OUP user menu

Evolution of peptidoglycan biosynthesis under the selective pressure of antibiotics in Gram-positive bacteria

Jean-Luc Mainardi , Régis Villet , Timothy D. Bugg , Claudine Mayer , Michel Arthur
DOI: http://dx.doi.org/10.1111/j.1574-6976.2007.00097.x 386-408 First published online: 1 March 2008

Abstract

Acquisition of resistance to the two classes of antibiotics therapeutically used against Gram-positive bacteria, the glycopeptides and the β-lactams, has revealed an unexpected flexibility in the peptidoglycan assembly pathway. Glycopeptides select for diversification of the fifth position of stem pentapeptides because replacement of d-Ala by d-lactate or d-Ser at this position prevents binding of the drugs to peptidoglycan precursors. The substitution is generally well tolerated by the classical d,d-transpeptidases belonging to the penicillin-binding protein family, except by low-affinity enzymes. Total elimination of the fifth residue by a d,d-carboxypeptidase requires a novel cross-linking enzyme able to process the resulting tetrapeptide stems. This enzyme, an l,d-transpeptidase, confers cross-resistance to β-lactams and glycopeptides. Diversification of the side chain of the precursors, presumably in response to the selective pressure of peptidoglycan endopeptidases, is controlled by aminoacyl transferases of the Fem family that redirect specific aminoacyl-tRNAs from translation to peptidoglycan synthesis. Diversification of the side chains has been accompanied by a parallel divergent evolution of the substrate specificity of the l,d-transpeptidases, in contrast to the d,d-transpeptidases, which display an unexpected broad specificity. This review focuses on the role of antibiotics in selecting or counter-selecting diversification of the structure of peptidoglycan precursors and their mode of polymerization.

Keywords
  • antibiotic resistance
  • β-lactam
  • glycopeptide
  • Gram-positive bacteria
  • peptidoglycan synthesis

Introduction

Peptidoglycan polymerization is inhibited by the two major classes of antibiotics, the β-lactams and the glycopeptides, that act by different mechanisms (Fig. 1). The β-lactams are structural analogues of the d-Ala4-d-Ala5 extremity of peptidoglycan precursors and act as suicide substrates of the d,d-transpeptidases that catalyze the last cross-linking step of peptidoglycan synthesis (Frere & Joris, 1985). The glycopeptides bind to the peptidyl-d-Ala4-d-Ala5 extremity of peptidoglycan precursors and block by steric hindrance the essential glycosyltransferase and d,d-transpeptidase activities of the penicillin-binding proteins (PBPs) (Reynolds et al., 1989). Because of these distinct modes of action, the selective pressure of β-lactams and glycopeptides imposes opposite constraints on the evolution of the peptidoglycan biosynthesis pathway. Substitutions in the peptidyl-d-Ala4-d-Ala5 extremity of peptidoglycan precursors are selected by glycopeptides because such modifications reduce the affinity of the antibiotics for the target (Nieto & Perkins, 1971). The substrate specificity of the d,d-transpeptidases may limit emergence of glycopeptide resistance by this mechanism because the modified precursors should be cross-linked by these enzymes. Conversely, modifications of the active site of the PBPs leading to β-lactam resistance appear to impair cross-linking of the modified precursors required for glycopeptide resistance. Likewise, the low-affinity PBPs appear to tolerate poorly modifications in the side chain of peptidoglycan precursors that are assembled by aminoacyl transferases of the Fem family. In this review, the focus will be on three polymorphisms that impact on the expression of glycopeptide and β-lactam resistance: (1) substitutions of the C-terminal residue (d-Ala5) of peptidoglycan precursors that lead to acquired or intrinsic resistance to glycopeptides, (2) by-pass of the d,d-transpeptidases by an l,d-transpeptidase due to production of precursors devoid of d-Ala5, and (3) diversification of the structure of the side chain of peptidoglycan precursors and the parallel evolution of the transpeptidases.

