Crystal Structure of Wild-type Penicillin-binding Protein 5 from Escherichia coli

Penicillin-binding protein 5 (PBP 5) of Escherichia coli functions as a d-alanine carboxypeptidase (CPase), cleaving d-alanine from the C terminus of cell wall peptides. Like all PBPs, PBP 5 forms a covalent acyl-enzyme complex with β-lactam antibiotics; however, PBP 5 is distinguished by its high rate of deacylation of the acylenzyme complex (t½ ≈ 10 min). A Gly105 → Asp mutation in PBP 5 markedly impairs deacylation with only minor effects on acylation, and abolishes CPase activity. We have determined the three-dimensional structure of a soluble form of wild-type PBP 5 at 1.85-Å resolution and have also refined the structure of the G105D mutant form of PBP 5 to 1.9-Å resolution. Comparison of the two structures reveals that the major effect of the mutation is to disorder a loop comprising residues 74–90 that sits atop the SXN motif of the active site. Deletion of the 74–90 loop in wild-type PBP 5 markedly diminished the deacylation rate of penicillin G with a minimal impact on acylation, and abolished CPase activity. These effects were very similar to those observed in the G105D mutant, reinforcing the idea that this mutation causes disordering of the 74–90 loop. Mutation of two consecutive serines within this loop, which hydrogen bond to Ser110 and Asn112 in the SXN motif, had marked effects on CPase activity, but not β-lactam antibiotic binding or hydrolysis. These data suggest a direct role for the SXN motif in deacylation of the acyl-enzyme complex and imply that the functioning of this motif is modulated by the 74–90 loop.

Bacterial cell wall peptidoglycan surrounds the cell and is essential for proper maintenance of cellular morphology and cell viability. In Escherichia coli, cell walls are composed of a repeating disaccharide, N-acetylglucosamine-␤-1,4-N-acetylmuramic acid, in which the muramic acid residues are substituted with the pentapeptide, L-Ala-D-␥-Glu-m-DAP-D-Ala-D-Ala (where m-DAP ϭ meso-diaminopimelic acid). During peptidoglycan synthesis, disaccharide-pentapeptide units are poly-merized onto nascent glycan chains (transglycosylation) and the peptide chains from different glycan strands are crosslinked (transpeptidation) by transpeptidases known as penicillin-binding proteins or PBPs. 1 Cross-linking of the peptide chains confers rigidity to the peptidoglycan and viability to the bacterial cell. The related carboxypeptidases, which hydrolyze the C-terminal D-Ala moiety from the peptide chain, may modulate the degree of cross-linking.
In E. coli at least 10 PBPs have been identified. These enzymes fall into two categories: the high molecular mass PBPs, which are essential for cell viability and catalyze transpeptidase and sometimes transglycosylase activity, and the low molecular mass PBPs, which are non-essential and catalyze D,D-carboxypeptidase (CPase) and sometimes D,D-endopeptidase activity (1). Regardless of the type of PBP, all of these enzymes react with peptide substrates and ␤-lactam antibiotics by a similar mechanism. The initial step in the reaction of PBPs with their peptide substrates is a nucleophilic attack of the D-Ala-D-Ala peptide bond by a conserved serine residue, leading to acylation of the serine hydroxyl side chain and the concomitant release of the C-terminal D-Ala. In the subsequent deacylation step, the acyl-enzyme complex can react with either an amino group (from m-DAP) of another peptide to form a cross-link (transpeptidation) or it can react with water to release the peptide (carboxypeptidation). Penicillin and other ␤-lactam antibiotics inactivate these enzymes by mimicking the structure of the D-Ala-D-Ala C terminus of the peptide chain (2,3) and reacting with the same serine nucleophile to form an analogous acyl-enzyme complex (4). Unlike the complex formed with peptide substrates, however, the ␤-lactam-PBP complex is long-lived and renders the enzyme inactive.
PBPs and other penicillin-interacting enzymes (e.g. class A ␤-lactamases) are characterized by a set of conserved motifs that are clustered in their respective active sites (5). These motifs include the Ser-X-X-Lys (SXXK) tetrad that contains the serine nucleophile, the Ser-X-Asn (SXN) triad, and the Lys-Thr(Ser)-Gly (KTG) triad. In all serine-based PBPs and ␤-lactamases of known structure, these motifs adopt a strikingly similar conformation to the extent that the active site of one PBP or ␤-lactamase can look very much like another. In addition to these three motifs, class A ␤-lactamases have a fourth motif, Glu-X-X-X-Asn, present on the so-called ⍀ loop, that is responsible for the extremely high rates of deacylation of the antibiotic-␤-lactamase complex (6 -8). Despite a wealth of structural and biochemical data, a consensus on the definitive mechanism of catalysis in ␤-lactamases has yet to emerge. Nevertheless, the picture of catalysis by ␤-lactamases is much clearer than in the related PBPs.
