. MIC 99 was determined as the lowest concentration of drug inhibiting visible growth. To determine whether the observed growth inhibition by SerB2 specific inhibitors was bactericidal or bacteriostatic, mycobacterial cultures were grown till an A 600 nm of 0.1 and subsequently exposed to drugs at 10ϫ MIC 99 concentration. For bacterial enumeration, samples were withdrawn at designated time points, and 100 l of 10-fold serial dilutions were plated on MB-7H11 plates and incubated at 37°C for 3 weeks. For intracellular killing experiments, THP-1 cells were seeded at a density of 5 ϫ 10 5 /well in a 24-well plate (Nunc) in RPMI medium supplemented with 10% FBS. Macrophages were infected with bacteria at a multiplicity of infection of 1:10 in triplicate wells for 4 h, washed with antibiotic free RPMI medium, and overlaid with RPMI medium containing 200 g/ml amikacin for 2 h. Subsequently, macrophages were washed twice with antibiotic free RPMI medium and overlaid with RPMI medium supplemented with 10% FBS. After 24 h postinfection, cells were overlaid with RPMI medium containing drugs at either 4ϫ or 16ϫ MIC 99 concentration for 4 days. For bacterial enumeration, macrophages were lysed by the addition of 1ϫ PBS, 0.1% Triton X-100, and the number of intracellular bacteria was determined by plating 100 l of 10-fold serial dilutions on MB-7H11 plates.
Molecular Docking Studies-All programs used for molecular docking studies were from Suite 2012 from Schrodinger. SerB1 and SerB2 models and HPSP crystal structure (Protein Data Bank code 1L8L) were optimized for docking studies with protein preparation wizard. These structures were finally minimized by converging it to a root mean square deviation (r.m.s.d.) tolerance of 0.3 Å using the OPLS 2005 force field. Three-dimensional ligand structures were generated using Lig-Prep module. Ligands were fitted to SerB1 and SerB2 models and HPSP crystal structure using the GLIDE extra precision method (35). Images were obtained and visualized through the Maestro interface.

RESULTS
Bioinformatic and Homology Modeling Studies-The enzymes involved in L-serine biosynthesis are highly conserved in various mycobacterial species ( Fig. 1A and Table 2). M. tuberculosis PGDH and PSAT enzymes catalyzing the first two steps of L-serine biosynthesis have been biochemically well characterized (13)(14)(15). Phylogenetic analysis among PSP proteins from various organisms revealed that PSP proteins from actinobacter and helicobacter species were more similar to each other in comparison with enterobacterial and human PSP enzymes (Fig.  1B). As shown in Fig. 1B, both SerB1 and SerB2 proteins were distantly related to each other and HPSP enzyme, with which they shared an overall sequence similarity of 18 and 27%, respectively. Multiple sequence alignment among PSP enzymes from various microorganisms revealed that these proteins share an overall homology of 15% among themselves, and both SerB1 and SerB2 possessed HAD specific motifs (motifs I-III; Fig (17)(18)(19)(20).
The best templates for homology modeling for SerB1 and SerB2 proteins were identified by PSI-BLAST analysis using Protein Data Bank. The closest homolog for SerB1 protein was PSP enzyme from Bordetella pertussis (Protein Data Bank code 3FVV) with 34% sequence identity, 41% query coverage (consisting only the HAD superfamily domain region), and an E-value of 7 e Ϫ16 . We were unable to build full-length structure of SerB1 because homologous template with high sequence coverage was not available. The closest homolog for SerB2 protein was SerB protein from Mycobacterium avium (Protein Data Bank code 3P96) with 84% sequence identity, 99% query coverage, and an E value of 0.0. The superimpositions of SerB1and SerB2-modeled proteins over 3FVV and 3P96, respectively, resulted in backbone r.m.s.d. of 0.449 and 0.451 Å, respectively. The superimpositions of SerB1 and SerB2 models over HPSP crystal structure resulted in r.m.s.d. of 6.216 and 2.565 Å, respectively (Fig. 2B). These computational studies predicted 11 ␣-helices and 7 ␤-sheets for SerB1 domain region and 16 ␣-helices and 15 ␤-sheets for SerB2 protein (Fig. 2B). The Ramachandran plot, a measure of the stereochemical parameters of modeled structures, revealed that 87 and 92% of the amino acid residues were in the favored region for SerB1 and SerB2 built models, respectively (data not shown). In addition, Verify_3D and Errat scores for SerB1 model were 81.02 and 85.85%, respectively, whereas these values were 96.21 and 94.06%, respectively, for SerB2 model. Overall, these results suggested that both models were of acceptable quality and suitable for molecular docking studies. Molecular docking of O-phospho-L-serine using Discovery Studios 2.5 revealed that amino acid residues Asp-127 (motif I), Thr-234 (motif II), Lys-279, and Asp-302 (motif III) in the case of SerB1 protein; residues Asp-185 (motif I), Ser-273 (motif II), Lys-318, and Asp-341 (motif III) in the case of SerB2 protein; and residues Asp-20 (motif I), Ser-109 (motif II), Lys-158, and Asp-179 (motif III) in the case of HPSP are part of their respective substrate binding pockets (Fig. 2C). In addition to these above mentioned critical interacting residues, docking studies also revealed that Val-186 and Ser-188 of SerB2 enzyme might also interact with Ophospho-L-serine.
