A PBP2x from a clinical isolate of Streptococcus pneumoniae exhibits an alternative mechanism for reduction of susceptibility to beta-lactam antibiotics.

The human pathogen Streptococcus pneumoniae is one of the main causative agents of respiratory tract infections. At present, clinical isolates of S. pneumoniae often exhibit decreased susceptibility toward beta-lactams, a phenomenon linked to multiple mutations within the penicillin-binding proteins (PBPs). PBP2x, one of the six PBPs of S. pneumoniae, is the first target to be modified under antibiotic pressure. By comparing 89 S. pneumoniae PBP2x sequences from clinical and public data bases, we have identified one major group of sequences from drug-sensitive strains as well as two distinct groups from drug-resistant strains. The first group includes proteins that display high similarity to PBP2x from the well characterized resistant strain Sp328. The second group includes sequences in which a signature mutation, Q552E, is found adjacent to the third catalytic motif. In this work, a PBP2x from a representative strain from the latter group (S. pneumoniae 5259) was biochemically and structurally characterized. Phenotypical analyses of transformed pneumococci show that the Q552E substitution is responsible for most of the reduction of strain susceptibility toward beta-lactams. The crystal structure of 5259-PBP2x reveals a change in polarity and charge distribution around the active site cavity, as well as rearrangement of strand beta3, emulating structural changes observed for other PBPs that confer drug resistance to Gram-positive pathogens. Interestingly, the active site of 5259-PBP2x is in closed conformation, whereas that of Sp328-PBP2x is open. Consequently, S. pneumoniae has evolved to employ the same protein in two distinct mechanisms of antibiotic resistance.

Streptococcus pneumoniae is a commensal, Gram-positive bacterium that colonizes the upper respiratory tract in humans and is responsible for over one million deaths per year, mostly among the elderly and young children in developing countries. S. pneumoniae is a leading cause of community-acquired respiratory infections including pneumonia, acute otitis, and sinusitis as well as invasive infections such as bacteremia and meningitis (1). ␤-Lactam antibiotics are presently the most widely used molecules employed in the fight against pneumococcal infection. However, since the introduction of penicillin in the late 40s, ␤-lactam-resistant strains of S. pneumoniae have emerged in various locations and have now spread worldwide (2).
The S. pneumoniae bacterium is surrounded by a thick cell wall of which the main component is the peptidoglycan, a highly cross-linked mesh that is essential for bacterial cell division and protection from osmotic shock and lysis (3). Peptidoglycan synthesis occurs through a biosynthetic pathway that is initiated in the bacterial cytoplasm; the last steps occur on the periplasmic side of the bacterial membrane. The cytoplasmic steps of the pathway are catalyzed by Mur enzymes and result in the synthesis of lipid II molecules, which harbor the peptidoglycan building units and are subsequently translocated across the cytoplasmic membrane (4). The final steps of synthesis, namely the polymerization of the glycan chains and their reticulation, are catalyzed by the glycosyltransferase and transpeptidase activities of penicillin-binding proteins (PBPs) 1 (5,6).
In S. pneumoniae, six PBPs have been identified and classified on the basis of their sequence similarities (7,8). High molecular mass PBPs are bifunctional (class A) enzymes harboring glycosyltransferase and transpeptidase (TP) activities or monofunctional enzymes (class B) with only the TP activity. Finally, a low molecular mass PBP displays a D,D-carboxypeptidase activity.
The TP activity of PBPs is responsible for cross-linking of peptidoglycan glycan chains through the formation of peptidic bridges (5). ␤-Lactam antibiotics specifically inhibit the TP activity by acylating the active site serine (9). These molecules are specific toward the TP domain because of the structural analogy between the ␤-lactam ring and the D-alanyl-D-alanine peptidoglycan moiety, the natural substrate of the PBPs. The inhibition of the TP activity causes a disruption in cell wall synthesis that often leads to cell death (10).
