The Apo-structure of the Low Molecular Weight Protein-tyrosine Phosphatase A (MptpA) from Mycobacterium tuberculosis Allows for Better Target-specific Drug Development*

Background: Low molecular weight protein-tyrosine phosphatase A, MptpA, is a key virulence factor of Mycobacterium tuberculosis. Results: We determined the apo-MptpA NMR structure and identified the binding site of kinase PtkA and of inorganic phosphate. Conclusion: There is a major rearrangement in the D-loop in the apo-state of MptpA. Significance: Detailed understanding of the intramolecular architecture and intermolecular interactions of bacterial apo-state phosphatases is crucial for design of novel anti-infectives. Protein-tyrosine phosphatases (PTPs) and protein-tyrosine kinases co-regulate cellular processes. In pathogenic bacteria, they are frequently exploited to act as key virulence factors for human diseases. Mycobacterium tuberculosis, the causative organism of tuberculosis, secretes a low molecular weight PTP (LMW-PTP), MptpA, which is required for its survival upon infection of host macrophages. Although there is otherwise no sequence similarity of LMW-PTPs to other classes of PTPs, the phosphate binding loop (P-loop) CX5R and the loop containing a critical aspartic acid residue (D-loop), required for the catalytic activity, are well conserved. In most high molecular weight PTPs, ligand binding to the P-loop triggers a large conformational reorientation of the D-loop, in which it moves ∼10 Å, from an “open” to a “closed” conformation. Until now, there have been no ligand-free structures of LMW-PTPs described, and hence the dynamics of the D-loop have remained largely unknown for these PTPs. Here, we present a high resolution solution NMR structure of the free form of the MptpA LMW-PTP. In the absence of ligand and phosphate ions, the D-loop adopts an open conformation. Furthermore, we characterized the binding site of phosphate, a competitive inhibitor of LMW-PTPs, on MptpA and elucidated the involvement of both the P- and D-loop in phosphate binding. Notably, in LMW-PTPs, the phosphorylation status of two well conserved tyrosine residues, typically located in the D-loop, regulates the enzyme activity. PtkA, the kinase complementary to MptpA, phosphorylates these two tyrosine residues in MptpA. We characterized the MptpA-PtkA interaction by NMR spectroscopy to show that both the P- and D-loop form part of the binding interface.

Tuberculosis is a chronic infectious disease caused by the facultative pathogen Mycobacterium tuberculosis, which is one of the oldest known pathogens, and has remained a major global health problem, with an estimated total of up to 8.8 million new cases and 1.4 million tuberculosis-related deaths annually (1). Although there are several treatment options available, the emergence of multidrug-resistant and extensive drug-resistant M. tuberculosis strains prioritizes the need for the development of new strategies to treat this widespread disease (2,3). One of the emerging strategies in the drug development process is the approach to target bacterial virulence factors, which play a key role in bacterial entry and survival in the host cell (4). Protein-tyrosine phosphatases (PTPs) 3 and protein-tyrosine kinases act as major virulence determinants by modifying host-pathogen signaling pathways (5)(6)(7)(8) and are considered to be potential drug targets for anti-tuberculosis therapeutics (9). The genome of M. tuberculosis encodes two PTPs, MptpA and MptpB, which are secreted during infection and act as key virulence factors important for bacterial survival during macrophage infection (6,10). Protein phosphorylation and dephosphorylation regulate various cellular functions, including pathogenicity (9). The dephosphorylation is catalyzed by different enzymes, which are categorized due to their specificity and structural properties into (i) serine/threonine and tyrosine phosphatases, (ii) dual specific phosphatases, and (iii) polymerase-histidinol phosphatases. The tyrosine phosphatases are represented by two major subclasses, the HMW-PTPs (ϳ30 kDa) and LMW-PTPs (ϳ20 kDa). MptpA (17.5 kDa) is a member of the LMW-PTP family. The MptpA deletion mutant of M. tuberculosis reveals that MptpA is essential for infection of host macrophages (11). The phagosome-lysosome fusion is important for macrophage microbicidal activity, and some organisms, including M. tuberculosis, evade this process for survival (12). MptpA dephosphorylates the host macrophage protein VPS33B (vacuolar protein sorting 33B), a protein that is responsible for the regulation of phagosomelysosome fusion and trafficking (11). Dephosphorylation of VPS33B by MptpA is mediated by subunit H of the macrophage vacuolar H ϩ -ATPase machinery, a multisubunit protein complex in the phagosome membrane-driven luminal acidification (13). Inactivation of VPS33B by dephosphorylation leads to arrest of the phagosome-lysosome fusion, thereby inhibiting the cellular response to infection. Recently, it has been shown that protein-tyrosine kinase A (PtkA) located immediately upstream from MptpA in the same operon, phosphorylates MptpA on two adjacent conserved tyrosine residues (14). Although the exact role of this phosphorylation is not yet determined, it was speculated that phosphorylation might increase the activity of MptpA or regulate its secretion (14). The important role of MptpA as mediator of M. tuberculosis virulence is evident, which makes it a potential drug target.
