Structural and Biochemical Analyses of Mycobacterium tuberculosis N-Acetylmuramyl-l-alanine Amidase Rv3717 Point to a Role in Peptidoglycan Fragment Recycling

Background: Peptidoglycan hydrolases help elongate, divide, and recycle bacterial cell walls.

Results: The structure and activity of Rv3717, a peptidoglycan hydrolase of Mycobacterium tuberculosis, are described.

Conclusion: Our data suggest a peptidoglycan-recycling role for Rv3717.

Significance: Functional specialization of homologous peptidoglycan hydrolases can be achieved by altering their substrate range via structural variation.

Abstract

Peptidoglycan hydrolases are key enzymes in bacterial cell wall homeostasis. Understanding the substrate specificity and biochemical activity of peptidoglycan hydrolases in Mycobacterium tuberculosis is of special interest as it can aid in the development of new cell wall targeting therapeutics. In this study, we report biochemical and structural characterization of the mycobacterial N-acetylmuramyl-l-alanine amidase, Rv3717. The crystal structure of Rv3717 in complex with a dipeptide product shows that, compared with previously characterized peptidoglycan amidases, the enzyme contains an extra disulfide-bonded β-hairpin adjacent to the active site. The structure of two intermediates in assembly reveal that Zn²⁺ binding rearranges active site residues, and disulfide formation promotes folding of the β-hairpin. Although Zn²⁺ is required for hydrolysis of muramyl dipeptide, disulfide oxidation is not required for activity on this substrate. The orientation of the product in the active site suggests a role for a conserved glutamate (Glu-200) in catalysis; mutation of this residue abolishes activity. The product binds at the head of a closed tunnel, and the enzyme showed no activity on polymerized peptidoglycan. These results point to a potential role for Rv3717 in peptidoglycan fragment recycling.

Introduction

Tuberculosis is among the most deadly infectious diseases worldwide. The current treatments are lengthy, taking 6–9 months to complete, whereas emergence of strains with resistance to all front-line drugs puts emphasis on developing new antimicrobials (1). A more detailed understanding of the physiology of Mycobacterium tuberculosis (Mtb), the etiologic agent of tuberculosis, can help improve current therapies and lay the foundation for developing new treatments. One area of bacterial physiology that can provide therapeutic targets is the mycobacterial cell wall.

Peptidoglycan forms the basis for all bacterial cell walls. A polymer constructed from disaccharide pentapeptide units, it forms a single molecule that surrounds the bacterial cell, allowing it to resist turgor pressure (2). Peptidoglycan also serves as a basis for attachment of all other cell wall-associated polysaccharides, lipids, and proteins. In mycobacteria, the resulting multilayered cell wall structure provides a highly selective barrier essential for survival (3). The mycobacterial cell wall is also a thoroughly validated drug target.

Peptidoglycan hydrolases are key enzymes in cell wall metabolism and have multiple functions in shape maintenance, cell growth, daughter cell separation, peptidoglycan maturation, and fragment recycling (4). Based on their chemical specificity, peptidoglycan hydrolases are divided into three classes: (i) glycosidases that cleave between the sugars, (ii) peptidases that target peptide bonds, and (iii) N-acetylmuramyl-l-alanine amidases that can separate peptides from sugar strands. Peptidoglycan hydrolases have been comparatively well studied in model organisms, Escherichia coli and Bacillus subtilis (5, 6). Recently, much attention has also been given to the peptidoglycan degradation machinery of Mtb (7, 8).

Peptidoglycan N-acetylmuramyl-l-alanine amidases fall into two families. Amidases from the Pfam family Amidase_2, in addition to bacterial and phage amidases, include eukaryotic proteins such as fly and human peptidoglycan recognition proteins with roles in anti-bacterial immunity (9). Members of the Pfam Amidase_3 zinc-dependent amidase family are limited to bacteria and phage and include the E. coli AmiA, AmiB, and AmiC and B. subtilis CwlB and CwlC enzymes. CwlB, previously called LytC, is the founding member of this family and was discovered as one of two major B. subtilis autolysins (10, 11). In Mtb, a peptidoglycan amidase CwlM was discovered by sequence similarity to B. subtilis amidase CwlB, and its muralytic activity was confirmed experimentally (12). Structures of one bacterial (13) and two phage (14, 15) N-acetylmuramyl-l-alanine amidases from the Amidase_3 family have been reported. However, the structural basis of substrate binding has not yet been investigated.

