Structure and Function of the First Full-Length Murein Peptide Ligase (Mpl) Cell Wall Recycling Protein

Bacterial cell walls contain peptidoglycan, an essential polymer made by enzymes in the Mur pathway. These proteins are specific to bacteria, which make them targets for drug discovery. MurC, MurD, MurE and MurF catalyze the synthesis of the peptidoglycan precursor UDP-N-acetylmuramoyl-L-alanyl-γ-D-glutamyl-meso-diaminopimelyl-D-alanyl-D-alanine by the sequential addition of amino acids onto UDP-N-acetylmuramic acid (UDP-MurNAc). MurC-F enzymes have been extensively studied by biochemistry and X-ray crystallography. In Gram-negative bacteria, ∼30–60% of the bacterial cell wall is recycled during each generation. Part of this recycling process involves the murein peptide ligase (Mpl), which attaches the breakdown product, the tripeptide L-alanyl-γ-D-glutamyl-meso-diaminopimelate, to UDP-MurNAc. We present the crystal structure at 1.65 Å resolution of a full-length Mpl from the permafrost bacterium Psychrobacter arcticus 273-4 (PaMpl). Although the Mpl structure has similarities to Mur enzymes, it has unique sequence and structure features that are likely related to its role in cell wall recycling, a function that differentiates it from the MurC-F enzymes. We have analyzed the sequence-structure relationships that are unique to Mpl proteins and compared them to MurC-F ligases. We have also characterized the biochemical properties of this enzyme (optimal temperature, pH and magnesium binding profiles and kinetic parameters). Although the structure does not contain any bound substrates, we have identified ∼30 residues that are likely to be important for recognition of the tripeptide and UDP-MurNAc substrates, as well as features that are unique to Psychrobacter Mpl proteins. These results provide the basis for future mutational studies for more extensive function characterization of the Mpl sequence-structure relationships.


Introduction
Bacterial cell walls are characterized by the presence of peptidoglycan (murein), a macromolecule built from sugar and peptide building blocks, with the sugars in an alternating arrangement of N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc). This peptidoglycan barrier helps to maintain turgidity of the cell and balance the internal osmotic pressure [1]. Disruption of proper bacterial cell wall formation is an effective strategy for combating bacterial infections using common antibiotics, such as penicillins [2].
De novo peptidoglycan synthesis is carried out by a series of enzymes that have been extensively studied biochemically and structurally [3][4][5]. Briefly, MurA and MurB are involved in the first steps of peptidoglycan assembly with the formation of UDP-GlcNAc-enolpyruvate and UDP-MurNAc, respectively. The Mur ligases, MurC, MurD, MurE and MurF, are then involved in sequential addition of L-and D-amino acids [L-Ala, D-Glu, mesodiaminopimelate (meso-A 2 pm) and D-Ala-D-Ala, respectively] to form UDP-MurNAc-L-Ala-c-D-Glu-meso-A 2 pm-D-Ala-D-Ala ( Figure 1) [6]. This polymer is further processed by several membrane and periplasmic enzymes to ultimately create the peptidoglycan. MurC-F enzymes share a common reaction mechanism, similar three-dimensional structures, and several invariant residues that classifies them in the Mur ligase family [7][8][9][10].
Recycling of the bacterial cell wall is believed to account for ,30-60% of peptidoglycan synthesis in Gram-negative bacteria [11]. During this process, the cell wall is catabolized using transglycosidases, amidases, endopeptidases and carboxypeptidases [12]. The Mpl protein is used to link the breakdown component, the tripeptide L-Ala-c-D-Glu-meso-A 2 pm (and also tetra-and pentapeptides), to UDP-MurNAc ( Figure 1). Mpl was first identified by sequence similarity with MurC (,20-25% identity) [6] and was also, therefore, assigned to the Mur ligase family. Mpl is functionally similar to MurC, as both mechanisms involve ligation of an amino acid or peptide to UDP-MurNAc; in MurC, a single amino acid (L-Ala) is attached and in Mpl, a tri-, tetra-or pentapeptide.
In this study, we have determined the crystal structure of a fulllength apo-Mpl protein from the permafrost bacterium Psychrobacter arcticus 273-4, PaMpl (UniProt accession code Q4FVQ2_PSYA2, locus name Psyc_0032, 505 amino acids) at 1.65 Å resolution and have also biochemically characterized this enzyme. We have analyzed the sequence features that differentiate Mpl from MurC-F as well as the common regions shared by different Mpl enzymes. Several features appear to be unique to Mpl from Psychrobacter species that may be of importance to permafrost bacteria.