The HMW-and LMW-PTPs share a common cysteine-catalyzed reaction mechanism. However, except for the catalytically important phosphate binding CX 5 R motif called the P-loop and the critical aspartic acid residue in the D-loop, they exhibit no other sequence similarity (15,16). The engagement of the phosphotyrosine substrate promotes a major conformational change in the PTP catalytic motifs (17), and notably, it has been speculated that modifications in the D-loop influence the catalytic activity of the protein (15,16,18). In most HMW-PTPs, the binding of ligand or substrate at the P-loop triggers a large structural change in the D-loop and swings its "open" conformation by ϳ10 Å to a "closed" conformation (18,19). In contrast, the role of the D-loop in LMW-PTPs is poorly understood (20). This is not surprising because all of the previously reported structures of LMW-PTPs (21)(22)(23)(24)(25)(26)(27), including the x-ray structure of MptpA (28), are obtained in the presence of ligands (phosphate/anions or substrate mimetics), and hence, the conformation (and dynamics) of the D-loop in the absence of ligand or substrate is unknown. In addition, a hallmark of the LMW-PTPs is the presence of two adjacent and well conserved tyrosine residues in the D-loop, whose phosphorylation status regulates its functional activity (29,30). Upon phosphorylation, two effects are observed for several LMW-PTPs: (i) increase in activity (29,30) and (ii) negative regulation (31). The regulation of the human LMW-PTP HCPTP-A, which has 37% sequence similarity to MptpA, is carried out by the phosphorylation on these two adjacent tyrosine residues (Tyr 131 and Tyr 132 ) (30). The in vitro phosphorylation of HCPTP-A Tyr 131 /Tyr 132 by pp60 v-Src has been reported and shows different effects on the enzyme activity. Whereas phosphorylation of Tyr 131 increases the enzyme activity 25-fold, phosphorylation of Tyr 132 does not affect the enzyme activity but leads to the recruitment of an adaptor protein Grb2 Src homology 2 domain, important for signal transduction. These findings indicate that both tyrosine residues are of crucial importance for the structural and functional regulation of HCPTP-A. Biochemical assays showed that the residues Tyr 128 and Tyr 129 in MptpA are phosphorylated by the kinase PtkA (14). Both the phosphorylation of these tyrosines and the structural conformation of the D-loop are supposed to play an important role in the structure-function relationship of MptpA.
Here, we present a high resolution solution nuclear magnetic resonance (NMR) structure of ligand-free MptpA, which reveals that the D-loop adopts an open conformation compared with the closed conformation observed in ligand-bound structures. Based on NMR chemical shift perturbations (CSPs), we were able to map the MptpA binding site for inorganic phosphate (P i ), a competitive inhibitor of LMW-PTPs, and to clarify the involvement of both the P-and D-loop in the catalytic mechanism. Furthermore, we show by mass spectrometry that PtkA can phosphorylate MptpA and show by NMR that the Pand D-loop of MptpA form part of the PtkA binding interface.
Our results report a previously unknown conformation of the D-loop in LMW-PTPs and refine our view of how the D-loop shuttles between the open and closed conformation upon ligand binding. The substantially different conformation of D-loop amino acids in the apo-structure, presenting the hydrophobic residues Tyr 128 and Tyr 129 in other conformations, allows for more specific structure-based drug design strategies to the apo-state of MptpA in the future.