␤-Lactam resistance in S. pneumoniae is a consequence of the generation of mosaic pbp genes, which result from intraand interspecies recombination with sequences from related streptoccoci (11). Such genes encode PBPs harboring tens of substitutions spread throughout the entire protein (12,13), leading to a loss of affinity for the antibiotics. Three of the streptococcal PBPs, namely PBP2x, PBP2b, and PBP1a, have been shown to be modified in penicillin-resistant strains isolated in a clinical setting (12,14). PBP2x, one of the two monofunctional high molecular mass PBPs in S. pneumoniae, is described as the primary PBP target in ␤-lactam-resistant strains (15,16). The appearance of a mosaic PBP2x is thus the first event occurring in the development of antibiotic resistance in S. pneumoniae (17).
We performed an analysis of 89 PBP2x sequences isolated from well characterized pneumococcal clinical isolates and found that they can be classified, based on cluster analysis, into three groups. The first group contains sequences closely related to the one isolated from the susceptible strain R6 (18). The two other groups contain sequences isolated from drug-resistant strains. The larger of these groups (53 sequences) contains mostly sequences related to PBP2x from ␤-lactam-resistant strain Sp328 (19). PBP2x from strains R6 and Sp328 have been well characterized, both biochemically (18 -21) and structurally (22)(23)(24). The second group of mutants, however, which is distinguished by the Q552E mutation within all of the PBP2x sequences, remains uncharacterized both on the biochemical and structural level. This category of PBP2x variants appears to present a molecular mechanism selected to reduce the reactivity toward ␤-lactams that is different from that of the Sp328like PBP2x proteins. One representative PBP2x, from strain 5259, was selected from this group for detailed study. In this work, we report the structural and functional analysis of 5259-PBP2x* and demonstrate the role of the Q552E substitution in the ␤-lactam resistance process.

Isolation of Drug-resistant Clones and Determination of Minimal Inhibitory Concentration (MIC) Values for ␤-Lactam
Antibiotics-The S. pneumoniae clinical strain 5259 of serotype 15 was isolated from a tracheal noninvasive sample at the University Hospital in Grenoble, France. S. pneumoniae 5259 was grown at 37°C in an atmosphere of 95% air, 5% CO 2 on Columbia blood agar plates (Biomérieux). Clones were isolated, and overnight liquid cultures were stored at Ϫ80°C.
The MIC values for penicillin G and cefotaxime were determined using the E-test method on Muller-Hinton agar plates supplemented with 5% sheep blood. S. pneumoniae 5259 transformants from overnight liquid cultures were plated on agar plates at 0.5 McFarland. E-test strips were laid on the plates, and the MIC was read after an incubation period of 18 to 24 h as the values at the intersection of the strip and the grown bacteria.
Construction of Expression Plasmids, Site-directed Mutagenesis, and Protein Purification-Genomic DNA from strain 5259 was extracted with the High Pure PCR preparation kit (Roche Applied Science). The region of pbp2x coding for residues 49 -750 (thus lacking the cytoplasmic region and the transmembrane helix; henceforth identified with an asterisk) was PCR-amplified as a BamHI-SalI fragment and cloned into pGEX-4T1 to create the pGEX-5259-pbp2x* plasmid, which expresses 5259-PBP2x*. The previously described pGEX-R6-pbp2x* plasmid was used to express PBP2x* from the R6 strain (25).
With the goal of generating PBP2x* mutants in position 552 of both R6-PBP2x* and 5259-PBP2x* proteins, the QuikChange Kit (Stratagene) was employed. The primer 5Ј-GCCCTCAAGTCTGGTACCGCTG-AGATTGCTGACG-3Ј and its reverse complement were used to generate the pGEX-R6-pbp2x*-Q552E plasmid. The primers introduced a KpnI restriction site within the mutation site for rapid screening of successful mutants. The pGEX-5259-pbp2x*-E552Q plasmid was created using the primer 5Ј-GCCCTTAAATCAGGTACCGCGCAGATTGC-GGATGAGAAAAAT-3Ј, and its reverse complement. The region of the expected mutation was sequenced for both constructs.
