Peptidoglycan Amidase MepA Is a LAS Metallopeptidase*

LAS enzymes are a group of metallopeptidases that share an active site architecture and a core folding motif and have been named according to the group members lysostaphin, d-Ala-d-Ala carboxypeptidase and sonic hedgehog. Escherichia coli MepA is a periplasmic, penicillin-insensitive murein endopeptidase that cleaves the d-alanyl-meso-2,6-diamino-pimelyl amide bond in E. coli peptidoglycan. The enzyme lacks sequence similarity with other peptidases, and is currently classified as a peptidase of unknown fold and catalytic class in all major data bases. Here, we build on our observation that two motifs, characteristic of the newly described LAS group of metallopeptidases, are conserved in MepA-type sequences. We demonstrate that recombinant E. coli MepA is sensitive to metal chelators and that mutations in the predicted Zn2+ ligands His-113, Asp-120, and His-211 inactivate the enzyme. Moreover, we present the crystal structure of MepA. The active site of the enzyme is most similar to the active sites of lysostaphin and d-Ala-d-Ala carboxypeptidase, and the fold is most closely related to the N-domain of sonic hedgehog. We conclude that MepA-type peptidases are LAS enzymes.

Peptidoglycan amidases are important for bacterial cell wall remodeling during cell growth and division (1). It has become clear in recent years that peptidoglycan hydrolases exist for nearly every amide linkage that occurs in bacterial cell walls (2). In many species of bacteria, the number of peptidolglycan hydrolases is even higher than the number of different amide linkages, pointing to a functional redundancy in peptidoglycan hydrolases or to a role of specific peptidoglycan hydrolases only at certain stages of the bacterial life cycle (2). Although the different peptidoglycan hydrolases have either no or no significant similarity at the sequence level, work in several laboratories has established that most of these peptidoglycan hydrolases are structurally related (3)(4)(5)(6).
We have recently proposed to classify lysostaphin-type enzymes, metallopeptidases with specificity for D-Ala-D-Ala and sonic hedgehog as LAS enzymes (6). LAS enzymes contain a single, tetrahedrally coordinated Zn 2ϩ in their active sites. Three Zn 2ϩ ligands are conserved and occur in the order histidine, aspartate, histidine in the sequence. They are part of two motifs, HX(3,6)D and HXH (EXXH in VanX-type enzymes), which are separated by 30 to 100 amino acids in the sequence. In the crystal structures, the histidine of the first motif coordinates the Zn 2ϩ via the N-⑀ atom and is oriented by a hydrogen bond from its N-␦ atom to a main chain carbonyl oxygen atom. The aspartate acts as a monodentate ligand, with one of its O-␦ atoms in contact with Zn 2ϩ and with the other exposed to solvent in the active enzymes. The glutamate of the EXXH motif of VanX-type enzymes and the first histidine of the HXH motif in all other LAS enzymes are close to the Zn 2ϩ in the crystal structures, but never in direct contact with it, suggesting that these residues may act as the catalytic base. The histidine of the EX(2)H motif and the second histidine of the HXH motif contact the Zn 2ϩ directly via their N-␦ atoms, and, except in VanX, are oriented by hydrogen bonds from their N-⑀ atoms to varying hydrogen bond acceptors. The identity of the fourth Zn 2ϩ ligand differs between structures. It is a water molecule in the structures of D-Ala-D-Ala carboxypeptidase (Protein Data Bank accession code 1LBU) (P. Wery, Ph.D thesis), VanX (Protein Data Bank accession code 1R44) (7), and sonic hedgehog (Protein Data Bank accession code 1VHH) (8).
In the case of the lysostaphin-type peptidase LytM (Protein Data Bank accession code 1QWY), a latent form of the enzyme was crystallized. In this structure, a poorly conserved asparagine residue from the N-terminal part of the enzyme contacts the metal via its O-␦. Biochemical data are consistent with a role of this residue as an "asparagine switch" (9).
Spatial superposition of LAS enzymes shows that the conservation of active-site architecture includes the presence of two additional residues that can be either glutamate or histidine in conserved locations. As these residues are not Zn 2ϩ ligands, it has been speculated that they play a role as general base/general acid in the catalytic mechanism. Beyond the catalytic machinery, LAS enzymes share a core folding motif of four antiparallel ␤-strands with conserved topology and strand order 1, 2, 4, 3 that is typically part of a larger central ␤-sheet with poorly conserved periphery. Outside of the core folding motif, LAS enzymes are very diverse, making the similarity between different LAS families undetectable by PSI-BLAST similarity searches (6).
