Loss of a Functionally and Structurally Distinct ld-Transpeptidase, LdtMt5, Compromises Cell Wall Integrity in Mycobacterium tuberculosis*

Background: M. tuberculosis LdtMt5 is an LdtMt2 paralog that cross-links peptidoglycan stem peptides. Results: LdtMt5 is structurally divergent, strains lacking LdtMt5 are more susceptible to chemical and environmental stresses, and LdtMt2 cannot compensate for its loss. Conclusion: LdtMt2 and LdtMt5 serve non-redundant roles in peptidoglycan maintenance. Significance: LdtMt5 is necessary for properly maintaining cell wall integrity and should be pursued as a drug target. The final step of peptidoglycan (PG) biosynthesis in bacteria involves cross-linking of peptide side chains. This step in Mycobacterium tuberculosis is catalyzed by ld- and dd-transpeptidases that generate 3→3 and 4→3 transpeptide linkages, respectively. M. tuberculosis PG is predominantly 3→3 cross-linked, and LdtMt2 is the dominant ld-transpeptidase. There are four additional sequence paralogs of LdtMt2 encoded by the genome of this pathogen, and the reason for this apparent redundancy is unknown. Here, we studied one of the paralogs, LdtMt5, and found it to be structurally and functionally distinct. The structures of apo-LdtMt5 and its meropenem adduct presented here demonstrate that, despite overall architectural similarity to LdtMt2, the LdtMt5 active site has marked differences. The presence of a structurally divergent catalytic site and a proline-rich C-terminal subdomain suggest that this protein may have a distinct role in PG metabolism, perhaps involving other cell wall-anchored proteins. Furthermore, M. tuberculosis lacking a functional copy of LdtMt5 displayed aberrant growth and was more susceptible to killing by crystal violet, osmotic shock, and select carbapenem antibiotics. Therefore, we conclude that LdtMt5 is not a functionally redundant ld-transpeptidase, but rather it serves a unique and important role in maintaining the integrity of the M. tuberculosis cell wall.

In M. tuberculosis, loss of Ldt Mt2 results in altered cell size, growth, and virulence as well as loss of the ability of the organism to secrete low molecular weight proteins and increased susceptibility to amoxicillin (8,14). The genome of M. tuberculosis encodes four additional paralogs of Ldt Mt2 . On the basis of in vitro cross-linking activity or sequence similarity, they have been annotated as Ldt Mt1 (Rv0116c), Ldt Mt3 (Rv1433), Ldt Mt4 (Rv0192), and Ldt Mt5 (Rv0483) and share amino acid sequence identity of 36,34,35, and 28% with Ldt Mt2 , respectively. It is unclear whether the five sequence paralogs are functionally redundant.
We used a combination of biophysical, biochemical, and genetic approaches to study Ldt Mt5. Here, we report the first crystal structures of apo-and meropenem-bound Ldt Mt5 and describe the phenotypic effects on M. tuberculosis lacking this enzyme. Our data indicate that Ldt Mt5 is structurally divergent compared with other M. tuberculosis LD-transpeptidases and that this protein serves a critical and distinct role in proper maintenance of M. tuberculosis cell wall integrity, highlighting its potential as a novel drug target.

Experimental Procedures
General Methods-All reagents were obtained from commercial sources. Spectrophotometric analyses were performed on a Shimadzu UV-1800 UV-visible spectrophotometer. Primers were purchased from Integrated DNA Technologies. Isothermal titration calorimetry (ITC) experiments were performed using a high precision VP-ITC titration calorimeter system (Microcal Inc.). Ultraperformance liquid chromatography (LC)-high resolution MS samples were analyzed on a Waters Acquity H-Class ultraperformance LC system equipped with a multiwavelength ultraviolet-visible diode array detector in conjunction with a Waters Acquity BEH-300 ultraperformance LC column packed with a C 4 stationary phase (2.1 ϫ 50 mm; 1.7 m) in tandem with high resolution MS analysis by a Waters Xevo-G2 quadrupole-TOF electrospray ionization mass spectrometer. Molecular graphics and analyses were performed with the UCSF Chimera package. Chimera is developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by National Institutes of Health NIGMS Grant P41-GM103311).