EXPERIMENTAL PROCEDURES
MptpA Cloning, Expression, and Purification-The plasmid pET16bTEV vector containing the MptpA sequence according to RV2234 was obtained from Dr. Anil Koul. From the pET16bTEV-MptpA vector, the base pairs corresponding to full-length MptpA (amino acid residues 1-163; see Fig. 1A) were amplified by PCR using 5Ј-AAA CCA TGG GGA TGT CTG ATC CG-3Ј and 5Ј-TTA TTG CTC AGC GGT GGC AGC A-3Ј as primers. The PCR product was double-digested with NcoI/BlpId and cloned into the modified pKM263 (His 6tagged ProtGB1-TEV between NdeI and XhoI) vector, which contains the ProtGB1 for enhancement of solubility of the expressed fusion partner. The authenticity of the clone was validated via nucleotide sequencing. Protein expression was performed in Escherichia coli strain BL21(DE3)pLysS (Novagen) containing the desired plasmid using Lysogeny broth (LB) medium at 37°C supplemented with 100 g/ml ampicillin and 34 g/ml chloramphenicol. For induction of protein expression, 1 mM isopropyl-1-thio-␤-D-galactopyranoside at A 600 of 0.6 -0.7 was added. The culture was incubated in 5-liter Erlen-meyer flasks for 12 h with aeration (150 rpm) at 16°C before harvesting the cells by centrifugation (6,000 ϫ g, 30 min, 4°C). The cell pellet was resuspended in lysis buffer (300 mM NaCl, 50 mM Tris-HCl (pH 8.0), 10 mM ␤-mercaptoethanol) with the addition of complete protease inhibitor (1 tablet/100 ml of EDTA-free, Roche Applied Science). After cell disruption (Microfluidizer, 15,000 p.s.i., 2 cycles) the soluble fraction was isolated from cell debris by centrifugation (18,000 ϫ g, 45 min, 4°C). The supernatant was applied to a Ni-NTA FastFlow column (GE Healthcare) following the manufacturer's recommendations. His 6 -tagged ProtGB1 was removed by tobacco etch virus (TEV)-protease cleavage via dialysis overnight at 4°C (300 mM NaCl, 50 mM Tris-HCl (pH 8.0), 10 mM ␤-mercaptoethanol) and excluded by a Ni-NTA column. Further purification of the protein was performed by gel filtration on a HiLoad 26/60 Superdex 75 preparative grade column (GE Healthcare) in 25 mM HEPES (pH 7.0), 150 mM NaCl, and 1 mM DTT. The protein-containing fractions were pooled, concentrated, and stored at Ϫ80°C or immediately used for experimental procedures. Doubly ( 13 C, 15 N) labeled protein was expressed in M9 (32) minimal medium containing 15 NH 4 Cl (1 g/liter) as nitrogen and [ 13 C]glucose (2 g/liter) as sole carbon source.
Phosphatase Assay-Phosphatase activity was determined by adding 20 M MptpA to the buffer (25 mM Tris-HCl (pH 7.0), 50 mM NaCl, 2 mM EDTA) containing 25 mM p-nitrophenyl phosphate as substrate. The absorbance at 410 nm in a time-depen-dent manner was followed using a UV spectrometer (Varian, Cary50Bio) (supplemental Fig. S2). In parallel, assays with samples lacking MptpA were performed.