Protein production and crystallization
Clones were generated using the Polymerase Incomplete Primer Extension (PIPE) cloning method [19]. The gene encoding Mpl (GenBank: YP_263340, gi|71064613; UniProt: Q4FVQ2) was amplified by polymerase chain reaction (PCR) from P. arcticus 273-4 genomic DNA using PfuTurbo DNA polymerase (Stratagene) and I-PIPE (Insert) primers that included sequences for the predicted 5' and 3' ends. The expression vector, pSpeedET, which encodes an amino-terminal tobacco etch virus (TEV) protease-cleavable expression and purification tag (MGSDKIHHHHHHENLYFQ/  [6]. Enzymes MurC, MurD and MurE are involved in de novo synthesis of the bacterial cell wall peptidoglycan. In Gram-negative bacteria, it is estimated that ,30-60% of the peptidoglycan is recycled, and involves the enzyme Mpl. doi:10.1371/journal.pone.0017624.g001 G), was PCR amplified with V-PIPE (Vector) primers. V-PIPE and I-PIPE PCR products were mixed to anneal the amplified DNA fragments together. E. coli GeneHogs (Invitrogen) competent cells were transformed with the I-PIPE/V-PIPE mixture and dispensed on selective LB-agar plates. The cloning junctions were confirmed by DNA sequencing.
Expression was performed in selenomethionine-containing medium at 310 K. Selenomethionine was incorporated via inhibition of methionine biosynthesis [20], which does not require a methionine-auxotrophic strain. At the end of fermentation, lysozyme was added to the culture to a final concentration of 250 mg/ml, and the cells were harvested and frozen. After one freeze/thaw cycle, the cells were sonicated in lysis buffer [ TCEP] and the resin was washed with the same buffer. The flow-through and wash fractions were combined and concentrated for crystallization trials to 12.8 mg/ ml by centrifugal ultrafiltration (Millipore). PaMpl was crystallized using the nanodroplet vapor diffusion method [21] with standard JCSG crystallization protocols [22]. Sitting drops composed of 200 nl protein mixed with 200 nl crystallization solution were equilibrated against a 50 ml reservoir at 277 K for 26 days prior to harvest. The crystallization reagent that produced the PaMpl crystal for structure determination consisted of 0.2 M sodium acetate, 30% PEG-4000, and 0.1 M Tris-HCl, pH 8.5. No cryoprotectant was added to the crystal. Initial screening for diffraction was carried out using the Stanford Automated Mounting system (SAM) [23] at the Stanford Synchrotron Radiation Lightsource (SSRL, SLAC National Accelerator Laboratory, Menlo Park, CA). The diffraction data were indexed in monoclinic space group C2. The oligomeric state of PaMpl in solution was determined using a 1630 cm Superdex 200 column (GE Healthcare) coupled with miniDAWN static light scattering (SEC/SLS) and Optilab differential refractive index detectors (Wyatt Technology). The mobile phase consisted of 20 mM Tris-HCl pH 8.0, 150 mM sodium chloride, and 0.02% (w/v) sodium azide. The molecular weight was calculated using ASTRA 5.1.5 software (Wyatt Technology).
X-ray data collection, structure solution, and refinement MAD data were collected at the SSRL on beamline 9-2 at wavelengths corresponding to the high-energy remote (l 1 ), inflection point (l 2 ) and peak (l 3 ) of a selenium MAD experiment using the BLU-ICE [24] data collection environment. The data sets were collected at 100 K using a MarMosaic 325 CCD detector (Rayonix, USA). The MAD data were integrated and reduced using MOSFLM [25] and scaled with the program SCALA [26]. The heavy atom sub-structure was determined with SHELXD [27]. Phasing was performed with autoSHARP [28], SOLOMON [29] (implemented in autoSHARP) was used for density modification and ARP/wARP [30] was used for automatic model building to 1.65 Å resolution. Model completion and crystallographic refinement were performed with the l 1 dataset using COOT [31] and REFMAC5 [32]. The refinement protocol included experimental phase restraints in the form of Hendrickson-Lattman coefficients from autoSHARP and TLS refinement with one TLS group per domain. Data and refinement statistics are summarized in Table 1 [33].

Validation and deposition
The quality of the crystal structure was analyzed using the JCSG Quality Control server (http://smb.slac.stanford.edu/jcsg/ QC). This server verifies: the stereochemical quality of the model using AutoDepInputTool [34], MolProbity [35], and WHATIF 5.0 [36]; agreement between the atomic model and the data using SFcheck 4.0 [37] and RESOLVE [38], the protein sequence using CLUSTALW [39], atom occupancies using MOLEMAN2 [40], consistency of NCS pairs, and evaluates difference in R cryst /R free , expected R free /R cryst and maximum/minimum B-factors by parsing the refinement log-file and PDB header. Protein quaternary structure analysis was performed using the PISA server [41]. The depiction of the protein sequence on the secondary structure was adapted from an analysis using PDBsum [42], and all other renditions of the protein structure were prepared with PyMOL [43]. Atomic coordinates and experimental structure factors for Mpl from P. arcticus 273-4 to 1.65 Å resolution (code 3hn7) have been deposited in the Protein Data Bank (www.wwpdb.org).