1

(a) Peptidoglycan synthesis in Gram-positive bacteria. The nucleotide UDP-MurNAc-pentapeptide is formed in the cytoplasm by sequential addition of l-Ala, d-Glu, l-Lys, and of the dipeptide d-Ala-d-Ala. The membrane steps of peptidoglycan synthesis are initiated by the transfer of the phospho-MurNAc-pentapeptide moiety of the nucleotide to the lipid carrier (undecaprenyl-phosphate). Synthesis of the precursor is completed by the addition of the second sugar (GlcNAc) and of the side chain by tRNA-dependent aminoacyl transferases of the Fem family. The complete precursor linked to the lipid carrier (lipid II) is translocated through the membrane and polymerized by the glycosyltransferases (formation of glycan strands) and by the d,d-transpeptidase activity of PBPs (formation of the cross-links). Glycopeptide antibiotics bind to the peptidyl-d-Ala4-d-Ala5 extremity of pentapeptide stems and block by steric hindrance the transpeptidation and the transglycosylation reactions. The d,d-transpeptidases cleave the d-Ala4-d-Ala5 bond of a pentapeptide donor (b) and link the carbonyl of d-Ala4 to the amine at the extremity of the side chain of the acceptor substrate (4–3 cross-links). β-Lactam antibiotics, ampicillin (c) and imipenem (d), are structural analogues of the d-Ala4-d-Ala5 extremity of the precursors and act as suicide substrates of the d,d-transpeptidases. (e) Activation of the l,d-transpeptidation pathway confers resistance to β-lactam and glycopeptide in Enterococcus faecium. Activation of a cryptic d,d-carboxypeptidase leads to conversion of pentapeptide into tetrapeptide in the cytoplasm. The resulting precursors lacking d-Ala5 do not form complexes with glycopeptides at the cell surface, leading to resistance to these antibiotics. Tetrapeptide stems are cross-linked by an l,d-transpeptidase (Ldtfm) that by-passes the d,d-transpeptidase activity of PBPs. This enzyme cleaves the l-Lys3-d-Ala4 bond of a tetrapeptide donor and links the carbonyl of l-Lys3 to the amine at the extremity of the side chain of the acceptor substrate (3–3 cross-links). Ldtfm is not inhibited by β-lactams, except by carbapenems (d) such as imipenem. (f) Binding of vancomycin to the peptidyl-d-Ala4-d-Ala5 extremity of peptidoglycan precursors involves five hydrogen interactions (dashed lines). Substitution of d-Ala by d-lactate (d-Lac) at the C-terminal extremity of peptidoglycan precursors prevents formation of one of these hydrogen bonds and leads to a 1000-fold reduction in the affinity of the drug for its target.

