Novel Mechanism of b -Lactam Resistance Due to Bypass of DD-Transpeptidation in Enterococcus faecium *

The peptidoglycan structure of in vitro selected ampi-cillin-resistant mutant Enterococcus faecium D344M512 and of the susceptible parental strain D344S was determined by reverse phase high performance liquid chromatography and mass spectrometry. The muropeptide monomers were almost identical in the two strains. The substantial majority (99.3%) of the oligomers from the susceptible strain D344S contained the usual D -alanyl 3 D -asparaginyl (or D -aspartyl)- L -lysyl cross-link ( D -Ala 3 D - Asx- L -Lys) generated by b -lactam-sensitive DD-transpep-tidation. The remaining oligomers (0.7%) were produced by b -lactam-insensitive LD-transpeptidation, because they contained L -Lys 3 D -Asx- L -Lys cross-links. The mu- ropeptide oligomers of the ampicillin-resistant mutant D344M512 contained only these L -Lys 3 D -Asx- L -Lys cross- links indicating that resistance was due to the bypass of the b -lactam-sensitive DD-transpeptidation reaction. The discovery of this novel resistance mechanism indicates that DD-transpeptidases 21). Muropeptides were separated by reversed-phase high performance liquid chromatography (HPLC) and identified by mass spectrometry (MS) as described (21, 22). Fragmentation Analysis of Selected Muropeptides— The structure of muropeptides B and E of D344M512 was determined by MS/MS per- formed on singly and doubly charged molecules with argon as collision gas (21) using a Waters 600 MS-HPLC pump system and a Waters PD1991 liquid chromatograph with a diode array detector system cou-pled to a Finningam TSQ7000 triple quadrupole mass spectrometer (San Jose, CA). Muropeptides 11 and 13 of D344S and muropeptides 8, C, D, G, and H of D344M512 were purified by reverse-phase high pressure liquid chromatography and analyzed by MS/MS using the nanoelectrospray source kit for the Finnigam TSQ 7000 Protonona A/S (San Jose, CA). Samples were resuspended in 5 m l of H 2 O, 30% aceto-nitrile, 0.045% trifluoroacetic acid and sprayed at 0.6–0.7 kV. Under these conditions, MS/MS was done on the doubly charged protonated molecules using argon (2.5 millitorr) as collision gas at 28–38 eV. Scan accumulation was done with ; 150 five-min scans.

The peptidoglycan of Escherichia coli is generated by polymerization of a precursor composed of N-acetylglucosamine (GlcNAc) 1 and N-acetylmuramic acid (MurNAc) substituted by a L-alanyl-D-isoglutamyl-meso-diaminopimelyl-D-alanyl-Dalanine pentapeptide stem (L-Ala 1 -D-iGlu 2 -meso-A2 pm 3 -D-Ala 4 -D-Ala 5 ) (1). The final steps of peptidoglycan synthesis involve polymerization of the glycan strands by glycosyltransferases and cross-linking of the peptide stems by DDtranspeptidases. The latter enzymes catalyze formation of a peptide bond between the ␣-carboxyl of D-Ala at the fourth position of a donor stem and the ⑀ amino group of meso-A2 pm at the third position of an acceptor stem generating a D-Ala 4 3 meso-A2 pm 3 cross-link (2).
The first step of the transpeptidation reaction leads to the release of the C-terminal D-Ala 5 of the donor peptide stem and to the formation of a covalent adduct between the penultimate residue (D-Ala 4 ) and a conserved catalytic serine residue of the DD-transpeptidases (2,3). Antibiotics of the ␤-lactam class, such as penicillin and ampicillin, are structural analogs of the C-terminal D-Ala 4 -D-Ala 5 end of peptidoglycan precursors and act as suicide substrates in a similar acylation reaction (4). The second step of the transpeptidation reaction results in a crosslinking and release of the DD-transpeptidases. In contrast, acylation of the DD-transpeptidases by ␤-lactams is nearly irreversible. The DD-transpeptidases are the killing target of the ␤-lactams, because transpeptidation is essential to maintain the integrity of the cell wall (2).