Preparation of nucleotide precursors and peptides
The nucleotide precursor UDP-MurNAc was chemically synthesized [44] and UDP-MurNAc-peptides were prepared as described previously [45]. Radiolabeled UDP-[ 14 C]MurNAc was prepared from UDP-[ 14 C]GlcNAc (11.5 GBq/mmol, Amersham Biosciences), PEP, NADPH and purified His 6 -tagged MurA and MurB enzymes, as described previously [9]. The dipeptide L-Ala-D-Glu was synthesized chemically according to Sachs and Brand [46] and the tripeptide L-Ala-c-D-Glu-L-Lys was synthesized by a modified procedure of Schmidt et al. [47]. The tetrapeptide L-Alac-D-Glu-L-Lys-D-Ala was generated by treatment of pentapeptide L-Ala-c-D-Glu-L-Lys-D-Ala-D-Ala (Bachem) with purified E. coli PBP5, as described earlier [48]. The latter enzyme was overexpressed as a soluble His 6 -tagged form from the pET-PBP5s plasmid and purified on Ni 2+ -NTA agarose using standard procedures. meso-A 2 pm-containing peptides were prepared from the corresponding UDP-MurNAc-peptides as follows: first, mild acid hydrolysis (10 min at 100uC in 0.1 M HCl) of UDP-MurNAc-peptides generated MurNAc-peptides, which were purified by HPLC as mixtures of the two anomers a and b [49]; then, MurNAc-peptides were hydrolyzed to MurNAc and peptides by incubation with purified N-acetylmuramoyl-L-alanine amidase (E. coli AmiD) [50]. The released peptides were subsequently purified by HPLC on a 3 mm ODS-Hypersil column (0.46625 cm). Elution was with 0.05% trifluoroacetic acid for the meso-A 2 pm-containing tripeptide, or with a gradient of acetonitrile (from 0 to 20% in 30 min) in the same buffer for other peptides, at a flow rate of 0.5 ml/min. Peptides were quantified by amino acid analysis with a Hitachi L8800 analyzer (ScienceTec) after hydrolysis in 6 M HCl for 16 h at 95uC.

Protein production for biochemical characterization
The pSpeedET:Pampl plasmid was used for over-expression of the His 6 -tagged PaMpl protein in E. coli cells. In order to avoid any potential contamination with the EcMpl protein, the E. coli strain MLD2502, which has a deletion of the chromosomal mpl gene (BW25113 Dmpl::Cm R ) [13], was used as the host strain for these experiments. Cells were grown at 37uC in 1 l of 2YT medium (BioRad) supplemented with chloramphenicol and kanamycin (25 and 50 mg/ml, respectively). Growth was monitored at 600 nm and the expression of Mpl was induced with 0.02% (w/v) arabinose when the optical density of the culture reached 0.7. Overproduction of PaMpl was observed following induction of cultures with arabinose, as judged by SDS-PAGE analysis of cell extracts performed as previously described using 13% polyacrylamide gels [51]. However, amounts of this protein recovered in the supernatant fraction following cell disruption and centrifugation were lower than those recovered previously with the EcMpl ortholog [13] (data not shown). Cells were harvested 3 h later and washed with cold 20 mM potassium phosphate buffer, pH 7.4, containing 0.5 mM MgCl 2 and 0.01% 2-mercaptoethanol (buffer A). The cell pellet was resuspended in 10 ml of buffer A containing protease inhibitors (complete EDTA-free, Roche) and disrupted by sonication. The suspension was then centrifuged at 200,0006 g with a Beckman TL100 centrifuge and the His 6 -tagged Mpl protein present in the soluble fraction (supernatant) was purified at 4uC on Ni 2+ -nitrilotriacetate (Ni 2+ -NTA) agarose (Qiagen). Binding of Mpl protein to the polymer was followed by extensive washing with 45 ml of buffer A containing 300 mM KCl and 10 mM imidazole, and elution was done in 15 ml using a discontinuous gradient of imidazole, from 10 to 300 mM. The His 6 -tagged Mpl was eluted with 100 mM imidazole. After dialysis of the corresponding fraction against 200 volumes of buffer A, the Mpl preparation was applied onto 0.25 ml of cobalt resin (TALON, Clontech) for a further purification-concentration step to eliminate a 30-kDa contaminating protein frequently copurified with His-tagged proteins, which likely is the SlyD protein [52]. After washing with buffer A containing 10 mM imidazole, the His 6 -tagged Mpl protein was eluted with 0.5 ml of buffer A containing 200 mM imidazole. The final purified fraction was dialyzed against 100 volumes of buffer A and stored at 220uC after addition of glycerol (10% final concentration). The final preparation of PaMpl was concentrated to 1.2 mg/ml and was at least 95% pure as judged by SDS-PAGE analysis. Mpl migrated as a protein of ,57 kDa, consistent with the calculated value (57,187 Da, including the His-tag). The yield was relatively good at ,2 mg/liter of culture, which is, nevertheless, ,50 fold lower than that routinely obtained with EcMpl. Protein concentrations were determined by the Bradford method, using bovine serum albumin as a standard [53], or by quantitative amino acid analysis after hydrolysis of a sample in 6.0 M HCl for 24 h at 105uC. To Highest resolution shell. ESU = Estimated overall coordinate error [33]. R sym = S|I i -,I i .|/S|I i |, where I i is the scaled intensity of the i th measurement and ,I i . is the mean intensity for that reflection. R cryst = S||F obs |-|F calc ||/S|F obs |, where F calc and F obs are the calculated and observed structure factor amplitudes, respectively. R free = as for R cryst , but for 5.1% of the total reflections chosen at random and omitted from refinement. b Typically, the number of unique reflections used in refinement is slightly less that the total number that were integrated and scaled. Reflections are excluded due to negative intensities and rounding errors in the resolution limits and cell parameters.