Escherichia coli MepA is a periplasmic murein endopeptidase that is believed to play a role in the removal of murein from the sacculus and could also play a role in the integration of nascent murein strands into the sacculus (10). Mechanistically, the enzyme is an amidase that cleaves the alanyl-meso-2,6-diaminopimelyl peptide bond that connects peptidoglycan strands in Gram-negative cell walls (11). Partially purified preparations of E. coli MepA were reported to be sensitive to metal-chelating agents, deoxyribonucleic acid and lipoteichoic acid, but insensitive to penicillin (12). Because of the low abun-dance of MepA in the E. coli periplasm, the previously reported purification procedure required large amounts of cells (12).
Homologues of E. coli MepA occur in many Gram-negative bacteria, consistent with the universal presence of amide bonds between D-alanine and meso-2,6-diaminopimelate in these species (13). MepA-type enzymes have not been detected in Grampositive organisms (14), even though some of them, such as Bacillus subtilis, contain alanyl-meso-2,6-diaminopimelyl peptide bonds in their peptidoglycan (13). Consistent with the penicillin insensitivity of the E. coli enzyme, MepA-type peptidases are not related in sequence to the largest class of murein endopeptidases, the penicillin-binding proteins. MepA-type enzymes are poorly understood at the mechanistic level. Their fold and catalytic class have not been determined. In the MEROPS peptidase data base, they are currently classified as family U6 ("unknown") (14).
In this article, we show that MepA-type peptidases are LAS enzymes. We demonstrate that the recombinant E. coli enzyme is inhibited by Zn 2ϩ chelators, and that mutations in any of the signature sequence residues strongly reduce or abolish activity. Moreover, we present the crystal structure of E. coli MepA to prove that the fold and active site architecture of the enzyme are consistent with the classification as a LAS enzyme.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Protein Purification-Standard PCR techniques were used to amplify mepA from genomic DNA of E. coli XL-1Blue. The amplified DNA fragment was cloned into pET15b with restriction enzymes NcoI and XhoI, and the construct was checked by double strand sequencing. For protein expression, the plasmid was transformed into E. coli BL21(DE3) cells. Bacteria were initially grown in 37°C to an A 600 of 0.5-0.8, induced with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside, and then grown for up to 3 h in 25°C. For selenomethionine containing protein, the plasmid was transformed into the methionine auxotroph E. coli strain B834(DE3). Cells were cultured in 1 liter of medium at 37°C to an A 600 of 1.0 and additionally up to 6 h after induction with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside, according to the published protocol (15).
Cells were harvested and resuspended in buffer A (50 mM Tris, pH 7.5). After sonication and high-speed centrifugation at 40,000 ϫ g, the supernatant was fractionated with ammonium sulfate. The material precipitating between 35 and 60% saturation was dissolved in buffer A and dialyzed overnight against 100 volumes of buffer A with one buffer change. The dialyzed supernatant was applied to a DEAE-Sepharose FF column (Amersham Biosciences) equilibrated with buffer A, and the flow-through fractions containing MepA were collected and subsequently applied on an SP-Sepharose column (Sigma). After extensive washing of the column with 100 mM NaCl, buffer A, the protein was eluted with 200 mM NaCl, buffer A. The pooled fractions with MepA were concentrated to 1 ml (YM-10 Centricons) and subjected to a Sephacryl S-300 (Amersham Biosciences) gel filtration column in 10 mM Tris-HCl (pH 7.5), 100 mM NaCl. The pure MepA was dialyzed overnight against 100 volumes of 5 mM Tris-HCl (pH 7.5) and concentrated to 10 -15 mg/ml (Amicon 10 kDa cut-off regenerated cellulose fibers).
Crystallization-All crystals were grown at room temperature (21°C) by sitting drop vapor diffusion with 0.5 ml of reservoir buffer. Orthorhombic, monoclinic, and triclinic crystals could be grown.