Expression and Purification of M. tuberculosis Phosphoserine Phosphatases-To biochemically characterize PSP enzymes, pET28b-serB1 and pET28b-serB2 were transformed into BL-21 (DE3, plysS), and expression of recombinant protein in transformants was induced by addition of 1 mM IPTG. As shown in Fig. 3A, His 6 -SerB2 was expressed at high levels in soluble fraction and purified at levels Ͼ95% using nickel-nitrilotriacetic acid chromatography with a total yield of 8 mg/liter. However, we observed that expression of His 6 -SerB1 in E. coli transformants was very minimal and only detectable by immunoblot analysis using ␣-His 6 antibody. As expected, both His 6 -SerB1 and His 6 -SerB2 migrated at their predicted molecular masses of 42.0 and 45.0 kDa, respectively, on 10% SDS-PAGE (data not shown). To achieve better expression, serB1 was also cloned in other prokaryotic expression vectors such as pMALc2x or pGEX4T-1. However, as observed in the case of His 6 -SerB1, expression of both MBP-SerB1 and GST-SerB1 in IPTG-induced transformants was very minimal and nondetectable in Coomassie Brilliant Blue-stained 10% SDS-PAGE (data not shown).
Biochemical Characterization of SerB2 Enzyme-To determine kinetic parameters for SerB2 enzyme, steady state kinetics was performed by varying O-phospho-L-serine concentration in the presence of 1 M His 6 -SerB2 enzyme. Conversion of O-phospho-L-serine to L-serine by SerB2 followed Michaelis-Menten kinetics with a K m of 92.68 M and k cat of 8.83 min Ϫ1 (Fig. 3B). The catalytic efficient constant (k cat /K m ) for SerB2 enzyme was 0.0952 min Ϫ1 M Ϫ1 . These kinetic constants were observed to be lower in comparison with those obtained for PSP enzymes characterized from either Hydrogenobacter thermophiles (K m of 1.6 mM), P. gingivalis (K m of 2.0 mM and 2.6 mM for phosphoserine peptides), or Pseudomonas aeruginosa (K m of 207 M) (26,36,37). We also observed that maximal SerB2 activity was observed in the initial 5 min and that inclusion of 0.01% Triton X-100 enhanced SerB2 activity by 15-20% (Fig.   FIGURE 1. A, schematic representation of L-serine biosynthetic pathway in M. tuberculosis. The gene identifier and genetic essentiality of enzymes involved in L-serine biosynthesis in M. tuberculosis have been mentioned. PGM, phosphoglyceratemutase; PGDH, phosphoglycerate dehydrogenase. B, phylogenetic analysis of PSP proteins. The evolutionary history was inferred using the neighbor joining method using MEGA5 software, and distance was computed using the POISSON correction method and is shown as the units of number of amino acid substitutions per site. The branches are labeled with the protein accession number along with organism name. The bootstrap consensus tree inferred from 100 replicates is taken to represent the evolutionary history of the taxa. 3C). As expected, SerB2 displayed substrate preference for Ophospho-L-serine over O-phospho-L-threonine (Fig. 3D). We did not observe any P i release even at 80 M of O-phospho-Lthreonine (Fig. 3D). The amino acid residues of SerB2 enzyme predicted to interact with O-phospho-L-serine were mutated, cloned into either pET28b or pMALc2x, and purified as His 6tagged or MBP-tagged proteins. We observed that mutation of aspartic acid at position 185 and lysine at position 318 completely abolished SerB2 activity (Fig. 3E). As shown in Fig. 3E, mutation of aspartic acid at position 341, serine at position 273, and valine at position 186 abolished SerB2 activity by 80, 60, and 50%, respectively, as compared with wild type protein activity.