Structure Calculation-Structure calculations were performed using the software packages CYANA (54 -56) and ARIA/CNS (57,58). Initial structure calculations and optimizations were done independently with CYANA version 3.9 and ARIA version 1.2. The final structure calculation was performed with CYANA version 3.9 and refined using ARIA version 1.2 protocols. The MptpA resonances were manually assigned and used as input together with unassigned NOESY peak lists for fully automated NOESY resonance assignment and calibration. The three-dimensional 15 N and 13 C separated NOESY spectra were peak-picked using the restricted peakpicking routine of SPARKY version 3.114 (59). The NOESY peak lists were refined by manual inspection of the resonances around crowded regions and the diagonal. Furthermore, all of the resonances between 4.72 and 4.76 ppm were excluded because of overlap with the water signal. The chemical shift tolerances were set to 0.10 ppm for the heavy atoms and 0.02 and 0.03 ppm for the bound protons and other protons, respectively. In addition to NOE data, hydrogen bond distances and amply defined dihedral angle restraints (based on TALOSϩ predictions (60,61) and confirmed with NOESY resonance assignments and initial structure calculations) as well as 1 D (H,N) residual dipolar coupling and 3 J(H N ,H ␣ ) coupling constant (Karplus relation) restraints were included in the standard structure calculation with CYANA version 3.9 (100 structures per iteration, 15,000 refinement steps). The final bundle of the 20 best structures was used as input for optimization with CNS version 1.1 (62) using the ARIA version 1.2 setup and protocols for refinement in explicit water (63). The NOE distance restraints were converted to ARIA/CNS format. Next to the 1 D(H,N) residual dipolar couplings (anisotropy, Ϫ18.72, rhombicity 0.17), 3 J(H N ,H ␣ ) coupling constants, hydrogen bonds, and dihedral angle restraints, T 1 /T 2 relaxation values were also included as diffusion anisotropy restraints (anisotropy, 1.25; rhombicity, 0.39; c ϭ 12.12) in this refinement stage. The standard protocols and allhdg5.3 force field were used with optimized potentials for liquid simulations non-bonded parameters. The final structure bundle was analyzed with PROCHECK version 3.5.4 (64,65) and CING (66).

NMR Resonance and NOE Cross-peak Assignment of MptpA-
The 1 H, 15  and completed the assignment of the aliphatic side chain resonances using (H)CC(CO)NH-TOCSY (37,38) and H(C)CH-TOCSY (39). These experiments were all performed at T ϭ 303 K, which allowed us also to extend the backbone amide resonance assignment of MptpA with a nearly complete side chain assignment. We were able to assign 96.8% of all proton resonances at pH 7.0 and T ϭ 303 K (Fig. 1B). In the two-dimensional 1 H, 15 N TROSY experiment, 151 backbone amide protons of 156 expected resonances were resolved, and 15 side chain resonances were detected. The resonances for a stretch of amino acids, including Thr 12 , Gly 13 , and Arg 17 (interestingly being part of the active site), as well as the single amino acids Thr 119 and His 132 were missing, although for all of those except Arg 17 the side chain assignment could be achieved. For the automated NOESY cross-peak assignment, peak lists containing intensities were generated from the three-dimensional 15 N and 13 C separated NOESY spectra by using the restricted peakpicking routine of SPARKY. In total, 11,735 peaks were selected as input data for CYANA, whereas peaks close to the H 2 O signal (4.72-4.76 ppm) were excluded from further analysis. The automated NOESY assignment routine assigned 7,764 peaks, resulting in 4,821 NOE-based distance restraints (on average about 29 per residue). After the structure calculation with CYANA and refinement in explicit water with CNS using ARIA protocols, a bundle of 20 structures ( Fig. 2A) with an average backbone RMSD of 0.22 Ϯ 0.04 Å and all heavy atom RMSD of 0.62 Ϯ 0.04 Å, for the core of the protein (residues 3-114 and 132-158) excluding the flexible termini and part of the D-loop, was calculated ( Table 1).
Comparison of Apo-and Ligand-bound MptpA-The available x-ray structures of MptpA are in complex with either chloride (PDB entry 1U2P) or glycerol (PDB entry 1U2Q) (28). Comparison of MptpA x-ray structure (PDB entries 1U2P and 1U2Q) and our solution structure bundle (PDB entry 2LUO) reveals an RMSD of 1.31 Ϯ 0.05 and 1.34 Ϯ 0.05 Å (all heavy atoms and backbone residues 4 -158), respectively. We performed a structural alignment, based on our structure, of only the P-and D-loop motifs, which confirmed the conserved P-loop structure (0.31 Ϯ 0.06 and 0.27 Ϯ 0.03 Å) but revealed major deviations in the D-loop orientation represented by an RMSD of 2.14 Ϯ 0.21 and 1.92 Ϯ 0.19 Å (Table 2). Detailed analysis of the x-ray structures of MptpA (PDB entries 1U2P and 1U2Q) (28) revealed crystal packing artifacts concerning the D-loop region. The crystal packing restricts the orientation of the D-loop residues Glu 125 and Asp 126 . The structural comparison of the D-loop conformation in solution and in the x-ray structures of MptpA (PDB entries 1U2P and 1U2Q) (28) reveals major deviations for the residues Glu 125 , Asp 126 , Pro 127 , Tyr 128 , Tyr 129 , and Gly 130 . Notably, the side chain orientation of residues Tyr 128 and Tyr 129 in the D-loop is different as well as for Trp 48 in the W-loop. Specific distance measurements between the residues in the D-loop and residues in the P-loop and W-loop clearly indicate this and show the more open confirmation for the binding pocket (Fig. 3, A and B). In particular, the distance between Trp 48 N ⑀1 and Tyr 128 O is nearly 13 Å dis-placed. Consequently, the surface surrounding the active site is influenced because the hydrophobic residues Tyr 128 /Tyr 129 are pointing in the opposite direction of the P-loop (Fig. 4, E and F). Furthermore, the key residue Asp 126 involved in catalysis is about 3 Å further apart from the active site as observed in the x-ray structures (Fig. 3, A and B).