To further our understanding of both acylation and deacylation in PBPs and to examine their mechanistic similarities with ␤-lactamases, we have focused on PBP 5 from E. coli. PBP 5 is one of the best characterized PBPs and provides an excellent system in which to study the catalytic mechanism of penicillininteracting enzymes (9 -11). PBP 5 is the most abundant PBP in E. coli and catalyzes the major D-alanine carboxypeptidase activity in vivo (12). Although it is not essential for cell viability, recent studies have shown that PBP 5 is required for proper cell morphology in E. coli missing certain PBPs (13) and that it may have a role in cell division (14). PBP 5 is unusual among PBPs in that it catalyzes a respectable ␤-lactamase activity (t1 ⁄2 Ϸ 10 min at pH 7.0) (15,16). A deacylation-defective mutant of PBP 5 containing a Gly 105 to Asp mutation (termed PBP 5Ј) was identified from a screen of bacterial strains lacking carboxypeptidase IA activity (17,18). Kinetic studies demonstrated that PBP 5Ј had near normal acylation rates with penicillin G, but displayed a 30-fold decrease in the rate of deacylation of the penicilloyl-PBP 5Ј complex (16,19). Recently, we reported the crystal structure of PBP 5Ј (20), which revealed a conventional arrangement of active site motifs but, in the absence of the wild-type structure, provided few clues regarding the underlying reason for the defect in deacylation. In this paper we present the structure of wild-type PBP 5 solved to 1.85-Å resolution. This structure is compared with that of PBP 5Ј now refined to the higher resolution of 1.9 Å. The principal difference in the wild-type structure is the ordering of a loop near the active site comprising residues 74 -90, leading to the hypothesis that the disordering of this loop in the mutant enzyme is responsible for its markedly decreased deacylation rate. Kinetic data from mutants in which this loop was removed are consistent with this hypothesis. We also identify two conserved serine residues on this loop that are important for CPase activity.

MATERIALS AND METHODS
Crystallization-A soluble form of wild-type PBP 5 missing its cleavable signal sequence was expressed in the cytoplasm of E. coli BL21* cells (Invitrogen, Carlsbad, CA) and purified by ampicillin affinity chromatography as described previously (16). For crystallization experiments the protein was concentrated to 5.5 mg/ml in 20 mM Tris, pH 7.5, containing 30 mM NaCl and 3 mM ␤-mercaptoethanol. Wild-type PBP 5 did not crystallize under the same conditions as those used for PBP 5Ј (20) and so a new search for conditions was undertaken. Initial trials used the vapor diffusion hanging drop technique in which 3 l of protein was mixed with 3 l of well solution. A wide variety of conditions were tested, including Crystal Screens I and II (Hampton Research, Laguna Niguel, CA). PBP 5Ј was purified and crystallized according to previously established conditions (16,20).
Data Collection-Crystals of both wild-type and G105D mutant PBP 5 were cryo-protected by passage through their respective stabilization solutions (wild-type PBP 5: 8% PEG 400, 50 mM Tris-HCl, pH 8.0; PBP 5Ј: 20% PEG 4000, 100 mM Tris, pH 7.0) containing increasing concentrations of glycerol, in 2% increments, up to a maximum of 25%. At each step the crystals were soaked for 5-10 min except at the final step where the crystals were soaked overnight. Data were collected on a RAXIS-IVϩϩ imaging plate system mounted on an RU-H3R x-ray generator operating at 50 kV and 100 mA and fitted with Osmic Confocal Optics (Rigaku/MSC, The Woodlands, TX). For the wild-type crystals, the crystal-to-detector distance was set at 140 mm and data were collected in 0.5 o oscillation frames with an exposure time of 5 min per frame. A total of 215 o of data were collected. For PBP 5Ј crystals, the crystal-to-detector distance was 150 mm and data were collected in 1 o -oscillation frames with an exposure time of 5 min per frame. A total of 200 o of data were obtained. Both data sets were processed with CrystalClear (21).
Structure Determination and Refinement-The starting model for both refinements was the previous 2.3-Å structure of PBP 5Ј that had been refined against data obtained at room temperature (20). The structure of wild-type PBP 5 was solved by molecular replacement using AMoRe (22) with a search model of PBP 5Ј in which Asp 105 had been replaced by glycine. Initially, both models were refined with XPLOR (23), and included several rounds of rigid body refinement, but later rounds used REFMAC5 (24). Manual refitting of the structures was performed as necessary with O (25). Water molecules were included as they became visible in both the ( F o Ϫ F c ) and 2( F o Ϫ F c ) electron density maps and if they had at least two potential hydrogen bonding partners. The quality of the models was assessed by PROCHECK (26) and superimpositions were performed using LSQKAB (27). The coordinates and structure factors for the wild-type PBP 5 (1NZO) and PBP 5Ј (1NJ4) structures have been deposited with the Protein Data Bank.
