Crystal Structure of a Peptidoglycan Synthesis Regulatory Factor (PBP3) from Streptococcus pneumoniae*

Penicillin-binding proteins (PBPs) are membrane-associated enzymes which perform critical functions in the bacterial cell division process. The single d-Ala,d-Ala (d,d)-carboxypeptidase in Streptococcus pneumoniae, PBP3, has been shown to play a key role in control of availability of the peptidoglycal substrate during cell growth. Here, we have biochemically characterized and solved the crystal structure of a soluble form of PBP3 to 2.8 Å resolution. PBP3 folds into an NH2-terminal, d,d-carboxypeptidase-like domain, and a COOH-terminal, elongated β-rich region. The carboxypeptidase domain harbors the classic signature of the penicilloyl serine transferase superfamily, in that it contains a central, five-stranded antiparallel β-sheet surrounded by α-helices. As in other carboxypeptidases, which are present in species whose peptidoglycan stem peptide has a lysine residue at the third position, PBP3 has a 14-residue insertion at the level of its omega loop, a feature that distinguishes it from carboxypeptidases from bacteria whose peptidoglycan harbors a diaminopimelate moiety at this position. PBP3 performs substrate acylation in a highly efficient manner (kcat/Km = 50,500 m–1·s–1), an event that may be linked to role in control of pneumococcal peptidoglycan reticulation. A model that places PBP3 poised vertically on the bacterial membrane suggests that its COOH-terminal region could act as a pedestal, placing the active site in proximity to the peptidoglycan and allowing the protein to “skid” on the surface of the membrane, trimming pentapeptides during the cell growth and division processes.

Bacterial division is a complex phenomenon that requires the coordination of diverse processes including chromosomal segregation, FtsZ ring-dependent membrane constriction, and cell wall synthesis at the site of septation. The latter process involves the polymerization of glycan chains and transpeptidation of pentapeptidic moieties within the structure of the peptidoglycan, a highly cross-linked mesh that is crucial for main-taining bacterial shape and providing protection from osmotic shock and lysis (1). Both reactions are catalyzed by penicillinbinding proteins (PBPs), 1 membrane-associated molecules, which can be classified as high molecular mass (hmm; often bifunctional) and low molecular mass (lmm; monofunctional) and play key roles in the bacterial life cycle. The pathogenic bacterium Streptococcus pneumoniae offers a unique opportunity for the study of the relationship between cell division and cell wall synthesis, since it carries a relatively simple set of six PBPs, compared with other well studied organisms which present much higher complexity (2). In this organism, PBP1a, -1b, and -2a catalyze both glycosyltransfer and transpeptidation; PBP2b and -2x only catalyze the latter reaction, and PBP3, the single lmm PBP in S. pneumoniae, has been shown to act as a D-Ala,D-Ala (D,D) carboxypeptidase (3).
The central role of hmm PBPs in the cell growth and division processes has been recently confirmed through the study of their localization within the cell cycle through the employment of immunofluorescence techniques (4). In S. pneumoniae, the constriction of the FtsZ-ring is spatially coupled to PBP2x-and PBP1a-mediated septal peptidoglycan synthesis, with the former process preceding the latter by approximately 5 min (4). At the beginning of the cell cycle, PBP3 localizes throughout the whole bacterial surface but seems to be absent from the future division site (5). Since the D,D-carboxypeptidase activity of PBP3 removes the COOH-terminal D-alanine of the peptidoglycan pentapeptide side chains, its hemispheric localization implies that the cellular region neighboring the future division site will be the only one where full-length pentapeptides will be available as substrates for other PBPs. Interestingly, a mutant pneumococcal strain which lacks PBP3 displays abnormal morphology and exhibits multiple septa initiated at aberrant locations (6). Thus, it is likely that the availability of intact pentapeptidic substrates dictates the localization of the hmm PBPs. Therefore, by guaranteeing that pentapeptides are available uniquely at the future division site, PBP3 may ensure the spatial coordination of the FtsZ-ring with the septum synthesis machinery.