Substitutions at the C-terminal position of peptidoglycan precursors

Glycopeptide antibiotics, vancomycin and teicoplanin, are widely used to treat severe infections due to Gram-positive bacteria resistant to β-lactam antibiotics, in particular methicillin-resistant Staphylococcus aureus. The drugs bind to the peptidyl-d-Ala4-d-Ala5 extremity of peptidoglycan precursors at the bacterial cell surface (Fig. 1) and block the transglycosylation and transpeptidation reactions by steric hindrance (Reynolds, 1989). The target is unusual as most other antibiotics interact with proteins and rRNA. Because glycopeptides do not penetrate the cytoplasmic membrane, the drugs have access to their target only after translocation of the lipid intermediate II to the cell surface. Binding of glycopeptides to this precursor is expected to block polymerization of the glycan strands because the antibiotics are sufficiently bulky to prevent access of glycosyltransferases to the disaccharide moiety of the precursors. Binding of glycopeptides to stem pentapeptides also inhibits the d,d-transpeptidase activity of PBPs that act directly on the d-Ala4-d-Ala5 peptide bond of their acyl donor substrate. The unique mode of action of glycopeptides implies unique mechanisms of resistance. Impermeability is not relevant in Gram-positive bacteria because of the extracellular location of the target. Detoxification of glycopeptides has not been reported. Decreased affinity for the targets implies a modification of the substrate specificity of peptidoglycan biosynthetic enzymes rather than the classical nucleotide or amino acid substitutions that prevent binding of antibiotics to protein or RNA targets. The critical enzyme for synthesis of the peptidyl-d-Ala-d-Ala target is the Ddl ligase, which catalyzes peptide bond formation between two d-Ala. Modification of the dipeptide is expected to affect multiple biosynthetic steps including the addition of the dipeptide onto UDP-MurNAc-tripeptide by the MurF ligase, the addition of side chain amino acid residues to the resulting pentapeptide by Fem transferases, and the final cross-linking step of peptidoglycan polymerization performed by the d,d-transpeptidases. An unusually long delay has separated the launching of vancomycin into clinical practice (1956) and the emergence of resistance in enterococci (Leclercq et al., 1988). During this period, glycopeptides were considered to be ‘irresistible’ antibiotics because remodeling of the d-Ala4-d-Ala5 motif would simultaneously require the modification of several biosynthetic enzymes (Reynolds, 1989). Resistance mediated by substitution of d-Ala5 by d-lactate (d-Lac) was therefore unexpected and not easily recognized (Dutka-Malen et al., 1990; Bugg, 1991a; Arthur et al., 1992a). Since the first reports on expression of the prototypic vanA gene cluster in the enterococci, it has been shown that modification of the d-Ala4-d-Ala5 extremity of peptidoglycan precursors is not universally well tolerated by the peptidoglycan biosynthetic enzymes in other bacteria. In this section, the authors will briefly present the mechanism of resistance conveyed by Tn1546, which has been reviewed previously (Arthur et al., 1996b), and discuss in more detail its evolution in Gram-positive bacteria and its integration in peptidoglycan metabolism.

The vanA gene cluster

Transposon Tn1546, harboring the vanA gene cluster, was initially detected in various plasmids in Enterococcus faecalis and Enterococcus faecium (Fig. 2) (Arthur et al., 1993). Tn1546 encodes two biosynthetic enzymes: the VanH dehydrogenase, which reduces pyruvate into d-Lac, and the VanA ligase, which catalyzes formation of an ester bond between d-Ala and d-Lac (Bugg, 1991b; Arthur et al., 1992a). The resulting depsipeptide, d-Ala-d-Lac, is added onto the UDP-MurNAc-tripeptide by the MurF ligase, leading to the production of peptidoglycan ending in d-Lac5 instead of d-Ala5. The substitution results in a 1000-fold reduction in the affinity of vancomycin for its target due to the loss of the hydrogen interaction connecting the NH group of the d-Ala4-d-Ala5 peptide bond to vancomycin (Fig. 1f). Elimination of precursors ending in d-Ala5 is required for resistance (Arthur et al., 1996a). This function is carried out by two Tn1546-encoded d,d-peptidases that hydrolyze d-Ala-d-Ala (VanX) and cytoplasmic precursors containing a stem pentapeptide (VanY) (Fig. 2) (Arthur et al., 1998). The two peptidases act in series to fully eliminate the precursors containing the target of glycopeptides. Hydrolysis of the dipeptide d-Ala-d-Ala formed by the host ligase Ddl almost completely prevents synthesis of UDP-MurNAc-pentapeptide ending in d-Ala4-d-Ala5. Precursors containing the target of glycopeptides, which escape the action of VanX, are subsequently eliminated by VanY. The insertion of the d,d-carboxypeptidase into the membrane is thought to optimize its interaction with the lipid intermediates at the inner surface of the cytoplasmic membrane (Arthur et al., 1998). VanX and VanY are Zn2+-dependent d,d-peptidases and are therefore unrelated to PBPs. VanX displays no esterase activity on the depsipeptide d-Ala-d-Lac (Reynolds et al., 1994; Wu et al., 1995) whereas VanY has a low carboxyesterase activity (Arthur et al., 1998), implying that elimination of the target of glycopeptides occurs at the expense of moderate hydrolysis of the pentadepsipeptide stems required for peptidoglycan polymerization in the presence of glycopeptides. The VanX dipeptidase introduces an energetic cost in the expression of resistance due to the futile cycle involving ATP-dependent ligation of two d-Ala by the host ligase and hydrolysis of the resulting dipeptide.