Construction and Purification of PBP 5 Variants-Constructs of PBP 5 were made in which residues 72-92 were deleted, termed "loop out" (LO). These constructs were made in both wild-type PBP 5 (PBP 5-LO) and in the G105D mutant (PBP 5Ј-LO). PBP 5-LO was constructed by 4-primer PCR as described previously (28). The outside primers were complementary to sequences surrounding the multiple cloning site in the pT7-7K vector (29). The LO-down internal primer was complementary to codons 64 -71 and 93-98, inclusive (5Ј-CGGAACCTGCATGC-CCGGGATAGTGACTAAATCAGTTTCTT-3Ј), whereas the LO-up internal primer (5Ј-CCGGGCATGCAGGTTCCG-3Ј) was complementary to codons 93-98. In the first round of PCR, pT7-PBP 5 was amplified with pT7-up and LO-down and with pT7-down and LO-up. Following the first round of PCR, the two fragments were isolated and added to a new tube with the outside primers only, and PCR was carried out as before. The overlapping regions allowed for the annealing of the two fragments and the full-length gene missing codons 72-92 of the mature PBP 5 sequence was generated. The final PCR product was digested with XbaI and HindIII and ligated into similarly digested pT7-7K to produce pT7-PBP 5-LO. To construct pT7-PBP 5Ј-LO, the XhoI-SphI fragment encompassing the 5Ј end of the gene up to codon 94 of pT7 PBP 5-LO was ligated with the SphI-BglII fragment from pT7-PBP 5Ј containing the G105D mutation and the BglII-XhoI vector fragment from pT7-7K. Individual mutations were constructed in a similar manner and cloned into the same vector as for the loop-out constructs. All constructs were verified by sequencing. Following transformation of the vectors into BL21* cells, the mutant proteins were purified by ampicillin affinity chromatography as described for PBP 5 and PBP 5Ј (16).
Kinetic Analysis of the Interaction of PBP 5 Variants with [ 14 C]Penicillin G and Peptide Substrate-The reaction mechanism for the interaction of PBPs with peptide and ␤-lactam antibiotics is, where, KЈ ϭ (k Ϫ1 ϩ k 2 )/k 1 , E⅐S is the non-covalent Michaelis complex, E-S is the covalent acyl-enzyme complex, and P is the released product. The constant k 2 /KЈ, which describes the formation of the acyl-enzyme complex (E-S) at low (subsaturating) concentrations of ␤-lactam antibiotics, was determined from time courses of the formation of E-S with [ 14 C]penicillin G essentially as described (30). PBP 5 (48 g; 1.2 nmol) was diluted into 150 l of binding buffer (50 mM sodium phosphate, pH 7.0, 10% glycerol) and mixed with an equal volume of 100 M [ 14 C]penicillin G in binding buffer. At timed intervals, 20-l aliquots were removed, mixed with 5 ml of 5% trichloroacetic acid (w/v), and incubated on ice for 15 min. The acidified proteins were passed through number 30 glass fiber filters (Schleicher and Schuell, Keene, NH) and the filters were washed twice with 5 ml each of 1% trichloroacetic acid, 33% methanol. The filters were then air dried, placed in scintillation vials with 3 ml of Scinti-safe scintillation fluid (Fisher Scientific, Pittsburg, PA), and counted. k 3 values were determined from semi-log plots of the % [ 14 C]penicillin G remaining versus time. PBP 5 proteins (11-37 g; 0.28 -0.92 nmol) were diluted into 150 l of binding buffer and mixed with an equal volume of 100 M [ 14 C]penicillin G for 10 min at 30°C. At t ϭ 0, penicillin G was added to 3 mM, and the amount of radioactivity remaining covalently bound to the protein was determined by removing 20-l aliquots at various times and quantitating the amount of bound [ 14 C]penicillin G as described above. The t1 ⁄2 values were determined from the plots and converted to k 3 values by dividing the t1 ⁄2 in seconds into 0.693.
CPase assays were carried out with the peptide N,NЈ-diacetyl-L-Lys-D-Ala-D-Ala and the amount of hydrolysis of the ultimate D-alanine residue was quantitated with the spectrophotometric microtiter plate protocol described by Gutheil et al. (10). Assays contained from 0 to 10 mM peptide and 0.83 M enzyme, with the exception that PBP 5Ј, PBP 5-LO, and PBP 5Ј-LO were also tested at 8.3 M.