The D-Ala 3 meso-A2 pm 3 cross-links generated by the DDtranspeptidases are prevalent in the peptidoglycan of E. coli although minor meso-A2 pm 3 3 meso-A2 pm 3 cross-links have been detected in the exponential (ϳ2%) and stationary (ϳ4%) phases of growth (5)(6)(7). The enzymes generating the minor meso-A2 pm 3 3 meso-A2 pm 3 cross-links have not been identified. By analogy with DD-transpeptidases, these putative LDtranspeptidases are thought to cleave the C-terminal D-Ala 4 of a donor tetrapeptide stem peptide before linking the ␣-carboxyl of meso-A2 pm 3 to the ⑀-amino group of meso-A2 pm 3 of an acceptor stem peptide (8,9). The ␤-lactam ring does not contain any LD-peptide bond indicating that LD-transpeptidases do not belong to the family of penicillin-binding proteins (4,10). In agreement with this notion, LD-carboxypeptidases, which cleave the C-terminal D-Ala 4 residue of tetrapeptide stems, are not acylated by ␤-lactams (11).
The overall structure and mode of synthesis of peptidoglycan is conserved in eubacteria, although variations have been detected, in particular in the sequence of the peptide stem. In the Gram-positive bacteria Enterococcus faecium and Lactobacillus casei, D-iGlu at the second position is amidated, meso-A2 pm at the third position is replaced by L-Lys, and the ⑀-amino group of the latter amino acid is substituted by D-Asn or D-Asp (12)(13)(14). Consequently, the DD-transpeptidases of E. faecium and L. casei catalyze formation of D-Ala 4 3 D-Asx-L-Lys 3 cross-links.
Several lines of evidence indicate that the DD-transpeptidase targets of ␤-lactam antibiotics are ubiquitous in eubacteria that produce peptidoglycan. First, penicillin-binding proteins have been reproducibly detected based on acylation with radiolabeled penicillin (15,16). More importantly, analyzes of peptidoglycan precursors and of the corresponding biosynthetic enzymes have shown that the C-terminal DD-configuration of residues at the C-terminal positions 4 and 5 of the peptide stem is uniformly conserved (12,17). This implies conservation of the structural analogy between ␤-lactams and the donor substrate of the DD-transpeptidases. Finally, analyses of peptidoglycan structure in a wide range of bacteria have invariably revealed that cross-linking was mainly or exclusively generated by DD-transpeptidase activities that catalyzed peptide bond formation between D-Ala at the fourth position of a donor stem and the amino group of the acceptor (12). In this report, we showed that emergence of high level ampicillin resistance in an in vitro selected mutant of E. faecium was associated with replacement of D-Ala 4 3 D-Asx-L-Lys 3 by L-Lys 3 3 D-Asx-L-Lys 3 cross-links establishing for the first time that bacteria can bypass the requirement for ␤-lactam-sensitive DD-transpeptidase activity.

EXPERIMENTAL PROCEDURES
Strains and Growth Conditions-E. faecium D344S is highly susceptible to ampicillin and derives from E. faecium D344 (18) by a spontaneous deletion of pbp5 encoding the low-affinity penicillin-binding protein 5. 2 This strain was chosen to avoid selection of ampicillin resistance because of penicillin-binding protein 5 alterations (19). E. faecium D344M512 is a spontaneous mutant of D344S obtained by five serial selection steps on agar containing increasing concentrations of ampicillin. All cultures were performed at 37°C in brain heart infusion (Difco Laboratories, Detroit, MI) agar or broth without shaking. Minimal inhibitory concentrations of ampicillin (Bristol-Myers, Paris, France) were determined by the agar dilution method (20).
