Balance between Two Transpeptidation Mechanisms Determines the Expression of (cid:1) -Lactam Resistance in Enterococcus faecium *

The D , D -transpeptidase activity of high molecular weight penicillin-binding proteins (PBPs) is essential to maintain cell wall integrity as it catalyzes the final cross-linking step of bacterial peptidoglycan synthesis. We investigated a novel (cid:1) -lactam resistance mechanism involving by-pass of the essential PBPs by L , D -transpep- tidation in Enterococcus faecium. Determination of the peptidoglycan structure by reverse phase high per-formance liquid chromatography coupled to mass spectrometry revealed that stepwise selection for ampicillin resistance led to the gradual replacement of the usual cross-links generated by the PBPs ( D -Ala 4 3 D -Asx-Lys 3 ) by cross-links resulting from L , D -transpep- tidation ( L -Lys 3 3 D -Asx-Lys the final ampicillin concentration. Preliminary experiments showed that the length of the preincubation in the presence of ampicillin (25, 45, or 120 min) did not affect v i /v o . Assay of D , D -Carboxypeptidase Activity— The activity was assayed by quantifying UDP-MurNAc- L -Ala- (cid:1) - D -Glu- L -Lys- D -Ala (UDP-MurNAc- tetrapeptide) formed by hydrolysis of the C-terminal D -Ala residue of UDP-MurNAc- L -Ala- (cid:1) - D -Glu- L -Lys- D -Ala- D -Ala (UDP-MurNAc-penta- peptide) prepared as previously described (2). The assay was performed at 37 °C in a 100- (cid:5) l mixture containing membrane or cytoplasmic extracts (10–60 (cid:5) g of proteins), UDP-MurNAc-pentapeptide (1.1 m M ), Tris-HCl (50 m M , pH 7.0), and MgCl 2 (1 m M ). The reaction was stopped by precipitating the proteins with sulfosalicylic acid (0.25 mg). After centrifugation (10,000 (cid:2) g , 3 min), 90 (cid:5) l of the supernatant was ana- lyzed by RP-HPLC. UDP-MurNAc-pentapeptide was separated from UDP-MurNAc-tetrapeptide using isocratic elution (50 m M ammonium formate, pH 5.0) at a flow rate of 2 ml/min on a (cid:5) -Bondapak C 18 column (7.8 (cid:2) 300 mm, Waters). The products were detected by the absorbance at 262 nm. the Peptidoglycan Precursors Pools— After to an A 650 0.7, vancomycin was added to a final concentration of 100 (cid:5) g/ml to block transglycosylation , and incubation was

The synthesis of bacterial cell wall peptidoglycan is a two-stage process. First, the disaccharide peptide monomer unit is assembled in a series of cytoplasmic and membrane reactions (1). In Enterococcus faecium, the resulting unit is composed of N-acetylglucosamine (GlcNAc) 1 and N-acetylmuramic acid  (2)(3)(4). The final steps of peptidoglycan synthesis involve its transfer through the cytoplasmic membrane, its polymerization to glycan strands by glycosyltransferases, and the cross-linking of stem peptides by D,D-transpeptidases. These latter enzymes catalyze the formation of a peptide bond between the carboxyl of D-Ala at position 4 of a donor stem peptide and the amino group of the D-asparagine or D-aspartate linked to the ⑀-amino group of L-Lys at position 3 of an acceptor peptide stem (3)(4)(5). The D,D-transpeptidases of E. faecium thus catalyze the formation of D-Ala 4 3 D-Asx-L-Lys 3 cross-links after release of the C-terminal D-Ala 5 of the donor peptide stem (Fig. 1A).