Structure Comparisons with Other LMW-PTPs-Our solution structure of M. tuberculosis MptpA in its apo-state was compared with the x-ray crystal and NMR solution structures of LMW-PTPs of other organisms: yeast (LTP1 from Saccharomyces cerevisiae), human (HCPTP-A), bovine (BPTP), eukaryotic parasite (TPTP from Tritrichomonas foetus), and bacterial (YwiE from Bacillus subtilis, cj1258 from Campylobacter jejuni and Wzb from Escherichia coli). The LMW-PTP sequences from these different organisms have identities between 26 and 38% (Fig. 5). The active site motif (H/V)CX 5 R(S/T) (P-loop) is well conserved in all species, as well as Ser 42 and the D-loop  OCTOBER 5, 2012 • VOLUME 287 • NUMBER 41 motif DP (Asp 126 and Pro 127 in MptpA), which is followed by hydrophobic residues (Tyr 128 and Tyr 129 in MptpA). Furthermore, Ser 42 , which is part of the hydrogen bonding network surrounding the P-loop, is conserved among all species. Additionally, Glu 22 is highly conserved and is also found to be in close spatial proximity to His 71 , which is involved in formation of the P-loop conformation. Although there is hardly any sequence similarity for the W-loop, around residue Trp 48 (only residues Gly 51 and Gly 55 are conserved to some extent), it shows a high structural resemblance to the other LMW-PTPs (Fig. 5, green). The overall fold of MptpA secondary structure elements is highly consistent with structures found in other eukaryotic and prokaryotic LMW-PTPs.

Structure and Function of Apo-MptpA by NMR
Comparison of P-and D-loop Motifs-Superimposing only P-loop (Cys 11 -Arg 17 ) and a part of the D-loop (Leu 122 -Asp 131 ) residues of all compared LMW-PTPs confirmed the conserved conformation of the P-loop and revealed a reorientation of the D-loop ( Table 2). The side chain orientation of residues Tyr 128 / Tyr 129 of MptpA in the x-ray structure is consistent with bovine BPTP (Fig. 4, A and B). Furthermore, similar deviations concerning the conformation of the D-loop are observed. Fig. 3, C and D, shows a detailed view of the P-loop electrostatic surface after superposition of the x-ray structures from bovine BPTP and MptpA with the here elucidated solution structure of MptpA. Both x-ray structures, in complex with chloride/phosphate, show the same orientation of the two adjacent tyrosine residues in the D-loop, pointing toward the active site motif (Fig. 3, C and D, magenta and green). However, the side chains of Tyr 128 /Tyr 129 represented by our NMR structure in the apostate are not oriented toward the active site cavity (Fig. 3, C and  D, cyan). We should also mention that although in bovine LMW-PTP (BPTP), in the presence of phosphate, a protein concentration-dependent dimerization induced by the interaction of W-and D-loop residues is observed (20), we do not . The side chain residues crucial for the catalytic activity (Cys 11 (P-loop) and Asp 126 (D-loop)) and substrate specificity (Trp 48 (W-loop)) are shown as sticks. The overall structure of MptpA consists of a four-stranded parallel folded ␤-sheet (␤ 1 (Leu 5 -Cys 11 ), ␤ 2 (Ala 37 -Ala 43 ), ␤ 3 (Leu 85 -Leu 89 ), ␤ 4 (Arg 106 -Leu 110 )) connected via five ␣-helices (␣ 1 (Cys 16 -Arg 32 ), ␣ 2 (Asp 55 -His 65 ), ␣ 3 (Gly 77 -Ala 82 ), ␣ 4 (Asp 90 -Leu 100 ), ␣ 5 (His 132 -Asn 160 )) also known as the Rossmann fold. The P-loop is flanked by the W-and D-loop. C, backbone dynamics of MptpA measured at 600 MHz and T ϭ 303 K. Mapping the S 2 values on the backbone bundle of 20 lowest energy structures. Colors range from blue (S 2 ϭ 1) to red (S 2 ϭ 0) to