PBP3 is associated to the bacterial membrane through a COOH-terminal amphiphilic helix. In the D,D-carboxypeptidase reaction catalyzed by PBP3, an active serine residue reacts with the D-Ala-D-Ala COOH terminus of a peptide chain of the peptidoglycan to form a transient acyl-enzyme complex that is subsequently hydrolyzed. The reaction results in the formation of a tetrapeptide that can only serve as an acceptor for a subsequent transpeptidation reaction by other PBPs (3). As in the case of the D,D-transpeptidase activity, D,D-carboxypeptidation is inhibited by penicillin and other ␤-lactam antibiotics that mimic the structure of the D-Ala-D-Ala carboxyl terminus of the pentapeptide chain (7). These antibiotics react with PBPs to form a stable acyl-enzyme complex, resulting in prolonged inhibition of the enzymes.
Recently, the structure of PBP5, the soluble, signal peptidelacking form of the D,D-carboxypeptidase from Escherichia coli, was reported in wild type and mutant forms to high resolution (1.85 and 1.9 Å, respectively; Refs. 8 and 9). Wild type PBP5 deacylates its acyl-enzyme complex at a very high rate, which is reminiscent of that of a class A ␤-lactamase (the latter with a poor substrate). Although these reports shed light on the enzymology of D,D-carboxypeptidation and the two-domain fold of the enzyme, several points still remain unclear, including the nature of the D,D-carboxypeptidase substrates and the enzymatic role of lmm PBPs in the cell division process.
In light of our previous reports on the localization of PBP3 in the cell cycle (5) and in an effort to answer some of the questions above, we performed the enzymatic characterization of pneumococcal PBP3 and solved the structure of a soluble form of the enzyme at 2.8 Å resolution. Although the general folds of the pneumococcal and E. coli enzymes are similar, PBP3 harbors a significantly longer omega-like loop, a feature subsequently identified as a telltale motif in enzymes present in bacteria whose peptidoglycan structures contain an L-lysine group in the third stem peptide position. Interestingly, its carboxypeptidase domain is highly reminiscent of that of transpeptidase K15 from Streptomyces spp. but shares structural resemblance with other peptidoglycan biosynthetic enzymes only in the immediate vicinity of the active site. PBP3 is a highly efficient D,D-carboxypeptidase, hydrolyzing a synthetic peptide substrate 180 times more efficiently than E. coli PBP5; however, acyl-enzyme deacylation is 20-fold slower than for its E. coli counterpart, suggesting that PBP3 plays a particular role in control of peptidoglycan reticulation in the Gram-positive cell wall. Last, the positioning of the active site on the opposite face of the molecule from the COOH-terminal, membrane-interacting region may place it in optimal position to contact the peptidoglycan layer throughout the two cellular hemispheres.

Measurement of Kinetic and Antibiotic Recognition Parameters-
The construction of plasmid pGEX-sPBP3* encoding the soluble form of wild type PBP3, which lacks both the COOH-terminal helix and the signal peptide (sPBP3*), was described previously (5). D,D-Carboxypeptidase activity was assayed with N,N-diacetyl-L-Lys-D-Ala-D-Ala (Ac 2 -KAA). sPBP3* (10 nM) was incubated at 37°C in 50 mM Tris-HCl (pH 8.5), 50 mM NaCl, 1 mM EDTA, 0.5 mg/ml bovine serum albumin, and Ac 2 -KAA at concentrations ranging from 37.5 to 24,000 M. After various time intervals, the reaction was stopped by addition of penicillin G to 0.1 mM, and released D-Ala was measured by the method described by Johnson et al. (10).
The functional homogeneity of the protein sample was determined by titrating the active sites present in the preparation using [ 3 H]ben-zylpenicillin (20 Ci/mmol, 1 mCi/ml; Amersham Biosciences) as a reporter. sPBP3* solutions at 2 and 5 M were incubated for 15 min at 37°C in 50 mM Tris-HCl (pH 8.0), 200 mM NaCl containing 0.01-20 M [ 3 H]benzylpenicillin. The samples were subsequently submitted to SDS-12% PAGE electrophoresis, and estimation of [ 3 H]benzylpenicillin bound to proteins was monitored by two different procedures. The gel was stained with Coomassie Blue, destained, incubated with Amplify (Amersham Biosciences), dried, and either exposed to film for 16 h or cut around the protein bands. In the latter case, the gel slices were mixed with 5 ml of LSC mixture (Picofluor 15, Packard), and their radioactivity was measured using a liquid scintillation analyzer (Packard model 2100TR).