Cloning, Overexpression, and Purification of Ldt Mt5 -A truncated version of ldt Mt5 (encoding amino acids 55-451) was amplified by PCR (1ϫ New England Biolabs GC reaction buffer, 200 M dNTPs, 2 ng/l CDC1551 M. tuberculosis genomic DNA, 500 nM primers (Table 1), 1 unit of Phusion polymerase, and 3% DMSO), digested with NdeI and XhoI, and cloned into a modified pET28a vector that encodes for a TEV-cleavable N-terminal His 6 tag (10). Escherichia coli BL21(DE3) cells harboring the ldt Mt5 -pET28a plasmid were grown to an A 600 of ϳ0.5 at 37°C. Flasks were then cooled on ice with periodic shaking. Protein overexpression was induced with 100 M isopropyl 1-thio-␤-D-galactopyranoside, and flasks were returned to an incubator shaker at 16°C for an additional 24 h. Cells were harvested at 4°C and stored overnight at Ϫ20°C. Thawed cells were resuspended in protein purification buffer (25 mM Tris, pH 8.0, 400 mM NaCl, 10% glycerol, and 1 mM tris(2-carboxyethyl)phosphine (TCEP)) and protease inhibitor mixture (Roche Applied Science) and lysed by ultrasonication, and cell debris was removed by centrifugation at 4°C. The supernatant was incubated with nickel-nitrilotriacetic acid resin at 4°C for 90 min, and His 6 -tagged Ldt Mt5 was eluted from the resin over a stepwise gradient of 5-500 mM imidazole. Fractions containing Ldt Mt5 (as determined by SDS-PAGE) were combined, and protein concentration was determined using the Bio-Rad Protein Assay with bovine serum albumin (BSA) as a standard. The sample was then subjected to dialysis overnight at 4°C against 1 liter of 25 mM Tris, pH 8.0, 100 mM NaCl, 10% glycerol, and 1 mM TCEP in the presence of TEV protease (1:100 TEV:Ldt Mt5 ). Following dialysis, the TEV-treated sample was incubated with fresh nickel-nitrilotriacetic acid resin at 4°C for 90 min. Cleaved Ldt Mt5 was collected as flow-through, whereas the Histagged TEV protease and cleaved His 6 tag remained bound to the resin. Ldt Mt5 was subjected to a second dialysis against 1 liter of 25 mM Tris, pH 8.0, 100 mM NaCl, and 1 mM TCEP for 4 h at 4°C. The concentration of Ldt Mt5 was determined using the Bio-Rad Protein Assay with BSA as a standard, and Ldt Mt5 was concentrated to 12.8 mg/ml prior to being flash frozen in liquid N 2 . Protein was stored at Ϫ80°C.
Site-directed Mutagenesis Studies-Site-directed mutagenesis reactions were performed as described previously with minor modifications (15). Briefly, two PCRs (25 l) containing either the forward or reverse primer were set up in parallel. Each PCR contained 1ϫ New England Biolabs GC reaction buffer, 200 M dNTPs, 1 ng/l template, a 500 nM concentration of either the forward or reverse primer, 1 unit of Phusion polymerase, and 3% DMSO. The ldt Mt5 -pET28a plasmid was used as the template to generate each Ldt Mt5 variant (see Table  1 for primers). Sufficient elongation of primer at 68°C occurred over 15 min. Forward and reverse PCRs were then combined (new volume of 50 l), and complementary strands were reannealed following gradual cooling (95°C, 5 min; 90°C, 1 min; 80°C, 1 min; 70°C, 30 s; 60°C, 30 s; 50°C, 30 s; and 40°C, 30 s). Samples were then incubated at 37°C, and template DNA was digested with 1 unit of DpnI for 3 h. All constructs were fully sequenced, and competent bacteria were transformed with mutagenic plasmid. All variants were purified as described above.
Kinetic Analyses-The nitrocefin hydrolytic activities of Ldt Mt2 , Ldt Mt5 , and Ldt Mt5 variants were measured spectrophotometrically as reported previously for Ldt Mt2 (10) but with modifications. Briefly, reaction mixtures containing 1ϫ tribuffer (100 mM MES, 50 mM N-ethylmorpholine, and 50 mM diethanolamine, pH 10), 0.1 mg/ml BSA, 100 mM NaCl, 1 mM TCEP, 5% DMSO, and 10 M Ldt Mt5 or Ldt Mt5 variant were preincubated at 37°C for 5 min. Nitrocefin (Calbiochem) was added to initiate each reaction, and the rate of nitrocefin hydrolysis was measured at 486 nm at 37°C. For each set of reactions, the rate of nitrocefin hydrolysis in the absence of enzyme was observed and was subtracted from the initial rate of nitrocefin hydrolysis in the presence of enzyme at each substrate concentration. Initial rates of nitrocefin hydrolysis were measured over 3 min. An extinction coefficient (⑀ M ) of 20,500 M Ϫ1 cm Ϫ1 was used in determining the concentration of hydrolyzed nitrocefin as it was experimentally determined that ⑀ M does not change with changing pH under these buffering conditions (data not shown). Non-linear regression analyses of initial velocities were performed using GraphPad Prism (version 5). Reaction mixtures containing nitrocefin (100 M) and Ldt Mt5 or Ldt Mt2 (5 M) at varying pH values in 1ϫ tribuffer and the conditions described above were monitored spectrophotometrically for determining the optimal pH for each enzyme. To rule out the possibility that residual TEV incompletely removed during protein purification could be contributing to nitrocefin hydrolysis at pH 10, we evaluated nitrocefin as a substrate for TEV at pH 10. Only baseline levels of hydrolysis were observed, indicating that TEV was not significantly contributing to the observed rates of hydrolysis by Ldt Mt5 and Ldt Mt5 variants (data not shown). Conversely, when we evaluated nitrocefin as a substrate for His 6 -Ldt Mt5 , we observed rates of hydrolysis that were comparable with TEV-cleaved Ldt Mt5 (data not shown).
Crystallization Conditions-Crystals of Ldt Mt5 were obtained by the hanging drop vapor diffusion method at 20°C. Drops of 2 l of protein (12.8 mg/ml) and 1 l of reservoir solution were equilibrated against a reservoir containing 85 mM sodium citrate, pH 5.6, 25.5% polyethylene glycol (PEG) 4,000, 170 mM ammonium acetate, and 15% glycerol. Crystals suitable for data collection grew within 1 week. Crystals of Ldt Mt5 in complex with meropenem were obtained with protein incubated with the ligand (645 M) using crystallization conditions identical to that of the apo crystal.