Transformation of S. pneumoniae Strains-The R6 strain of S. pneumoniae was used as recipient for transformations. Competent cells, generated as described previously (27), were diluted 10-fold in C-medium with 0.018% albumin (8%, boiled), and were mixed with 50 ng of plasmid DNA. Cells were incubated for 30 min on ice and subsequently for 120 min at 37°C before being plated on Columbia blood agar Base EH (BD Biosciences) plates enriched with 4% horse blood and cefotaxime at concentrations ranging from 0 to 0.3 g.ml Ϫ1 . Isolated clones were picked after 24 h of incubation at 37°C in an atmosphere of 95% air, 5% CO 2 and then grown in glucose-buffered broth (Diagnostic Pasteur). Overnight cultures were then stored at Ϫ80°C until the MIC determination was carried out. The MIC values were measured using the E-test method as described above for three independent isolates. All transformed colonies were confirmed as S. pneumoniae by agglutination and optochine susceptibility assays. Finally, genomic DNA of each transformant was extracted, and the pbp2x genes were PCR-amplified and sequenced.
Biochemical Characterization of the Purified Proteins-Acylation efficiencies were measured for R6-PBP2x*, R6-PBP2x*-Q552E, 5259-PBP2x*, and 5259-PBP2x*-E552Q for both penicillin G and cefotaxime. The k 2 /K parameter was determined by following the intrinsic fluorescence of the proteins during antibiotic binding (9). Measurements were made at 37°C using a spectrofluorimeter coupled to an SFM-400 stopped-flow apparatus (Bio-Logic). Proteins were used at a concentration of 0.6 M and antibiotics at excess concentrations ranging from 25 M to 2.5 mM in 10 mM sodium phosphate, pH 7.0. Excitation was measured at 280 nm, and emission was recorded above 305 nm. At a given concentration of antibiotic, the pseudo first-order rate constant, k app , was determined by nonlinear least squares fitting to the equation Fluo t ϭ Fluo 0 exp(Ϫk app ⅐t). The k 2 /K parameter, which accounts for the acylation efficiency, was determined by linear square fitting to the equation k app ϭ (k 2 /K)/[antibiotic].
The deacylation rate was determined for all four proteins using [ 3 H]penicillin G as previously described (28). The deacylation reaction obeys the following equation: k 3 t ϭ ln[EI*] t /[EI*] 0 , where [EI*] 0 is the initial concentration of acyl enzyme and [EI*] t is its concentration at time t. To initiate acylation, 2 M protein was incubated with 1 M radiolabeled penicillin G in a buffer composed of 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 1 mM EDTA. After 10 or 30 min of incubation for R6-PBP2x* and 5259-PBP2x*, respectively, 15 mM unlabeled penicillin G was added, defining the beginning of the kinetic assay. Protein samples were withdrawn at various times and immediately boiled with 1% SDS prior to analysis by SDS-polyacrylamide gel. The amount of radioactivity was measured in the protein bands by liquid scintillation counting. The rate constant k 3 was determined by nonlinear least squares fitting to the equation cpm t ϭ cpm 0 ⅐exp(Ϫk 3 ⅐t).
Crystallization, Data Collection, and Processing-Screening experiments were performed by a TECAN Genesis robot at 8°C using commercial Hampton crystallization screens kits and employing the sitting drop vapor diffusion method (Hampton Research). Crystallization drops were composed of 1 l of protein solution and an equal amount of reservoir solution, which was 200 l in volume. A first hit was identified with 15% polyethylene glycol 1500, and the optimization experiments were set up by hand using the hanging-drop technique. Crystals grew under the following conditions. 1 l of precipitating solution (14% polyethylene glycol 1500, 50 mM sodium acetate, pH 4.3, ammonium sulfate 200 mM) was added to 1 l of the protein solution at a concentration of 10 mg.ml Ϫ1 (in a buffer containing 50 mM Tris-HCl, pH 8, 100 mM NaCl, 1 mM EDTA). The drops were equilibrated over a reservoir of 500 l of the precipitating solution and kept at 8°C. Crystals grew as bipyramids to maximum dimensions of 0.15 ϫ 0.17 ϫ 0.12 mm after 2 weeks. They belong to the tetragonal space group, P4 2 22, with unit cell dimensions of a ϭ b ϭ 195.0 Å, c ϭ 154.2 Å. The asymmetric unit accommodates two monomers, corresponding to a crystal packing parameter (V M ) of 4.76 Å 3 .Da Ϫ1 and a solvent content of 74%.