To analyze the kinetics of the deacylation reaction, 2 M purified sPBP3* was labeled with 1 M [ 3 H]benzylpenicillin at 37°C during 15 min in 50 mM Tris-HCl (pH 8.0), 200 mM NaCl. Excess of cold benzylpenicillin (15 mM) was then added, and the reaction was continued at 37°C. Aliquots were regularly removed, submitted to SDS-PAGE electrophoresis, and the amount of radioactivity was measured in the protein bands as mentioned above.
The ability of sPBP3* to hydrolyze the pseudo substrate N-benzoyl-D-alanylmercaptoacetic acid (S2d), which is a thioester analog of the stem wall peptide, was explored to generate a comparison profile of hydrolysis rates for other, previously characterized pneumococcal PBPs. Hydrolysis of S2d was followed by monitoring the amount of thiol group released using the method described by Zhao et al. (11).
Crystallization and Structure Solution-Selenomethionine (SeMet)substituted sPBP3* was expressed in E. coli B834. Cells were grown in LeMaster medium (12) containing 40 mg⅐l Ϫ1 methionine that was progressively replaced by SeMet. Expression was induced at A 600 ϳ0.3 and the purification of the protein was carried out as described previously (5), except that all of the buffers were supplemented with 10 mM dithiothreitol. Complete replacement of methionine residues by selenomethionine was confirmed by electrospray mass spectrometry.
SeMet-labeled sPBP3* crystals were grown by hanging drop vapor diffusion using 1.5 l of protein solution (4 mg⅐ml Ϫ1 ), 1.5 l of well Multiwavelength anomalous diffraction data were collected at 100 K from a single SeMet crystal at peak, inflection, and remote wavelengths, on an ADSC Quantum 4R CCD detector at beamline ID14EH4 at the European Synchrotron Radiation Facility, Grenoble, France. 140 degrees of data collected for each wavelength were processed using MOSFLM (13) and CCP4 (14). Using the peak anomalous data, the selenium sites were located with ShelX-d (15) and refined with SHARP (16). The resulting 44 sites were used for phasing with SHARP. Phase improvement with DM and SOLOMON to 2.8 Å produced a clearly interpretable electron density map, from which an initial model was built using ARPwARP (17). The four molecules in the orthorhombic asymmetric unit were sufficiently different so that non-crystallographic symmetry averaging was not helpful. The model was improved by iterative rounds of manual fitting using O and QUANTA and refinement in CNS (18). Structural superpositions were performed using LSQKAB (14). Figs. 1-3 were prepared with Molscript (19) and Raster3D (20).

sPBP3* Is a Highly Efficient D,D-Carboxypeptidase-
The interaction between PBPs and peptidoglycan substrates or ␤-lactam antibiotics obeys a three-step reaction that is represented as follows, where E is the active PBP enzyme, I is the substrate, EI is the non-covalent Michaelis-Menten complex, EI* is the acyl-enzyme covalent complex, and P is the product of the reaction (21). Enzymatic parameters for carboxypeptidation were estimated by measuring the initial velocities () at various concentrations of Ac 2 -KAA. Inhibition of sPBP3* by its own substrate was observed above a ligand concentration of 15 mM, a phenomenon that has been previously reported for a the D,D-carboxypeptidase from Neisseria gonorrhoeae (22). The values and standard error of k cat ϭ 110 Ϯ 10 s Ϫ1 and K m ϭ 19 Ϯ 3 mM (thus, k cat /K m ϭ 5689 M Ϫ1 ⅐s Ϫ1 ) were obtained by fitting the data points, from zero to peak activity, against the equation ; these estimates may be lower than the true k cat and K m due to the inhibition of the enzyme at high ligand concentrations. By comparison, the k cat /K m value measured with this substrate for The k 3 deacylation rate calculated for sPBP3* is comparable with those reported for other S. pneumoniae PBPs (Table I) but is 20 -30-fold lower than that of E. coli PBP5 (9). In addition, sPBP3* possesses an efficiency of hydrolysis of 50 500 Ϯ 2500 M Ϫ1 ⅐s Ϫ1 for the pseudo substrate S2d. This value is 200 -1000fold higher than for PBP1a, PBP2a, and PBP2b and 20-fold higher than for PBP2x (Table II) (23)(24)(25)(26)(27). These high hydrolytic efficiency values suggest that PBP3 may play an important hydrolytic role during the peptidoglycan biosynthetic process.