␤-Lactam antibiotics, which are structural analogues of the C-terminal D-Ala 4 -D-Ala 5 end of the peptide stem, act as suicide substrates of the D,D-transpeptidases in an acylation reaction (6). Because transpeptidation is essential to the integrity of the cell wall, these enzymes are the killing target of ␤-lactams (7). D,D-Transpeptidases are multimodular enzymes that combine a C-terminal penicillin-binding domain to an N-terminal glycosyltransferase (class A) or morphogenic (class B) domain (8). Penicillin binding-proteins (PBPs) also include monomodular enzymes with D,D-carboxypeptidase or D,D-endopeptidase activity (8). Among the D,D-transpeptidases of E. faecium, low-affinity PBP5 (class B) is responsible for intrinsic low-level ␤-lactam resistance. In clinical isolates, acquired high-level resistance to these antibiotics is generally associated with increased production of PBP5 or with amino acid substitutions near the conserved motifs of this protein (9 -13). Recently, we searched for other resistance mechanisms and obtained after five selection steps a highly ampicillin-resistant mutant, designated D344M512, or briefly M512, from the hypersusceptible E. faecium D344S that does not harbor the pbp5 gene. Analysis of the peptidoglycan structure by reverse-phase HPLC (RP-HPLC) coupled to mass spectrometry revealed substitution of D-Ala 4 3 D-Asx-L-Lys 3 cross-links (Fig. 1A) by L-Lys 3 3 D-Asx-L-Lys 3 cross-links (Fig. 1B) establishing for the first time that L,Dtranspeptidation could by-pass the essential ␤-lactam-sensitive D,D-transpeptidases (14).
The presence of the unusual L-Lys 3 3 D-Asx-L-Lys 3 crosslinks in M512 implies that an L,D-transpeptidase cleaves the L-Lys 3 -D-Ala 4 peptide bound of a donor peptide stem and links the ␣-carboxyl of its L-Lys 3 to the amino group of the D-Asx residue of an acceptor peptide stem (Fig. 1B). Knowledge of this type of enzyme is limited. In Escherichia coli, cross-links generated by L,D-transpeptidation are present in a minority of the muropeptides (Ͻ8%) (15)(16)(17)(18)(19)(20) but the enzyme responsible for their formation has not been studied. An L,D-transpeptidase activity was detected in crude membrane preparations of Enterococcus hirae (21). This enzyme catalyzed in vitro the exchange of the C-terminal D-Ala residue of the dipeptide N ␣ , N ⑀ -acetyl-L-Lys-D-Ala (Ac 2 -L-Lys-D-Ala) for radioactive D-Ala and, to a lesser extent, for D-Asp, which is the normal in vivo acceptor residue. Although the peptidoglycan was not analyzed (21), the L,D-transpeptidase was thought to form L-Lys 3 3 D-Asx-L-Lys 3 cross-links in vivo.
We have now characterized specific cellular and biochemical aspects of the peptidoglycan metabolism of the highly resistant mutant E. faecium M512 and of the four intermediary mutants M1, M2, M3, and M4. This included the identification of ampicillin-resistant D,D-carboxypeptidase and L,D-transpeptidase activities, the HPLC and mass spectrometry analyses of the peptidoglycan and of the cytoplasmic precursor pools, the examination of cells by electron microscopy, and the study of their proneness to autolysis. The contribution of Lys 3 3 D-Asx-L-Lys 3 cross-links to peptidoglycan synthesis was found to increase with the level of ampicillin resistance, although no variation of the L,D-transpeptidase activity was detected. Selection for high-level resistance led to production of a ␤-lactam-insensitive D,D-carboxypeptidase, indicating that the availability of tetrapeptide donor stems was one of the limiting factors for L,D-transpeptidation.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Parental strain E. faecium D344S is highly susceptible to ampicillin and derives from E. faecium D344 (10) by a spontaneous deletion of pbp5 encoding low-affinity PBP5 (22). E. faecium M1, M2, M3, M4, and M512 are spontaneous mutants of D344S obtained by five successive selection steps on brain heart infusion (Difco) agar containing increasing concentrations of ampicillin as follows. An inoculum of 4 ϫ 10 9 colony forming units of D344S was plated on agar containing 2-fold increasing concentrations of ampicillin (0.06 -4 g/ml). Mutants appeared after 72 h of incubation on plates containing 0.06, 0.12, 0.25, and 0.5 g/ml ampicillin (2-10 colonies per plate, frequency of about 10 Ϫ8 ). The selection procedure was repeated for one of these mutants, designated M1, that was obtained on the highest ampicillin concentration (0.5 g/ml). Second step mutants derived from M1 were observed at the same frequency (about 10 Ϫ8 ) up to 1 g/ml ampicillin. One mutant, M2, growing at the latter concentration, was chosen for further selection steps. Using this approach, mutants M3, M4, and M512 were sequentially obtained from plates containing 2, 4, and 512 g/ml, respectively. The frequencies were about 10 Ϫ8 , 10 Ϫ8 , and 10 Ϫ6 , at the third, fourth, and fifth selection steps, respectively.