Orthorhombic crystals appeared after 4 weeks by mixing equal amounts of reservoir buffer containing 0.2 M ammonium acetate, 0.1 M trisodium citrate dihydrate (pH 5.69), 30% (w/v) polyethylene glycol 4000, and 10 mg/ml protein solution in 5 mM Tris-HCl (pH 7.5). One asymmetric unit consisted of two molecules. For cryoprotection crystals were soaked in the mixture containing 85% reservoir buffer and 15% polyethylene glycol 400 (Table I).
Monoclinic crystals were grown against a reservoir containing 30% (w/v) polyethylene glycol monomethyl ether 5000, 200 mM ammonium sulfate, 100 mM MES 1 (pH 6.5), by mixing 2 l of the reservoir buffer with 2 l of protein solution (15 mg/ml) and 0.4 l of 1,4-butanediol. Crystals with two monomers of MepA in the asymmetric unit appeared within 3-4 days. They could be flash cryocooled directly from the mother liquor (Table II).
Triclinic crystals were grown under the same conditions as monoclinic ones if the additive 1,4-butanediol was omitted from the drop, by mixing 2 l of the reservoir buffer with 2 l of protein solution (10 mg/ml). Crystals contained 6 molecules per asymmetric unit, and could be reproduced also with the selenomethionine version of the protein (Table II).
Structure Determination-All datasets were collected at beamline BW6 at DESY synchrotron in Hamburg (Germany). Fluorescence scanning showed that the orthorhombic and triclinic forms contained Zn 2ϩ , whereas Zn 2ϩ was unexpectedly absent from the monoclinic form.
The orthorhombic crystal form was solved by multiple anomalous diffraction at the Zn 2ϩ edge. Three highly redundant datasets with both Friedel mates were collected to a resolution of about 2.6 Å (Tab. I), and the dataset at the remote wavelength was then extended in a final pass with long exposures to a resolution of 1.9 Å. AUTOSHARP (16) was used to identify the two Zn 2ϩ sites and to calculate initial phases with an overall figure of merit of 0.57 to 2.6 Å that were subsequently improved by SOLOMON (17). These phases combined with the high resolution dataset served as an input for ARP/WARP (18) that delivered an ϳ50% complete model (without sequence assignment) for subunit A, and almost no model for subunit B. No further building was attempted in this crystal form.
Two copies of the partial model for subunit A were subsequently located in the monoclinic crystal form using MOLREP (19) and used to calculate starting phases for an ARP/WARP-run at 1.4-Å resolution, which yielded near complete models and a confident sequence assignment for both monomers in this crystal form.
The model from the monoclinic crystal form was then used to solve the triclinic crystal form, again by molecular replacement using MOL-REP (19). The MOLREP search for six monomers worked with default settings, probably because of the high quality of the model from the monoclinic crystal form and because of the presence of pseudotranslation vectors relating several monomers in identical or near identical orientations. As the success of the molecular replacement procedure was not expected at the stage of data collection, we had collected anomalous data at both the Zn 2ϩ and selenium edges for a selenomethinonine crystal of this form. With the help of the model phases, the Zn 2ϩ and selenium scatterers could be identified with excellent signal in anomalous difference Fourier maps. This assignment confirmed Zn 2ϩ in the active sites, identified three additional Zn 2ϩ sites, and confirmed all selenium sites predicted by the model, strongly supporting the sequence assignment. Heavy atom sites were then used to calculate a  (20), and this map was improved by combined solvent flattening and 6-fold averaging with DM (20) using model-derived NCS operators and a model-derived averaging mask. The resultant averaged map was very clear and allowed to model differences between the triclinic and monoclinic forms with confidence. Structure Refinement-The orthorhombic crystal form proved unexpectedly difficult to refine even with structural information from the other crystal forms. We were unable to lower the free R factor significantly below 30% (for 1.9 Å data), suggesting that changes or more likely disorder in the poorly defined subunit B in this crystal form are not adequately described by the model.
The monoclinic form has well defined electron density for both subunits, and could be satisfactorily refined with REFMAC (21) to an effective resolution of 1.4 Å with good stereochemistry (Table II). The final model comprises residues 20 to 244 and 260 to 274 for subunit A, residues 23 to 244 and 259 to 272 for subunit B, and 560 water molecules and 9 sulfate anions from the crystallization buffer. The assignment of one molecule of 1,4-butanediol from the crystallization buffer is very tentative.