RESULTS
Structure Determination-Crystals of wild-type PBP 5 were obtained over wells containing 8% PEG 400, 100 mM Tris, pH 8.0. This condition is considerably different from that for PBP 5Ј (20% PEG 4000, 50 mM Tris, pH 7.0 (16)) and indicates some difference in the protein structure. Diffraction studies revealed that the crystals belonged to space group C2 with cell dimensions a ϭ 109.4 Å, b ϭ 50.3 Å, c ϭ 84.5 Å, and ␤ ϭ 120.9 o . Electron density maps showed that the region between residues 74 and 90, which is absent in the structure of PBP 5Ј because of apparent flexibility, is ordered in the wild-type structure. The R factor of the final model is 20.9% (R free ϭ 24.5%).
For the purposes of structure comparison a new data set under cryo conditions was obtained from crystals of PBP 5Ј. This resulted in a slight improvement in resolution (from 2.3 to 1.9 Å) compared with previous data collected at room temperature (20), but a significant improvement in completeness and redundancy. The cell dimensions under cryo conditions were a ϭ b ϭ 50.2 Å and c ϭ 136.8 Å, which differs principally from the room temperature cell through a 3.5-Å reduction in the c dimension. The structure was refined against the 1.9 Å data to a final R factor of 20.9% (R free ϭ 24.4%). Compared with the room temperature structure, residues 85-90 are now visible in the electron density map but residues 74 -84 remain disordered. In both structures difficulty was encountered in fitting the region 154 -157 because this appears to exhibit some disordering, thus weakening the electron density. The data collection and final refinement statistics of both models are shown in Table I.
Structure Description-The overall structure of wild-type PBP 5 exhibits the same fold as that previously described for PBP 5Ј (20). Briefly, PBP 5 is composed of two domains that are oriented approximately at right angles to each other (Fig. 1A). Domain 1, which is a transpeptidase-like domain typical of penicillin-interacting enzymes, comprises a five-stranded antiparallel ␤ sheet packed on both sides by ␣ helices. Domain 2 is unique to PBP 5 and is formed by a sandwich of two antiparallel ␤ sheets, one three-stranded and the other twostranded. The active site of PBP 5 is located at the interface of the five-stranded anti-parallel ␤ sheet and the large ␣ helical cluster within domain 1.
Structure Comparison-The common main chain atoms of domain 1 of wild-type PBP 5 and PBP 5Ј superimpose very closely with an r.m.s. deviation of 0.88 Å, indicating that the mutation has little effect on the overall structure. After superimposition a slight shift in the relative position of domain 2 is observed (Fig. 1A). This difference arises from a hinging motion centered at the domain boundary (residue 262) resulting in a relative 6-Å shift of residues at the base of domain 2. The shifted position of domain 2 is likely the result of crystal packing interactions, which are different in the two structures. In the wild-type structure the tip of domain 2 packs against the 74 -90 loop, whereas in PBP 5Ј this region is largely exposed.
Superimposition reveals two regions within domain 1 in which significant structural differences are observed between the wild-type and mutant enzymes: 1) residues 74 -90, and 2) residues 190 -193 of the ␤7-␤8 loop (Fig. 1B). Slight differences are also seen in the so-called ⍀-like loop, but because the electron density for this region is weak in both structures, this is most likely because of the difficulties in fitting the model.
74 -90 Region-In the structure of PBP 5Ј reported previously, which was solved using data collected at room tempera-ture, the region between residues 74 and 90 (inclusive) was not visible in the electron density map (20). In the new structure of PBP 5Ј refined against data collected under cryo conditions, residues 85-90 are now visible and adopt an irregular loop structure, whereas residues 74 to 84 remain disordered (Fig.  1B). The crystallographic B factors for residues 85-90, however, are high (in the 50s) and some of the side chains are poorly resolved, indicating a high degree of flexibility. The increased ordering of residues 85-90 in PBP 5Ј is likely because of the improvement in the quality of data obtained under cryogenic conditions. In contrast, the entire region encompassing 74 -90 in the wild-type structure is highly ordered and has refined with low B factors (in the 30s). The secondary structure of this region, which is stabilized by numerous non-bonding interactions to the surrounding protein, is mostly an irregular loop but starts with a helical turn and finishes in an extended conformation. In the middle is a striking "zig zag" motif in which there are tight turns made by Pro 81 and Gly 85 . Interestingly, residues 85-90 adopt a different fold in each structure with the divergence beginning at residue 90 (reverse direction) (Fig. 1B). In the wild-type structure this loop remains close to the protein domain, whereas in PBP 5Ј these residues point out toward solvent.