gray (no data/no reliable fit). The very low order parameters for the N and C terminus (white, Met 1 , Ser 2 , Gly 161 , and Ser 163 ) are not shown. The figure was generated by PyMOL. Analysis of the Active Site-stabilizing Network-In addition to the aromatic residues Trp 48 , Tyr 128 , and Tyr 129 surrounding the active site, the amino acids His 49 , Asp 123 , and Arg 91 place a unique charge distribution around the active site (28). The interaction between Thr 12 O ␥1 and His 93 N ⑀2 further stabilizes the P-loop. This interaction is a unique feature of bacterial LMW-PTPs, whereas His 93 is found to be replaced by aspartic acid in all known mammalian LMW-PTPs. Distance analysis of Thr 12 O ␥1 and His 93 N ⑀2 revealed a close spatial proximity (3.0 Ϯ 0.5 Å) of these atoms, confirming a potential hydrogen bond interaction. In turn, the backbone amide proton of His 93 interacts with Asp 90 O ␦2 , which is found to be conserved among bacterial LMW-PTPs. Hence, the interaction between His 93 and the side chain of Asp 90 is of crucial importance for the substrate specificity and confirmed in our structure. Distance measurements based on the apo-state solution structure show that Glu 22 is in spatial proximity to the imidazole ring of His 71 . Multiple sequence alignment identifies Glu 22 and His 71 as highly conserved among LMW-PTPs (Fig. 5). Therefore, the electrostatic interaction between Glu 22

JOURNAL OF BIOLOGICAL CHEMISTRY 34575
Dynamic Properties of MptpA-Relaxation analysis of MptpA revealed order parameters S 2 of 0.7-0.9 for the rigid core residues of the protein, with the exception of the flexible loop regions and the termini. Mapping of the order parameter S 2 onto the backbone structure reveals a few distinct regions and residues with different dynamical properties (Fig. 2C). The N-and C-terminal residues Met 1 , Ser 2 , Gly 161 , and Ser 163 have low order parameters of S 2 Յ 0.3, indicating an unrestricted motion that is commonly observed at the termini of a protein (supplemental Table S1 and Fig. S3). Residue Asn 14 , located in the P-loop, has a high order parameter of S 2 ϭ 0.861 Ϯ 0.005, exhibiting Asn 14 as relatively rigid. Due to exchange broadening the P-loop, residues Cys 11 -Gly 13 and Ile 15 -Ser 18 were not detectable. The D-loop motif (Arg 111 -Gly 131 ) contains two residues with a low order parameter, indicating that those are more flexible, Gly 118 S 2 ϭ 0.662 Ϯ 0.046 and Val 124 S 2 ϭ 0.656 Ϯ 0.004. Notably, the ( 1 H)-15 N hetNOE relaxation experiment (supplemental Table S1 and Fig. S3) reveals a high flexibility for the D-loop residues Arg 116 -Asp 123 . These residues are located in the D-loop region, which undergoes structural reorientations during ligand binding. Furthermore, residues Thr 119 -Ala 121 as well as Glu 125 and Pro 127 located in the D-loop are not observable due to exchange broadening. Interestingly, residue Asp 126 , which is critical during dephosphorylation, has a high order parameter of ϳ0.8, which indicates that this residue has a strict orientation.  comparable with the results of a phosphate titration reported for the TPTP LMW-PTP from T. foetus (68). However, in contrast to our observations for MptpA, phosphate binding to TPTP at pH 5.2 was reported to lead to the detection of amide backbone signals that were invisible in the absence of phosphate (68), indicating a slow exchange rate. Based on our experiments, which show broadening of signals, we propose that the binding of phosphate to MptpA at pH 7.0 can be described by an intermediate exchange rate.