To examine the changes in secondary structural content (such as ␣-helix, ␤ sheet, or random coil) of SerB2 upon amino acid changes, both wild type and mutant proteins were analyzed using far-UV CD spectroscopy between wavelength range of 195 and 250 nm. As shown in Fig. 3F, SerB2 showed characteristic spectra of a protein consisting of a mixture of secondary structure elements such as ␣ helix and ␤ sheet. We observed that mutant proteins S273A and D341G showed similar spectra as that of the wild type protein. However, D185G and K318E proteins showed decreased and increased secondary structural content, respectively, in comparison with that of the wild type protein (Fig. 3F). These data suggest that reduced enzymatic activity of S273A and D341G proteins was not due to changes in their secondary structures but could be due to their altered interaction with O-phospho-L-serine as predicted by molecular docking. However, a decrease in activity of both D185G and K318E proteins could be due to the combined effect of both altered structure and interaction with O-phospho-L-serine.
Optimization of Assay Conditions for HTS-To optimize the assay conditions for HTS to identify novel SerB2 inhibitors, the influence of various parameters such as cations and buffer pH on its activity was evaluated in saturating O-phospho-L-serine concentration. We observed that inclusion of divalent ions such as Mg 2ϩ and Mn 2ϩ significantly enhanced SerB2 activity in comparison with inclusion of Ca 2ϩ , Zn 2ϩ , and Fe 3ϩ in the assay buffer. The optimum SerB2 activity was observed upon inclusion of 5 mM Mg 2ϩ or Mn 2ϩ ions in the assay buffer (Fig. 4A). In our assay conditions, increasing Fe 3ϩ ion concentration in the buffer did not affect SerB2 activity, whereas significant reduction in activity was observed upon inclusion of Zn 2ϩ ion in assay buffer in a dose-dependent manner (Fig. 4A). We also observed a slight decrease in SerB2 dephosphorylating activity upon increasing concentration of Ca 2ϩ ion in the assay buffer (Fig. 4A). To determine pH optima for SerB2 enzyme, assays were performed in buffers of pH ranging from 6.0 -9.0 and optimum activity was achieved at pH 7.5 (Fig. 4B). Therefore, the optimum conditions for HTS assays were 100 mM Tris, pH 7.5, 5 mM MgCl 2 , 5 mM DTT, and 0.01% Triton X-100. For further optimization and validation of HTS assays, SerB2 activity was determined in the presence of increasing concentration of known PSP inhibitors such as DL-AP 3 and sodium orthovandate (26,33,34). We observed that both DL-AP 3 and sodium orthovandate inhibited SerB2 activity by 50% at the concentration of 458 and 670 M, respectively (data not shown). The IC 50 values obtained for TABLE 2 The locus of genes for enzymes involved in L-serine biosynthesis in various mycobacterial genomes. these two inhibitors for SerB2 enzyme were comparable with those obtained for other PSP enzymes (26,33,34).

HTS Assays and Identification of Inhibitors for SerB2
Enzyme-In our preliminary screening, we determined the percentages of inhibition for each compound (belonging to either NCI diversity set or mechanistic set or natural product set) at 100 M concentration, and the compounds that inhibited SerB2 activity by at least 50% were considered to be primary hits. In our initial assays, the majority of the compounds were inactive, whereas 21 compounds inhibited SerB2 activity by Ͼ50% in vitro (Fig. 4, C and D). We observed that SerB2 enzymatic activity was inhibited by Ͼ80, 60 -80, and 40 -60% in the presence of 5 compounds, 4 compounds, and 12 compounds, respectively (Fig. 4, C and D). However, in our repeat in vitro assays, only 10 of these 21 compounds inhibited SerB2 enzyme by at least 50% at 100 M concentration. These 10 active scaffolds were structurally different from known PSP inhibitors (Fig. 5).