Data Collection, Structure Determination, and Refinement-All diffraction data were collected at beamline X6A of the National Synchrotron Light Source of the Brookhaven National Laboratory (Table 2). X-ray diffraction experiments were carried out with crystals frozen in their respective mother liquor without addition of cryoprotectant. The crystal structure of apo-Ldt Mt5 was determined by molecular replacement with the program MOLREP (16) using previously determined Ldt Mt2 structures of individual domains as searching models (Protein Data Bank code 3TUR, catalytic domain (CD) and bacterial Ig-like (BIg) B domains; Protein Data Bank code 4HU2, BIgA domain). After 20 cycles of refinement of the three separate domains as rigid bodies with REFMAC (CCP4), the structure was rebuilt with the molecular modeling program Coot (17) and further refined with the program PHENIX using restrained and translation, libration, and screw anisotropic refinement protocols with four translation, libration, and screw groups for each BIg domain and two for the CD (18). The apo-Ldt Mt5 structure was refined to a final R work of 0.21 and an R free of 0.255 with 96.1% of the residues in favored stereochemistry regions (Table 2) and solved to 1.98 Å.
The crystal structure of meropenem-bound Ldt Mt5 was determined by molecular replacement and refined using a protocol similar to that used when solving the apo-Ldt Mt5 structure ( Table 2). The partial meropenem adduct was built inside the positive A (mF o Ϫ DF c ) electron density difference map. Weak electron density for the sulfur atom of meropenem was observed, and no electron density for the pyrrolidine ring extension of meropenem was observed. The meropenem adduct structure was refined to an R work of 0.23 and an R free of 0.275 with 93% of the residues in Ramachandran favored regions (  (19). Buried surface area calculations were performed using the Protein Interfaces, Surfaces, and Assemblies' Service (PISA) at the European Bioinformatics Institute (20). Coordinates and structure factors were deposited in the Protein Data Bank under the codes 4Z7A (apo-Ldt Mt5 ) and 4ZFQ (meropenem-bound Ldt Mt5 ).
Calorimetric Studies-Freshly thawed Ldt Mt5 protein was dialyzed overnight in 1 liter of buffer containing 25 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 1 mM TCEP at 4°C. Dialyzed protein solution was then filtered through a 0.22-m filter and diluted to 10 M. The protein concentration was determined by UV absorption at 280 nm using a calculated extinction coefficient of 78,840 M Ϫ1 cm Ϫ1 . Freshly thawed aliquots of 20 mM carbapenem stock solutions were diluted to 1 mM in protein dialysis buffer. Protein and ligand solutions were degassed for 20 min in a ThermaVac. Ligand injections (10 l) into the cell containing Ldt Mt5 were performed with 240-s equilibrations between injections. Data were analyzed with Origin 7 software (OriginLab). All calorimetry experiments were carried out at 27°C.
M. tuberculosis Strains and Culture Conditions-M. tuberculosis CDC1551 (21) (herein referred to as "wild type") was used as the host strain to generate a transposon insertion mutant in the MT0501 (ldt Mt5 ::Tn) gene as described previously (22). All strains were grown in Middlebrook 7H9 liquid medium supplemented with 0.2% glycerol, 0.05% Tween 80, 10% (v/v) oleic acid/albumin/dextrose/catalase, 50 g/ml cycloheximide (herein referred to as 7H9 complete medium), and when indi-cated 2.0 g/ml crystal violet. The ldt Mt5 ::Tn strain was grown in the presence of 20 g/ml kanamycin. Middlebrook selective 7H11 solid medium (BD Biosciences) was used for enumerating colony forming units (cfus) in in vitro growth studies. M. tuberculosis H37Rv was used in generating meropenem-resistant strains.
In Vitro Growth and Crystal Violet Studies-All M. tuberculosis strains were grown to midexponential phase with an A 600 of ϳ0.8 in 7H9 complete medium at 37°C. Cultures were then diluted to an A 600 of 0.1 in 7H9 complete medium in the presence or absence of 2.0 g/ml crystal violet, and turbidity of the cultures was determined daily. Appropriate dilutions of each strain were cultured on Middlebrook 7H11 medium to determine surviving bacilli by enumerating the cfus.
Osmotic Tolerance Studies-Wild-type or ldt Mt5 ::Tn strains were grown to late exponential phase (A 600 ϳ 2-3) in 7H9 complete medium. Cultures were diluted to an A 600 of 0.5, and cells were pelleted at low speed. Cells were resuspended in 150 mM NaCl or double distilled H 2 O (0 mM NaCl) solutions containing 0.05% Tween 80. Cells were incubated in these conditions for 1 h with shaking at 37°C. Viability was determined by culturing and enumerating the cfus on 7H11 selective agar.
Determination of Minimum Inhibitory Concentration-Carbapenem minimum inhibitory concentrations were determined using the standard broth dilution method (23). Briefly, 10 5 M. tuberculosis bacilli were inoculated into 2.5 ml of 7H9 broth supplemented with 0.2% glycerol, 10% (v/v) oleic acid/ albumin/dextrose/catalase, and 50 g/ml cycloheximide, and the drug was added at different concentrations in the M-mM range. The cultures were incubated at 37°C without shaking and evaluated for growth by visual inspection of the broth at 14 and 21 days. Minimum inhibitory concentration values are representative of three independent experiments.

JOURNAL OF BIOLOGICAL CHEMISTRY 25673
Electron Microscopy Experiments-Field emission scanning electron microscopy and transmission electron microscopy experiments were performed as described previously (14).