2

Mechanism of glycopeptide resistance conveyed by Tn1546. Transposon Tn1546 encodes two biosynthetic enzymes (VanH and VanA) for production of d-Lac-ending precursors, two d,d-peptidases (VanX and VanY) for elimination of d-Ala-ending precursors, and a two-component regulatory system (VanRS) for glycopeptide-inducible expression of the resistance genes. VanZ confers teicoplanin resistance by an unknown mechanism. The transposition functions include 38-bp inverted repeats (IRL and IRR), a transposase (ORF1), and a resolvase (ORF2).

Together, VanA, VanH, VanX, and VanY allow for the complete replacement of d-Ala5 by d-Lac5 (Arthur et al., 1998). Construction of Enterococcus faecalis strains that coproduce the two types of precursors in various proportions indicated that the level of vancomycin resistance gradually increases from 4 to 2000 μg mL−1 with a gradual elimination of precursors ending in d-Ala5 (from 43% to 2%) (Arthur et al., 1996a). Teicoplanin resistance requires an even more drastic elimination of the pentapeptide precursor. Production of d-Lac-ending precursors in sufficient amounts to sustain peptidoglycan synthesis is therefore not sufficient for resistance probably because translocation of small amounts of pentapeptides ending in d-Ala-d-Ala is sufficient to poison the polymerization complexes. This ‘dominant’ effect likely results from binding of glycopeptides to lipid intermediate II, because formation of this complex sequesters the undecaprenyl lipid carrier. The latter molecules are present in a very small number of copies per cell because they function in a ‘catalytic’ manner through recycling by dephosphorylation after each cycle of incorporation of a disaccharide-peptide unit into nascent peptidoglycan. Production of small amounts of lipid intermediate II ending in d-Ala4-d-Ala5 is therefore expected to lead rapidly to the accumulation of complexes that trap the lipid carrier in dead-end intermediates, thereby blocking processing of any disaccharide-peptide subunits even though they do not interact with glycopeptides (Arthur et al., 1998).

The vanA gene cluster of Tn1546 encodes a fifth resistance protein, VanZ, which alone confers low-level resistance to teicoplanin (Arthur et al., 1995). The underlying mechanism is unknown, but is clearly distinct from the mechanism conveyed by the rest of the vanA cluster, as teicoplanin resistance is not associated with any modification of the cytoplasmic pool of peptidoglycan precursors. VanZ and VanY increase the level of teicoplanin and vancomycin resistance mediated by the VanHAX proteins, respectively (Arthur et al., 1996a).

The glycopeptide resistance genes of Tn1546 are regulated by a two-component regulatory system that comprises a membrane-associated sensor kinase (VanS) and a response regulator (VanR) related to transcriptional activators of the OmpR-PhoB subclass (Arthur et al., 1992b). In the presence of glycopeptides, the VanS sensor acts as a kinase, leading to phosphorylation of VanR, and activation of the promoter (PH) for cotranscription of the vanHAXYZ resistance genes. In the absence of glycopeptide, VanS acts as a phosphatase and prevents transcriptional activation of the resistance genes by the phosphorylated form of VanR (Arthur et al., 1997). This negative control is required because VanR is activated independently from VanS, presumably by a heterologous kinase encoded by the host chromosome. Alternatively, VanR could catalyze its own phosphorylation using acetyl-phosphate as the substrate (Wright et al., 1993). The phosphorylated form of VanR activates a second promoter (PR), located upstream from the vanR gene, indicating that induction of the vanA gene cluster in response to glycopeptides involves not only a modulation of the relative concentration of VanR and its phosphorylated form but also of the absolute concentration of the regulatory protein (Arthur et al., 1999a). In the absence of glycopeptides, there is an energetic cost for regulated expression of the resistance genes because VanS-independent phosphorylation of VanR and its dephosphorylation by its partner sensor kinase constitutes a futile cycle. This energy cost is, however, expected to be limited by the relatively low concentrations of VanR and VanS present under noninducing conditions for transcription from the PR promoter.