190 -193 Loop-The only other region that is altered significantly between the two structures is the connecting loop between ␤7 and ␤8, comprising residues 190 -193 (Fig. 1B). In the wild-type structure, this loop is displaced slightly away from the active site because of the presence of Met 89 , which is central to the hydrophobic core in this region (see below). In the PBP 5Ј structure, the absence of this residue has enabled Phe 190 to move in and fill the void in the hydrophobic core. Effect of the G105D Mutation-Given that glycine residues tend to act as helix breakers, it might have been expected that the ␣4 helix containing the mutation would terminate earlier in wild-type PBP 5 than in the mutant PBP 5Ј structure, thereby affecting the conformation of the SXN active site motif that directly follows the ␣4 helix. Surprisingly, little or no structural difference in this helix or its residue side chains is seen between the two structures (Fig. 2). In PBP 5Ј only the side chain of Gln 109 has moved slightly to avoid a clash with Asp 105 , which lies directly below this residue on the helix. In the wild-type structure, the side chain of Gln 109 is within hydrogen bonding distance of the carbonyl group of Leu 88 , a contact that is absent in PBP 5Ј. Instead, the primary effect of the G105D mutation in PBP 5Ј is to disrupt a hydrophobic core region. In the wild-type structure, this core is formed by the hydrophobic faces of ␣4, ␣5, ␤7, ␤8, and the 74 -90 loop (Fig. 3). At the heart of the hydrophobic core is Met 89 , with additional contributions of Phe 188 , Phe 190 , Leu 91 , Leu 102 , and Ala 114 . In the PBP 5Ј structure, a glycine to aspartic acid mutation at position 105 disrupts this region by projecting directly into the space occupied by Met 89 . As a result, Met 89 is displaced along with Phe 90 and residues 85-88. The space vacated by Met 89 is partially occupied by Phe 190 , hence the movement of residues 190 -193.
A buried aspartate (Asp 113 ) is also present within the hydrophobic core and each of its ␥ oxygen is within hydrogen bonding distance of three amide nitrogens (Fig. 4). O ␦1 of Asp 113 contacts the amides of residues 89 -91, whereas O ␦2 contacts the amides of 89, 113, and 114. The rarity of buried aspartates together with its role in stabilizing the structure of the 74 -90 region suggests that the conformation of this loop is critical for  (20). The starting and ending points of the disordered region in PBP 5Ј are denoted by the small spheres at the C␣ atoms of residues 73 and 91, and are colored orange in PBP 5Ј, and green in wild-type. To emphasize the disparity between the two structures, the position of Gly 85 is also shown in the wild-type structure (green sphere), which is located far from its counterpart in PBP 5Ј. For reference, C␣ atoms of important active site residues are shown as small red spheres. This figure was prepared using MOLSCRIPT (36) and RASTER3D (37). function. As discussed below, the change in conformation of residues 85 to 90 and the disordering of residues 74 -84 is likely responsible for the altered kinetic properties of PBP 5Ј.
Active Site Comparison-A side-by-side comparison of the two active sites is shown in Fig. 5. The key residues are as follows: Ser 44 and Lys 47 of the Ser-X-X-Lys tetrad are located on one face of the ␣2 helix, Ser 110 and Asn 112 of the Ser-X-Asn triad are present on the loop connecting helices ␣4 and ␣5, and Lys 213 , one of three residues in the Lys-Thr(Ser)-Gly triad, extends into the cavity. One other residue of note is Arg 198 , which is on the loop at the top of the cavity. It is immediately obvious that the principal change in the active site in the wild-type structure results from the ordering of the loop encompassing residues 74 -90 (shown as a red loop in Fig. 5A). This loop is oriented over the active site and interacts with the SXN motif via two serine residues, Ser 86 and Ser 87 , both of which are highly conserved in PBP 5-related CPases (20). Ser 87 makes two potential hydrogen bonds, one between its side chain ␥-oxygen and the carbonyl oxygen of Gln 109 and the other from its carbonyl group to N ␦2 of Asn 112 (Figs. 4 and 5). There is also a potential hydrogen bond between the ␥-oxygen of Ser 86 and N ␦2 of Asn 112 . Thus Asn 112 is contacted by Ser 86 , Ser 87 , and Lys 47 . In PBP 5Ј these serine-mediated interactions are absent because of the disordering of the 74 -90 loop.
The relative positions of the residues comprising the conserved active site motifs of penicillin-interacting enzymes are essentially identical. The network of hydrogen bonding interactions observed in the PBP 5Ј structure (20), those between Ser 110 and Lys 213 , and from Lys 47 to Asn 112 and the carbonyl of His 151 , are all retained. One notable difference arises through a rotation of Ser 44 , such that in the PBP 5Ј structure this residue no longer forms a hydrogen bond with Lys 47 (Figs. 5  and 6). The electron density for this residue in both structures, however, indicates some rotational freedom of the serine side chain. Arg 198 is also oriented slightly different in the two structures because the presence of Ser 87 on the 74 -90 loop has pushed this residue upwards in the wild-type enzyme. The likely impact of these structural differences on catalytic function is discussed below.