The triclinic form at 2.4-Å resolution was refined with CNS (22) and REFMAC (21) without NCS restraints in the final stage. Models are similarly complete as for the P21 form, but all six copies of MepA contain Zn 2ϩ in their active sites. Three additional Zn 2ϩ ions are located at the interfaces between monomers that are related by local 2-fold symmetry. Sulfates from the buffer are bound to sites near the active centers that also accommodate sulfates in the P21 form.
Generation and Purification of Mutants-MepA mutants H113A, D120A, H209A, and H211A were obtained by PCR-based site-directed mutagenesis according to the Stratagene protocol with Pfu Turbo DNA polymerase (Stratagene). The engineered MepA variants were all soluble and expressed at least at the same level as wild type MepA. For their purification, the same protocol as for the wild type MepA could be used. Alternatively, a simplified procedure was applied, with the periplasmic fraction isolation as a first step. This included growing cells at 30°C to an optical density (A 600 ) of 0.6, induction with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside, and further incubation for 2 h at the same temperature. Harvested cells were then incubated in Buffer B (20% sucrose, 30 mM Tris-HCl, pH 7.5) for 10 min at room temperature and centrifuged to collect cells. Osmotic shock was accomplished by the rapid addition of 5 mM MgSO 4 to washed cells and gentle mixing. The sample was then incubated on ice for an additional 5 min and centrifuged at 12,000 ϫ g to remove the spheroplasts and intact cells from the supernatant, which was the periplasmic fraction. MepA mutants were the dominant band on SDS-polyacrylamide gels in this solution. For final purification, the solution was subjected to ion-ex-change chromatography on SP-Sepharose resin as described above for native MepA. After dialysis to 5 mM Tris-HCl buffer (pH 7.5), the protein could be concentrated and stored in Ϫ70°C.
Peptidoglycan Isolation-Gram-negative peptidoglycans were isolated according to Ref. 23 with modifications. Briefly, Pseudomonas putida were grown in 30°C on LB medium and harvested at A 600 of ϳ1.5. The cell pellet was resuspended in cold water and added slowly to 8% boiling SDS. Samples were boiled for another 30 min and then incubated 12 h at room temperature. Polymeric peptidoglycan was recovered by high-speed centrifugation (100,000 ϫ g) and washed several times to remove traces of SDS. Harvested, insoluble material was then treated with ␣-amylase at 37°C and subsequently with Pronase at 60°C. After dilution with water, the sacculi were washed several times and finally resuspended in 10 mM Tris-HCl buffer (pH 7.5).
Zymography-The activity of MepA was detected on 12% polyacrylamide-sodium dodecyl sulfate gels containing 0.05% P. putida PCM 2124 murein sacculi, according to the described protocol (24) with modifications. Before the gels were cast, peptidoglycans were homogenized by sonication. Samples containing whole cell supernatants (cytoplasmic and periplasmic fraction) were prepared from cells grown as described above. The supernatants were loaded on the gel in amounts corresponding to 5 g of whole protein content. In the case of purified proteins, about 0.2 g of protein was loaded per lane. Electrophoresis was done in the cold room. Gels were incubated in renaturation buffer (0.5% Triton X-100, 5 mM MgCl 2 , 20 mM BisTris buffer, pH 6.0) for 16 h, with one buffer exchange. After incubation, the zymograms were rinsed with Milli-Q water, stained with 0.1% methylene blue in 0.01% KOH, and destained in deionized water. Lytic zones indicating the enzyme hydrolytic activity appeared as clear zones on the blue background. For testing the dependence of MepA activity on pH of the buffer, zymography gels were incubated in renaturation buffers of different pH values, using 20 mM sodium citrate buffer for pH 4 and 5, and 20 mM Tris-HCl buffer for pH 7, 8, and 9.