Susceptibility Tests-Minimal inhibitory concentrations (MICs) were determined on brain heart infusion agar containing 2-fold dilutions of ampicillin (Bristol-Myers Squibb) (10). Plates were incubated at 37°C for 24 h. MICs were reproducible (four independent experiments) and did not vary after a longer incubation (48 h instead of 24 h).
Analysis of Peptidoglycan Structure-Peptidoglycan was extracted at 100°C with SDS (4%) from exponentially growing bacteria (A 650 ϭ 0.7), purified after treatment with Pronase and trypsin, and digested with lysozyme and mutanolysin (14). The resulting muropeptides were reduced with sodium borohydride and separated by RP-HPLC coupled to mass spectrometry (MS and MS/MS) as previously described (14,23). The linear RP-HPLC gradient (0 -100% B) was applied between 5 and 45 min and elution in buffer B was continued for an additional 5 min (buffer A, 0.05% trifluoroacetic acid in water; buffer B, 0.035% trifluoroacetic acid and 20% acetonitrile in water) at a flow rate of 0.5 ml/min.
Preparation of Cytoplasmic and Membrane Extracts-Bacteria were grown to an A 650 of 0.7, harvested by centrifugation (4,000 ϫ g for 10 min at 4°C), and washed twice in 10 mM sodium phosphate (pH 7.0). Bacteria were disrupted with glass beads in a cell disintegrator (The Mickle Laboratory Engineering Co., Gromshall, United Kingdom) for 2 h at 4°C. The extract was centrifuged (5,000 ϫ g for 10 min at 4°C) to remove cell debris and the supernatant was ultracentrifuged at 40,000 ϫ g for 30 min at 4°C. The supernatant was saved (cytoplasmic fraction) and the pellet was washed twice in 10 mM sodium phosphate buffer (pH 7.0) (membrane fraction). The protein contents were determined with the Bio-Rad protein assay (Bio-Rad) with bovine serum albumin as standard.
Assay of L,D-Transpeptidase Activity-The activity was determined by quantifying the Ac 2 -L-Lys-D-[ 14 C]Ala formed by the exchange reaction between nonradioactive Ac 2 -L-Lys-D-Ala and D-[ 14 C]Ala (21) because D-[ 14 C]Asp was not available. The standard assay (50 l) contained E. faecium membrane or cytoplasmic extracts (75-500 g of proteins), Ac 2 -L-Lys-D-Ala (5 mM), D-[ 14 C]Ala (0.15 mM; 2.0 Gbq/mmol, ICN Pharmaceuticals, Orsay, France), 10 mM sodium cacodylate buffer (pH 6.0), and 0.1% Triton X-100 (v/v). The reaction was allowed to proceed at 37°C and stopped by boiling the samples for 3 min. After centrifugation (10,000 ϫ g, 2 min), 45 l of the supernatant was analyzed by RP-HPLC at 25°C on a -Bondapak C 18 column (3.9 by 300 mm, Waters) with isocratic elution (0.05% trifluoroacetic acid in water/ methanol, 9:1 (v/v)) at a flow rate of 0.5 ml/min. Products were detected by the absorbance at 220 nm and by scintillation with a Radioflow Detector (LB508, PerkinElmer Life Sciences) coupled to the HPLC device.
Preparation and Analysis of the Peptidoglycan Precursors Pools-After growth to an A 650 of 0.7, vancomycin was added to a final concentration of 100 g/ml to block transglycosylation, and incubation was continued for 30 min. Peptidoglycan precursors were extracted with diluted formic acid (25) and analyzed by RP-HPLC as described above.
Expression of the vanY Gene in the Strains of E. faecium-The expression vector pNJ3 (to be described elsewhere) carries a promoter active in enterococci (P 2 ), two origins of replication active in Gramnegative (oriR pUC) and positive (oriR pAM␤1) bacteria, a gentamicin resistance marker, and the origin of transfer of transposon Tn916. Plasmid pJC1 was constructed by inserting the 1.1-kb fragment of transposon Tn1546 into pNJ3 to place the vanY D,D-carboxypeptidase gene (26) under control of the heterologous promoter P 2 . Plasmid pJC1(P 2 vanY) was introduced by electroporation into JH2-2::Tn916 and subsequently transferred by conjugation to E. faecium D344S and M3.