Here, we report the structural and biochemical analysis of Rv3717, an Mtb peptidoglycan amidase. We found that the recombinant Rv3717 protein is active on muramyl dipeptide (MDP),² a model substrate (Fig. 1A), but not on polymerized peptidoglycan. Structures of two partially assembled forms of Rv3717 allowed us to establish the contributions of Zn²⁺ binding and disulfide formation to assembly and catalysis. Structure of the mature Rv3717 bound to l-alanine-iso-d-glutamine dipeptide, a product of MDP hydrolysis, helped identify an additional catalytic glutamate residue (Glu-200) that does not participate in binding to Zn²⁺ ions. This product-bound structure demonstrates the mode of substrate binding likely shared by the Amidase_3 family proteins.

FIGURE 1.

Rv3717 is a functional zinc-dependent mycobacterial N-acetylmuramyl l-alanine amidase. A, amidase-catalyzed hydrolysis of muramyl dipeptide produces N-acetylmuramic acid and an l-alanine-iso-d-glutamine dipeptide. B, phylogenetic relationships among peptidoglycan amidases of E. coli, B. subtilis, M. tuberculosis, M. smegmatis, and a distantly related Listeria monocytogenes bacteriophage PSA endolysin, PlyPSA. Maximum-likelihood tree of Amidase_3 family peptidoglycan amidases. Bootstrapping values at nodes show high confidence in the mycobacterial (dark gray) and E. coli (medium gray) clades. C, muramyl dipeptide hydrolysis by Rv3717 depends on the presence of Zn²⁺. Increase in primary amine concentration following 40-min degradation by metal-bound and metal-free forms of Rv3717 amidase was measured using ortho-phthalaldehyde detection and corresponds to production of the dipeptide product. The error bars represent 1 S.D. above the mean of three replicate measurements.

EXPERIMENTAL PROCEDURES

Sequence Analysis of Amidase_3 Family Proteins

Hidden Markov model (HMM)-based searches of bacterial proteomes were performed with the program hmmscan, part of the HMMER 3.0 package (16). The search model for Amidase_3 was downloaded from the Pfam server (17) and the predicted proteomes of E. coli K12 substrain DH10B, B. subtilis 168, M. tuberculosis H37Rv, and Mycobacterium smegmatis mc² 155 were obtained from the NCBI (GenBank^TM CP000948, NC_000964, NC_000962, CP000480). Protein sequences of peptidoglycan amidases identified with hmmscan were trimmed to their catalytic domains and aligned using the hmmalign program. Protein alignment was analyzed using the Phylip package to produce a maximum-likelihood tree with 100 bootstrap replicates. Graphical representation of the tree was created with the FigTree program.

Cloning and Expression of Rv3717

Further sequence analysis with the SignalP web service (18) revealed that Rv3717 contained a secretion signal peptide sequence at its N terminus. Consequently, an N-terminal truncation of Rv3717 matching the sequence of the predicted mature protein (residues 20–241) was cloned into pDEST15 vector using the Gateway system (Invitrogen). N-terminal GST-fused Rv3717 was expressed by autoinduction (19) in E. coli BL-21 Codon Plus cells. Cell pellets were harvested by centrifugation and stored at −80 °C. Cells were resuspended in buffer A (300 mm NaCl, 20 mm HEPES, pH 7.5, 5% glycerol, 0.5 mm TCEP) supplemented with protease inhibitors 4-(2-aminoethyl)benzenesulfonyl fluoride and E-64 (0.25 mm and 1 μm, respectively). Cells were lysed by sonication, lysates were cleared by centrifugation, and glutathione affinity chromatography was carried out at room temperature using 5-ml GST-affinity columns (GE Healthcare). After elution with 30 mm glutathione in buffer A, protein was cleaved with 0.1 mg of TEV protease/liter of culture while being dialyzed against buffer B (30 mm NaCl, 20 mm HEPES, pH 7.5, 5% glycerol, 0.5 mm TCEP). The sample was passed through GST-affinity and anion exchange Capto-Q columns (GE Healthcare) attached in tandem to achieve complete removal of the GST tag. The flow-through fraction was oxidized by addition of one-tenth final volume of oxidizing buffer C (100 mm reduced glutathione, 10 mm oxidized glutathione, 300 mm Bistris propane, pH 9, 10% glycerol, 300 mm NaCl, 10 mm zinc acetate). The sample was filtered through a 0.2-μm syringe filter and concentrated using centrifugal filters with a 10-kDa cutoff (Amicon). Concentrated oxidized protein was applied to a Superdex-75 column mounted on an FPLC instrument and preparative size-exclusion chromatography was performed against a non-reducing buffer containing 100 mm NaCl, 20 mm HEPES, pH 7.5, and 10% glycerol.