Results
Ldt Mt5 Structure-The apo and meropenem adduct structures of an N-terminally truncated Ldt Mt5 protein lacking the hydrophobic domain predicted to be a membrane anchor for this protein (amino acids 55-451) were determined using x-ray crystallography (Fig. 1). This truncated protein displays higher sequence identity to Ldt Mt2 (31%) than the full-length protein does (28%) and includes the proline-rich extension of the C-terminal subdomain (ex-CTSD) comprising residues 417-451 that is absent in all other M. tuberculosis LD-transpeptidases ( Fig. 2A).
The overall structural overlap between apo-Ldt Mt5 and Ldt Mt2 (Protein Data Bank code 3VYN) has an r.m.s. deviation of 2.2 Å for 225 aligned C␣ atoms, including 79 identical residues, highlighting their overall structural similarity (Fig. 2, B and C). There are small differences with regard to the orientation of the domains among other structurally characterized LDtranspeptidases. The equivalent BIg domains of Ldt Mt5 and Ldt Mt2 are similar; the BIgA domains display a small r.m.s. deviation of 1.0 Å among 65 pairs of C␣ atoms aligned, and the BIgB Tryptophan residues of the CTSD are also represented as sticks and interact with a hydrophobic patch at the interface of the BIgB domain. The prominent outer cavity that is observed in Ldt Mt2 is absent in apo-Ldt Mt5 but is indicated as a reference (10). The Ldt Mt5 secondary structure schematic is colored as a rainbow from blue (N terminus) to red (C-terminus). Orange dashes represent the disordered portion of loop L C , and red dashes represent the disordered ex-CTSD. This figure was made using Chimera (19). The largest changes among structurally characterized LD-transpeptidases are observed in the ␤-hairpin, and in the case of Ldt Mt5 , there is a dramatic displacement of loop L C that occurs after adduct formation (indicated with red curved arrows). The C-terminal portion of the CTSDs was excluded for clarity. D, accessible surface map of apo-Ldt Mt5 colored by the magnitude of the observed atomic temperature factors from low (green) to high (magenta) motility. The flexibility of the ␤-hairpin as indicated by the high atomic temperature factor correlates with its large displacement upon adduct formation. These images, the sequence alignment, and structural superpositions were performed using the program MOE. The sequence representation was performed using ESPript3 (48). affects placement of the ␤-hairpin in comparison with its position in apo-Ldt Mt1 , whereas the meropenem adduct-Ldt Mt5 structure displays the largest changes in the ␤-hairpin (Fig. 2C).

Structural and Biochemical Characterization of Ldt
The Conformation of the Ldt Mt5 BIg Domains Is Mostly Maintained by Unspecified Hydrophobic Interactions-A small, solvent-accessible area (312 Å 2 ) is buried in the interface between the BIg domains in Ldt Mt5 . A short, three-residue linker (Ala 153 , Pro 154 , and Val 155 ) joins BIgA and BIgB. BIgA is rotated ϳ30°a round the axis passing through the interdomain linker compared with BIgA of Ldt Mt2 (Fig. 2B). Two hydrophobic patches comprising Pro 80 , Tyr 125 , and Pro 154 of BIgA and Tyr 239 and Tyr 248 from BIgB and main-chain atoms of ␤-barrel loops from Surface zones related to the ␤-hairpin and loop L C that display the largest structural differences among apo and holo structures are colored purple. Acylation of Ldt Mt5 by meropenem causes displacements of these structural elements as indicated by the green arrows (right panel) that "restore" the outer cavity.
both of the domains make an interdomain contact. These proline and tyrosine residues are well conserved among three-domain LD-transpeptidases ( Fig. 2A). Tyr 125 (from BIgA) and Tyr 239 (from BIgB) exchange hydrogen bonds with main-chain atoms of the opposite domain. This greasy and weak contact may provide flexibility in the orientation of the domains. In the Ldt Mt5 crystal form, a PEG molecule from the crystallization buffer is bound to an exposed hydrophobic patch (Pro 122 and Tyr 239 ) at the interface of the BIgA and BIgB domains, apparently stabilizing the observed relative orientations of the domains.
The Conformations of the BIgB Domain and EYY-fold Are Maintained by the CTSD-The L E loop of the Ldt Mt5 CTSD is slightly longer in comparison with that of Ldt Mt2 (Fig. 2, B and  C). This loop wedges between the EYY-fold and the BIgB domain (Fig. 1). The Ldt Mt5 CTSD is rich in tryptophan residues (Trp 398 , Trp 400 , Trp 404 , and Trp 407 ). Extensive hydrophobic contacts among the CTSD, BIgB, and CD domains increase the rigidity of the BIgB/CD assembly. In addition to regular contacts between Tyr 392 and Trp 398 in loop L E with Leu 209 and Pro 210 in BIgB, the aromatic rings of Trp 400 , Trp 404 , and Trp 407 in the ␣ 3 helix of the CTSD form a "zipper-like" interaction with the aromatic ring of Tyr 225 and aliphatic portions of the side chains of Arg 223 and Arg 221 of the BIgB domain. This structure provides 1336 Å 2 of additional area buried in the BIgB/EYYfold assembly, which itself only contributes 433 Å 2 .