LAS Motifs in MepA
Enzymes-Our recent definition of LAS enzymes (6) prompted us to search for additional protein families that could belong to this group. Therefore, we searched alignments of poorly characterized protease families for the HX(3,6)D and HXH motifs that are characteristic for LAS enzymes and found that both are present in MepA peptidases. We further inferred from consensus secondary structure predictions that the characteristic histidines and the aspartate MepA Purification-The E. coli mepA gene was amplified by PCR and various E. coli expression constructs were prepared. A number of attempts to express the protein in the cytoplasm, both alone and as a fusion protein, yielded large amounts of insoluble protein. However, when overexpressed MepA was targeted to the periplasm by its own native leader sequence, soluble protein in moderate yield could be recovered from whole cell lysates. As recombinant overexpression of MepA rendered cells prone to lysis, direct isolation of MepA from the periplasm was not possible. Nevertheless, mass spectrometry analysis of MepA purified from whole cell lysates showed that the overexpressed protein had a mass of 28,295 Da, corresponding to the protein after cleavage of the predicted 19-residue periplasmic leader sequence. The protein was purified as described under "Experimental Procedures," and obtained in an overall yield of ϳ2 mg per liter of culture.
Zymography Assay for MepA Activity-With sufficient amounts of enzyme in hand, we next attempted to develop a simple, non-radioactive assay for MepA activity. Previous experience with peptidoglycan peptidases suggested that the activity could be conveniently assayed by zymography with purified peptidoglycans polymerized as a substrate into denaturing SDS-PAGE gels. As E. coli peptidoglycan is difficult to isolate in quantity, we settled for peptidoglycan from P. putida that is similar or identical to E. coli peptidoglycan and gave clear signals for MepA activity (Fig. 1A).
Inhibitor Sensitivity of MepA-Our hypothesis about the similarity of MepA to LAS enzymes suggested that the enzyme should be sensitive to the metal chelating agents EDTA and 1,10-phenanthroline, but not to serine and cysteine peptidase inhibitors such as phenylmethylsulfonyl fluoride and E-64. This turned out to be the case. Zn 2ϩ at 10 mM concentration was also inhibitory to the enzyme, consistent with previous reports about the inhibition of zinc-dependent metallopeptidases by unphysiologically high concentrations of Zn 2ϩ (9, 25) (Fig. 1B).

Role of Conserved Residues in the HX(3,6)D and HXH Motifs-
The LAS hypothesis further implied that the conserved histidines and the aspartate in the HX(3,6)D and HXH motifs of MepA should be catalytically important (9,26). To test this assumption for MepA, each of the four conserved residues was separately mutated to alanine. E. coli cells overexpressing any of the four mutants were significantly less prone to lysis than cells overexpressing the wild-type protein. As described under "Experimental Procedures," this allowed the isolation of periplasmic fractions and thus considerably simplified the purification procedure. More importantly, it suggested a difference in activity between the wild-type and mutant proteins. This difference was also clearly apparent from the zymography results. As the mutants were expressed in mepAϩ background, the preparations of mutant protein were contaminated with traces of wild-type protein, which could be responsible for the residual activity that was observed in these lanes (Fig. 1C).
Determination of the MepA Structure-The definition of LAS enzymes requires the presence of a conserved catalytic metal center in the context of a core folding motif of four antiparallel ␤-strands with characteristic topology. To investigate the fold and detailed active site architecture of MepA enzymes, we decided to solve the crystal structure of E. coli MepA. To increase the chances that the structure of the enzyme in the crystal would be physiologically relevant, we first tested the dependence of the enzyme activity on the pH of the buffer. Although the zymography requires proper refolding and activity, we can be sure from the results in Fig. 2 that the enzyme is active at least in the range from pH 5 to 8. Gratifyingly, the enzyme could be crystallized in three different crystal forms in this pH range ("Experimental Procedures"). Scans of the x-ray fluorescence showed that the triclinic and orthorhombic crystal forms contained Zn 2ϩ . Surprisingly, fluorescence scans of the best diffracting monoclinic form demonstrated that Zn 2ϩ was entirely absent from these crystals. They were grown from a different batch of protein, and we suspect that the Zn 2ϩ was already lost at this stage of protein purification, because no exogenous Zn 2ϩ was added to the chromatography buffers.
The crystal structures were solved as described under "Experimental Procedures," exploiting information from all three crystal forms at different stages of the structure solution process. In all forms, MepA was present as a dimer, and one dimer each was present in the P21 and P21212 asymmetric units, whereas the P1 form contained three dimers that were essentially translationally related. Two crystal forms were fully refined: the P1 form with Zn 2ϩ and the P21 form without Zn 2ϩ in the active center. The effects of the presence or absence of Zn 2ϩ are confined to the region around the active site and to the dimer-dimer interface. Although remote from both the active sites and the dimer interface, there is also substantial variation in the order and location of the C-terminal helix downstream of the disordered region of MepA. If this region is excluded from the superposition, monomers superimpose with a root mean square deviation of ϳ0.5 Å (Tables I and II).