Prior to data collection, crystals were briefly immersed in a cryoprotectant solution having the same composition as the mother liquor but with a slight increase of polyethylene glycol 1500 (to 16%) and containing 15% 2-methyl-2,4-pentanediol as cryoprotectant. Crystals were subsequently flash-cooled by being plunged into liquid nitrogen. X-ray diffraction data were collected on the BM30 beamline at the European Synchrotron Radiation Facility (Grenoble, France) operating at a wavelength of 0.999 Å. Angular increments between diffraction images of 1°a nd a distance crystal-to-detector of 210 mm were used. Raw diffraction images were indexed and integrated with MOSFLM, version 6.2.2 (29). Data scaling, merging, and reduction was carried out with programs of the CCP4 suite (30). The self-rotation function displays peaks showing a 2-fold axis perpendicular to the c axis.
Structure Solution and Refinement-The crystal structure of 5259-PBP2x* was determined by molecular replacement using AMoRe (31) with data between 30 and 3.0 Å and the R6-PBP2x* structure as a search model (22) (Protein Data Bank accession code 1QME). Residues differing from the 5259-PBP2x* sequence were replaced by alanines prior to the molecular replacement procedure. The rotation and the translation functions gave two clear single solutions, with an R-factor of 42.8% and a correlation coefficient of 68.1%, which corresponded to two molecules in the asymmetric unit related by a 2-fold noncrystallographic axis. These two solutions were employed in the refinement procedure performed with CNS 1.1 (32). The first cycle of the refinement procedure consisted of 20 cycles of rigid-body refinement with data between 31 and 4 Å. At this stage the N-terminal domain of the model was completely built in the electron density map for both molecules in the asymmetric unit. Subsequently, several cycles of simulated annealing, energy minimization, and grouped temperature factor refinement were performed. Cycles of refinement were interchanged with manual rebuilding sessions using the molecular graphics program O (33). Water molecules were positioned progressively if they were observed to be in an environment susceptible to providing hydrogen bonds. In the final stages of the refinement, individual restrained B-factor refinement was performed. These iterative cycles of refinement and map interpretation led to a model with R-factor ϭ 23.5% and R-free ϭ 25.6% (34). The stereochemistry of the model was analyzed with PROCHECK (35); a total of 83.1% of the non-glycine residues were found in the "most favored" region of the Ramachandran plot.

RESULTS
Characterization of pbp2x Sequences-Eighty-nine strains displaying a wide range of ␤-lactam resistance levels were selected from a large collection of well characterized clinical strains kept at the University Hospital in Grenoble, France. The MIC values for penicillin G and cefotaxime were measured for all of strains, and the pbp2x fragments coding for each TP domain were sequenced. Among the 26 drug-sensitive strains (MIC PenG Ͻ 0.1 g.ml Ϫ1 ), only 10 PBP2x sequences displayed differences from that of PBP2x from strain R6. Similarly, the 53 remaining drug-resistant strains (MIC Ͼ 0.1 g.ml Ϫ1 ) displayed 12 distinct PBP2x sequences. When the 22 different sequences (10 from sensitive strains, 12 from resistant strains) were submitted to a clustering algorithm (www.genebee. msu.su), they were found to fall into three groups: one contained 10 PBP2x sequences with major similarities to the R6-PBP2x sequence and originated from the susceptible strains; a second group displayed 11 PBP2x sequences that were similar to the previously characterized PBP2x from drug-resistant strain Sp328 (19,23); and a third differing PBP2x sequence displayed similarities to PBP2x from strain F2. The existence of two distinct groups of sequences from resistant strains was further confirmed by analysis of public data bases. Because our laboratories had already biochemically and structurally characterized a PBP2x molecule from strain Sp328 (23, 26), we initiated the characterization of a representative PBP2x from the second group of strains. Two strains from our hospitalderived collection displayed this particular PBP2x sequence, and one of them, strain 5259, was further characterized in this work.