Autolysis Assay-Bacteria from exponential (A 650 ϭ 0.7) and stationary phases were harvested by centrifugation (4,000 ϫ g for 10 min at 4°C), washed three times with ice-cold distilled water, and incubated at 37°C in 0.3 M sodium phosphate (pH 7.0) (28). Turbidity was monitored at 650 nm for 36 h.
Detection of Autolysins-Autolytic enzymes were detected according to the method of Beliveau et al. (29). Bacteria were grown to an A 650 of 0.6 in 10 ml of brain heart infusion broth, washed with distilled water, resuspended in 300 l of phosphate buffer (50 mM, pH 7.0), treated with 20 g of mutanolysin and 40 g of lysozyme, and resuspended in 500 l of denaturing buffer (2% dithiothreitol, 15% sucrose, 3.8% SDS, w/v). Samples were boiled for 3 min and 60 l were applied to a SDSpolyacrylamide gel containing 1 mg/ml dry heat-inactivated E. faecium cells. Renaturation of lytic enzymes was obtained by overnight incubation at 37°C with gentle shaking in 25 mM Tris-HCl (pH 8.0) containing 1% (v/v) Triton X-100.
Electron Microscopy-Bacteria were grown to exponential phase in the absence of ampicillin, harvested at a same A 650 of 0.7, fixed, and stained with 1% uranyl acetate as previously described (30). Ultrathin sections were contrasted with lead nitrate (30).  Table I indicates the proportions of muropeptides 13, E, and G that were the most abundant dimers. Muropeptide 13 was the major dimer generated by D,D-transpeptidation and contained a donor tetrapeptide stem and an acceptor tripeptide stem with a D-asparagine branched on the ⑀-amino group of both lysine residues (Asn-tetra-Asn-tri). Peaks E (Asn-tri-Asn-tri) and G (Asn-tri-Asn-tetra) were the major dimers generated by L,D-transpeptidation.

Muropeptide Composition of the
In the absence of ampicillin (Table I), muropeptides E and G generated by L,D-transpeptidation were detected in small amounts (3.1%) in D344S, indicating that this mode of transpeptidation was pre-existing in the parental strain.
Stepwise increases in the proportion of muropeptides E and G and in the MICs of ampicillin were only detected for the 1st (M1), 4th (M4), and 5th (M512) selection steps. Specifically, the proportion of muropeptides E and G increased from 3.1 (D344S) to 11.4% (M1) at the 1st step, from 11.7 (M3) to 28.7% (M4) at the a Disaccharide-asparagine-tetrapeptide-asparagine-tripeptide-disaccharide. b Disaccharide-asparagine-tripeptide-asparagine-tripeptide-disaccharide. c Disaccharide-asparagine-tripeptide-asparagine-tetrapeptide-disaccharide. d These conditions reproduce those reported in our previous study (14) that were mistakenly described as growth in the absence of antibiotic since the carry over of the antibiotic from the preculture was not taken into account.
4th step, and from 28.7 (M4) to 71.5% (M512) at the 5th step. Selection led to parallel large increases of the ampicillin MICs at each of these steps (8-, 64-, and Ն32-fold, respectively). In contrast, marginal increases of the MICs (2-fold) were observed for the 2nd and 3rd selection steps and the muropeptide composition of mutants M1, M2, and M3 were similar. These observations indicate that three of the five selection steps led to increased L,D-transpeptidation to the detriment of D,Dtranspeptidation. Activation of the L,D-transpeptidation pathway at these steps was associated with large increases in the ampicillin-resistance level.