MDP Hydrolysis by Rv3717

Reactions included 100 mm sodium phosphate buffer, pH 6.5, MDP was used at 500 μm, and Rv3717 at 5 μm. Reactions were mixed and incubated at room temperature for 40 min and stopped by centrifugation through a 10-kDa cutoff filter. Sample aliquots of 20 μl were mixed with 100 μl of o-phthalaldehyde developing solution (1.5 mm o-phthalaldehyde, 350 mm sodium borate, pH 9.5, 1% (v/v) β-mercaptoethanol). Following a 15-min incubation, absorbance reading at 340 nm was recorded and results were analyzed in Excel. For the EDTA-treated sample, the protein was washed with 10 mm EDTA in 100 mm sodium phosphate over the concentrator, and EDTA was removed by washing two more times with EDTA-free buffer. For the EDTA-treated, zinc-reconstituted protein, the enzyme was treated as above and 1 mm final concentration of zinc acetate was added to the assay buffer. For the comparison of oxidized and reduced Rv3717, the protein was preincubated in 100 mm sodium phosphate buffer, pH 8.0, with or without 10 mm TCEP for 2 h under nitrogen atmosphere; 1 mm final concentration TCEP was included in the assay buffer for the reduced sample.

Crystallographic Analysis of Rv3717

All forms of Rv3717 were crystallized in 200 mm NaCl, 100 mm Tris, pH 8.5, 25% PEG 3350 by vapor diffusion in a hanging drop format. The crystals were harvested from the drops and frozen in liquid nitrogen. Data were collected at Beamline 8.3.1 at the Advanced Light Source (20) and reduced using HKL2000. Molecular replacement using Bartonella henselae AmiB as the search model (PDB code 3NE8) was performed using Phenix (21). Model building and refinement were carried out using Phenix and Coot (22). The data collection and model refinement statistics are listed in Table 1. Molecular images were generated using Chimera (23). Mapping of the secondary structure to the protein alignment was performed using ESPript. For surface conservation analysis of Rv3717, we used BLAST to gather 100 of the highest scoring unique protein sequences from phylum Actinobacteria (Taxonomy ID 201174). The sequences all had greater than 95% query coverage, yet ranged in sequence identity from 38 to 100%. They were aligned using ClastalW2 algorithm on the EBI server with default parameters. Chimera (23) was used to map the percent residue conservation scores onto the protein surface.

TABLE 1.

Data collection and refinement statistics

	Rv3717 l-Ala-iso-d-Gln complex	Rv3717 reduced, metal-free form	Rv3717 reduced, Zn2+-bound form
Data collection
Wavelength (Å)	1.12	1.12	1.28
Temperature (K)	100	100	100
Space group	P 21212	C2	C2
Cell parameters
a, b, c (Å)	56.0/76.0/49.5	125.4/45.7/67.8	125.4/45.9/68.1
α, β, γ (o)	90/90/90	90/116.0/90	90/116.3/90
Asymmetric unit copies	1	2	2
Resolution (Å)a	50.0–2.10 (2.14–2.10)	50.0–2.20 (2.24 –2.20)	50.0–2.67 (2.72–2.67)
Rsym (%)	15.8 (52.7)	9.1 (51.9)	14.4 (77.7)
I/σI	8.65 (3.27)	9.8 (1.4)	13.1 (1.5)
Completeness (%)	97.54 (87.31)	97.13 (94.67)	95.99 (82.89)
Redundancy	6.8 (5.5)	2.3 (2.3)	7.5 (5.1)
Refinement
Resolution (Å)	45.07–2.10	42.33–2.19	42.52–2.67
Number of reflections	12,431	16,314	9,721
Rwork/Rfree (%)	20.63/24.46	20.46/23.28	19.7/22.32
Number of atoms
Protein	1565	2599	2660
Solvent	66	79	18
B factors
Protein (Å2)	32.3	37.5	35.6
Solvent (Å2)	33.4	36.6	22.6
Root mean square deviations
Bond lengths (Å)	0.008	0.005	0.005
Bond angles (°)	0.850	1.186	1.102
Ramachandran plot
Favored (%)	97	98	98
Disallowed (%)	0	0	0
PDB ID	4M6G	4M6H	4M6I

^a Values in parentheses refer to the highest resolution shell.