Peptidoglycan Structure Analysis-Peptidoglycan was obtained after 4% sodium dodecyl sulfate treatment at 100°C. The pellet was treated with Pronase (200 g/ml) for 16 h at 37°C in 10 mM Tris-Cl, pH 7.4, and then after centrifugation (40,000 ϫ g, 4°C), the pellet was treated with trypsin (200 g/ml) for 16 h at 37°C in 20 mM potassium phosphate buffer (pH 7.8) to remove the contaminating proteins from the proteaseresistant proteoglycan. The pellet was washed twice with water and treated with lysozyme (200 g/ml) and mutanolysin (250 g/ml) for 16 h at 37°C in 1 ml of 25 mM potassium phosphate buffer (pH 6.5), 10 mM MgCl 2 (14,21). Muropeptides were separated by reversed-phase high performance liquid chromatography (HPLC) and identified by mass spectrometry (MS) as described (21,22).
Fragmentation Analysis of Selected Muropeptides-The structure of muropeptides B and E of D344M512 was determined by MS/MS performed on singly and doubly charged molecules with argon as collision gas (21) using a Waters 600 MS-HPLC pump system and a Waters PD1991 liquid chromatograph with a diode array detector system coupled to a Finningam TSQ7000 triple quadrupole mass spectrometer (San Jose, CA). Muropeptides 11 and 13 of D344S and muropeptides 8, C, D, G, and H of D344M512 were purified by reverse-phase high pressure liquid chromatography and analyzed by MS/MS using the nanoelectrospray source kit for the Finnigam TSQ 7000 Protonona A/S (San Jose, CA). Samples were resuspended in 5 l of H 2 O, 30% acetonitrile, 0.045% trifluoroacetic acid and sprayed at 0.6 -0.7 kV. Under these conditions, MS/MS was done on the doubly charged protonated molecules using argon (2.5 millitorr) as collision gas at 28 -38 eV. Scan accumulation was done with ϳ150 five-min scans.

RESULTS
Characteristics of D344S and D344M512-The minimal inhibitory concentration of ampicillin for the parental strain D344S was 0.06 g/ml. Growth of the ampicillin-resistant mutant D344M512 was not inhibited at the highest drug concentration tested (minimal inhibitory concentrations Ͼ 2000 g/ml).
Muropeptide Composition of Peptidoglycan from D344S and D344M512-The peaks in the HPLC muropeptide profiles of peptidoglycan from D344S (Fig. 1A) and D344M512 (Fig. 1B) were identified, and their relative amounts were determined (Table I). Monomers were eluted first between 49 and 69 min followed by oligomers (68 -84 min). The monomers accounted for the same proportion of all muropeptides in the two strains (37-38%), and the monomer profiles were almost identical (peaks 1-10). The molecular mass and deduced structure of these monomers were in accordance with previous analyses of different strains of E. faecium (13,14). The oligomer profile of the parental strain D344S was also in agreement with previous studies (13,14). In contrast, the peptidoglycan of D344M512 contained a novel oligomer muropeptide species (peaks A-M). The muropeptide profiles of D344M512 grown in the absence of antibiotic (Fig. 1B) or in the presence of 32 g/ml ampicillin (data not shown) were very similar indicating that cross-linking of these novel oligomers was not inhibited by the drug. We focused on the determination of the structure of the most abundant peaks of D344M512 and comparison with the oligomers of D344S. This required a definitive assignment of the structure of the main D344S dimer by MS/MS (peak 13) and MS analysis of peaks 15, 17, 18, 19, 21, and 22 that have not been resolved in previous analyses.