Assay for tripeptide ligase activity
The standard assay mixture (40 ml) contained 100 mM Tris-HCl buffer, pH 8.4, 5 mM ATP, 5 mM MgCl 2 , 0.25 mM L-Ala-c-D-Glu-meso-A 2 pm, 0.4 mM UDP-[ 14 C]MurNAc (500 Bq), and purified PaMpl enzyme (70 ng of protein). Mixtures were incubated for 30 min and reactions were stopped by the addition of 10 ml of acetic acid, followed by lyophilization. The residues were dissolved in 150 ml of 50 mM ammonium formate, pH 3.2, and 130 ml-aliquots were injected onto a Nucleosil 100C 18 5 mm column (0.46615 cm, Alltech-France) using the same buffer at 0.6 ml/min as the mobile phase. Detection was performed with a radioactive flow detector (model LB506-C1, Berthold) using the Quicksafe Flow 2 scintillator (Zinsser Analytic) at 0.6 ml/min. Quantitation was carried out with the Radiostar software (Berthold).
For determination of the kinetic constants, the same assay was used with various concentrations of one substrate and fixed concentrations of the others. In all cases, the substrate consumption was ,20%, the linearity being ensured within this time interval even at the lowest substrate concentration. The data were fitted to the equation v = V max S/(K m +S) using the MDFitt software developed by M. Desmadril (IBBMC, Orsay, France).
Identical assay conditions were used when other peptides were tested as substrates. The buffer and pH used for the separation of the radiolabeled substrate and product slightly varied depending on substrate used: ammonium formate at pH 3.9 for L-Ala-D-Glu, ammonium formate at pH 4.

Sequence analysis
Mpl and MurC-F enzymes have three domains that are classified in Pfam families [54], PF01225 (Mur ligase catalytic domain), PF08245 (Mur ligase middle domain) and PF02875 (Mur ligase glutamate binding/C-terminal domain). Mpl was reported to exhibit significant sequence similarity with MurC when first isolated and identified [6]. A multiple sequence alignment (calculated using CLUSTALW [55] and depicted using ESPRIPT [56]) between PaMpl, NmMpl, EcMurC, HiMurC, TmMurC with MurD-F (EcMurD, EcMurE and EcMurF), shows that significantly more residues are, indeed, conserved between Mpl and MurC (,21-25% overall pair-wise sequence-based identity) than between Mpl and MurD-F (,15-17% overall pair-wise sequence identity) (Figure 2). The highest variation is found in the Cterminal, peptide-binding domain. Functionally, Mpl and MurC cannot substitute for each other under normal physiological conditions. However, overexpression of the E. coli mpl gene on a multi-copy plasmid was shown to complement an E. coli thermosensitive murC mutant, which indicated that Mpl could also accept L-alanine as a substrate [13].

Overall structure
The cloning, expression, purification and crystallization of PaMpl was carried out using standard Joint Center for Structural Genomics (JCSG; http://www.jcsg.org) protocols as detailed in Materials and Methods. The crystal structure of PaMpl was determined by Multi-wavelength Anomalous Diffraction (MAD) phasing to a resolution of 1.65 Å . Data collection, model and refinement statistics are summarized in Table 1. PaMpl is present as a monomer in the crystal asymmetric unit (Figure 3). The final model includes Gly0 (a remnant from cleavage of the expression and purification tag); residues 1-209, 216-265, and 285-504 of the protein (the full-length protein contains 505 residues); and 588 water molecules. Residues 210-215, 266-284 and 505 are disordered in the crystal structure and have not been modeled. The Matthews' coefficient [57] is ,2.2 Å 3 /Da, with an estimated solvent content of ,45%. The Ramachandran plot produced by Molprobity [35] shows that 98.1% of the residues are in the favored regions with none in disallowed regions.