As expected, these primary hits inhibited SerB2 activity in a dose-dependent manner with at least 50% inhibition observed at the highest concentration. As shown in Table 3, CID 4 and CID 8 were most potent in our in vitro SerB2 inhibition assays with an IC 50 value of 4.12 Ϯ 1.2 and 3.89 Ϯ 1.2 M, respectively. The IC 50 values for remaining primary hits varied between 10 and 40 M with CID 2 showing least inhibitory activity in our in vitro enzymatic assays (Table 3). To determine specificity of these primary hits, we next evaluated their ability to inhibit either M. tuberculosis PstP (Rv 0018c) or alkaline phosphatase enzyme. Both PstP and alkaline phosphatase were suitable control for these assays because they belong to different phosphatase families. Alkaline phosphatase is a hydrolase responsible for removing phosphate group from various molecules including nucleotides, proteins, and alkaloids. PstP belongs to PP2C phosphatase family that strictly requires Mn 2ϩ ion for binding. We observed that CID 1, CID 5, CID 6, CID 7, CID 8, CID 9, and FIGURE 2. A, multiple sequence alignment among PSP proteins. Multiple sequence alignment among PSP proteins from various microorganisms was performed using Clustal Omega software. Highly conserved residues among PSP enzymes from various bacterial species are shaded in black, whereas light gray shading denotes a level of conservation among these proteins. The accession numbers for these proteins are EFC04663, Staphylococcus aureus; YP_177732, M. tuberculosis SerB1; EDN61796, Saccharomyces cerevisiae; AAA97284, E. coli; ZP_18026828, Vibrio cholerae; YP_005780390, Helicobacter pylori; NP_217558, M. tuberculosis SerB2; YP_002342576, Neisseria meningitidis; YP_004510496, P. gingivalis; and NP_253647, P. aeruginosa. B and C, superimposition of modeled structures of SerB1, SerB2, and HPSP (Protein Data Bank code 1L8L). B, the built SerB1 model (green), SerB2 model (pink), and HPSP (yellow) were superimposed and visualized using PyMOL. C, the binding site for O-phospho-L-serine in superimposed models is magnified. The critical amino acid residues of SerB1 (green), SerB2 (pink), and HPSP (yellow) predicted to interact with O-phospho-L-serine have been highlighted. SEPTEMBER 5, 2014 • VOLUME 289 • NUMBER 36

JOURNAL OF BIOLOGICAL CHEMISTRY 25155
CID 10 failed to significantly inhibit PstP enzyme even at 100 M concentration, whereas CID 2, CID 3, and CID 4 displayed equal inhibitory activity (ϳ80 -90%) for both SerB2 and PstP enzymes, in vitro (Fig. 6A). We observed that CID 1, CID 5, CID 6, CID 7, CID 8, CID 9, and CID 10 inhibited enzymatic activity of SerB2 and PstP enzymes by ϳ70 and 0 -20%, respectively. We also observed that none of these inhibitors were able to inhibit alkaline phosphatase activity even at 100 M concentration (Fig. 6A). We next performed SPR experiments to confirm binding of CID 1, CID 7, CID 8, CID 9, and CID 10 with SerB2 enzyme. The protein immobilization over the flow cell varied between 8,000 and 10,000 RU, and binding responses were recorded at 1 mM drug concentration. At this tested concentration, all the compounds showed increase in RU indicative of interaction with SerB2; however, the increase in RU varied for different compounds (Fig. 6B). We observed that CID 1, CID 7, and CID 8 displayed higher binding to SerB2 enzyme in comparison with CID 9 and CID 10 (Fig. 6B).
Cell Cytotoxicity, Antimycobacterial Activity, and HPSP Inhibition Studies-In our lactate dehydrogenase-based cell viability assays, we observed that primary hits CID 1, CID 5, CID 6, CID 7, CID 8, and CID 10 that specifically inhibited SerB2 enzyme in vitro were noncytotoxic to THP-1 macrophages even at 50 M concentration (Table 3). Scaffolds such as CID 2, CID 3, and CID 4 that inhibited both PstP and SerB2 activity by Ͼ90% were highly cytotoxic to THP-1 macrophages with TC 50 values of 5, 5, and 2.5 M, respectively (Table 3). A similar pattern of cell cytotoxicity for these scaffolds was also observed in Vero cell lines (Table 3). Next, these inhibitors were also evaluated for their ability to inhibit mycobacterial growth in vitro. In our MIC 99 determination assays, most of these compounds possessed modest activity (ranging from 2 to 25 M) against both M. tuberculosis H 37 Rv and M. bovis BCG Danish strains ( Table 3). As shown in Table 3, CID 5, CID 6, CID 8, and CID 10 were less active against mycobacteria in vitro in our whole cellbased assays, which might be attributed to lower intracellular concentration of drugs because of (i) their poor penetration, (ii) their effluxing out by various pumps, or (iii) their modification by intracellular enzymes. The most potent inhibitors in our whole cell-based assays were CID 1 and CID 7, both of which displayed MIC 99 values of 2 M against both M. tuberculosis and M. bovis BCG in vitro.