Structure of the Main Oligomers of Parental Strain D344S-
The main dimer of D344S, peak 13, accounted for 21.7% of all peaks and had a molecular mass of 1932.5 (Table I). Nanoelectrospray MS/MS analysis (Fig. 2) indicated that muropeptide 13 was a dimer containing 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) (Fig.  3). The cross-link was of the expected D-Ala 4 3 D-Asn-L-Lys 3 type. The structure of dimers 11, 12, 14, and 16 and of trimer 20 (Table I and Fig. 3) has been reported in E. faecium (13,14). The structures of the other previously unidentified peaks (15, 17, 18, 19, 21, and 22) were deduced from their retention times and molecular masses ( Fig. 1A and Table I). Muropeptides differing by an increase of 42 or 84 mass units from other dimers were considered to harbor one or two O-acetylated N-acetyl muramyl residues (21). Four dimers (15,17,19, and 21) that contained an O-acetylated N-acetylmuramyl residue have been previously identified in L. casei (21). The cross-link in all these oligomers (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) was that expected to be found after DD-transpeptidation involving the cleavage of C-terminal D-Ala 5 of the donor pentapeptide and formation of a cross-link between the penultimate D-Ala 4  Purified peptidoglycan was digested with lysozyme and mutanolysin, and muropeptides were resolved by reverse-phase high pressure liquid chromatography. Numbers and letters correspond to peaks identified in Table I. work and previous studies (13,14), it can be estimated that at least 98% of all the oligomer muropeptides present in D344S harbored the D-Ala 4 3 D-Asx-L-Lys 3 cross-link generated by DD-transpeptidation.

Structure of Dimers E, C, B, and A of Mutant D344M512-
The most prevalent dimer of D344M512, muropeptide E (retention time 73.6 min, Fig. 1B), accounted for 17.6% of all peaks and had a molecular mass of 1861.8 (Table I). The molecular mass was consistent with a single dimer structure (Asn-tri-Asntri) containing a novel cross-link of the L-Lys 3 3 D-Asx-L-Lys 3 type (Fig. 3). This structure was confirmed by MS/MS (Fig. 4).
Muropeptide C (M r 1862.7) eluted 1.5 min before muropeptide E (M r 1861.8). MS/MS analysis (data not shown) indicated that the difference of one mass unit resulted from the presence of an aspartate instead of an asparagine residue on the ⑀-amino group of the L-lys 3 of the donor peptide stem (Asp-tri-Asn-tri) (Fig. 3). This observation and previous analyses of E. faecium and L. casei peptidoglycans indicated that aspartate-branched muropeptides eluted before the asparagine-containing analogs (14,21).
Muropeptide B (M r 1747.8) was the second most abundant dimer (8%). MS/MS (Fig. 5) indicated that this muropeptide was a tri-Asn-tri dimer containing the unusual L-Lys 3 3 D-Asn-L-Lys 3 cross-link and an unsubstituted L-Lys 3 in the donor stem peptide (Fig. 3). Dimer A (M r 1748.7) was most probably the D-aspartate-containing analog of dimer B because it eluted 1.2 min before dimer B and differed by one mass unit from dimer B.
Structure of Dimers D and G of D344M512 Compared with That of Dimers 11 and 13 of D344S-Comparison of the muropeptides from D344S and D344M512 revealed the presence of dimers that had similar molecular masses but did not elute exactly with the same retention times. For example, muropeptides 13 (see above) and G present in D344S and D344M512, respectively, had a molecular mass of 1932.5 and 1932.0 and a retention time of 75.4 and 76.4 min (Table I, Fig. 1). For this pair of muropeptides, that contained two D-asparagine residues (Table I), the molecular mass was compatible with two alternative structures containing either a donor tetrapeptide stem and an acceptor tripeptide stem with a D-Ala 4 3 D-Asn-L-Lys 3 cross-link or a donor tripeptide stem and an acceptor tetrapeptide stem with a L-Lys 3 3 D-Asn-L-Lys 3 cross-link (Fig. 3). The two structures were differentiated on the basis of the presence or absence of a C-terminal alanine residue. The complete anal- The molecular composition was determined by reverse-phase HPLC, mass spectrometry, and MS/MS. The relative amounts of the compounds are expressed as percentages calculated from the UV absorbance of peaks in HPLC elution profiles of D344S (Fig. 1A) and D344M512 (Fig. 1B)  a The values are presented as a percentage of the sum of all peaks presented in the Table. b Muropeptide peak designation as in Fig. 1.  