The peptidoglycan structure analysis was repeated for the strains grown in the presence of subinhibitory concentrations of ampicillin to test the effect of PBP inhibition (Table I). For D344S, M1, M2, and M3, the concentration of ampicillin added to culture medium corresponded to the maximum concentration allowing growth. Partial inhibition of the D,D-transpeptidases by ampicillin increased the proportion of dimers generated by L,D-transpeptidation in D344S (from 3.1 to 25%, 8-fold) and in mutants M1, M2, and M3 (6-fold). However, the PBPs remained essential targets in these mutants because higher concentrations of ampicillin inhibited growth. For M4 and M512, the D,D-transpeptidation pathway was almost completely inhibited by ampicillin at 0.5 g/ml, which corresponds to the subinhibitory concentration tested for M3. Under these conditions, and in contrast to M3, the proportion of muropeptides generated by L,D-transpeptidation reached 82.  (21). L,D-Transpeptidase activity was detected in mem-brane preparations of D344S (Fig. 2), which contained minor amounts of dimers with a L-Lys 3 3 D-Asx-L-Lys 3 cross-link in its peptidoglycan. This activity was similar for D344S and M512 (23 Ϯ 4 and 32 Ϯ 6 pmol/min/mg of protein, respectively). The concentrations of ampicillin required to inhibit the L,Dtranspeptidase activity by 50% (IC 50 ) were also similar for D344S and M512 (about 105 and 110 g/ml, respectively). Residual activity (about 15-25%) was detected at 1600 and 3200 g/ml for both strains. Thus, increased synthesis of L-Lys 3 3 D-Asx-L-Lys 3 cross-links in mutant M512 was not associated with increased L,D-transpeptidase activity. Neither the L,D-transpeptidase produced by susceptible strain D344S nor that of M512 were inhibited by low concentrations of ampicillin and the IC 50 of the antibiotic were similar for the two enzyme preparations. D,D-Carboxypeptidase Activity-In the presence of ampicillin (20 g/ml), D,D-carboxypeptidase activity was detected only in membrane preparations from M4 and M512 (Table II). This enzyme was not inhibited by ampicillin at 2000 g/ml (data not shown). Membrane preparations from D344S, M1, M2, and M3 contained a 10-fold lower D,D-carboxypeptidase activity that was totally (Ͼ95%) inhibited by ampicillin at 20 g/ml. These results indicate that the fourth selection step, which generated mutant M4, resulted in high-level production of a ␤-lactaminsensitive D,D-carboxypeptidase in addition to the D,D-carboxypeptidase activity of the putative monofunctional PBPs.
Expression was originally detected in transposon Tn1546 that confers glycopeptide resistance in enterococci (26). Plasmid pJC1 was constructed by inserting the vanY gene under the control of the P 2 promoter of the shuttle vector pNJ3 to test the influence of elevated D,D-carboxypeptidase activity on ␤-lactam resistance. Introduction of pJC1(P 2 vanY) into mutant M3 led to an increase in the MIC of ampicillin (from 2 to 256 g/ml). A similar increase in the MIC (from 2 to 128 g/ml) was obtained at the fourth selection step that generated M4 from M3 (Table I). The vector alone had no effect on the level of resistance. Thus, elevated D,D-carboxypeptidase activity was responsible for increased resistance observed at the fourth selection step.
Plasmid pJC1(P 2 vanY) did not increase the MIC of ampicillin in D344S. Selection of ampicillin-resistant mutants from D344S/pJC1(P 2 vanY) resulted in a highly resistant mutant in three steps (MIC, 64 g/ml) instead of the four steps necessary to obtain M4 (MIC, 128 g/ml) from D344S. These observations suggest that production of VanY did not by-pass the requirement for mutations selected at the 1st, 2nd, and 3rd selection steps.
Binding of ␤-Lactams to PBPs-PBP labeling with benzylpenicillin revealed similar SDS-PAGE patterns for D344S and M512, indicating that their level of production was not modified (data not shown). Based on the competition assay, ampicillin at 4 g/ml saturated all PBPs of D344S and M512. Partial saturation at lower drug concentrations showed that the affinity of the PBPs were not modified in the highly resistant mutant M512.
Cell Autolysis and Detection of Autolysins-Autolysis of D344S cells collected from exponential and stationary phases led to a 50% decrease of the A 650 after 9 and 24 h of incubation at 37°C in phosphate buffer, respectively. In contrast, no significant autolysis of M512 was observed after 36 h. Crude extracts from D344S and M512 contained two major autolysins active on heat-killed E. faecium D344S (data not shown). These enzymes were not active on heat-killed M512 cells. These ob-servations suggested that the mutant M512 was not prone to autolysis because its peptidoglycan containing cross-links generated by L,D-transpeptidation was no longer well recognized by the major E. faecium autolysins.