S. pneumoniae strain 5259 has MICs of 0.19 and 0.094 g.ml Ϫ1 for penicillin G and cefotaxime, respectively. The 5259-PBP2x* sequence displays 28 mutations when compared with the R6-PBP2x* sequence. These mutations are as follows: three in the N-terminal domain (A172T, R254Q, and M256V), three in the C-terminal domain (L710F, Q721E, and T745K) and 22 in the TP domain. The latter 22 substitutions are within the C-terminal part of the TP domain, from amino acids 447 to 616. Among these substitutions, only 9 (I462L, T490S, A491V, D506E, N514H, L565S, D567N, A572V, and D616E) are common to both Sp328-PBP2x* and 5259-PBP2x* sequences. Among the 13 other substitutions in the TP domain, 9 are unique to 5259-PBP2x* (Q447M, S449A, I483L, L517M, P535A, T538N, Q552E,  V563T, and Y568N), and 4 occur in both sequences, but the substituted amino acid residue is different (L510V, T513D, L523T, and N576H). The 5259-PBP2x sequence does not display the T338A substitution, which was shown to be an important factor in the development of drug resistance in strains such as Sp328 (26) and is located adjacent to the active site serine; but it does contain a Q552E mutation adjacent to the third catalytic motif of the active site.
Notably, public data bases contain 21 PBP2x* sequences harboring the Q552E substitution in the TP domain. When submitted to a cluster analysis, these sequences were found to differ markedly from those related to that from strain Sp328. Among these sequences, only four also harbor the T338A mutation, whereas one displays a T338P modification. Among the 16 remaining sequences, the ones from strains F2, G54, and Sp1465 are 100% identical from amino acids 266 to 616 to PBP2x from strain 5259.
In Vitro Analysis of 5259-PBP2x*, 5259-PBP2x*E552Q, and R6-PBP2x*Q552E-To gain insight into the molecular mechanism of resistance generated by 5259-PBP2x and to assess the role of the Q552E substitution in PBP2x catalysis and structure, different PBP2x* proteins were produced for structural analysis and in vitro characterization assays. In addition to the production of R6-PBP2x* and 5259-PBP2x*, site-directed mutagenesis was performed at position 552 of both proteins to generate the R6-PBP2x*-Q552E and the 5259-PBP2x*-E552Q mutants. All four proteins were expressed and purified analogously.
The kinetics of the interaction between PBPs and ␤-lactams can be described by a three-step reaction represented by the following equation (36).
The first step represents the formation of a noncovalent complex EI with the dissociation constant K ϭ k Ϫ1 /k 1 followed by the acylation of the active site serine by the ␤-lactam molecule with the rate constant k 2 . Those two steps are termed the "acylation efficiency" and are characterized by the second order rate constant k 2 /K. In this work, the acylation efficiency was measured at pH 7.0 by monitoring the decrease of the intrinsic fluorescence of each protein upon antibiotic binding. The acylation step is a very fast reaction and was thus analyzed with a stopped-flow apparatus coupled to a spectrofluorimeter.
The results are presented in Table I and Fig. 1. The introduction of the substitution Q552E into R6-PBP2x* is responsible for a 4.2-and 5.6-fold decrease of the acylation efficiency for cefotaxime and penicillin G, respectively. These values obtained for R6-PBP2x* and for R6-PBP2x*-Q552E are in agreement with the previously reported data (25). The acylation efficiency of 5259-PBP2x* is reduced 15-fold for cefotaxime and 23-fold for penicillin G when compared with R6-PBP2x*. The reversion of the substitution in position 552 within 5259-PBP2x* from Glu to Gln increases the acylation efficiency of the modified protein 2.7-fold for cefotaxime and 4.3-fold for penicillin G. The retained reactivities of R6-PBP2x*-Q552E and 5259-PBP2x*-E552Q toward both of the antibiotics tested are about 20% of the reactivity of the wild type R6-PBP2x* (as shown in Fig. 1).
The final step of the reaction between PBPs and ␤-lactams is deacylation. This step regenerates an active PBP, releasing an inactivated compound, and is characterized by the rate constant k 3 . The deacylation rate was measured at pH 8.0 by fluorography as described previously (28); results are given in Table I. A slight modification of the deacylation rate was observed in 5259-PBP2x* as well as in the other point mutants.
In Vivo Analysis of 5259-PBP2x*-To assess the role of the 5259-pbp2x gene in the ␤-lactam resistance process in S. pneumoniae and to demonstrate the effect of the Q552E substitution, the various modified pbp2x genes were introduced into the recipient genome of the susceptible strain R6 by homologous recombination. The MICs for penicillin G and cefotaxime were determined using the E-test method; results are presented in Table II and Fig. 2. The introduction of any of the mutant pbp2x genes always led to a higher resistance for cefotaxime than for penicillin G. Note that the MIC CTX of the 5259-pbp2x transformant is thus the same as the MIC CTX of the originating strain 5259.