MepA Active Center-The crystal structure of the Zn 2ϩ containing form of MepA shows an active site that fits the LAS consensus perfectly well. The Zn 2ϩ is tetrahedrally coordinated by histidine 113 and aspartate 120 of the HX(3,6)D motif and histidine 211, the second histidine of the HXH motif. His-113 contacts the Zn 2ϩ with its N-⑀ and donates a hydrogen bond from the N-␦ to the main chain carbonyl oxygen of methionine 98. For histidine 211, it is the N-␦ that contacts the metal. The N-⑀ is within hydrogen bonding distance and properly oriented to donate a hydrogen bond to aspartate 118 O-␦, a strictly conserved residue in the MepA family (Fig. 3).
As MepA is active at the crystallization pH, we would have expected a solvent molecule as the fourth Zn 2ϩ ligand. This is clearly not the case in our crystal structure, where histidine 110 N-␦ plays this role. Although the Zn 2ϩ site resembles a structural Zn 2ϩ site, the analogy with other LAS enzymes suggests that histidine 110 would be displaced by a substrate.
Several features of the MepA crystal structure support this idea. First, histidine 110 appears more mobile than the other histidine zinc ligands, because its imidazole side chain is not fixed in space by a hydrogen bond to an ancillary residue. Second, histidine 110 is defined only moderately well in electron density, and anchored on a loop that is among the most flexible regions in the structure. Third, this residue and the loop that anchors it are among the residues in the MepA structure that differ most between the Zn 2ϩ -bound and Zn 2ϩfree forms.
Histidines 206 and 209 are in close proximity to the active site, but they do not coordinate the Zn 2ϩ directly. Superposition shows that these two residues are spatially equivalent to two histidine residues that are present both in the structures of LytM (Protein Data Bank accession code 1QWY) and D-Ala-D-Ala-carboxypeptidase (Protein Data Bank accession code 1LBU) and have been considered as possible general base residues. In proximity to these two histidines, a sulfate molecule from the crystallization buffer is bound to all copies of MepA, both in the Zn 2ϩ bound and Zn 2ϩ free crystal forms, where it apparently balances some of the charge of the histidines (Fig. 3).
MepA Fold-The MepA fold is arranged around a central, six-stranded, mixed ␤-sheet. The four central strands of this sheet are antiparallel and connected as expected for LAS enzymes, with strand order 1, 2, 4, 3 in LAS nomenclature and ϩ1, ϩ2x, Ϫ1 connectivities in Richardson nomenclature (27,28). Active site residues are anchored on this core motif as expected for LAS enzymes. Quantitative DALI (29) structure comparisons of MepA with all proteins in the Protein Data Bank confirm the similarity to LAS enzymes. The highest DALI scores are obtained for the superposition of MepA with the LAS proteins sonic hedgehog (DALI score 8.3) and D-Ala-D-Ala-carboxypeptidase (DALI score 5.2). The similarity between MepA and these two LAS enzymes goes beyond the previously defined consensus, and includes a helix upstream of strand 3 and another helix downstream of strand 4. The central ␤-sheet and these two helices are the most rigid parts of the MepA structure. High B-factors are found for residues on the periphery, especially residues 123 to 158 and for the last ϳ60 residues. Within this stretch of 60 residues at the C terminus, around 20 residues between two helices are disordered in all crystal forms (Fig. 4A).
MepA Disulfide Topology-In the crystal structures, and presumably also in the E. coli periplasm, MepA contains three disulfide bonds per protomer. All three cysteine disulfide bridges are strategically placed to hold the otherwise very loosely packed C-terminal residues in place. This is particu- larly true for the isolated helix at the C terminus, which lies downstream of the disordered region in the structure (Fig. 4B).