Fig. 1. a, % of total muropeptides; b, Rt, retention time (min); c, assignment of Asp to either stem peptide is arbitrary. of one N-acetylglucosamine residue (m/z 1730.5). Similarly, the peak at m/z 1437.9 was because of the loss of one alanine residue at the C-terminal end of the tetrapeptide stem after the loss of both N-acetylglucosamine residues (m/z 1527.3). The loss of the alanine residue was also found from another major fragmentation product (m/z 1178.7) yielding ions at m/z 1089.6 (Fig. 6). Thus, muropeptide G had a cross-link through an asparagine between two lysine residues (Fig. 3). In contrast, the fragmentation profile of muropeptide 13 of D344M512 did not reveal the presence of any structure generated by the cleavage of a C-terminal alanine residue confirming that the C-terminal D-alanine of the donor tetrapeptide stem was participating in the formation of a D-Ala 4 /z 1343.7). The loss of both GlcNAc residues, one MurNAc residue, and NH 3 led to an ion at m/z 1048.2; the loss of an additional alanine and isoglutamine residues gave an ion at m/z 847.3. The loss of both GlcNAc, one MurNAc, one alanine, one isoglutamine, and one lysine gave an ion at m/z 737.5; the loss of an additional asparagine residue gave an ion at m/z 625.2. The peak at m/z 261.3 corresponds to dipeptide asparaginyl-lysine. The peak at m/z 328.5 corresponds to the tripeptide alanyl-isoglutaminyl-lysine. 1818.7) and D of D344M512 (M r 1818.0) were found to correspond to a tetra-Asn-tri dimer with a D-Ala 4 3 D-Asn-L-Lys 3 cross-link and to a tri-Asn-tetra dimer with a L-Lys 3 3 D-Asn-L-Lys 3 cross link, respectively (data not shown). above), most likely dimers F and G differed only by the presence of a D-asparagine or a D-aspartate residue. Finally, MS/MS analysis of muropeptide H revealed an Asn-tri-Asn-penta dimer containing the L-Lys 3 3 D-Asn-L-Lys 3 cross-link-present in the other dimers of D344M512 (data not shown).

Analogs of Dimer G-Dimer
Overview of the Oligomers from D344S and D344M512-All the oligomers of mutant D344M512 that were analyzed contained the unusual L-Lys 3 3 D-Asx-L-Lys 3 cross-link instead of the D-Ala 4 3 D-Asx-L-Lys 3 cross-link present in the parental susceptible strain. Despite this major difference, the crosslinked peptidoglycan generated by DD-transpeptidation in D344S or by LD-transpeptidation in D344M512 displayed striking similarities. In particular, tripeptide stems were detected at the acceptor position in the majority of the oligomer muropeptides from both strains (Table I and Fig. 3). Moreover, L-Lys 3 in the muropeptides of both strains was more frequently substituted by D-Asn than by D-Asp. Finally, as mentioned above, the extent of the cross-link was similar in D344S and D344M512 because oligomers, mainly dimers, represented 62-63% of the muropeptides (Table I).
Further Attempts to Find Common Muropeptides Oligomers in D344S and D344M512-As shown in Fig. 1 and Table I, all the major muropeptide oligomers of D344M512 that were produced in sufficient amount for structural analysis contained the unusual L-Lys 3 3 D-Asx-L-Lys 3 cross-link. Further analysis of very minor peaks did not provide any evidence for the presence of muropeptide containing the usual D-Ala 4 3 D-Asx-L-Lys 3 cross-link present in D344S (data not shown). In contrast, further analysis of minor peaks of D344S revealed two dimers (A and B) that contained the unusual L-Lys 3 3 D-Asx-L-Lys 3 cross-link. These peaks were very minor components of the peptidoglycan of D344S accounting for only 0.7% of the total muropeptides (Table I and Fig. 1A). Thus, the unusual L-Lys 3 3 D-Asx-L-Lys 3 cross-links preexisted in the parental strain.