Overall Structure of 5259-PBP2x*-To gain insight into a possible novel mechanism of ␤-lactam resistance displayed by 5259-PBP2x* in atomic detail, we solved its crystal structure by molecular replacement at a resolution of 3.0 Å. The relevant statistics of data collection and model refinement are given in Table III. The overall architecture of 5259-PBP2x* is very similar to that of R6-PBP2x* (22,24) in that it consists of three domains: an elongated "sugar tong"-like N-terminal domain (residues 50 -265) and a central TP domain (residues 266 -616) that is connected to a C-terminal region (residues 635-750) by a linker segment (residues 617-634). For both monomers in the asymmetric unit of the crystal, the electron density map in the N-terminal domain was clear and interpretable and allowed the construction of the structural model from residue Lys 64 onward.
The two monomers in the 5259-PBP2x asymmetric unit are related by a 2-fold noncrystallographic axis that is perpendicular to the c axis and makes an angle of 12°with the a axis. Superposition of the two monomers (using LSQKAB from the CCP4 suite of programs) resulted in an r.m.s.d. of 1.2 Å calculated on the 687 C␣ positions of residues 64 -750. The secondary structure elements are similar, but some differences are observed in the coil regions. The superposition of a The error is the standard error obtained from the fit of the k app versus ͓antibiotic͔ data to the equation k app ϭ (k 2 /K)⅐͓antibiotic͔. b The error is the standard error obtained from the fit of ͓EI*͔ data to the equation ͓EI*͔ t ϭ ͓EI*͔ 0 ⅐exp(Ϫk 3 ⅐t).    (Fig. 3). The monomer-monomer interface involves 17 residues from each monomer, of which a majority have hydrophilic side chains. The interaction is stabilized by eight hydrogen bonds and two salt bridges (Arg 138 from one monomer to Asp 313 from a neighboring monomer). We suggest that these intermolecular interactions stabilize the Nterminal domains, which may explain why they are less disordered than in other crystal structures of PBP2x. During model building, the electron density map corresponding to the side chains of residues that had been mutated into alanines was clearly interpretable.
A least squares fit of the main-chain residues of the TP domain between R6-PBP2x* and 5259-PBP2x* structures generates an r.m.s.d. of 0.6 Å for both monomers. The largest differences in backbone positions are observed in the loops 313-325, 354 -361, 374 -391, 517-535, and 555-568, which are known to be flexible. Similarly, a least-squares fit with the main-chain residues of the TP domain of Sp328-PBP2x* results in an r.m.s.d. of 0.9 Å. Again, the main differences are observed in the loop regions 353-360, 516 -536, and 554 -568.
The 5259-PBP2x* Active Site-In the active site of 5259-PBP2x*, the three conserved motifs involved in PBP catalytic activity were clearly traceable (Fig. 4). The first motif, Ser 337 -X-X-Lys, is located at the bottom of the cleft on the first turn of helix ␣2. The second element, Ser 395 -X-Asn, is on a short loop between the ␣4 and ␣5 helices. The side chains of Ser 395 and Asn 397 point into the active site, forming one wall of the active site cavity. The third motif, Lys 547 -Ser-Gly, is situated on the innermost strand (␤3) of the central sheet. These three latter residues form the opposite wall of the active site cavity.
The majority of the mutations (17 of 22) that are observed in 5259-PBP2x* are concentrated in the active site area of the TP domain (Fig. 3). Only a few point mutations occur in the N-and C-terminal domains. A similar observation was made in the crystal structure of PBP2x from drug-resistant strain Sp328. The lack of mutations within the sugar tong region in both mutant structures thus reinforces the hypothesis that residues located within this domain may play important roles in proteinprotein recognition (23). In the TP domain, most of the mutations are exposed to the surface and surround the active site cavity. Other groups of mutations are located above the active site cleft in helix ␣9 and in the loop connecting helix ␣9 to strand ␤3, as well as on strand ␤4 and in the preceding loop (E552-N568), which borders the right side of the active site. Except for strand ␤4, these segments are flexible and display different conformations in the various PBP2x crystal structures reported.