MepA Dimers in All Crystal Forms-In all crystal forms, MepA packs as a dimer and buries 1300 Å 2 of mostly hydrophilic surface. The extent of the contact area, 650 Å 2 , places the interaction between MepA monomers in the twilight zone of biologically significant homodimer interactions (30). The dimerization in the crystals brings two arginine aspartate pairs into close proximity. This favorable dipole-dipole interaction appears to compensate for the penalty of stacking the two arginine guanidinium groups. In the Zn 2ϩ containing triclinic form, a Zn 2ϩ that is distinct from the catalytic Zn 2ϩ is bound at the dimer interface. This Zn 2ϩ is tetrahedrally coordinated by aspartates 147 and histidines 150 from both monomers. In the Zn 2ϩ -free form, a water molecule takes the place of the Zn 2ϩ and hydrogen bonds to both aspartates. The two histidines move out and capture two additional water molecules.
MepA from E. coli Versus MepA from Other Species-Sequence alignments reveal that residues aspartate 147 and histidine 150 that form the Zn 2ϩ binding site in the E. coli MepA crystals are not conserved in other species. Mapping of MepA alignment scores on the molecular surface with Consurf (31) reveals an interesting pattern: the most conserved surface patches are the walls of the active site cleft, and two patches on the surface that are connected to the active site cleft and run essentially perpendicular to it (Fig. 5). The branched structure of the conserved surface patches is reminiscent of the branched structure of the MepA substrate, and thus it appears likely that the conservation pattern maps out the active site. This inter-pretation is also consistent with the location of a trapped sulfate in the crystal that is likely to exploit a binding site for a negatively charged carboxylate of the substrate. residue is taken by a water molecule in the structures of D-Ala-D-Ala carboxypeptidase and sonic hedgehog, and by the O-␦ of an asparagine residue in the structure of latent LytM. Based on the analogy with the "cysteine switches" that keep matrix metallopeptidases proteolytically inactive (32), we have previously proposed that the asparagine in the LytM structure plays a similar role as an asparagine switch. By analogy with the matrix metallopeptidases, we have further shown that proteolytic removal of a LytM "proregion" activates the enzyme, at least in vitro (9). Removal of a proregion as the activation mechanism in vivo remains to be proven for LytM, and appears highly unlikely in the case of MepA for a number of reasons. First, the proteolytic cut in MepA would have to be very precise, because the occluding histidine 110 occurs only three residues upstream of the first Zn 2ϩ ligating histidine in the sequence. Second, the cut would not separate the N-and Cterminal cleavage products that would remain anchored to each other via a disulfide linkage. Third, even if the N-terminal fragment would dissociate, it would take part of the LAS core folding motif with it, and most probably destroy the integrity of the enzyme. Thus, it appears that despite the high conservation of the active site in LAS enzymes, there is substantial variation in the identity and role of the fourth Zn 2ϩ ligand. Examples have so far been found for the absence of a fourth amino acid ligand to the Zn 2ϩ , for the presence of an amino acid ligand that is likely to be displaced by a substrate, and for amino acid ligands that require the proteolytic removal of the anchoring profragments for their activation.
Convergent or Divergent Evolution of LAS Enzymes?-The high similarity between LAS enzymes at the structural level ( Fig. 6) contrasts sharply with the lack of sequence similarity between different families of LAS enzymes, even in iterative PSI-BLAST searches, and raises the question of whether the similar active sites in different families of LAS enzymes have arisen by convergent or divergent evolution. Several arguments support the existence of a common ancestor of LAS enzymes. First, the active sites occur in the context of similar folds. The MepA structure shows that the similarity between some LAS enzymes goes beyond the previously described consensus and includes two helices that are spatially separated from the active site by the central ␤-sheet and are thus unlikely to be required for function (Fig. 6). Second, the peptidoglycan amidase activity of MepA is consistent with the role of other LAS enzymes in bacterial cell wall hydrolysis, again reinforcing with the idea of a common ancestor. So far, the N-domain of sonic hedgehog is the only LAS protein without a role in peptidoglycan hydrolysis, and indeed without any enzymatic activity, because the signaling function of sonic hedgehog is not mediated by proteolysis (33).
With the classification of MepA as a LAS enzyme, five protein families, namely lysostaphin-type enzymes, D-Ala-D-Ala amino-and carboxypeptidases, MepA like enzymes, and the N-domains of sonic hedgehog proteins are now known to belong to the LAS superfamily. It remains to be seen whether more families of LAS enzymes can be found, and, if so, whether these families would also have a role in peptidoglycan hydrolysis.