Relatively few water molecules are observed in either active site and few are structurally conserved. The most significant water molecule is one within hydrogen bonding distance of Ser 44 in both structures. In PBP 5Ј, a water molecule (Wat 130 ) bridges the ␥ oxygen of Ser 44 and the carbonyl group of His 216 . In the wild-type enzyme, a similar water molecule is present (called Wat 89 ) but because Ser 44 is rotated slightly with respect to PBP 5Ј, the water is too distant to hydrogen bond with His 216 .
Functional Analysis of PBP 5 Mutants and Variants Lacking Residues 72-90 -As discussed above, the structure of wild-type PBP 5 reported here differs from PBP 5Ј primarily by the ordering of the region encompassing residues 74 -90. This result suggests that the functional phenotype of PBP 5Ј, i.e. a markedly impaired deacylation rate and lack of CPase activity, is due directly to the disruption of this region by the G105D mutation. To test this hypothesis and to determine the role of the 74 -90 loop in the activity of PBP 5, we made constructs in which this loop was deleted. In the wild-type structure, the ␣ carbons of residues 71 and 93 are separated by 4.7 Å. Because the residue following Pro 93 is a glycine, we predicted that FIG. 2. Superimposition of PBP 5 wild-type and PBP 5 structures in the region of the mutation. Both backbone and side chain residues show very little change in helix ␣4 containing the G105D mutation. The one exception is the side chain of Gln 109 , which moves away in the mutant structure to accommodate the presence of the Asp mutation (see text for details). The color scheme is identical to Fig. 1. This figure was prepared using MOLSCRIPT (36) and RASTER3D (37) .   FIG. 3. Stereo representation of the hydrophobic core region in wild-type PBP 5. This region is formed in part by the 74 -90 loop, which is altered in PBP 5Ј. In this superimposition the wild-type structure is shown as solid bonds, whereas the PBP 5Ј structure is shown as dashed lines. The ␣ carbon positions are shown as dots and are labeled. In areas where the structures diverge, the equivalent atoms in PBP 5Ј are also labeled. This figure was prepared using MOLSCRIPT (36).
fusing Ile 71 to Pro 93 would promote the formation of a ␤-turn with minimal disturbance to the overall structure of the protein.
PBP 5, PBP 5Ј, and both proteins missing residues 72-92 (termed PBP 5-LO and PBP 5Ј-LO, respectively) were expressed in E. coli and purified by ampicillin affinity chromatography. Like PBP 5 and PBP 5Ј, the loop-out constructs were soluble and bound stoichiometric amounts of [ 14 C]penicillin G (data not shown). Internal deletion of residues 72-92 had only a small effect on the specificity constant for penicillin G. Wildtype PBP 5 had a k 2 /KЈ of 390 M Ϫ1 s Ϫ1 with [ 14 C]penicillin G, whereas PBP 5Ј, PBP 5-LO, and PBP 5Ј-LO had k 2 /KЈ values ϳ4to 6-fold lower (Table II). Moreover, the k 2 /KЈ values for the two loop-out constructs were very similar to the k 2 /KЈ constant for PBP 5Ј, and effects of the loop-out were not additive with the G105D mutation.
In contrast to the small effect on the acylation rate constants in the loop-out constructs, there was a much greater effect on the deacylation rates of the penicilloyl-PBP 5 complex. PBP 5 and PBP 5Ј hydrolyzed bound [ 14 C]penicillin G with rate constants of 78 ϫ 10 Ϫ4 s Ϫ1 (t1 ⁄2 ϭ 14.4 min) and 3 ϫ 10 Ϫ5 s Ϫ1 (t1 ⁄2 ϭ 380 min), respectively, which were very similar to the constants reported previously (16, 31) (Table II). Consistent with our hypothesis, PBP 5-LO had a deacylation rate constant of 3.7 ϫ 10 Ϫ5 s Ϫ1 (t1 ⁄2 ϭ 315 min), very similar to the k 3 rate constant of PBP 5Ј. Thus, deletion of residues 72-92 in PBP 5 had essentially the same affect on deacylation as mutation of Gly 105 to Asp. When residues 72-92 were deleted in PBP 5Ј, the k 3 constant decreased another 3-fold to 1.1 ϫ 10 Ϫ5 s Ϫ1 (t1 ⁄2 ϭ 1050 min).