Among these substitutions, Q552E, Q447M, S449A, and Y568N are located within a distance of 10 to 12 Å from Ser 337 and are positioned strategically to play important roles in the drug resistance phenomenon. Q552E is located after the third catalytic motif Lys 547 -Ser-Gly at the end of strand ␤3 and borders the active site cavity. This mutation replaces a polar residue and introduces a negative net charge at the entry of the active site (Fig. 5), an event that clearly disfavors the binding of ␤-lactams, which bear a global negative charge. The side chain of Glu 552 in 5259-PBP2x* adopts the same orientation as Gln 552 in R6-PBP2x* and points toward the outside of the active site.
Another active site mutation, Y568N, is located at the N terminus of ␤4; the Asn side chain thus lies across from that of Glu 552 . Interestingly, in the recently reported crystal structure of PBP2a from a methicillin-resistant Staphylococcus aureus strain, two identical residues (Glu 602 and Asn 613 ) are found in equivalent structural positions (the end of strand ␤3/begining of strand ␤4) (37). The third and fourth mutations, Q447M and S449A, are located at the entry of the active site and buried in the vicinity of the active site serine, respectively. These two substitutions are absent from Sp328-PBP2x*. Whereas the Q552E substitution is found in 21 publicly available sequences and can be considered as a signature mutation, substitutions at positions 447, 449, and 568 are observed only in a subset of sequences.  The most noticeable difference between the 5259-PBP2x* active site and that of PBP2x* from the drug-sensitive R6 strain lies on the positioning of ␤3, especially at the level of residues Ser 548 to Thr 550 (r.m.s.d. ϭ 0.5 Å; Fig. 5, the overlaid yellow strand). It is interesting that, in PBP2a from S. aureus (37), ␤3 also displays a noticeable bend (away from the central region of the sheet); these authors have suggested that such positioning plays an important role in the development of drug resistance by strains harboring PBP2a.
Interestingly, the 5259-PBP2x* active site is very different from the one observed for PBP2x* from pneumococcal strain Sp328. Although 5259-PBP2x* displays a closed active site that is reminiscent of the PBP2x* from the drug-sensitive strain R6, the active site of Sp328-PBP2x* is in an open configuration due to the presence of clashing mutations (23). The fact that the same protein may, by homologous recombination, acquire completely different active sites suggests that S. pneumoniae employs PBP2x in the development of drug resistance via two very different mechanisms. DISCUSSION ␤-Lactam resistance in S. pneumoniae is achieved by the acquisition of mosaic PBPs with decreased affinities for ␤-lactam antibiotics through intra-and interspecies homologous recombination (38,39), which generate tens of substitutions within distinct sequences (12). PBP2x, the primary resistance determinant in S. pneumoniae (20), exists under different forms depending on the phenotypic background from which it comes (14). In this work, we have classified PBP2x proteins from clinical isolates into three main groups depending on their primary structures. Sequences coming from susceptible strains are closely related to the sequence of the drug-sensitive strain R6. Mosaic sequences isolated from resistant strains can be divided in two groups, those that harbor the T338A substitution and are related to PBP2x from the highly resistant strain Sp328 and those that can be defined by the landmark presence of the Q552E substitution. In this work, we have performed the biochemical and structural characterization of a PBP2x from drug-resistant strain 5259, which belongs to the latter group of sequences.
In previous work (25), the introduction of the Q552E mutation in R6-PBP2x was reported to play an important role in the modulation of the sensitivity of PBP2x to ␤-lactams by reducing the acylation efficiency. The role of this substitution is reinforced by the fact that the reversion from Glu to Gln in position 552 of the 5259-PBP2x* led to an increase of the acylation efficiency by a factor 2.7 for cefotaxime and 4.3 for penicillin G. The kinetic parameters of the reverted 5259-PBP2x* in position 552 are close to the ones determined for R6-PBP2x*-Q552E. The MICs of the transformant cells harboring either the 5259-pbp2x*-E552Q or the R6-pbp2x*-Q552E genes are identical and are 0.047 and 0.032 g.ml Ϫ1 for cefotaxime and penicillin G, respectively. Thus, the effect of the single Q552E substitution is equivalent to the global effect of the 27 other substitutions spread throughout the 5259-PBP2x*. The Q552E mutation, the key determinant substitution for this group of PBP2x proteins, is thus responsible for a 4.1-and 5.6-fold decrease of the acylation efficiency for cefotaxime and penicillin G, respectively, when compared with the R6-PBP2x*. These values are in good agreement with the previously published data (25). The kinetic effect of this substitution in the molecular background of the R6-PBP2x* is 2-fold higher than that of the T338A substitution, which is the land- mark mutation for PBP2x molecules from Sp328-like strains (26).