Integration into the metabolism of the host

Dissemination of glycopeptide resistance genes involves mainly transposition of Tn1546 between different replicons, including self-transferable plasmids that promote the horizontal transfer of the glycopeptide resistance genes among enterococci (Arthur et al., 1993). Glycopeptide resistance is expressed in the recipient bacteria in the absence of mutational alteration of host genes. This was established by selecting for markers other than glycopeptide resistance in gene transfer experiments involving conjugation of natural plasmids as well as electro-transformation of recombinant plasmids. These experiments demonstrated that the peptidoglycan biosynthetic enzymes of Enterococcus faecalis and Enterococcus faecium tolerate the substitution of d-Ala by d-Lac at the extremity of peptidoglycan precursors. In particular, analysis of mature peptidoglycan showed that the BppA1 and BppA2 aminoacyl transferases efficiently transfer l-Ala from Ala-tRNAAla to peptidoglycan precursors containing d-Lac (Bouhss et al., 2002). Pentadepsipeptides ending in d-Lac are efficiently used by the d,d-transpeptidases as induction of glycopeptide resistance and production of the modified precursors do not modify the degree of peptidoglycan cross-linking (Bouhss et al., 2002). Because enterococci produce six d,d-transpeptidases that perform partially redundant functions (Arbeloa et al., 2004b), this observation does not demonstrate that all PBPs function equally well with precursors ending in d-Ala and d-Lac. In fact, one of these six enzymes, PBP5, might not be functional with the modified precursors (al-Obeid, 1992). This low-affinity PBP is responsible for intrinsic resistance to low levels of ampicillin and to high levels of third-generation cephalosporins such as ceftriaxone. Induction of the glycopeptide resistance abolishes PBP5-mediated β-lactam resistance, suggesting that this d,d-transpeptidase does not function with peptidoglycan ending in d-Lac. Induction of the glycopeptide resistance genes may also indirectly enhance the activity of the d,d-carboxypeptidases because all free stem pentapeptides ending in d-Ala4-d-Ala5, which are abundant under noninducing conditions in Enterococcus faecalis, are replaced by stem tetrapeptides ending in d-Ala4 rather than pentadepsipeptides ending in d-Ala4-d-Lac5 (Bouhss et al., 2002). This observation suggests that pentadepsipeptides are better substrates than pentapeptides for the d,d-carboxypeptidase DacA, which is the only low-molecular-weight PBP produced by Enterococcus faecalis (Arbeloa et al., 2004b). Interestingly, depsipeptides had been considered to be nonphysiological substrates before the characterization of glycopeptide resistance (Rasmussen & Strominger, 1978). Substitution of d-Ala by d-Lac was, however, already known to accelerate the rate of acyl-enzyme formation by several bacterial d,d-carboxypeptidases (Rasmussen & Strominger, 1978). Peptidoglycan precursor analogues ending in d-Lac have therefore been used to demonstrate the formation of acyl-enzyme intermediates. The increased rate of acylation of the PBPs allowed accumulation of sufficient amounts of the covalent intermediary to obtain biochemical evidence of its existence.