The Ldt Mt5 CD Displays Large Structural Differences Relative to Ldt Mt1 and Ldt Mt2 -The CD of Ldt Mt5 displays marked differences in comparison with those of Ldt Mt1 and Ldt Mt2 . The largest differences are seen 1) within the fold and placement of a ␤-hairpin flap that includes loop L F (the shortest among homologs), 2) in the conformation and partial disorder of loop L C , and 3) in the size of loop L D and loop L E (of the CTSD) (Fig. 2). All of these structural differences are in close proximity to the Ldt Mt5 active site. The ␤-hairpin flap covers the active site and is the structural feature that distinguishes some LD-transpeptidases (10, 24, 28) from the first structurally characterized protein containing the EYY-fold (27). This flap displays the largest temperature factors relative to the remainder of Ldt Mt5 (Fig.  2D), is nine residues shorter in Ldt Mt5 , and displays low homology to other M. tuberculosis LD-transpeptidase ␤-hairpin flaps (Fig. 2C). In Ldt Mt5 , loop L C displays considerable disorder. Electron density for residues 347-353 in apo-Ldt Mt5 and residues 348 -356 in the meropenem adduct structure was not observed. However, the residues of loop L C that are ordered display fold differences relative to Ldt Mt1 and Ldt Mt2 (Fig. 2C).
In previously solved LD-transpeptidase structures, the catalytic site is exposed through two connected cavities, the outer and inner cavities (Fig. 2, B and C). Compared with Ldt Mt1 and Ldt Mt2 , the small footprint and placement of this ␤-hairpin in Ldt Mt5 lead to greater exposure of the catalytic site from the inner cavity (Fig. 2). The ␤-hairpin flap and loop L C of apo-Ldt Mt5 are shifted toward the outer cavity, closing it (Figs. 2C and 3A). In our meropenem adduct-Ldt Mt5 structure, the hairpin and loop are partially disordered; however, the ordered portions appear to shift away from the catalytic site, thereby exposing it (Fig. 3B). Thus, the acylation of Ldt Mt5 by meropenem

Structural and Biochemical Characterization of Ldt Mt5
OCTOBER 16, 2015 • VOLUME 290 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 25677 appears to "create" an outer cavity reminiscent of that observed in Ldt Mt1 and Ldt Mt2 (Fig. 3C).
Loop L D (residues 289 -302) within the CD core of Ldt Mt5 is larger compared with Loop L D of Ldt Mt1 and Ldt Mt2 ( Fig.  2A-C). The Ldt Mt5 L D loop has a three-residue insertion that includes a bulky arginine residue (Arg 297 ) and forms a protruding insertion (Fig. 2C). Although most of this loop fold remains unperturbed, the insertion displaces the adjacent L C loop (residues 338 -358), thereby closing the active site outer cavity and dramatically modifying the fold and placement of the L C loop.
The Ldt Mt5 Active Site Is Structurally Divergent Relative to Ldt Mt1 and Ldt Mt2 -The structural differences within the Ldt Mt5 CD have dramatic effects on the active site architecture and the readiness of catalytic residues to participate in enzymatic reactions and presumably PG stem recognition. In Ldt Mt1 and Ldt Mt2 , a conserved methionine residue (Met 175 and Met 303 , respectively) on the internal side of the ␤-hairpin flap limits the space around the catalytic cysteine. The unique placement of this loop in Ldt Mt5 results in displacement of this methionine (Met 316 in Ldt Mt5 ) by the non-conserved Glu 328 (Fig. 3). Glu 328 is substituted with smaller hydrophobic residues in Ldt Mt2 (Val 322 ) and Ldt Mt1 (Ala 195 ). Interestingly, the electron density surrounding this glutamate residue indicates that Glu 328 is present in three alternative conformations in the apo-Ldt Mt5 structure but shows only one conformation in the meropenem adduct structure (Fig. 3, A and B). The most populated conformation of Glu 328 and the conserved motif Asn 362 (implicated in PG stem recognition) form hydrogen bonds with His 342 , thereby orienting His 342 in such a way that it is rotated 180°from the orientation required to deprotonate Cys 360 (Fig.  4, A and B). Furthermore, Cys 360 and His 342 make a strong contact with one another; electron density connecting the sulfur atom to the imidazole ring carbon is visible in the experimental electron density map (Fig. 4A). However, no covalent bond is present: the coordination geometry of the C⑀1-S␥-C␤ bond angle is much smaller than 109°, which is expected for a direct bond. It is likely that this strong contact and coordination of the histidine ring by Asn 362 and Glu 328 make it difficult for the imidazole ring to rotate into a position that is better poised for catalysis (Fig. 4B). Thus, it is clear that His 342 is not optimally poised to act as a catalytic base in Ldt Mt5 as is the equivalent residue in Ldt Mt2 (His 336 ) is (Fig. 3C).
The L C loop of Ldt Mt5 is fully embedded in the conserved HXX 14 -17 (S/T)HGChN motif that characterizes this family of transpeptidases. Ldt Mt5 has two variations in the conserved motif: a motif alternative Thr 357 of Ldt Mt5 replaces the Ldt Mt2 serine (Ser 351 ), and Asn 358 replaces the characteristic motif histidine (His 352 in Ldt Mt2 ). The C-terminus of loop L C forms the "anion hole" at the catalytic site of Ldt Mt2 . Thr 357 occludes the outer entrance to the active site (Figs. 3A and 4C) where the PG stem binds to Ldt Mt2 (Fig. 3C) (10). The loop, which contains the anion hole that comprises a large quantity of positively charged atoms, is folded differently relative to Ldt Mt2 , and Met 346 of loop L C replaces a tryptophan residue that is conserved in all other M. tuberculosis LD-transpeptidases (Fig. 2A).