Whole B. subtilis Peptidoglycan Degradation

0.1 mg of B. subtilis peptidoglycan (Sigma) in aqueous suspension was treated by 0.01 mg of either Rv3717 or mutanolysin in 50 mm sodium phosphate buffer, pH 6.5, for 60 h with shaking. Mutanolysin samples were supplemented with 1 mm magnesium chloride and amidase samples with 1 mm zinc acetate. Absorbance measurements at 595 nm were used to measure the decrease in turbidity.

RESULTS

As part of a comprehensive annotation of Mtb peptidoglycan hydrolases, we identified Rv3717 and Rv3915, two enzymes that belong to the Pfam family Amidase_3. The founding member of this Pfam family is B. subtilis amidase CwlB, previously LytC, the family also includes E. coli AmiA, AmiB, and AmiC proteins. The Mtb hydrolase Rv3915 has been previously identified and named CwlM by Deng and colleagues (12), whereas Rv3717 has not yet been investigated. Phylogenetic analysis indicated that the two mycobacterial amidases form a separate clade suggesting that their diversification occurred after divergence from E. coli and B. subtilis homologs (Fig. 1B).

We purified recombinant Rv3717 protein and showed that it was able to hydrolyze MDP, releasing N-acetyl muramide and l-Ala-iso-d-Gln dipeptide products (Fig. 1, A and C). Rv3717 hydrolytic activity was dependent on the presence of Zn²⁺ ions retained by the enzyme during gel filtration against a zinc-free buffer. Washing the enzyme with 10 mm EDTA abolished hydrolytic activity, and this effect was reversed by the addition of excess Zn²⁺ to the reaction buffer (Fig. 1C). We concluded that Rv3717 is a functional zinc-dependent peptidoglycan amidase.

To investigate the structure of Rv3717 and its mode of ligand binding, we determined the crystal structure of the recombinant protein. Rv3717 crystallized in the presence of MDP in space group P2₁2₁2 with one enzyme-ligand complex per asymmetric unit. X-ray data were collected at cryogenic temperature using synchrotron radiation (Table 1). The structure was solved with molecular replacement using B. henselae AmiB (PDB 3NE8). Rather than the MDP substrate that was used in the crystallization, the electron density clearly revealed one of the products, l-Ala-iso-d-Gln, bound in the active site. The Rv3717 structure contains the typical features of the Amidase_3 fold: a central six-stranded β-sheet, six surrounding α-helices, and a Zn²⁺ ion coordinated by two histidines and one glutamate in the active site (Fig. 2A). In addition, Rv3717 has a unique feature, a 20-amino acid insertion that forms a short 3₁₀-helix and a β-hairpin linked to the core enzyme with a disulfide bond (Fig. 2B). This insertion is absent in homologs from model organisms as well as the paralogous Rv3915 amidase in Mtb (Fig. 2C).

FIGURE 2.

The structure of Rv3717 is an elaboration on the Amidase_3 fold. A, ribbon diagram of Rv3717 in complex with dipeptide l-Ala-iso-d-Gln, one of two products formed during muramyl dipeptide hydrolysis. The amidase catalytic center contains a Zn²⁺ cation that is coordinated by His-35, His-125, and Glu-70. B, comparison of the Rv3717 structure (yellow) to another Amidase_3 protein, CwlV from Paenibacillus polymyxa (PDB 1JWQ, blue) highlights the intra-domain insertion of a short 3₁₀-helix and a β-hairpin characteristic of Rv3717 and its actinomycete orthologs. The added features are held close to the enzyme core by disulfide-bonded cysteine residues (spheres). C, section of the protein sequence alignment highlighting the amino acid insertion (residues 41–60 in the alignment) responsible for this Rv3717 structural feature.

Because this unique disulfide-bonded insertion is positioned close to the active site, we hypothesized that disulfide oxidation could affect enzyme folding and catalysis. To address this question, we solved crystal structures of reduced Rv3717 in zinc-bound and zinc-free forms. Reduced protein crystallized under the same conditions but in space group C2 with two protein molecules per asymmetric unit. Compared with the oxidized protein structure, the model for the reduced protein contains internal gaps due to lack of electron density. In the zinc-free form, two of the zinc-binding residues, His-35 and Glu-70, were disordered in the B molecule, whereas the whole His-35 residue including backbone atoms was in the wrong conformation to coordinate Zn²⁺ in the A molecule (Fig. 3A, top panel).

FIGURE 3.