Structure comparisons
A superimposition of PaMpl onto the partial structure of the NmMpl (ND and MD, PDB id 3eag) shows that the corresponding domains are quite similar with an r.m.s.d. of 2.2 Å over 311 Ca residues and a structure-based sequence identity of 57% ( Figure 4). However, a few notable differences, for example, include PaMpl residues 141-152 (loop and helix, green), 262-285 (orange), and  [55]) of Mpl with MurC-F, rendered using ESPRIPT [56]. The pairwise sequence identity between Mpl and MurC is ,21-25%, and is less between Mpl and MurD-F, ,15-17%. The PaMpl segment of residues 184-192 (black bar), comprised of the sequence motif (S/T)AFFDKRSK, is not present in Mur ligases and may be functionally important in Mpl. This motif packs against residues 90-92 (green bar), which are unique to and conserved in Mpl proteins (from analysis of 17 annotated Mpl enzymes in the UniProt database, August 2009, data not shown). Residues 7-13 constituting the GI(C/G)GTFM motif (blue bar) and Asp31 (blue circle) are likely to interact with UDP. Tyr182 (blue circle) is unique to Mpl and may be involved in substrate recognition. Residues that are likely to be important in ATP binding are: Asn205, His323 and Asn327 (adenine-binding, magenta circles); Asp373 (ribose-binding, orange circle); and Gly113, Lys114, Thr115, Thr116, Arg358 and His376 (tri-phosphate binding, cyan and orange circles). Metal-binding residues are likely to be Thr115, Glu178 and His210. The Psychrobacter Mpl enzymes have several insertions (residues 141-152, 161-171, 262-285 and 444-450, yellow bars), which may be related to cell wall recycling in these bacteria that survive in permafrost conditions. doi:10.1371/journal.pone.0017624.g002 444-450 (helix, cyan) that represent insertions into the core structure that are not present in NmMpl, the 161-171 loop (yellow), which adopts a different conformation compared to the corresponding loop (153-162) in NmMpl, and a disordered region in PaMpl (210-215) that is ordered (loop 199-208, red) in NmMpl.
As Mpl is most similar in sequence and function to MurC and since MurC-F enzymes have been compared in detail elsewhere [7,8,58], we have restricted our comparisons to MurC structures (EcMurC with bound Mg 2+ , PDB id 2f00 and HiMurC with substrate analogs UDP-MurNAc-L-Ala (UMA) and a nonhydrolyzable ATP analog AMPPNP (ANP) bound, PDB id 1p3d). PaMpl and EcMurC can be superposed with an r.m.s.d. of 2.2 Å over 283 Ca atoms (representing the most structurally conserved regions, ND and MD, as there is variation in the orientation of CD). Residues in EcMurC that correspond to the 3 domains of PaMpl are 1-118, 119-325 and 326-483, respectively. PaMpl is larger than EcMurC, with major insertions in the MD (255 vs. 207 residues in the MD). In a domain pair-wise comparison, the respective ND, MD and CD superpose with r.m.s.d.s of 1.4 Å (26% structure-based sequence identity), 2.1 Å (33% sequence identity) and 2.1 Å (20% sequence identity) over 99, 195 and 125 Ca residues respectively. Flexible, multiple structure alignment of PaMpl with EcMurC (PDB 2f00), HiMurC (PDB 1p31), and TmMurC (PDB 1j6u) using POSA [59] shows that these proteins share a common structural core ( Figure 5) consisting of 355 Ca residues with an r.m.s.d. of 2.7 Å . Moreover, conformational flexibility is inherent in the overall architecture of Mur enzymes as illustrated by comparing the substrate-bound and apo structures of MurC and MurF ( Figure 6, C and D, respectively). Superposition of the apo PaMpl structure with the substrate-bound structures of HiMurC (PDB 1p3d and 1gqy) results in r.m.s.d.'s of 4.2 Å over 411 Ca residues and 4.9 Å over 414 Ca residues, as a result of significant domain movements. The largest conformational difference is in the disposition of the CD in Mpl, which is rotated 30u relative to ND and MD, when compared to its disposition in MurC ( Figure 6, A and C, respectively). Detailed inspection of the individual domains reveals that the region with the least sequence conservation between Mpl and MurC proteins is in the CD as discussed later.