Overview of the sPBP3* Structure-The structure of sPBP3* was determined by multiwavelength anomalous diffraction using seleno-methionyl-substituted protein; data collection, phasing, and refinement statistics are shown in Table III. The final structure, which includes four 360-amino acid monomers in the asymmetric unit, has an R factor of 21.2% (R free ϭ 26.2%) at 2.8 Å; 83.1% of the residues lie within the most favored region of the Ramachandran plot and 477 water molecules are included in the model.
The sPBP3* monomer consists of a single polypeptide chain organized into two domains that are orientated approximately at right angles to each other. The spatial relationship between the domains is reminiscent of that observed for its E. coli homolog, the PBP5 D,D-carboxypeptidase, with which PBP3 shares 27% sequence identity (Fig. 1). Domain I (cyan in Fig. 1,  A and B) comprises residues 25-292 and bears the signature fold topology of the penicilloyl-serine transferase superfamily, thus harboring the active site of PBP3. Using the standard secondary structure classification of the class A and class C ␤-lactamases as well as PBP2x (27,28), the first domain is principally constituted by a central five-stranded antiparallel ␤-sheet (␤3/␤4/␤5/␤1/␤2) and two main helices (␣8/␣11). In addition, this domain contains three two-stranded antiparallel ␤-sheets (␤2a/␤2d, ␤2b/␤2c, and ␤2e/␤2f) and three helices (␣2a/␣4/␣5).
The COOH-terminal domain II bears an elongated structure which comprises residues 293-393 and is formed by a sandwich between two anti-parallel ␤-sheets. Comparison of the four molecules, which exist in the asymmetric unit, reveals that the greatest differences at the C␣ level map to loop regions within the COOH-terminal domain, where the electron density is of poor quality in certain regions. Notably, the large area of interaction between the surfaces of domains I and II (ϳ800 Å 2 , including six potential hydrogen bonds) guarantees stability for the full-length molecule, which is reflected by the slight variation (Ͻ1°) of the angle between domains I and II in all four molecules of the asymmetric unit.
The sPBP3* Active Site-The sPBP3* active site is at the distal end from the COOH terminus of the molecule. As observed with the other carboxypeptidases and penicillinmetabolyzing enzymes, the active site is mainly defined by three conserved structural motifs: SXXK (Ser 56 -Ile 57 -Thr 58 -Lys 59 ), which includes the nucleophilic Ser 56 residue, positioned at the NH 2 -terminal end of helix ␣2; SXN (Ser 119 -Ala 120 -Asn 121 ), which forms the turn between helices ␣4 and ␣5 on the left side of the cavity, and K(T/S)G (Lys 239 -Thr 240 -Gly 241 ), which lines strand ␤3 (Fig. 2A). In addition, the backbone NH groups of the essential Ser 56 and Thr 242 residues occupy positions that are compatible with the oxyanion hole function required for catalysis. The NH 2 terminus of helix ␣11 and the loop between ␣6 and ␤2d also contribute residues to the active site. These include Arg 278 , located at the right top angle of the cavity, Thr 160 , and the structural Gly 161 , present on the extended loop at the bottom of the cavity.
The hydrogen bonding network within the active site is extensive ( Fig. 2A) and is identical in all four molecules in the asymmetric unit. The ⑀-NH 2 group of Lys 59 plays a central role in this network, forming hydrogen bonds with the hydroxyl group of Ser 56 and Ser 119 , the side chain carbonyl group of Asn 121 and the backbone carbonyl group from Thr 160 . Two water molecules are observed within the hydrogen bonding network, one of which (O-26) is conserved in the K15 transpeptidase (29), as well as in E. coli PBP5 (8,9). Although the architecture of the active site of the three enzymes is similar (compare Fig. 2, A-C), some significant differences can be observed in the orientation of three important catalytic residues: the side chains of Ser 110 , Lys 213 , and Thr 214 of PBP5 are oriented differently from the equivalent residues in sPBP3*, Ser 119 , Lys 239 , and Thr 240 , respectively. In particular, in PBP5, Ser 110 and Lys 213 point away from the active site, and consequently the classical hydrogen bonding network within the active site is not formed in this molecule. It is of interest that all three molecules, which recognize peptidic substrates, harbor a conserved glycine residue at the bottom of the cleft (Gly 161 , Gly 152 , and Gly 144 in sPBP3*, PBP5, and K15, respectively). The importance of the absence of a side chain at this position becomes evident if one considers that it is part of a binding pocket which could accommodate the penultimate D-alanine residue of the peptidic substrate, as is the case for the well studied R61 D,D-peptidase (30).