Phosphorylation of MptpA by PtkA-Phosphorylation of MptpA Tyr 128 and Tyr 129 is catalyzed by the kinase PtkA, which was previously determined via TLC (14). We performed luciferase assays (Fig. 6) that confirmed the autophosphorylation activity expected for PtkA as well as the lack of autophosphorylation activity proposed for MptpA. Phosphorylation of MptpA Tyr 128 /Tyr 129 was confirmed by 31 P NMR spectroscopy as well as MALDI-MS (supplemental Figs. S4 and S5). The phosphorylation reaction did not, however, yield a sufficient amount of monophosphorylated MptpA for further investigations by multidimensional NMR spectroscopy. We therefore can only speculate that an intermolecular dephosphorylation of the monophosphorylated MptpA species occurs because the phosphorylation of one of the two adjacent tyrosine residues is proposed to increase the activity of the enzyme 25-fold (30), at least in other LMW-PTPs. Previous studies concerning the phosphorylation of human LMW-PTP by pp60 v-src demonstrated that in the presence of phenylarsine oxide, a selective PTP inhibitor, the phosphorylation level is notably increased (70). This assumption suggests that the autodephosphorylation activity of MptpA is in accordance with the low yield of phosphorylated MptpA.

DISCUSSION
All known classical PTPs, vaccinia virus H1-like dual specific phosphatases, and the LMW-PTPs share a common active site motif that is located in a crevice on the molecular surface. The  OCTOBER 5, 2012 • VOLUME 287 • NUMBER 41 signature motif (H/V)CX 5 R(S/T) (P-loop) binds not only phosphorylated protein substrates but also oxyanions, including phosphate, tungstate, or sulfate. The active site motif is flanked by the D-loop, containing the catalytic aspartate residue, which is opposite to the nucleophilic cysteine residue. The superposition of the ligand-free and ligand-bound structures of HMW-PTPs has provided insight into the movement of the highly conserved WPD (Trp-Pro-Asp)-loop toward the catalytic center covering the active site like a "flap." In the ligand-bound state, the aspartate residue is pointing toward the bound oxyanion and donates a proton to the tyrosine leaving group. The tungstate-bound state of the Yersinia HMW-PTP describes a movement of the aspartate (Asp 356 ) by 6 Å toward the active site, thus positioning the carboxylate in spatial proximity to the oxyanion oxygen (19). This conformational change has also been described in the human PTP1B C215S mutant during the binding of phosphotyrosine, bringing the aspartate (Asp 181 ) into the catalytic active site (71). The ligand-induced loop closure observed for HMW-PTPs has thus far been assumed to be operative for LMW-PTPs as well.

Structure and Function of Apo-MptpA by NMR
However, up to now, the structural characterization of apo-LMW-PTPs has not been reported, and a conclusive statement about the movement of the loop containing the catalytic aspartate residue therefore could not be made. In previous studies, the structural characteristics of LMW-PTPs in solution were exclusively described in complex with inorganic phosphate (27,68,72,73), which serves as a competitive inhibitor and stabilizes the conformational plasticity of the phosphate binding loop (P-loop). Here, we present the solution structure of a LMW-PTP in its apo-state. Fig. 3, A and B, demonstrates the spatial differences of the D-loop orientation between our apo-MptpA solution structure (PDB entry 2LUO) and the x-ray structure (PDB entry 1U2P) of MptpA in complex with chloride. The substrate-binding pocket (P-loop) is flanked by the D-loop containing Asp 126 and Tyr 128 /Tyr 129 . The difference in distance between Asp 126 C ␣ and Arg 17 C ␣ in the chloridebound and apo-MptpA is about 3 Å. Therefore, a movement of the D-loop similar to the one observed in HMW-PTPs is also observed in LMW-PTPs. Furthermore, the distances from the active site residue Arg 17 C ␣ to the Tyr 128 and Tyr 129 C ␣ backbone carbons are about 3 and 5 Å larger, respectively, in apo-MptpA than in holo-MptpA. Besides the D-loop, the W-loop containing Trp 48 , which is important for substrate specificity, also flanks the active site and is supposed to interact with substrates. The indole side chain of Trp 48 reorients upon ligand binding and therefore modulates the surface of the binding pocket between the apo-and the holo-state. The distance between Trp 48 N ⑀1 (W-loop) and Tyr 128 O (D-loop) in apo-MptpA is significantly increased as compared to the holostructure (a displacement of nearly 13 Å). The concerted movement of the D-loop and residue Trp 48 results in a more open conformation in the absence of ligands. Examination of LMW-PTPs from other species in complex with phosphate also reveals this deviation in the D-loop conformation. On the basis of our first apo-structure of a LMW-PTP, we propose comparable D-loop dynamics during ligand binding as observed for the Yersinia PTP.