Growth Rate and Abnormal Morphology-The stepwise increase in ampicillin resistance was accompanied by a decrease in growth rate. In the absence of ampicillin in the culture medium, the generation time was 27, 48, 60, 72, 84, and 90 min for D344S, M1, M2, M3, M4, and M512, respectively. For M512, the generation time increased from 90 to 180 min when the ampicillin concentration increased from 750 to 4000 g/ml. In comparison with parental strain D344S (Fig 3a), electron microscopy of thin sections of M1 and M2 grown in antibiotic-free medium showed no difference in morphology (data not shown). For M3, thickening of the cell wall and abnormalities of the septa were present in about 30% of the cells (Fig. 3b). For M4 and M512, severe abnormalities were observed in more than 80% of the cells, including cell wall thickening and formation of cell aggregates resulting from disordered septation (Fig. 3,  c and d). DISCUSSION In this report, we show that the level of ampicillin resistance in the mutants derived from E. faecium D344S is determined by a balance between the D,D-and L,D-transpeptidation pathways for peptidoglycan cross-linking (Table I). The 1st, 4th, and 5th selection steps that led to mutants M1, M4, and M512, respectively, resulted in large increases in the proportion of muropeptides containing the L-Lys 3 3 D-Asx-Lys 3 cross-links generated by L,D-transpeptidation. Large increases of the ampicillin MICs were observed at these steps.
Partial inhibition of the D,D-transpeptidases by ampicillin  1 mM) in the presence or absence of ampicillin at 20 g/ml. UDP-MurNAc-tetrapeptide was separated from UDP-MurNAc-pentapeptide by RP-HPLC. The ␤-lactam sensitive and insensitive activities were mainly present in the membrane fractions, and only low levels were detected in the cytoplasmic fractions (data not shown).  (less or equal to 0.5 g/ml) increased the proportion of muropeptides with an L-Lys 3 3 D-Asx-Lys 3 cross-link in all strains. Higher drug concentrations inhibited growth of D344S, M1, M2, and M3, indicating that at least one PBP remained essential in these strains. In contrast, the PBPs were not contributing to peptidoglycan cross-linking in M512 as growth in the presence of a saturating concentration of ampicillin (32 or 1000 g/ml) led to exclusive synthesis of L-Lys 3 3 D-Asx-Lys 3 cross-links. In the absence of ampicillin, the two types of crosslinks were detected, indicating that the L,D-and D,D-transpeptidation pathways were functional in all strains although their relative contribution varied. Several factors that could affect this balance were examined, revealing that activation of the resistance pathway was not associated with any variation of the L,D-transpeptidase activity or of the PBPs, but to production of a D,D-carboxypeptidase that generated the tetrapeptide donor stems for L,D-transpeptidation.
The L,D-transpeptidase activity measured by the exchange reaction was similar in D344S and M512 (Fig. 2), indicating that increased production of the enzyme was not required for by-pass of the D,D-transpeptidases. The L,D-transpeptidase activity was only partially inhibited by high concentrations of ampicillin. The drug concentrations required for 50% inhibition were also similar for the enzyme present in membrane preparations from D344S and M512 (IC 50 of about 105 and 110 g/ml, respectively). Thus, emergence of high-level resistance to ampicillin did not appear to involve any alteration of the L,D-transpeptidase. The following observations may account for the apparent discrepancy between the IC 50 (110 g/ml) and the MIC (Ͼ2,000 g/ml) observed for M512. The generation time of this strain increased from 90 to 180 min when the ampicillin concentration increased from 750 to 4,000 g/ml, indicating that the drug significantly slowed growth despite the high MIC. A residual L,D-transpeptidase activity was detected in vitro at ampicillin concentrations of 1600 (25%) and 3,200 g/ml (15%) in membrane preparations of M512. This residual activity could be sufficient to sustain bacterial growth in M512 although at a lower rate. As previously discussed (14,21,31), the D,D-transpeptidase and L,D-transpeptidase are not expected to be inhibited by ␤-lactams by the same mechanism. The drugs are suicide substrates of the D,D-transpeptidases because the ␤-lactam ring is structurally related to the D-Ala-D-Ala moiety of peptidoglycan precursors (32). The L,D-transpeptidases, which cleave the L-Lys-D-Ala rather than D-Ala-D-Ala peptide bond, are therefore not expected to be acylated by the same mechanism.