Oligomerization
PaMpl is present as a monomer in the crystal asymmetric unit and crystal packing analysis using PISA [41] did not identify any potential higher order oligomers in the crystal lattice. Analytical SEC suggested a tetramer in solution ( Figure 7C, highest peak in the continuous blue curve). However, static light scattering measurements indicated a molecular weight of 120.5 kDa ( Figure 7C, discontinuous blue curve) with an oligomeric state of 2.09, indicating that a dimer is the prominent oligomeric form in solution (n.b. in cases of discrepancy between SEC and light   scattering, the light scattering results are more likely to represent the dominant solution state). SEC also indicated that EcMpl is dimeric in solution. In contrast, EcMurD (PDB 1uag), EcMurE (PDB 1e8c) and EcMurF (PDB 1gg4), have been reported to be monomers in the crystal and in solution. Interestingly, EcMurC (PDB 2f00) is also a dimer in the crystal structure but, in solution, displays a dynamic equilibrium between monomeric and dimeric forms. Both forms were shown to be active [16,60]. The PFYG  (residues 222-225) motif in EcMurC (corresponding to PSTG, residues 232-235, in PaMpl), which is conserved in ,50% of MurC proteins, is involved in dimerization. Phe223 and Tyr224 (in the H7-b11 loop) from MD of one EcMurC monomer interact with Met16 and Val19 (not found in Mpl) and Val81, Ile106 and Met111 (corresponding positions in Mpl are Leu64, Thr89 and Phe94) in a hydrophobic pocket in the ND of its dimer partner (Figure 7, A and B). Also, Arg17 and Arg18 of EcMurC, which form salt bridges with Glu306 and Glu307 respectively (Figure 7, A and B), are not found in PaMpl. Thus, residues involved in dimerization in PaMpl (and Mpl proteins in general) differ, suggesting that the mode of dimerization may also be different.

ND:UDP-MurNAc binding region and MD:ATP and magnesium binding region
Residues in the ND and MD that are important for interaction with UDP-MurNAc, ATP and metal have been described for Mur enzymes, and most of these residues are conserved in Mpl ( Figure 2). Analysis of protein-ligand interactions observed in the HiMurC/UMA/ANP complex structure using LigPlot [61] shows that many of the important residues are also structurally conserved in PaMpl (Figure 8).
The glycine-rich loop GI(C/G)GTFM conserved across all Mpl enzymes ( Figure 2, blue bar, residues 7-13 in PaMpl) is also present in MurC and other oxidoreductases [62], where it interacts with UDP. However, the Phe at position 12 in PaMpl is considerably larger than the corresponding glycine in MurC enzymes. Although, the identity of the three residues that interact with the uracil ring of UDP (HiMurC Ile50, Ile87 and His70; EcMurC Leu51, Ile88 and His71 [16]) is not conserved in Mpl (Ala32, Met71 and Tyr53 in PaMpl), the nature of the interactions is conserved (main chain, hydrophobic and side-chain hydrogen bond interactions with these 3 residues, respectively, Figure 8), as the PaMpl does also bind to UDP as demonstrated by biochemical characterization. Asp31 in PaMpl is conserved in almost all Mur enzymes and is important for interactions with the diphosphate of UDP (Figure 2, blue circle). Conserved residues in Mpl proteins that are also in the vicinity of the UDP-MurNAc-binding pocket, but are not structurally equivalent to those observed to interact with ligands in the HiMurC structure (Figure 8), are Tyr35, Asn69, Arg73 and Gly74, which could also be involved in recognition and binding of UDP-MurNAc in Mpl (Figure 9). In addition, the Mplspecific region, which is conserved in the ND, comprises of SGP (residues 90-92 in PaMpl) that packs against another conserved loop (S/T)AFFDKRSK (residues 184-192, Figure 2, green and black bars). These interactions could play a role in the different positioning of ND and MD (so as to interact with both the UDP and MurNAc moieties) compared to that in Mur enzymes ( Figure 9). Tyr182, which is unique to Mpl (Figure 2, blue circle and Figure 9) and located just prior to the start of this loop, could also be involved in substrate recognition. Thus, the various Mpl residues described above are potentially involved in UDP-MurNAc interactions and provide a guide for site-directed mutagenesis to test their functional roles.
As for all other Mur enzymes, it is assumed that the ligation of peptide to UDP-MurNAc should also be ATP-driven in Mpl and that phosphorylation of the C-terminal carboxylate forms an acyl phosphate intermediate, followed by the nucleophilic attack of amino group of the amino acid or peptide substrate. Thus, in PaMpl, the corresponding adenine binding residues would be Asn205, His323 and Asn327 (Figure 2, magenta circles), the ribose binding residue would be Asp373 (Figure 2, orange circle), and residues that stabilize the tri-phosphate would be the GKTT motif (P-loop found in kinases and ATPases, residues 113-116; Lys 114 corresponding to Lys129 in HiMurC and Lys130 in EcMurC) (Figure 2, cyan circles), and Arg358 and His376 (Figure 2, orange circles) [10]. These residues are all structurally conserved when compared to the HiMurC ligand complex (Figure 8). Like MurC, PaMpl activity (see Optimal conditions for enzyme activity) is dependent on magnesium (in other structures of Mur ligases, two Mg ions (Mg1, Mg2) per molecule are found), although no Mg 2+ is observed in our structure. Mg1 would interact with the band cphosphates of ATP and Mg2 would interact with the c-phosphate based on available MurC structures. In EcMurC (PDB 2f00), only Mg2 was observed due to the absence of ATP, whereas both metals were observed in the complex structure of HiMurC (PDB 1p3d). Thus, the binding site of Mg1 in PaMpl is likely to be comprised of Thr115 and Glu178 (corresponding to Thr130 and Glu173 in HiMurC) with the remainder of the coordination sphere completed by two water molecules and two oxygen atoms from the phosphate groups of the ATP. The second metal binding site in EcMurC and HiMurC is around His199 and His198, respectively, with waters making up the remainder of the coordination sphere. The corresponding residue in PaMpl is His210, but this region (residues 210-215) is disordered in our structure, probably due to the absence of metal ions. A loop (residues 404-414) from CD in PaMpl is folded into part of the ATP and UDP-MurNAc binding site (Figure 9). A significant conformational change is likely to occur in this region upon binding substrates and cofactors and involves ordering and closing of the His210 region towards Mg2 and opening of the CD. In the truncated NmMpl, which lacks CD, the region corresponding to missing residues 210-215 in PaMpl is ordered even in the absence of bound metal. Conserved residues in Mpl proteins that are in the vicinity of the ATP-binding pocket, but are not equivalent to those observed to interact with ANP in HiMurC (Figure 8), are Tyr182, Gln223 and Phe224 (which is a phenylalanine or tyrosine in MurC) ( Figure 9). Phe224 corresponds to a lysine in MurD (Lys198 in EcMurD), MurE and MurF that is post-translationally carbamoylated. Several experiments have demonstrated that this carbamoylated lysine is essential [63] as it aids in the stability of Mg2 binding. In MurC, absence of this lysine is compensated by presence of a glutamate that is also conserved in Mpl enzymes (PaMpl Glu181) (Figure 9).