Another characteristic of PBP 5Ј is its lack of CPase activity. If the major effect of the G105D mutation is the disruption of the region encompassing residues 74 -90, we would expect that the loop-out constructs would also lack CPase activity. Indeed, both PBP 5-LO and PBP 5Ј-LO lack any detectable CPase activity with N,NЈ-diacetyl-L-Lys-D-Ala-D-Ala, even at peptide concentrations as high as 10 mM and enzyme concentrations as high as 8 M. The loss of CPase activity in both PBP 5Ј and PBP 5-LO could result from one of two scenarios: 1) the peptide forms an acylenzyme complex but is unable to undergo deacylation (i.e. no turnover), or 2) the peptide is incapable of forming an acylenzyme complex, either because of a lack of substrate binding or acylation. To distinguish between these two possibilities, we incubated PBP 5 and PBP 5-LO with or without 20 mM peptide substrate at pH 8.5 for 20 min prior to assessing their acylation rate with [ 14 C]penicillin G. There was no difference in the k 2 /KЈ constants between buffer and peptide preincubation conditions (data not shown), precluding the possibility that the peptide

FIG. 5. A side-by-side comparison of active site residues in (A) wild-type and (B) G105D mutant PBP 5.
The active site is formed by five structural elements. Three of these contain the well known sequence motifs typical of penicillin-interacting enzymes: the SXXK motif (yellow), the SXN motif (green), and the KTG motif (orange). Other motifs are specific to PBP 5 and close homologues: the ⍀-like loop (blue), a loop containing Arg 198 (purple) and the 74 -90 region (red). Note the absence of the latter in PBP 5Ј.
substrate forms a stable acyl-enzyme complex.
We also investigated the role of the two serine residues (Ser 86 and Ser 87 ) in the 74 -90 loop, which are conserved in PBP 5-related CPases (20). As discussed above, these residues sit atop the SXN active site motif and contact both the main chain and side chain amides of Asn 112 and the main chain carbonyl of Gln 109 (Fig. 4). Ser 86 and Ser 87 were mutated to alanine, and the enzymatic and kinetic activities of the resulting proteins were determined. Mutation of these serine residues either individually or together had little effect on the interaction of the protein with [ 14 C]penicillin G (Table II). The acylation rate constants of the mutants were nearly identical to wild-type PBP 5, suggesting these residues serve no role in binding or acylation of ␤-lactam antibiotics. The mutants showed only a slight decrease in k 3 with [ 14 C]penicillin G, between 1.4-and 2.6-fold lower than wild-type (Table II). In contrast, all the serine mutants showed a much greater decrease (between 8-and 16-fold) in the k cat /K m constants for CPase activity measured with peptide substrate. These data suggest that these two serine residues are important for CPase activity of PBP 5.

DISCUSSION
Our structural studies of PBP 5 began with PBP 5-G105D (called PBP 5Ј), a deacylation-defective mutant of PBP 5 (20). This structure revealed that PBP 5 has two domains: a penicillin-binding domain showing high structural similarity to other PBPs and class A ␤-lactamases, and a second domain of unknown function. In the absence of structural information for the wild-type enzyme, however, the mutant structure provided little insight into how the G105D mutation leads to a catalytic deficiency in deacylation. In this study, we present the structure of wild-type PBP 5 determined to 1.85-Å resolution and compare it with the previous structure of PBP 5Ј now refined to 1.9-Å resolution. The striking difference between the two structures is the order/disorder in residues 74 -90, suggesting that this region is critical for catalytic function. This hypothesis is supported by the apparent precision in its structure and by its close proximity to the active site. Most striking is the presence of a buried aspartate (Asp 113 ) that orients the loop via interactions with the amide nitrogens of residues 89 -91, an interaction that is lost in PBP 5Ј (see Fig. 4). Buried aspartates are rare in proteins and their presence usually signifies a region important for function. Another unusual feature is the zig zag structure that immediately precedes Ser 86 and Ser 87 .
Disordering of the loop encompassing residues 74 -90 in PBP 5Ј could arise in two ways: 1) through differences in crystal packing between the two crystal forms or 2) as a direct effect of the G105D mutation. In essence, this is a chicken and egg problem. Did the crystal packing in the P3 2 form afford the loop greater flexibility, leading to its disorder, or did the mutation disrupt the loop and thereby prevent crystallization of PBP 5Ј in the C2 space group? Several lines of evidence suggest the latter explanation is most likely. First, the wild-type protein does not crystallize under the same conditions obtained for PBP 5Ј, which suggests that the protein structure itself determines the crystal form and not vice versa. Second, examination of the wild-type structure shows an obvious mechanism for how the mutation site disrupts the 74 -90 loop. In the mutant enzyme Asp 105 displaces Met 89 from a hydrophobic core and in doing so alters the loop structure and hence the crystallization behavior of the protein. Finally, removal of the loop by deleting the codons encoding residues 72-92 results in a protein with a nearly identical functional phenotype as PBP 5-G105D, suggesting that the disordering of the loop leads directly to the decrease in deacylation of [ 14 C]penicillin G and abrogation of CPase activity.