Further Characterization of Monomer Muropeptides-Five monomers (peaks 4, 5, 9, and 10) present in the muropeptide profiles of D344S and D344M512 had not been fully characterized in previous studies of the E. faecium peptidoglycan. MS/MS analysis indicated that muropeptide 5 was a monomer containing a tripeptide stem with a D-asparagine branched on L-Lys 3 , whereas muropeptide 4, which eluted 1.2 min before muropeptide 5 and differed by one mass unit, was the D-aspartate-containing analog of muropeptide 5. Based on the 42 mass unit difference, muropeptide 10 differed from muropeptide 5 by O-acetylation of the MurNAc residue. In agreement, peak 10 had the same retention time and molecular mass as the Oacetylated tripeptide monomer characterized by MS/MS in L. casei (21). According to the same criteria, muropeptide 9 was the D-aspartate-and O-acetyl-containing analogue of muropeptide 5. Finally, MS/MS analysis indicated that muropeptide 8, only present in a small amount, contained a D-asparaginesubstituted pentapeptide stem (data not shown).

DISCUSSION
Analysis of the peptidoglycan of the ampicillin-resistant mutant D344M512 revealed the presence of a novel type of crosslink that connected two lysine residues (L-Lys 3 3 D-Asx-L-Lys 3 ) instead of the usual cross-link generated by the DD-transpeptidases (D-Ala 4 3 D-Asx-L-Lys 3 ) in susceptible strains. Qualitatively, this finding is not entirely novel, because replacement D-Ala 4 3 meso-A2 pm 3 by meso-A2 pm 3 3 meso-A2 pm 3 has been previously reported in E. coli for a minority (Ͻ4%) of the cross-links in the stationary phase of growth (1,(5)(6)(7). Quantitatively, the results obtained for D344M512 were unprecedented because the L-Lys 3 3 D-Asx-L-Lys 3 cross-link was present in 100% of the muropeptide oligomers that were analyzed (Table I and Fig. 1). As detailed under "Results" extensive analysis was performed to rule out the presence of the D-Ala 4 3 D-Asx-L-Lys 3 cross-links even in minor muropeptide oligomer species of D344M512. The absence of this type of structure indicates that DD-transpeptidase activity was playing no role in peptidoglycan cross-linking. In agreement, the increase in the proportion of L-Lys 3 3 D-Asx-L-Lys 3 cross-links from 0.7% in D344S to 100% in D344M512 was associated with a Ͼ3 ϫ 10 4 -fold increase in the minimal inhibitory concentration of ampicillin, and the resistant mutant was not inhibited by the highest drug concentration tested (2000 g/ml). Moreover, the muropeptide profile of resistant mutant D344M512 was unaffected by ampicillin (data not shown) indicating that synthesis of the novel cross-link was unaffected by the drug. Taken together, these results indicate that the requirement for the essential ␤-lactam-sensitive DD-transpeptidase activity was bypassed by a ␤-lactam-insensitive LD-transpeptidase activity that preexisted to a low extent in the susceptible parental strain.
The bypass of the essential DD-transpeptidases is a novel resistance mechanism to ␤-lactam antibiotics. Previously described bacterial strategies for acquisition of ␤-lactam resistance include decreased outer membrane permeability in Gramnegative bacteria (23), production of ␤-lactamases that inactivate the drugs (24), and production of DD-transpeptidases with reduced affinity for ␤-lactams (16,19,25). Despite the wide distribution of these resistant mechanisms in pathogenic bacteria, ␤-lactams are still the most broadly used antibiotic class because several strategies have been successfully developed to overcome resistance, including modification of ␤-lactam structure to prevent hydrolysis by ␤-lactamase, association of ␤-lactams with ␤-lactamase inhibitors, and design of new drugs that display increased affinity for DD-transpeptidases from resistant bacteria. The emergence of resistance by LD-transpeptidation bypass mechanism described in this study is worrisome because it is expected to confer cross-resistance to all ␤-lactams.