The overall structure of sPBP3* is highly reminiscent of that of PBP5 (9) (Fig. 3A). Domains I of the two structures superimpose with an r.m.s. deviation of 1.09 Å; the main differences include helix ␣2a, which is replaced by an irregular loop in PBP5, and the omega-like loop (residues 156 -181 in sPBP3* and residues 147-158 in PBP5), which is longer in sPBP3* due to an insertion of 14 residues (yellow in Fig. 3A). A comparison of domain II for both molecules reveals greater differences including an r.m.s. deviation between C␣ atoms of 2.83 Å. In sPBP3*, this region is clearly less compact than in PBP5, with shorter secondary structure elements.
A particularly outstanding feature of the sPBP3* structure is the close similarity of domain I with the general fold of the Streptomyces K15 transpeptidase (Ref. 29; Fig. 3B); this observation is surprising, since PBP3 catalyzes only carboxypeptidation, while K15 is a transpeptidase (and must first catalyze a carboxypeptidation of the terminal D-Ala residues before transpeptidating with an amino acceptor group). Domain I of sPBP3* can be superimposed onto the structure of the transpeptidase K15 (Ref. 29; Protein Data Bank code 1SKF) with an r.m.s. deviation of 1.54 Å. The same calculation performed with the structure of TEM-1 ␤-lactamase (Protein Data Bank code 1BTL) reveals a C␣ rms deviation of 1.70 Å. As is the case for PBP5, both K15 and TEM-1 harbor shorter omega-like loops than sPBP3* (compare Fig. 3, A-C).
The secondary structural elements that harbor the conserved catalytic residues of the penicillin-binding domain of the transpeptidase PBP2x of S. pneumoniae (Ref. 28; Protein Data Bank code 1PMD) PBP2a of Staphylococcus aureus (31; Protein Data Bank code 1MWR) and domain I of sPBP3* are similarly positioned (Fig. 3), with ␤3, ␣2, ␣4, and ␣5 playing important roles. However, outside of the active site, all of the other re-gions display large differences, as can be observed from the poor superposition results for these molecules (Fig. 3, D and E).
The Omega-like Loop; a Key Structural Feature within D,D-Carboxypeptidases-The omega-like loop of class A ␤-lactamases harbors residues that are required both for maintenance of active site topology and for enzymatic activity. The large structural deviations observed between sPBP3* and PBP5 at the level of the omega loop prompted us to explore this region within the sequences of other putative "PBP5-like" D,D-carboxypeptidase homologues in different bacterial species. Our genomic search focused on proteins with potential or demonstrated D,D-carboxypeptidase activity and for which a topology similar to sPBP3* or PBP5 was predicted, including the presence of a signal peptide, a penicillin-binding domain (analogous to domain I), a COOH-terminal extension of ϳ100 residues (analogous to domain II), and an amphiphilic helix at the COOH terminus (Fig. 4). In addition, only carboxypeptidases harbored by bacteria whose cell wall composition is known were chosen for the study.
This primary sequence analysis revealed the existence of two groups of PBP5-like D,D-carboxypeptidases that could be differentiated based uniquely on the number of residues of their omega-like loops. Interestingly, this variation in length could be correlated to the chemical composition of the bacterial peptidoglycan. Indeed, PBP5-like D,D-carboxypeptidases from bacterial species whose peptidoglycan harbors stem peptides with a diaminopimelate moiety in the third position, such as PBP5 from E. coli or the PBP from B. subtilis, are characterized by a short omega-like loop. On the other hand, the primary sequences of PBP5-like D,D-carboxypeptidases from species whose peptidoglycan possesses stem peptides with a lysine residue in the third position, such as PBP3 from S. pneumoniae, display a much longer omega-like loop (in this case, corresponding to an insertion of 14 residues). Interestingly, the same carboxypeptidases that carry the longer omega-like loops also harbor a 6-amino acid insertion downstream from the third catalytic motif (KTG; see Fig. 4), which, in the tertiary fold, is in proximity to the omega loop. It is thus conceivable that such structural characteristics provide differential substrate recognition within the peptidoglycan biosynthetic machinery.