CD: tripeptide binding region
The CD of Mpl is functionally unique from MurC-F proteins since it primarily attaches a tripeptide to UDP-MurNAc (n.b. it can also attach a tetra or pentapeptide, but less efficiently). MurC residues believed to be involved in binding to the incoming amino acid are His376, Arg377, Arg380, Tyr346 and His348 as observed in the HiMurC-product complex (PDB 1p3d) (Figure 8, n.b. Tyr346 is not seen in this representation). In the case of MurE, it is believed that Arg416 should be the main determinant for peptide selectivity [64]. The various Mur enzymes, as well as Mpl, have most variation in this CD domain (Figure 2), which reflects that they are all likely to have different interactions with the incoming Figure 8. PaMpl residues potentially involved in interactions with ATP and UDP-MurNAc. Schematic representation rendered using LigPlot [61] of the interactions observed between HiMurC and ligands UDP-MurNAc-L-Ala (UMA) and AMPPNP (ANP) in the crystal structure (PDB id 1p3d). The majority of these residues in HiMurC (red labels) ND and MD are structurally conserved in PaMpl (blue labels), suggesting that Mpl and MurC have similar interactions with ATP and UDP-MurNAc. doi:10.1371/journal.pone.0017624.g008 peptide substrate, but this aspect of Mur function remains the least characterized. Conserved, solvent-exposed, Mpl-specific residues in the CD are likely to be instrumental in defining substrate specificity and include Arg357, Arg358, Phe374, Ala375, His376, His377, Glu402, Pro403, Arg404, Ser405, Asn406, Thr407, Ser483, Asn484 and Gly485, which may be important in binding tri, tetra, and pentapeptide substrates (Figure 9). Based on their location in the PaMpl structure, it seems likely that, if these residues were to interact with peptide, the CD should undergo a conformational change with respect to the ND and MD. Figure 9. Additional Mpl-specific residues in ND, MD and CD that may be involved in substrate interactions. Model of PaMpl bound to ANP, UMA and metal based on the superimposition of the ND and MD of PaMpl onto the corresponding domains from HiMurC (ND and MD are the most conserved domains between the proteins) bound to ANP and UMA (PDB id 1p3d). After superimposition, the ligand coordinates were transferred from HiMurC to generate the model of PaMpl bound to ligands. Numerous residues in the binding pocket are unique and conserved in Mpl proteins and could be involved in additional interactions with substrates. These include ND Tyr35, Asn69, Arg73 and Gly74 (not shown). Ser90, Gly91 and Pro92 pack against Mpl-specific loop consisting of Ser184 (Thr or Cys in different Mpl proteins), Ala185, Phe186, Phe187, Asp188 and Lys189. Tyr182, which is directly upstream of this loop, could also be involved in the recognition. Mpl-specific residues in CD are likely to play a crucial role in substrate specificity and include Arg357, Arg358, Phe374, Ala375, His376, His377, Glu402, Pro403, Arg404, Ser405 (in dual conformation), Asn406, Thr407, Ser483 (in dual conformation), Asn484 and Gly485. doi:10.1371/journal.pone.0017624.g009 Kinetic parameters and substrate specificity of PaMpl The expression and purification of PaMpl for biochemical characterization was carried out as outlined in Materials and Methods. The kinetic parameters of the PaMpl enzyme towards its three substrates were determined (Table 2). These values (k cat = 260620 min 21 ; K m = 1.0060.13 mM, 0.1160.05 mM and 0.3660.08 mM for UDP-MurNAc, ATP and L-Ala-c-D-Glumeso-A 2 pm, respectively) were generally similar to those previously determined with the E. coli ortholog (k cat = 290 min 21 ; K m = 0.2 mM and 0.1 mM for ATP and L-Ala-c-D-Glu-meso-A 2 pm, respectively). The K m value obtained for UDP-MurNAc, however, was significantly higher (K m EcMpl = 0.25 mM).
The substrate specificity of PaMpl was also investigated. As shown in Table 3, tetrapeptide L-Ala-c-D-Glu-meso-A 2 pm-D-Ala and pentapeptide L-Ala-c-D-Glu-meso-A 2 pm-D-Ala-D-Ala were also accepted as substrates, but less efficiently. As the Michaelis-Menten plots were linear up to 0.25 mM, it was only possible to determine enzyme velocities at this concentration (k values for these substrates were 5.960.6 min 21 and 5.360.6 min 21 , respectively). PaMpl activity was comparatively much lower with smaller substrates L-Ala and L-Ala-D-Glu (k = 0.760.1 min 21 and 1.460.2 min 21 , respectively) and substitution of meso-A 2 pm by L-Lys in the tripeptide dramatically decreased enzyme activity (k = 0.2560.05 min 21 and 260620 min 21 for L-Lys-and meso-A 2 pm-containing tripeptides, respectively). A clear preference for A 2 pm at the third position of the peptide had already been observed with EcMpl, although to a lesser extent (k cat values of the E. coli enzyme for these two tripeptides were 16 min 21 and 290 min 21 , respectively) [13].