In conclusion, the substitution of d-Ala4 by d-Lac5 is tolerated by all enzymes in the peptidoglycan assembly pathway of the enterococci, although the production of the modified precursors clearly has an impact on the in vivo activity of the PBPs. Regulated expression of the van gene cluster is selectively advantageous because production of precursors ending in d-Ala and in d-Lac are alternatively required for resistance to β-lactam and glycopeptide antibiotics, respectively. In addition, replacement of d-Ala by d-Lac increases the tetrapeptide content of mature peptidoglycan presumably because the DacA d,d-carboxypeptidase hydrolyzes ester bonds more efficiently than peptide bonds. These observations imply that production of alternate peptidoglycan precursors provides a means to regulate the relative contribution of different transpeptidases to peptidoglycan cross-linking if the enzymes use the two types of precursors with different catalytic efficiencies. It remains to be determined whether such putative control could play a role in morphogenesis. Substitution of d-Ala5 by other d-2-hydroxyacids, such as d-2-hydroxybutyrate and d-2-hydroxyvalerate, has been obtained under laboratory conditions by inactivating the gene encoding the VanH dehydrogenase and supplementing the culture medium with these d-2-hydroxyacids (Arthur et al., 1992a, 1999b). Thus, the peptidoglycan biosynthesis enzymes of Enterococcus faecalis tolerate modifications in the side chain of the fifth residue in addition to the replacement of a d-2-amino- by a d-2-hydroxy-acid. It appears, therefore, that the narrow specificity of these enzymes is probably not the critical factor that limits emergence of glycopeptide by target modification in enterococci.

Mutations altering the specificity of d-Ala-d-Ala ligases

The Ddl ligases from glycopeptide-susceptible bacteria do not accept d-Lac as a substrate at physiological pH (Bugg et al., 1991b). Ligases from glycopeptide-resistant bacteria preferentially use d-Lac, but can also form small amounts of d-Ala-d-Ala. Single amino acid substitutions in both types of enzyme are sufficient to switch these specificities (Park et al., 1996). However, evidence for alteration of the specificity of a d-Ala-d-Ala ligase under the selective pressure of glycopeptides has never been obtained, probably because this is not sufficient for resistance. Mixed production of d-Ala-d-Ala and d-Ala-d-Lac would require a d,d-dipeptidase to prevent coproduction of precursors ending in d-Ala and in d-Lac. In addition, enterococci produce l-Lac as the end product of glycolysis, accounting for the essential role of the d-specific VanH dehydrogenase in expression of glycopeptide resistance. It is clear from these observations that resistance to glycopeptides by production of d-Lac-ending precursors cannot emerge following acquisition of a single mutation. This could account for the considerable difficulty in selecting for glycopeptide-resistant mutants under laboratory conditions, which led to a different resistance mechanism.

Diversity and modular organization of van gene clusters

In Tn1546, the genes encoding the VanH dehydrogenase, the VanA ligase, and the VanX dipeptidase are essential for glycopeptide resistance. All van clusters mediating resistance by production of d-Lac-ending precursors contain homologues of the vanHAX genes, which are always present in that order and cotranscribed (Courvalin et al., 2006). Self-defense is mediated by VanHAX-related proteins in the actinomycetes that produce glycopeptide antibiotics (Marshall et al., 1997). The high amino acid identity (ca. 60%) indicates that this is a likely origin for the portion of Tn1546 encoding these proteins, although substantial divergent evolution has occurred including the adaptation of codon usage to different hosts.

All van gene clusters encode a d,d-carboxypeptidase. These enzymes are more diverse than VanHAX as illustrated by comparison of the enterococcal vanA and vanB gene clusters (24% and 60% sequence identity for comparison of VanY and VanHAX, respectively) (Evers & Courvalin, 1996). In addition, the d,d-carboxypeptidase genes occupy different positions in the vanA and vanB clusters, respectively, downstream and upstream of the vanHAX element. The vanD cluster is unusual because it encodes a d,d-carboxypeptidase belonging to the PBP family instead of a metallo-d,d-carboxypeptidase (Reynolds et al., 2001).