The crystal structure of 5259-PBP2x* displays high similarity to the overall fold presented by the structure of PBP2x* from drug-sensitive strain R6 except within the active site region. In this area of the molecule, strand ␤3, notably within the segment Ser 548 -Thr 550 , is displaced by 0.5 Å (with a maximum of 0.7 Å for Thr 550 ) from the position occupied by the same strand in R6-PBP2x. It should be noted that strand ␤3 is not particularly disordered (residues 544 -552 display average temperature factors of ϳ60 Å 2 in both monomers). Moreover, the other strands, ␤4 and ␤5, superimpose very well with the corresponding strand in the R6-PBP2x structure. Consequently, we suggest that the movement of ␤3 could be related to the development of ␤-lactam resistance in pneumococci. Interestingly, the poor acylation rate of PBP2a from another Grampositive pathogen, S. aureus, is known to be correlated to a distorted active site (37). The active site cavity of PBP2a is in a closed conformation, thus restricting the accessibility of the catalytic serine Ser 403 . In the structure of PBP2a, strand ␤3 adopts a twisted conformation to accommodate the N terminus of the ␣2 helix. Furthermore, for acylation to occur, the N terminus of helix ␣2 undergoes a conformational change to position Ser 403 for nucleophilic attack. Concomitantly, this conformational rearrangement engenders a twisting movement on ␤3 and a consequent displacement of the central ␤-sheet (37). It is thus conceivable that the mechanism of resistance displayed by pneumococcal strains that carry PBP2x molecules with similarities to 5259-PBP2x* involves local movement of strand ␤3, as is the case for PBP2a.
Most mutations in the 5259-PBP2x* structure are located within the active site region, with four mutations, namely Q447M, Q552E, S449A, and Y568N, located within a 12-Å radius. It is clear that the introduction of a negative charge in position 552 greatly affects the entry of the active site cavity, producing a negative effect for the binding of ␤-lactam antibiotics, which are negatively charged. Interestingly, PBP5fm from Enterococcus faecalis, which is responsible for the natural resistance of these pathogens to ␤-lactams, displays a Glu residue in a position equivalent to Glu 552 in 5259-PBP2x* (40). It is thus conceivable that a negative charge in this position plays an important role in the process of antibiotic recognition by PBPs in a variety of drug-resistant bacteria.
In the structure of the acyl enzyme intermediate of R6-PBP2x* with cefuroxime, it was shown that, upon binding of the antibiotic, the side chains of Gln 452 and Gln 552 adopt new rotameric conformations and point toward the outside of the active site, thus opening the ground of the active site cavity (22). In particular, a hydrogen bond with the Gln 447 side chain stabilizes the new conformation adopted by the Gln 452 side chain. The Q447M mutation in 5259-PBP2x* abrogates this interaction and therefore disfavors the opening of the active site, and it may also affect the ␤-lactam resistance mechanism.
The crystal structure of PBP2x from the highly drug-resistant strain Sp328 showed that mutation of threonine 338 into an alanine led to the abrogation of a crucial hydrogen bond between the hydroxyl group of the threonine residue and a buried water molecule. In addition, two other key mutations, S389L and N514H, led to an "open" active site due to the positioning of the His 514 side chain within the region where ␣4 should be located in order to close the catalytic cleft. This clash could thus be responsible for the loss of affinity of the protein toward ␤-lactams. In the case of 5259-PBP2x, although the active site remains closed, the drug resistance profile of transformed and clinical strains is still marked. Consequently, it is conceivable that PBP2x proteins can participate in drug resistance phenomena by employing two distinct mechanisms, which are evident through sequence analyses and structural characterizations. In the first mechanism, an open active site causes a deficiency in ester bond formation and substrate acylation (Sp328 (23)). In the second, a closed active site undergoes a local modification of ␤3 and introduces a negative charge at its entrance, thus providing a less optimal cleft for recognition of antibiotic molecules.