Ldt Mt5 Is Acylated by Meropenem during Crystallization-
The crystal form grown in the presence of meropenem shows electron density for Ldt Mt5 residues 56 -317, 327-348, and 356 -416. Electron density for most of the ␤-hairpin flap is missing in this crystal form, and like apo-Ldt Mt5 , most of loop L C and the ex-CTSD are disordered. The carbapenem core of meropenem was fitted in additional electron density near the catalytic cysteine, which forms an adduct with Cys 360 (Fig. 3B); however, no electron density for the 3-[5-(dimethylcarbamoyl)pyrrolidin-2-yl] group of meropenem was observed. Interestingly, the presence of this adduct restores the anion hole and other portions of L C to a similar fold previously observed in other LD-transpeptidases, placing residues with a probable role in catalysis (Met 346 , Asn 358 , and Thr 357 ) in positions equivalent to those observed in active LD-transpeptidases (Fig. 4C). Thus, meropenem binding induces a conformational change that enables access to the catalytic site from the outer cavity as observed in other LD-transpeptidases (Fig. 3). In addition, this change promotes release of His 342 from its nonproductive contact such that it now hydrogen bonds with Cys 360 (distance of N⑀-S␥, 3.2 Å; Figs. 3B and 4C).
The most stable tautomer of the carbapenem core is observed where the ring nitrogen is deprotonated (double bond between C3 and N4) and C2 is sp 3 hybridized and is in agreement with previously reported Ldt Mt2 -meropenem adduct structures (24,25). The meropenem core lies with its most apolar side facing a hydrophobic patch formed by Gly 338 , the aliphatic portion of the side chain of Glu 339 , and Phe 340 at the inner cavity. The C-terminal portion of the main chain of loop L C (Gly 359 ) provides apolar contacts with the other side of the carbapenem core. Four hydrophilic interactions are also observed between the carbapenem core and Ldt Mt5 : 1) Asn 358 and 2) the main chain nitrogen atom of Cys 360 hydrogen bond to the carbonyl of the opened penem ring, 3) Glu 328 hydrogen bonds to the meropenem hydroxyethyl group, and 4) a water molecule (W 601 ) mediates interaction between the meropenem core carboxylate and the carboxylate of Glu 339 (Fig. 3B).
We evaluated a series of ␤-lactams, including the carbapenems listed in Table 3, and measured the thermodynamics of ␤-lactam binding to Ldt Mt5 using ITC. Despite the presence of a meropenem adduct on Ldt Mt5 , no significant heat exchange associated with binding was measured by ITC, and no adduct was detected by mass spectrometry after a 5-h incubation of meropenem and Ldt Mt5 (Table 3).
Ldt Mt5 -catalyzed Nitrocefin Hydrolysis Is Optimal at Basic pH-Ldt Mt5 was probed for transpeptidase/␤-lactamase activity using nitrocefin as a substrate. A pH rate profile analysis revealed that Ldt Mt5 is optimally active at pH Ͼ9 (Fig. 5) even after correcting for spontaneous ring opening at basic pH in the absence of enzyme, although its activity was not significantly different from that observed for Ldt Mt2 , which optimally catalyzes nitrocefin hydrolysis at pH 7 (10) (Fig. 5). We also measured nitrocefin binding at pH 8 where little Ldt Mt5catalyzed nitrocefin hydrolysis was observed (Fig. 5), but no detectable heat of exchange was observed using ITC (data not shown).
Conserved Active Site Residues Are Not Required for Nitrocefin Hydrolysis-On the basis of the apo-Ldt Mt5 structure, we rationally designed and purified Ldt Mt5 putative active site variants in an attempt to identify residues responsible for nitrocefin hydrolysis at pH Ͼ9. Surprisingly, all of the Ldt Mt5 variants evaluated, including C360A Ldt Mt5 , hydrolyzed nitrocefin (Table 4 and Fig. 6) with the rates of hydrolysis ordered as follows: N358H Ͼ wild type Ͼ H342Q Ͼ H342A Ͼ T357V Ͼ N358A Ͼ N362A Ͼ C360A. Although the rates of hydrolysis were relatively low, conservative mutations (N358H and H342Q) had the lowest impact on k cat /K m specificity constants, whereas C360A Ldt Mt5 was the least active variant we tested. Interestingly, the K m(nitrocefin) for the N358H variant was 2-fold lower than that of wild type, whereas the k cat values were relatively comparable.
Two residues at the ends of loop L C that interact with the PG stem in the outer cavity of Ldt Mt2 (10), His 352 and Trp 340 , are substituted with Asn 358 and Met 346 , respectively, in Ldt Mt5 .

JOURNAL OF BIOLOGICAL CHEMISTRY 25679
Asn 358 replaces this conserved motif histidine (His 352 ) in Ldt Mt2 that participates in recognition of the donor PG stem (10) and in Ldt Mt5 that participates in recognition of the meropenem adduct (Fig. 3B). Trp 340 in Ldt Mt2 is proposed to provide tetrapeptide substrate specificity (10). In an attempt to make the Ldt Mt5 active site more Ldt Mt2 -like, we generated the Ldt Mt5 N358H/M346W double variant. Contrary to our expectations, mass spectrometry data indicate that the double variant was not acylated by the carbapenems tested (Table 3), and the double variant behaved like wild-type Ldt Mt5 when probed for its ability to catalyze nitrocefin hydrolysis in a pH rate profile analysis (data not shown).