DISCUSSION
Lmm PBPs are membrane-associated enzymes that play important roles in the maintenance of cell shape and in cellular growth and division processes (5,6,(32)(33)(34). Here, we have biochemically and structurally characterized a soluble form of PBP3 from the human pathogen S. pneumoniae. This work has confirmed that the single lmm PBP from S. pneumoniae catalyzes a D,D-carboxypeptidation reaction, although its catalytic domain is highly reminiscent of the structure of the K15 transpeptidase from Streptomyces. In addition, sPBP3* is also able to recognize ␤-lactam antibiotics, as expected from the presence of the three conserved penicilloyl serine transferase superfamily motifs present in domain I (SXN, SXXK, K(T/S)G). Notably, both D,D-carboxypeptidase activity and ␤-lactam processing proceed through the formation of an acyl-enzyme intermediate followed by deacylation of the enzyme. Based on these data as well as structural and mechanistic information that is available for other PBPs and ␤-lactam-recognizing enzymes (8,9,(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)36), we propose that the acylation mechanism of sPBP3* may occur in four steps. Initially, there is formation of a non-covalent complex that may be stabilized by interactions between the carboxylate group of the ligand and conserved side chains of Thr 240 and Arg 278 . Backbone nitrogen atoms of Thr 242 and Ser 56 could play the role of the oxanion hole required for stabilization of the intermediate reactive species. Ser 56 is activated to form the ester bond with the substrate through abstraction of a proton from neighboring Ser 119 by the COOH of the ligand. This is followed by protonation of the nitrogen of the ␤-lactam ring by the carboxylate and subsequently ring breakage (35). An alternative concerted mechanism is possible, whereby the hydrogen from Ser 119 is transferred to the nitrogen of the ␤-lactam ring at the same time as the hydrogen of Ser 56 is transferred to Ser 119 , and the former forms the acyl bond with the antibiotic (37). sPBP3* is exceptionally efficient in hydrolyzing the pseudo substrate S2d, over 20 times more active than the most efficient pneumococcal enzyme measured to date, PBP2x (Table II) The numbering scheme follows that of the longest sequence, the E. faecalis enzyme (and thus does not relate to the one employed in the description of the sPBP3* structure). In the S. pneumoniae enzyme, the carboxypeptidase domain ranges from residues 55 to 335. Note that S. pneumoniae, Streptococcus mutans, Enterococcus faecalis, and S. aureus, all of which posses a lysine residue in the third position of the stem peptide, harbor an insertion at the level of the omega-loop, as well as between amino acids 314 and 319. Strictly conserved residues are in red, residues conserved in at least half of the species analyzed are in green, and residues that display side chain similarity are in blue.
greater than the value measured for PBP5 with the same substrate (32 M Ϫ1 ⅐s Ϫ1 ; (9)), indicating a greater catalytic efficiency for the pneumococcal enzyme. This observation suggests that PBP3 plays an important role in control of peptidoglycan reticulation in S. pneumoniae. Indeed, Gram-positive organisms have several layers of peptidoglycan, while Gram-negative bacteria, such as E. coli, have much smaller amounts. In addition, E. coli possesses at least three D,D-carboxypeptidases (PBP5, PBP6, PBP6b), and although the precise function of each one in the cell cycle is unknown, it is conceivable that they could share peptidoglycan processing duties within the Gram-negative cell wall. Hence, it is not surprising that PBP3, the only D,D-carboxypeptidase in S. pneumoniae, must be an enzyme with a very high catalytic activity, since it must limit the amount of pentapeptidic stem peptides in the peptidoglycan throughout the entire bacterial cell surface, with the exception of the division site, as shown by immunofluorescence localization studies (5).