Next we performed kill kinetics assays in vitro by exposing actively growing mycobacteria to 10ϫ MIC 99 concentration of either CID 1 or CID 7 or isoniazid. As shown in Fig. 6C, both CID 1 and CID 7 were bactericidal in their mode of killing with ϳ10-fold killing observed after exposure of mycobacteria for 3 days. In our time kill experiments, 7 days of exposure to CID 1 and CID 7 led to approximately ϳ80and 20-fold killing, respectively (Fig. 6C). As expected, no significant growth inhi- bition was observed in the presence of DMSO during the course of experiment (Fig. 6C). Because M. tuberculosis is an intracellular pathogen, we next evaluated the ability of both CID 1 and CID 7 to kill bacteria in THP-1 macrophages at either 4ϫ or 16ϫ MIC 99 concentration. In our macrophage experiments, both CID 1 and CID 7 were able to arrest M. bovis BCG replication in a dose-dependent manner. As shown in Fig. 6D, at 4 days after drug exposure, ϳ100and 50-fold intracellular kill-ing was observed in the presence of 16ϫ MIC 99 concentration of CID 7 and CID 1, respectively.
Cross-reactivity of these inhibitors with the human homolog would be a major concern in further validation of SerB2 as a drug target. Therefore, we compared the ability of these active and noncytotoxic primary hits to inhibit HPSP enzyme in vitro at 100 M concentration. As shown in Fig.  7A, both CID 7 and CID 10 were nonselective PSP inhibitors,   SEPTEMBER 5, 2014 • VOLUME 289 • NUMBER 36

High Throughput Screen Identifies PSP Specific Inhibitor
inhibiting both HPSP and SerB2 enzymes to a similar extent of 70% at 100 M concentration. However, the rest of the compounds CID 1, CID 5, CID 8, and CID 9 were highly specific in their ability to inhibit SerB2 enzyme in compari-son with HPSP enzyme even at 100 M concentration (Fig.  7A). These results suggest that despite the presence of a human homolog, subtle differences exist between secondary structures of these two enzymes, which could be explored  further for identification of inhibitors with more potency and specificity toward SerB2 enzyme. Molecular Docking Studies of Scaffolds on M. tuberculosis SerB2 Model Protein and HPSP Enzyme-Molecular docking studies were performed for CID 1 (clorobiocin), novobiocin, CID 5, CID 7 (rosaniline), CID 8, CID 9, and CID 10 using SerB2 built model and HPSP protein as described under "Experimental Procedures." The binding free energies for interaction of these inhibitors with SerB2 protein ranged from Ϫ4.44 to Ϫ5.14 kcal/mol, which is comparable with the binding energy for interaction of O-phospho-L-serine with SerB2 (Ϫ6.89 kcal/ mol; Table 3). Because amino coumarins are clinically utilized antibiotics with tolerable toxicity, we were particularly intrigued by the ability of clorobiocin to inhibit SerB2 enzyme in vitro. This class of compounds that includes novobiocin, clorobiocin, and coumermycin is known to inhibit DNA topoisomerase and heat shock proteins by binding to their nucleotide binding pockets (38,39). Interestingly, novobiocin, a compound structurally similar to clorobiocin, did not inhibit SerB2 activity even at 10-fold higher concentration in our in vitro assays (Fig. 7B). Molecular docking of a SerB2 built model with clorobiocin and novobiocin predicted that the difference in ability of these structural analogs to inhibit SerB2 enzyme could be attributed to (i) interaction of the Asp-341 residue of SerB2 enzyme with clorobiocin via hydrogen bond formation and (ii) better fit of the pyrrole ring of clorobiocin in comparison with the substituted amino group of novobiocin in the substrate binding pocket. In addition, Lys-361 and Arg-365 residues of SerB2 enzyme might interact with clorobiocin via hydrogen bond formation and Asp-187 and Glu-197 residues were observed to be closely associated (2.2 and 2.8 Å, respectively) with clorobiocin (Fig. 7, C and D). In concordance with our in vitro activity results, molecular docking of clorobiocin in HPSP revealed that Asp-179 (identical to Asp-341 in SerB2) is interacting with O-phospho-L-serine but not with clorobiocin ( Figs. 2C and 7E). In addition, further analysis revealed that the conserved FDVDST motif forms a different secondary structure (right-handed helix in case of HPSP) as compared with SerB2 protein, which might block the accessibility of clorobiocin to the substrate binding pocket of HPSP (Fig. 7E).