Penicillin-binding studies did not reveal any modification of the PBP patterns in M512 (data not shown). This observation implies that in this strain the resistance pathway is dominant as the D,D-transpeptidase appeared to remain functional based on ␤-lactam acylation. Competition experiments indicated that the acylation of PBPs by ampicillin was not affected by the five-step selection for resistance to this antibiotic because all PBPs were saturated by low concentrations of ampicillin (Ͻ4 g/ml). This observation implies that the ␤-lactam-sensitive D,D-transpeptidase and D,D-carboxypeptidase activities of the PBPs were not playing any essential role in the high level of resistance expressed by M512 (MIC Ͼ 2000 g/ml). In Staphylococcus aureus, methicillin resistance mediated by class B PBP2A requires the glycosyltransferase activity of class A PBP2 (33). Site-directed mutagenesis of the active-site serine residue of PBP2 showed that the glycosyltransferase module of this protein can function in the absence of its catalytically active D,D-transpeptidase module (33). In our system, peptidoglycan synthesis may similarly involve cooperation between the L,D-transpeptidase and the glycosyltransferase module of a class A PBP.
The fourth selection step that generated mutant M4 led to production of a ␤-lactam-insensitive D,D-carboxypeptidase (Table II) that was active in vivo as shown by the accumulation of UDP-MurNAc-tetrapeptide in the cytoplasm of M4 and M512 (Table III). Selection may have activated a cryptic gene encoding a metallo-D,D-carboxypeptidase because this type of enzyme is not inhibited by ␤-lactams (26,34,35) and no modification of the patterns of the PBP was observed. In agreement, production of the heterologous D,D-carboxypeptidase VanY increased the level of ampicillin resistance in mutant M3. This effect was not observed in D344S, suggesting that elevated D,D-carboxypeptidase activity can only increase the level of resistance after modification of other unknown function(s). High-level accumulation of UDP-MurNAc-tetrapeptide (Table III) suggests that the enzyme hydrolyzes the C-terminal D-Ala residue of UDP-MurNAc-pentapeptide in vivo. However, because formation of lipid intermediate I is reversible (1) and translocation of this intermediate may also be reversible (36), hydrolysis of pentapeptide stems at the outer surface of the cytoplasmic membrane cannot be excluded, as previously discussed for a D,D-carboxypeptidase involved in glycopeptide resistance (36).
Taking in account all these observations, production of the ␤-lactam-insensitive D,D-carboxypeptidase is expected to increase synthesis of L-Lys 3 3 D-Asx-Lys 3 to the detriment of D-Ala 4 3 D-Asx-Lys 3 cross-links by two mechanisms (Fig. 4). First, hydrolysis of the C-terminal D-Ala 5 reduces D,D-transpeptidation because the PBPs require a pentapeptide donor stem. Second, the D,D-carboxypeptidase by generating the tetrapeptide stem donor substrate acts as a new source to increase L,D-transpeptidation. The large increase in the MIC of ampicillin observed at the fourth selection step indicates that the availability of tetrapeptide stems is one of the limiting factors for activation of the resistance pathway. In low-level resistant mutants, the majority of the pentapeptide monomers were used in the D-Ala 4 3 D-Asx-Lys 3 cross-link by the D,D-transpeptidases. Because only small amounts of tetrapeptide stems were generated by the D,D-carboxypeptidase activity of monofunctional PBPs, only a few L-Lys 3 3 D-Asx-Lys 3 cross-links were formed. Because these latter PBPs were inhibited by ampicillin, the L,D-transpeptidation pathway remains also susceptible to the drug in the absence of another source of tetrapeptide substrate (Table II and Fig. 4).
By-pass of D,D-transpeptidation is a remarkable example of the flexibility of peptidoglycan synthesis although it was associated with severe defects in peptidoglycan metabolism as shown by increased generation times (up to 4-fold for M512) and cell wall abnormalities (Fig. 3). Impaired activity of the autolysins could be involved in these defects because peptidoglycan generated by L,D-transpeptidation was a poor substrate for hydrolytic enzymes both in vivo and in vitro. Peptidoglycan polymerization is thought to involve multienzyme complexes that include glycosyltransferases, D,D-transpeptidases, and hydrolases (20). Therefore, it is perhaps not surprising that acquisition of resistance led to impaired growth and required multiple mutations that did not directly affect the catalytic activity of the transpeptidases.