In the structures of class A ␤-lactamases, the omega-like loop is a 19-residue stretch comprised of amino acids which participate both in catalysis and maintenance of local topology. Insertional mutagenesis studies showed that enzymes with larger omega-like loops were still able to perform catalysis, and in addition, displayed expanded substrate specificities (38). In this study, we have identified that organisms which display a lysine residue as the third component of the peptidoglycan pentapeptide harbor D,D-carboxypeptidases with large omegalike loops as well as an insertion of six amino acids ϳ30 residues downstream from the catalytic KTG motif. Conversely, bacteria whose peptidoglycan displays a diaminopimelate residue at the same position carry D,D-carboxypeptidases with short omega loops (and lack the 6-residue insertion). The insertion of the amino acid at the third position in the stem peptide is catalyzed by MurE ligases, which show high specificity for their respective substrates (39); notably, if the incorrect residue is inserted at this position, one of the subsequent reactions, transpeptidation, does not take place (40). Thus, it is conceivable that, much like the transpeptidase enzymes that will not bind the incorrect amino acid at the third position of the stem peptide, D,D-carboxypeptidases also require specificity at this position for catalysis. Considering that the carboxypeptidation reaction may require the recognition of a large portion of the stem peptide, including the third moiety and maybe beyond, it is conceivable that the binding region generated by the omega-like loop, and the 6-residue insertion may be able to participate in substrate discrimination. Hence, carboxypeptidases involved in peptidoglycan metabolism may have evolved to perform optimal recognition of a large portion of the stem peptide.
Interaction with and Accessibility to Peptidic Substrates-Although lipid II, the natural substrate for PBP enzymes, carries only pentapeptides in its stem moiety, it is well documented that the pre-existent (available) murein can also harbor multimeric peptides (41). Although the reaction of D,Dcarboxypeptidation of stem pentapeptides has been suggested as being necessary to regulate the degree of peptidoglycan reticulation, it is still unclear whether carboxypeptidases are able to interact with more complex substrates, i.e. with reticulated peptidic molecules. The close similarity between the general folds as well as the active sites of K15 and sPBP3* suggests that it may be the case. The fact that the K15 active site cleft must harbor two peptidic chains to catalyze the transpeptidation reaction suggests that the highly similar sPBP3* may be able to recognize stem peptides that are more complex than the regular pentapeptide. This idea is in agreement with the observation that PBP3 is present throughout the entire bacterial cell surface, except at the site where hmm PBPs are posi-tioned, where peptidoglycan synthesis, which requires substrates with intact stem peptides, occurs (5). PBP3 thus must be a highly efficient processing enzyme, which eliminates the COOH-terminal D-Ala group from all available stem peptides, which are localized outside from the cell division site (independent of their configuration), thus ensuring that intact peptidoglycan will be present uniquely at its initial synthesis site.
PBP5 from E. coli has been suggested as being poised vertically on the bacterial membrane, i.e. with the longer axis of the protein placed perpendicularly to the bilayer (8), an orientation that could be shared by pneumococcal PBP3. Hence, domain II would serve as a pedestal that would bring the active site close enough to the peptidoglycan layer to interact with peptidic side chains that may extend in the direction of the inner membrane (Fig. 5). If PBP3 is placed vertically on the membrane, it would cover an approximate distance of 80 Å (8 nm) from the membrane and thus could interact with the stem or reticulated peptidic chains carried by the lower layers of peptidoglycan. It is conceivable that the COOH-terminal amphiphilic helix could add a certain flexibility to this orientation, enhancing the accessibility of the enzyme to its substrates.
In summary, PBP3 is a highly efficient D,D-carboxypeptidase, which, at the beginning of the cell cycle, trims stem and/or reticulated pentapeptides harbored by the pre-existent murein to generate inert regions of cell envelope. Its detection, by immunofluorescence, throughout the bacterial surface (5), coupled to the abovementioned model where the amphiphilic helix may permit superficial association to the membrane, suggests that it may "skid" on the surface of the membrane, trimming pentapeptides. Donor pentapeptides for the transpeptidation reaction, required for peptidoglycan synthesis, would then only be available at the future division site. Once murein synthesis is initiated, the D,D-carboxypeptidase could regain the regions with young peptidoglycan where it could again interact with stem or/and reticulated pentapeptides to regulate the reticulation degree of the future mature murein.