Loss of ldt Mt5 Modestly Enhances Susceptibility to Doripenem and Faropenem-Recent studies have reported that, in addition to inhibiting DD-transpeptidase and carboxypeptidase activities, carbapenems and penems bind to and inhibit LD-transpeptidases (10, 24, 25, 29 -31). We hypothesized that loss of Ldt Mt5 may alter sensitivity to carbapenems as the mutant lacking this protein would have one less target for carbapenems to inhibit. Minimum inhibitory concentration studies were performed to evaluate whether or not loss of ldt Mt5 affected the susceptibility of M. tuberculosis to carbapenems ( Table 5). The ldt Mt5 ::Tn strain reproducibly had modestly enhanced susceptibility to doripenem and faropenem (a penem) compared with wild type, but neither strain was susceptible to ertapenem or meropenem under the conditions that were tested. Both strains displayed similar susceptibilities to tebipenem pivoxil.
Mutations in the ldt Mt5 Locus Could Not be Detected in Meropenem-resistant Mutants-We tested the hypothesis that Ldt Mt5 is a target of meropenem and that meropenem-resistant M. tuberculosis strains would harbor a mutation in the gene encoding this enzyme. Toward this end, we generated genetically resistant strains by isolating mutants grown in the presence of 400 g/ml meropenem. Nine independent strains were isolated, their genomic DNA was purified, and the ldt Mt5 loci (which included ϳ100 bp in both 5Ј-and 3Ј-UTRs) were PCR-amplified and sequenced. In addition, we sequenced the locus of the parent M. tuberculosis H37Rv that was used to generate the mutants. The DNA sequences of the ldt Mt5 loci in all nine meropenem-resistant strains were identical to the parent M. tuberculosis H37Rv and to the sequence of the reference M. tuberculosis H37Rv genome (data not shown) (32).
Loss of ldt Mt5 Enhances Sensitivity to Crystal Violet and Osmotic Shock-To determine the effects of loss of functional Ldt Mt5 , we probed the cell wall integrity of wild-type and ldt Mt5 ::Tn strains using crystal violet and osmotic shock. In comparison with wild type, ldt Mt5 ::Tn M. tuberculosis displayed a minor growth defect when grown in complete medium (Fig.  7A). When complete medium was supplemented with crystal violet, ldt Mt5 ::Tn M. tuberculosis behaved similarly to cells lack-ing the dominant LD-transpeptidase Ldt Mt2 (ldt Mt2 ::Tn) as both strains were more susceptible to killing by the dye (Fig. 7B). Furthermore, ldt Mt5 ::Tn cells or ldt Mt2 ::Tn cells were 2-3 times less viable than wild-type cells when subjected to osmotic challenge (Fig. 7C). These findings suggest that loss of ldt Mt5 alters  cell wall permeability and sensitivity to crystal violet and compromises cell wall integrity. We also examined the cell morphology of M. tuberculosis lacking ldt Mt5 by electron microscopy. Interestingly, no observable changes in cell size and morphology between wild-type and ldt Mt5 ::Tn strains were observed (Fig. 8). The gene encoding Ldt Mt5 is in an operon downstream of murB, another PG biosynthetic enzyme. We attempted to complement our ldt Mt5 ::Tn strain with a wild-type copy of ldt Mt5 under the control of its native promoter. We designed and tested eight different comple-mented strains, but none were able to restore growth phenotypes (data not shown).

Discussion
Recently, 333 cross-links have been identified in the PG of a variety of bacterial species (5,(33)(34)(35)(36), and LD-transpeptidases have been identified as the enzymes that catalyze synthesis of this type of transpeptide bond (6 -8, 12, 33, 37, 38). Ldt Mt5 is a paralog of Ldt Mt2 from M. tuberculosis and has been reported to
There are significant structural differences within the CD of Ldt Mt5 and consequently the active site despite overall architectural similarity to Ldt Mt2 . The ␤-hairpin flap that covers the active site is the smallest among paralogs and exhibits high mobility (high B factors in the apo structure and partial disorder observed in the meropenem-bound structure; Fig. 2D). This hairpin and loop L C display the largest structural differences among M. tuberculosis LD-transpeptidases ( Fig. 2A), and the distinctive structural changes observed upon meropenem adduct formation suggest that their mobility and flexibility could play a role in the catalytic mechanism. The outer cavity that is present in other characterized LD-transpeptidases is absent in apo-Ldt Mt5 ; however, such an outer cavity is restored when meropenem acylates Ldt Mt5 . Upon acylation by meropenem, residues from the conserved motif, including His 342 , Asn 358 , Cys 360 , and Asn 362 , shift to positions resembling those occupied by equivalent residues in Ldt Mt2 (Figs. 3, B and C, and 4A) and Ldt Mt1 , lining a cavity that could accommodate a PG stem (Fig. 3).
We observed catalytic residues Cys 360 and His 342 in a nonproductive contact in our apo crystal form, and His 342 is not optimally poised for catalysis (Fig. 4A); however, the nonpro-ductive contact appears to be released upon adduct formation (Fig. 4B). We did not observe acylated Ldt Mt5 by mass spectrometry, likely the result of the presence of this nonproductive contact. However, we clearly observed a meropenem adduct on Ldt Mt5 in our crystal form. Thus, if given enough time, meropenem will acylate Ldt Mt5 over the extended incubation period that is required for co-crystallization. Alternatively, a component of the crystallization buffer may promote acylation of Ldt Mt5 by meropenem.