The conformational dynamics of kinases are very pronounced in solution (74 -77), and therefore it is interesting to compare the structural dynamics of our LMW-PTP with those dynamics observed for the catalytic domains of other kinases. In particular, the dynamics of the so-called DFG (Asp-Phe-Gly)-loop have been exploited to generate very potent new classes of inhibitors that are non-ATP-competitive. In the apostate of the protein, the DFG-loop is in equilibrium between the DFG-in and -out state (74). Ligands known to bind the DFG-in conformation do not influence this equilibrium, whereas DFGout ligands shift the equilibrium to the DFG-out conformation. The observation of similar dynamics and large conformational rearrangements upon substrate binding for phosphatases might open a strategy where the unique situation of the two activity-related adjacent tyrosine residues pointing away from the binding pocket could be utilized for drug targeting. Thanks to the structure solved here, the approach of designing inhibitors for different binding modes can now also be used for LMW-PTPs. Because the development of tight binding inhibitors for LMW-PTPs based on the ligand-bound structures has not yet resulted in potential drug candidates (69), the ligandfree structure of MptpA might now lead to a more successful design of potent inhibitors.
Furthermore, in order to design potential inhibitors, the structure-function relationship has to be taken into account, by employing strategies that influence the activation status of a potential drug target. LMW-PTPs are regulated by different mechanisms: (i) phosphate binding, (ii) oxidation of catalytic active cysteine residues, and (iii) phosphorylation. While binding of phosphate to the phosphatase leads to a non-covalent inactivation, it at the same time protects the enzyme against oxidation of the nucleophilic cysteine by reactive oxygen species. The phosphorylation of tyrosine residues located in the D-loop is of high importance for the regulation as observed for the human LMW-PTP HCPTP-A. As shown in vitro, phosphorylation of HCPTP-A Tyr 131 by pp60 v-src increases the enzyme activity 25-fold. However, the direct phosphorylation of active HCPTP-A did not lead to detectable yields of phosphorylated Tyr 131 . In order to induce such phosphorylation, the inhibitor phenylarsine oxide was added to the reaction mixture and resulted in inactivated HCPTP-A. In these experiments, autodephosphorylation was inhibited, and Tyr 131 -phosphorylated HCPTP-A could be detected (70). Previous studies have shown that the residues Tyr 128 and Tyr 129 of MptpA are phosphorylated by the kinase PtkA (14). Based on evidence for the HCPTP-A LMW-PTP, it can be assumed that phosphorylation of MptpA Tyr 128 /Tyr 129 increases the enzyme activity considerably. NMR titration experiments, provided here, support the previously assumed (14) protein-protein complex and resolve the binding surface with atomic resolution. The observed binding sites in the apo-state solution structure of MptpA involve the P-, W-, and D-loop as well as two additional adjacent regions. Because the mechanism of MptpA secretion into the host macrophages remains unclear, it might be of crucial interest to elucidate the consequences of phosphorylation driven by PtkA. Furthermore, the inhibition of the interaction between MptpA and PtkA might serve as a new strategy for the design of potential drug candidates by developing inhibitors of this protein-protein complex vital for pathogen survival.
In summary, our findings reveal the first apo-structure of an LMW-PTP elucidating the rearrangement of the D-loop upon ligand binding. The apo-structure can be described as more open compared with the structures in complex with P i or ligand mimetic, which seem to have a more closed conformation. We were able to map the binding site of PtkA on MptpA, elucidating the protein-protein interaction. Knowledge of the binding site for PtkA and the apo-structure of MptpA leads to further understanding of the regulation and ligand binding behavior of MptpA, which might serve as a basis for the successful design of new anti-tuberculosis therapeutics.