Molecular docking studies revealed that CID 5, CID 7, CID 8, CID 9, and CID 10 also interact with critical Asp-341 and Asp-187 residues of SerB2 protein (Fig. 8, A-E). In addition to these interactions, CID 5 is also interacting with Val-186 and Glu-197 via hydrogen bond and salt bridge interactions. Molecular docking studies revealed that CID 7 might also interact with Glu-197 and Val-186 residues of SerB2 protein, and there might be some electrostatic interactions with Glu-214, because it resides in close proximity of 3.8 Å (Fig. 8B). As shown in Fig. 8C, we observed that CID 8 might also interact with the Lys-361 residue of SerB2 enzyme through -cation interaction, whereas CID 9 might interact with Glu-194, Ser-188, and Val-186 residues of SerB2 protein through hydrogen bond formation (Fig.  8D). As shown in Fig. 8E, CID 10 might also be possibly interacting with Glu-197 and Glu-214 residues through electrostatic interaction because these two residues seem to be in close contact (2.23 and 1.99 Å, respectively). Despite the lack of structural similarities among these primary hits, these scaffolds possess a substructure that fits well in the SerB2 modeled protein. Molecular docking of CID 7 with HPSP crystal structure revealed that Asp-20, Asp-22, Asp-179, and Ala-71 might interact with rosaniline via hydrogen bond formation, and binding free energy for this interaction was Ϫ5.7 kcal/mol (Fig. 9B). In concordance with our in vitro activity assays, we observed that CID 5, CID 8, and CID 9 are not interacting with known HPSP critical residues, thereby explaining their inability to inhibit HPSP enzyme (Fig. 9, A, C, and D). Molecular docking of CID 10 with HPSP revealed that it might interact with Val-56 and Thr-182 residues via hydrogen bond formation and with Lys-158 through cationinteraction (Fig. 9E). Even though Val-56 and Thr-182 of HPSP are not conserved with SerB2 enzyme, Lys-158 is a conserved active site residue between SerB2 and HPSP (as per pair wise sequence alignment studies).
Validation of SerB2 Small Molecule Interactions Predicted by Molecular Docking Studies-To validate SerB2-small molecule interactions predicted by molecular docking, Lys-361, Arg-365, Glu-214, and Asp-187 were mutated to alanine residue as described under "Experimental Procedures." As shown in Fig.  10A, mutation of Asp-187 and Glu-214 reduced SerB2 activity by 50% as compared with wild type protein, whereas the enzymatic activity of K361A-SerB2, R365A-SerB2, and E197A-SerB2 were almost similar to that of wild type protein. Far-UV CD studies revealed that mutation of these amino acids did not significantly alter the folded state of these mutant proteins except for E197A-SerB2 where we observed increase in secondary structure content (Fig. 10B). These results suggest that reduction in activity of D187A-SerB2 and E214A-SerB2 could be due to their altered interaction with O-phospho-L-serine (Fig. 10, A and B). In our enzymatic assays, we observed that mutation of Asp-187 and Asp-341 prevented efficient binding of CID 1, CID 5, CID 7, CID 8, CID 9, and CID 10 to the binding pocket of these mutant proteins (Fig. 10, C-H). We also observed that mutation of Glu-197 to alanine residue reduced the ability of CID 1, CID 5, and CID 7 to inhibit dephosphorylation activity of SerB2 enzyme. As shown in Fig. 10 (E and F), Val-191 and Lys-361 were important for interaction of SerB2 with CID 7 and CID 8, respectively. In concordance with our docking studies, we report that Val-186 is important for interaction of SerB2 with CID 9, whereas Glu-214 is critical for interaction of CID 7 and CID 10 with SerB2 enzyme (Fig. 10, E, G,  and H).