The pK a of a cysteine side chain is 8.3. Nitrocefin is a poor substrate for Ldt Mt5 , but we observed Ldt Mt5 -catalyzed nitrocefin hydrolysis at pH Ͼ9 (Fig. 5). It is feasible that the nonproductive contact of the catalytic residues may be released at pH Ն9 by weakening of the hydrogen bonds holding the residues in this conformation (Fig. 4A). None of the active site variants we designed fully abolished this activity, including C360A Ldt Mt5 ; however,C360A Ldt Mt5 was the least active variant. Furthermore, the N358H substitution affected nitrocefin recognition (K m N358H Ͻ K m WT ). It has been demonstrated previously that substitutions to any of the catalytic residues of serine proteases significantly reduce the rate of peptide bond cleavage but do not completely abolish it (39), indicating that the remaining catalytic site environment after residue substitutions can still promote turn- over albeit slowly. It is feasible that, under basic conditions, Cys 360 is deprotonated and can hydrolyze nitrocefin and that excess hydroxide in the Ldt Mt5 active site will still promote turnover of this unnatural substrate even in the absence of the catalytic cysteine. Alternatively, different Ldt Mt5 residues may mediate nitrocefin hydrolysis.
In addition to catalyzing 333 transpeptidation in PG, LDtranspeptidases incorporate non-canonical D-amino acids into PG during stationary growth phase and catalyze attachment of Braun lipoprotein in some Gram-negative bacteria (12,40). Unlike Ldt Mt2 , Ldt Mt5 has a 33-residue ex-CTSD (residues 417-451). The ex-CTSD (Fig. 1) is disordered and contains prolinerich stretches ( Fig. 2A). Proline-rich regions have been observed in other mycobacterial PG biosynthetic enzymes, including putative DD-transpeptidases PonA1, PonA2, and PonA3 and Ldt Mt4 , another paralog of Ldt Mt2 (41,42). Although these proline-rich regions are seemingly common among these M. tuberculosis cell wall biosynthetic enzymes, their role in M. tuberculosis physiology is still largely unknown. Interestingly, proline-rich sequence stretches frequently mediate protein-protein interactions (43). The proline-rich ex-CTSD of Ldt Mt5 is in close proximity to the catalytic site. Thus, it is plausible that the Ldt Mt5 ex-CTSD participates in the recognition of protein substrates and/or binding partners, and these interactions may drive the conformational changes required to release His 342 and Cys 360 from their nonproductive contact. Likewise, it is reasonable to speculate that the active site of Ldt Mt5 may have evolved to accommodate large substrates like proteins and play a role in anchoring them to the PG reminiscent of the role LD-transpeptidases serve in Gram-negative species in anchoring Braun lipoprotein (12,40). Taken together, the major structural differences and divergent catalytic site suggest that Ldt Mt5 and Ldt Mt2 evolved to serve different functions in M. tuberculosis (Fig. 9).
It has been demonstrated that YbiS, an E. coli LD-transpeptidase, is a substrate of the thioreductase DsbG (44). In E. coli, DsbG reduces the catalytic cysteine of YbiS that is prone to sulfenylation in the periplasm. We have previously reported a crystal structure of Ldt Mt2 that shows Cys 354 oxidized to the sulfenic acid (10), suggesting that M. tuberculosis LD-transpeptidases are also susceptible to oxidation. Although we did not observe any sulfur adducts in our apo-Ldt Mt5 structure, it is conceivable that Ldt Mt5 requires binding of a protein partner to maintain the correct oxidation state of its catalytic cysteine in vivo.
Although all ␤-lactam antibiotics target DD-transpeptidases involved in 433 cross-link formation in PG maturation, only the carbapenem class of ␤-lactams (and faropenem, a penem) inhibit LD-transpeptidases. Furthermore, the genome of M. tuberculosis encodes for BlaC, an extended spectrum class A ␤-lactamase (45,46). For these historical reasons, ␤-lactams are seldom considered for treatment of M. tuberculosis infection. However, carbapenems have been recently identified as poor substrates for BlaC (47). We have previously demonstrated that M. tuberculosis lacking Ldt Mt2 is more susceptible to killing by ␤- lactams (8, 14). Sanders et al. (42) have reported that LdtC (homologous to Ldt Mt5 in M. tuberculosis on the basis of sequence) is the primary LD-transpeptidase in Mycobacterium smegmatis. Strains lacking ldtC are hypersusceptible to imipenem, and ldt Mt5 from M tuberculosis fully complements this phenotype in an ldtC mutant, suggesting that these enzymes are equivalent (42). We observed a modest enhancement in susceptibility of the ldt Mt5 ::Tn strain to select carbapenems (Table 5) presumably due to synthetic lethality as these ␤-lactams may inactivate other targets. Although our meropenem adduct crystal form supported very slow acylation of Ldt Mt5 over many days, we cannot rule out the possibility that Ldt Mt5 was more rapidly inactivated by this class of ␤-lactams in vivo, particularly in the event that Ldt Mt5 requires a protein-protein interaction for productive catalysis. To date, studies examining acylation of Ldt Mt5 by carbapenems, including the data presented here, have been in vitro, and interestingly, Ldt Mt5 is the only Ldt Mt2 paralog that is not inactivated by carbapenems. The increased susceptibility of ldt Mt5 ::Tn strains to osmotic shock and crystal violet coupled with the observed modest enhancement in susceptibility to carbapenems and our meropenem-Ldt Mt5 crystal form suggest that Ldt Mt5 is worth pursuing as a drug target.