Zn²⁺ binding and disulfide oxidation complete Rv3717 folding. A, in the absence of metal, the catalytic center residues of reduced Rv3717 are in the wrong conformation to coordinate Zn²⁺ (top panel), and they are reoriented following addition of the metal (bottom panel). B, the characteristic insertion that is folded in oxidized Rv3717 (yellow) is disordered in both metal-free (purple) and Zn²⁺-bound (salmon) structures of the reduced protein, leaving the folded core of a six-stranded β-sheet and six α-helices. C, disulfide reduction in Rv3717 does not slow MDP hydrolysis.

When the crystals of the reduced Rv3717 were soaked with Zn²⁺-containing cryoprotectant solution, crystallographic analysis revealed reordering of the catalytic residues and reappearance of Zn²⁺ density (bottom panel in Fig. 3A). The β-hairpin and the 3₁₀-helix that comprise the Rv3717 unique insertion were disordered in both forms (Fig. 3B). Thus, disulfide oxidation appears dispensable for folding of the core of the enzyme and of its catalytic center. Consistent with these results, reducing the disulfide bond did not alter the efficiency of MDP hydrolysis (Fig. 3C).

Although mature Rv3717 was crystallized in the presence of muramyl dipeptide, the electron density envelope in the active site corresponded to the dipeptide, l-Ala-iso-d-Gln (Fig. 4A). No density features that could accommodate muramic acid were present. We concluded that MDP was hydrolyzed during the crystallization process and that only the dipeptide product remained bound to the enzyme. The leaving group of the peptide was coordinated by a water molecule adjacent to a conserved amino acid, Glu-200, suggesting a role for this residue in catalysis (Fig. 4B). Consistent with this idea, mutating Glu-200 to either alanine or glutamine abolished MDP hydrolysis (Fig. 4C).

FIGURE 4.

l-Ala-iso-d-Gln dipeptide product bound to the Rv3717 active site identifies catalytic and substrate-binding residues. A, electron density envelope for the dipeptide product, but not the full MDP or N-acetylmuramic acid, is present in the active site. B, residues involved in catalysis include the Zn²⁺-binding triad of His-35, Glu-70, and His-125 along with the conserved Glu-200. C, Glu-200 is required for substrate hydrolysis. D, surface of Rv3717 amidase colored by residue conservation among 100 actinobacterial homologs highlights the conserved extended substrate-binding surface around the active site.

As a guide to define other functional sites, we focused on the conservation of the ligand-binding surface among likely Rv3717 orthologs. We used BLASTP to collect and ClustalW2 to align 100 highest scoring Rv3717 homologous sequences from the phylum Actinobacteria (95–100% query coverage, 38–100% amino acid sequence identity). We mapped amino acid sequence conservation to the Rv3717 protein surface and discovered that residues in the extended active site that interact with the bound product are highly conserved across Rv3717 homologs (Fig. 4D). Thus the mode of substrate binding is likely similar among these proteins.

Because other amidases have been reported to break down whole peptidoglycan sacculi, we tested the activity of Rv3717 in a standard assay based on turbidity of a peptidoglycan solution. Compared with the mutanolysin positive control, Rv3717 did not clear B. subtilis peptidoglycan suspension (Fig. 5A). This inability to process polymerized peptidoglycan is associated with a distinct feature of the substrate-binding site. The dipeptide product binds at the head of a large blind tunnel that extends into the body of the enzyme (Fig. 5B). The tunnel is long enough to accommodate 2–3 additional amino acids beyond the iso-d-Gln, but not an entire peptide cross-link. In contrast, the structures of Amidase_3 enzymes that hydrolyze polymerized peptidoglycan, including Listeria phage PSA endolysin (Fig. 5C), contain equivalent Zn²⁺-bound catalytic centers within more open and accessible active sites.

FIGURE 5.

Rv3717 amidase activity is restricted to peptidoglycan fragments. A, B. subtilis peptidoglycan sacculi are solubilized by mutanolysin but not Rv3717, detected by the drop in optical density. B, Rv3717 substrate binding site extends toward a closed tunnel that could limit substrate selection. Section through the Rv3717 active site shows the bound dipeptide extending away from the catalytic center toward a hydrophobic well in the enzyme surface. Hydrophobic surfaces are shown in orange, hydrophilic in blue. C, the same section through the catalytic domain of a homologous phage amidase (Listeria phage PSA endolysin, PDB 1XOV) shows a substrate-binding site that is more solvent exposed.