DISCUSSION
The development of novel inhibitors for essential and conserved M. tuberculosis pathways is one plausible solution to shorten duration of TB chemotherapy and eradicate drug-resistant TB. The advent of new computational methods, combinatorial synthetic chemistry approach and whole cell-and target-based HTS assays, have led to identification of several antitubercular scaffolds that are currently being evaluated in different stages of clinical trial. Numerous studies have shown that M. tuberculosis strains deficient in enzymes involved in various amino acid biosynthetic pathways are compromised in their ability to infect mice in comparison with the ability of parental strain (40 -42). In addition, several of these enzymes involved in amino acid biosynthesis are being currently explored for development of more potent antitubercular scaffolds (43)(44)(45)(46)(47)(48). L-Serine biosynthesis is an attractive and unexplored anti-microbial drug target because L-serine is not only involved in protein synthesis but also acts as precursor for various cellular metabolites (15, 49 -54). The enzymes involved in L-serine biosynthetic pathway are widely conserved across various mycobacterial species including Mycobacterium leprae, an organism that has undergone massive gene decay, thereby suggesting that these enzymes are indispensable and are essential for survival of bacteria in the host. This is the first study where a detailed biochemical characterization of PSP homolog from M. tuberculosis has been performed. Phylogenetic and sequence alignment analysis revealed that both SerB1 and SerB2 proteins are distantly related to each other and share an identity of 18 and 27%, respectively, with HPSP enzyme. Multiple sequence alignment analysis revealed that HAD specific motifs responsible for Mg 2ϩ ion binding, phosphoprotein formation, and stabilization are present in both SerB1 and SerB2. Despite several attempts, we were unable to express SerB1 in detectable amounts as either His 6tagged, GST-tagged, or MBP-tagged proteins. We observed that SerB2 had a substrate preference for O-phospho-L-serine over O-phospho-L-threonine and displayed lower kinetic constants in comparison with PSP enzymes previously characterized from H. thermophiles, P. gingivalis, and P. aeruginosa (26, , and CID 10 (E) was performed as described under "Experimental Procedures." We did not observe any significant interactions between CID 5, CID 8, or CID 9 and HPSP enzyme. The H-bond interactions between CID 7, CID 10, and amino acid residues of HPSP enzyme have been highlighted with yellow dotted lines. 36,37). The optimum SerB2 activity was observed in the pH range of 7.0 -8.5 and in the presence of 5 mM Mg 2ϩ or Mn 2ϩ ions. The activity of SerB2 enzyme was unaltered by the inclusion of Fe 3ϩ ion in the assay conditions. However, inclusion of Zn 2ϩ ion in the assay buffer abolished SerB2 enzymatic activity in a dose-dependent manner that might be attributed to disruption of the SerB2 secondary structure, thereby preventing binding of O-phospho-L-serine to its active site. We also observed reduction in SerB2 activity upon inclusion of Ca 2ϩ ion in the assay buffer, which might be attributed to disruption of nucleophilic attack by the conserved Asp residue (motif I) on the phosphate group of O-phospho-L-serine as reported earlier in the case of HPSP (22). The activity of SerB2 enzyme was enhanced by 15-20% by inclusion of nonionic detergent in assay conditions that might be attributed to stabilization of SerB2 secondary structure. Molecular docking and in vitro enzymatic assays using purified mutant proteins revealed that amino acid resi-dues Asp-185, Asp-187, Val-186, Ser-273, Lys-318, Glu-214, and Asp-341 are critical for SerB2 dephosphorylation activity.