Optimal conditions for enzyme activity
Assessment of activity as a function of pH showed an optimal pH value of 8.4 ( Figure 10). Like EcMpl and all other Mur ligases [13,65], PaMpl requires a divalent cation for activity. The optimal concentration determined for Mg 2+ was 5 mM ( Figure 10). As P. arcticus is a psychrophilic Siberian permafrost bacterium, a comparative study of PaMpl and EcMpl activities as a function of temperature was performed (Figure 10). At a low temperature ,15uC, the activity of PaMpl was almost twice that of EcMpl. The optimal temperatures for these two enzymes were 30uC and 37uC, respectively. The PaMpl activity was almost completely abolished at 42uC, although EcMpl still retained 95% of its optimal activity at this temperature.

Discussion
The high-resolution crystal structure of the full-length PaMpl provides insights into the structure and relative orientations of the three domains in this cell wall recycling enzyme. Analysis of residues likely involved in binding UDP-MurNAc, ATP, metals and tri, tetra and pentapeptide substrates combined with enzyme kinetics and activity characterization, provide a basis for further experimentation to explore the structure and function of Mpl proteins. As a recycling enzyme, the role of Mpl is to attach the tripeptide L-Ala-c-D-Glu-meso-A 2 pm to UDP-MurNAc. Structure analyses indicate that the orientation of the CD domain of PaMpl relative to the other domains is flexible and changes on substrate binding. The pre-catalytic form of Mpl must open up to allow UDP-MurNAc to bind to the ND, and the tripeptide has to approach and be recognized by the CD and positioned at the active site. PaMpl-specific residues in helix H16 may affect how the CD interacts with peptide substrate. Loop 161-171 that is near the domain boundary of the ND and MD is not found in EcMpl. As the protein domains separate to bind UDP-MurNAc, this PaMpl-specific loop may control the extent to which the ND can open and, in turn, may affect UDP-MurNAc binding. This conformational rearrangement may explain the higher K m values of PaMpl for UDP-MurNAc and tripeptide. The PaMpl-specific segments 141-152, 161-171 and 262-285, which are not present in EcMpl, might be responsible for its relative sensitivity to higher temperature compared to EcMpl.
The Mur enzymes are established drug discovery targets as disruption of peptidoglycan biosynthesis is a validated path to bacterial cell death. The genes coding for the Mur ligases are essential for bacterial survival, as demonstrated by knockout experiments [3]. The recycling enzyme Mpl, although not essential for growth, could also be considered an interesting target for drug discovery. Indeed, deletion of the mpl gene and other genes involved in peptidoglycan recycling increases antibiotic susceptibility of some pathogenic strains. For instance, disruption of these non-essential genes in Acinetobacter baylyi results in at least a 10-fold reduction in the MIC of b-lactams [66]. The identification of the Mpl active site and the structural determinants that correlate with its function may lead to the design of better Mur inhibitors through structure-based drug design. Such compounds could then shutdown both the de novo and the recycling pathways for cell wall synthesis and act as novel antimicrobial agents. Alternatively, as recently discussed [13], the   Table 3. Substrate specificity of PaMpl. broad substrate specificity displayed by Mpl enzymes could potentially be exploited as an Achilles' heel for incorporation of toxic peptides into the peptidoglycan network.