DISCUSSION

Peptidoglycan hydrolases are central players in bacterial cell wall homeostasis. Significant advances have been made investigating peptidoglycan hydrolases of model organisms, and extension of these studies to medically important species is a priority. In this work, we identified an Mtb peptidoglycan hydrolase, Rv3717, a homolog of well studied enzymes AmiA, AmiB, and AmiC of E. coli and CwlB of B. subtilis. We confirmed that Rv3717 is a zinc-dependent peptidoglycan amidase and have determined crystal structures of the mature enzyme in complex with a dipeptide product and two assembly intermediates.

Our analysis shows that Rv3717 is a single domain protein that is likely targeted to the pseudo-periplasmic space by the signal peptide. Once outside the cytoplasm, folding is completed by Zn²⁺ binding and disulfide oxidation necessary to lock down a 20-residue insertion that forms a raised rim around the active site. The core of the enzyme shared with other Amidase_3 family proteins includes the central six-stranded β-sheet, two α-helices on the face containing the active site, and four α-helices on the opposing face. Conserved catalytic center includes the Zn²⁺ ion that activates a bound water molecule for hydrolysis of the amide bond and three Zn²⁺-coordinating residues: His-35, His-125, and Glu-70. The Rv3717 characteristic 20-residue insertion forms a short 3₁₀-helix followed by a β-hairpin. This structural addition is anchored to the core of the enzyme by a disulfide bond. Disulfide oxidation and folding of the insertion are not required for folding of the enzyme core or in vitro MDP hydrolysis. These observations do not rule out potential contributions of this feature of the enzyme to substrate selection and in vivo activity.

The structure of Rv3717 bound to the l-alanine-iso-d-glutamine dipeptide provides previously unobserved details of substrate recognition and catalysis by this protein family. We identified an additional residue in the active site, Glu-200 in Rv3717, that is positioned behind the substrate and could function as a general acid/base in the reaction. This residue is conserved in Rv3717 homologs and mutations to either alanine or glutamine block catalysis. Other residues in contact with the bound dipeptide are highly conserved among the related actinobacterial amidases, including those with as little as 38% overall sequence identity, indicating that the mode of substrate binding is likely conserved among these enzymes. Structural alignment of Rv3717 to the three reported amidases shows that the dipeptide fits the surfaces of two of them, Listeria phage PSA and Clostridium phage phicd27 endolysins, without changes in position or conformation. In the third protein, B. henselae AmiB, the available structure is of the inactive protein with the active site occluded by an α-helix (13). This observation suggests that the position of the peptide portion of the substrate relative to the catalytic Zn²⁺ ion is conserved and allows prediction of ligand interacting residues in other Amidase_3 family proteins.

Unlike its Mtb paralog, Rv3915, which is active on cell wall fragments and whole peptidoglycan sacculi (12), Rv3717 activity appears limited to peptidoglycan monomers. This difference could be explained by the shape of the Rv3717 catalytic pocket. A deep tunnel next to the distal end of the bound dipeptide product is formed in part by the Rv3717 characteristic insertion. The size of the tunnel provides an elegant steric mechanism for substrate selection. Although the catalytic center is accessible, polymerized peptidoglycan cannot engage the enzyme, because the peptide cross-links cannot enter the tunnel. On the other hand, because the dipeptide product does not fill the tunnel, we speculate that substrates with longer stem peptides could be accommodated. In this case, Rv3717 substrate binding would depend on a peptidase autolysin first cleaving the peptide cross-links. This hypothesis is strengthened by the observation by Griffin et al. (24) that Mtb tolerates transposon insertions in the rv3717 gene, but not the rv3915 gene. Thus, Rv3717 enzymatic activity is not required for bacterial in vitro growth. These observations together with the fact that Rv3717 is active on peptidoglycan fragments suggest a role for this enzyme in peptidoglycan fragment recycling.

The concept of peptidoglycan recycling is well established in the Gram-negative bacteria (25), with many dedicated enzymes including AmpG, an E. coli importer of N-acetylmuramyl peptides. Because Gram-positive bacteria shed large amounts of peptidoglycan fragments into the medium, it has been assumed that they do not recycle peptidoglycan. This view, however, is currently being challenged (26). In the proposed Gram-positive recycling pathway, amidase activity is required for uptake of N-acetylmuramyl peptides due to the absence of an AmpG-like transporter. Because Mtb, more closely related to Gram-positive bacteria, possesses a nearly impermeable outer membrane the peptidoglycan fragments produced during cell wall turnover may well be recycled. The substrate specificity of Rv3717 defined by the tunnel that limits the length of stem peptides provides a guide for studies of the role of this enzyme in vivo.