Despite the importance for PSP enzymes in biosynthesis of L-serine, these enzymes have not been extensively studied as an antimicrobial drug target except for a report where dihydroquinolinone derivatives were shown to inhibit SerB enzymes from P. gingivalis (55). In the present study, HTS was performed using 2300 compounds belonging to a library received from National Cancer Institute Developmental Therapeutic Program, and we identified 10 scaffolds that inhibited dephosphorylation activity of SerB2 enzyme. The interaction of some of these chemical entities with SerB2 protein was further confirmed using SPR studies. A subset of these primary hits inhibited M. tuberculosis growth in vitro and displayed low cytotoxicity toward both THP-1 and Vero cell lines. The best two chemical entities in our primary hits, clorobiocin and rosaniline with a therapeutic index (T i , ratio of TC 50 /MIC 99 values) of FIGURE 10. A, phosphoserine phosphatase activity of wild type and mutant SerB2 proteins. P i release assays were performed using either 1 M of wild type or mutant SerB2 proteins in the presence of 100 M of O-phospho-L-serine. P i release was measured in enzymatic reactions, and the data are depicted as the means Ϯ S.E. obtained from three independent experiments. B, secondary structural analysis of wild type and mutant SerB2 protein using far-UV CD spectroscopy. The CD studies were performed as described above in the legend for Fig. 3F. We observed that except E197A, other mutant proteins did not show much significant alterations in their secondary structure as compared with the wild type protein. C-H, inhibition assays in the presence of SerB2 specific inhibitors. Activity assays for wild type and mutant proteins were performed in the absence and presence of CID 1 (C), CID 5 (D), CID 7 (E), CID 8 (F), CID 9 (G), or CID 10 (H) at 200 M concentration. The data presented in each panel are the means Ϯ S.E. of percentages of inhibition obtained from three independent experiments. Ͼ21, were evaluated for their ability to kill mycobacteria in liquid cultures and in infected macrophages. Both clorobiocin and rosaniline were bactericidal in their mode of killing and inhibited bacterial growth in infected macrophages in a dosedependent manner. NSC 76027, the most potent compound in our in vitro enzymatic assays with an IC 50 value of 3.8 M, was highly specific for SerB2 enzyme but displayed MIC 99 value of 150 M against both M. tuberculosis H 37 Rv and M. bovis BCG, which might be attributed to its poor penetration or intracellular stability. NSC 693172 displayed an IC 50 value of 4 M in our enzymatic assays but was highly cytotoxic in THP-1 cells, which might be due to its ability to inhibit other phosphatases in a nonspecific manner. Therefore, future experiments would involve designing of their structural analogs with more potent in vitro SerB2 specific and intracellular activity.
To the best of our knowledge, this is the first study where we show that clorobiocin inhibits SerB2 enzyme of M. tuberculosis, a non-ATP binding protein in addition to its other known bacterial targets such as DNA gyrase, heat shock protein, and UDPgalactose-4Ј-epimerase (38,39,56). Interestingly, novobiocin, a close structural analog of clorobiocin failed to inhibit SerB2 enzymatic activity even at 10-fold higher concentration, which might be attributed to hydrogen bond formation between Asp-341 residue of SerB2 protein with clorobiocin and a better fit of its pyrrole ring in the SerB2 substrate binding pocket. Similar subtle differences were also observed in inhibition of UDP-galactose 4Ј-epimerase by clorobiocin and novobiocin, respectively (56). Further experiments to solve crystal and co-crystal structure of SerB2 enzyme either alone or with clorobiocin would be useful to understand such subtle differences in the ability of these aminocoumarins to inhibit SerB2 enzyme. We propose that use of scaffolds like clorobiocin rather than novobiocin might be more effective for eradication of drug-resistant TB. Molecular docking of CID 5, CID 7, CID 9, and CID 10 in the SerB2 built model revealed that these scaffolds interact with common critical residues Asp-187 and Asp-341 of SerB2 protein via hydrogen bond formation. Because we did not observe much difference in the far-UV CD spectra for K361A-SerB2, E214A-SerB2, D187A-SerB2, D341G-SerB2, and wild type protein, we speculate that the observed reduced inhibition for these mutant proteins could be due to loss of interaction of these mutant proteins with their respective scaffolds. However, the altered folded state of E197A-SerB2 could have contributed to the loss of inhibition observed in the case of CID 1, CID 5, and CID 7.
Validation of enzymes with human homologs is hampered by lack of safe and highly specific inhibitors. Interestingly except CID 7 and CID 10 all other scaffolds were highly specific in their ability to inhibit M. tuberculosis SerB2 enzyme. As expected, we did not observe any significant interactions between CID 5, CID 8, and CID 9 with HPSP enzyme, therefore explaining their ability to specifically inhibit SerB2 enzyme. Because RNAi-mediated inhibition of PSAT and PSP enzymes reduces tumor formation in breast cancer, we propose that nonspecific inhibitors (NSC 93739 and NSC 165701) could be exploited further for treating such disorders or infections caused by P. gingivalis (57). Based on our observations, future experiments would involve (i) screening of more libraries to identify novel PSP inhibitors and (ii) structure-activity relationship studies involving these SerB2 specific scaffolds in an attempt to design analogs that display enhanced potency, specificity toward SerB2 enzyme, and better intracellular activity. These findings suggest that despite being widely conserved, enzymes involved in energy metabolism can be targeted to combat the problem of drug resistance in intracellular pathogens. Collectively, these results demonstrate feasibility of HTS to obtain novel PSP inhibitors that would be useful for development of anti-mycobacterial agents.