Comparison of the PBP 5 and PBP 5-G105D structures does not reveal an obvious mechanism for the loss of CPase activity in PBP 5Ј, especially because the 74 -90 loop appears too far removed from the active site to participate directly in catalysis. One explanation is that PBP 5Ј retains the ability to bind and react with peptide and that the loss of CPase activity is because of the marked decrease in the rate of deacylation, resulting in the accumulation of covalently bound peptide. However, this is unlikely because preincubation of PBP 5-LO with peptide had no effect on the rate of acylation with [ 14 C]penicillin G. A more likely explanation is that this region is important for peptide binding (and/or acylation) but to test this will require a complex of PBP 5 with peptide substrate.
One contributing factor to the loss of peptide binding in PBP 5Ј and PBP 5-LO may be the movement of Arg 198 , which sits at the top of the active site (Fig. 5). In PBP 5Ј, Arg 198 forms a hydrogen bond with the carbonyl of Gln 109 , whereas in the wild-type enzyme it is hydrogen-bonded to Ser 87 . The recent structure of the Streptomyces R61 D,D-peptidase in complex with its peptide substrate revealed a strong interaction between Arg 285 , the equivalent of Arg 198 in the D,D-peptidase, and the C-terminal carboxylate of the peptide substrate (32). Thus, the altered position of Arg 198 in PBP 5Ј (and presumably PBP 5-LO) may hamper its ability to interact with the C terminus of the peptide substrate in PBP 5, leading to a marked decrease in substrate binding and CPase activity.  Our results shed new light on the mechanism of deacylation in PBP 5. The wild-type and G105D mutant PBP 5 structures, together with kinetic analysis of the loop-out constructs, suggest that disordering of the 74 -90 loop is the primary reason for the defect in the rate of hydrolysis of the acyl-enzyme complex in PBP 5Ј. Furthermore, the precise interaction of this region with the active site via contacts from serines 86 and 87 to the SXN motif suggests that this motif has a major role in deacylation of the acyl-enzyme complex. How it does so is unclear. Its absence or presence makes almost no difference to the conformation of the SXN motif (Fig. 6), suggesting a subtle mechanism for the defect in deacylation in PBP 5Ј. Moreover, mutation of these residues to alanine had no effect on acylation with ␤-lactam antibiotics and only minor effects on deacylation. In contrast, we observed a ϳ10-fold decrease in k cat /K m for hydrolysis of peptide substrate, clearly implicating these residues in CPase activity. Interestingly, the functional phenotype resulting from mutation of these two serine residues in PBP 5 was much less severe than that obtained by deleting the 74 -90 loop, which resulted in a 30-fold decrease in deacylation activity and the total absence of CPase activity. At first glance this appears to suggest a role for this loop beyond that of orienting these two serine residues. However, because several of the hydrogen bonds from these two serines to residues in the SXN motif emanate from the peptide backbone and would be retained in the mutants, the severity of the functional phenotype in PBP 5Ј and PBP 5-LO may simply be because of the absence of these backbone-mediated interactions.
How then does the SXN motif function in deacylation? The two obvious candidates for a functional role are Asn 112 and Ser 110 : Asn 112 is contacted directly by both Ser 86 and Ser 87 , whereas Ser 110 is contacted indirectly via interactions from Ser 87 to Gln 109 and Gly 111 (see Fig. 4). In class C ␤-lactamases and the D,D-peptidase from Streptomyces R61, the equivalent residue of Ser 110 is a tyrosine and the phenoxide anion of this residue has been proposed to act as a general base in deacylation (33)(34)(35). For Ser 110 in PBP 5 to act in an analogous manner, however, seems unlikely because the pK a of the serine hydroxyl is too high for this residue to form an anion. The same argument applies to Asn 112 and so it is doubtful that either of these residues act directly in deacylation. Instead we have assigned Lys 47 as the general base to polarize the hydrolytic water molecule, which is located next to Ser 44 (11,20). A more likely role for either Ser 110 or Asn 112 would be to help orient the hydrolytic water molecule near Lys 47 . Asn 112 is suitably positioned immediately adjacent to Lys 47 (Fig. 6) and for Ser 110 , its side chain could rotate away from Lys 213 and toward Lys 47 (see Fig. 5A). Regardless of which residue is involved, it is clear that the absence of the 74 -90 loop in PBP 5Ј hinders deacylation. To investigate the role of the 74 -90 loop in deacylation further will require structures of both PBP 5 and PBP 5Ј covalently bound to penicillin G in which the hydrolytic water is apparent. These studies are underway.