Most gene clusters encode one additional protein that is specific for each cluster and has generally not been functionally explored, except for VanZ encoded by Tn1546 (above), and VanK (Hong et al., 2005) encoded by a van cluster from Streptomyces coelicolor, which does not produce any glycopeptide. VanK is a member of the Fem family that transfers a glycyl residue from Gly-tRNAGly to the ɛ amino group of meso-diaminopimelic acid (mDAP) at the third position of the cytoplasmic precursor UDP-MurNAc-pentadepsipeptide. Production of VanK is coregulated with the VanHAX-related enzymes and, therefore, is only produced under conditions of induction by glycopeptides. Streptomyces coelicolor produces a second aminoacyl transferase, FemX, that transfers a glycyl residue to UDP-MurNAc-pentapeptide under noninducing conditions. VanK is an essential enzyme under inducing conditions because FemX cannot use precursors ending in d-Lac. Similarly, Actinoplanes teichomyceticus, which produces teicoplanin, has two MurF ligases presumably because the host-keeping enzyme cannot add d-Ala-d-Lac onto UDP-MurNAc-tripeptide (Serina et al., 2004). Expression of the vanA gene cluster of Tn1546 in methicillin-resistant Staphylococcus aureus leads to impaired synthesis of the side chain of peptidoglycan precursors probably because the Fem transferases of Staphylococcus aureus display decreased catalytic efficiency with precursors ending in d-Lac (Severin et al., 2004). Staphylococcus aureus, therefore, represents an intermediary situation between Streptomyces coelicolor, in which an additional Fem transferase is required for resistance (VanK) because the host-keeping enzyme (FemX) does not function with d-Lac-ending precursors (Serina et al., 2004), and Enterococcus faecalis, in which the host keeping BppA1 and BppA2 transferases appear to function equally well with d-Ala- and d-Lac-ending precursors (Bouhss et al., 2002). Together, these data indicate that highly diverse proteins encoded by van gene clusters, in addition to the ubiquitous VanHAXY-related proteins, include enzymes that functionally replace host-keeping enzymes deficient in the processing of d-Lac-ending precursors. These enzymes are expected to be critical in determining the host spectrum of glycopeptide resistance expression following horizontal transfer of the gene clusters.

All vancomycin resistance gene clusters encode two-component regulatory systems comprising a response regulator of the OmpR-PhoB subclass (Arthur & Quintiliani, 2001; Depardieu et al., 2007). The receiver domain of the regulators and the kinase domain of the sensors contain the conserved amino acid motifs present in ‘classical’ two-component regulatory systems, which catalyze the same phosphotransfer reactions. In contrast, the N-terminal sensor domains of the kinases are unrelated, except for two closely related clusters detected in Streptomyces spp. (Hutchings et al., 2006). Despite the absence of primary sequence similarity, the sensor domains most probably have the same membrane topology, as they contain two clusters of hydrophobic amino acids potentially delineating periplasmic segments of various sizes. The lack of sequence similarity in the sensor domains of the kinases reflects different modes of recognition of the glycopeptides. Because all antibiotics that cause accumulation of lipid intermediate II act as inducers irrespective of their structure and of mode of action, the VanS sensor may recognize accumulation of this precursor. In contrast, the vanB cluster from the enterococci (Evers & Courvalin, 1996) and two clusters from Streptomyces spp. (Hutchings et al., 2006) are inducible by a subset of the glycopeptides that were tested. This selectivity suggests that direct interaction of the corresponding sensor kinases with the drug, perhaps bound to lipid intermediate II, is responsible for induction. The ubiquitous presence of regulatory systems in van clusters indicates that there is a selective advantage for the production of alternate types of precursors in response to environmental conditions.

Several lines of evidence indicate that the van gene clusters have been generated by the recruitment of different genes or set of genes, including the variations in the guanosine plus cytosine content of different portions of the clusters, the variable percentage of sequence identity between the encoded proteins, the variation in the order of the genes, and the presence of unrelated accessory proteins. The DNA recombination mechanisms operating for the recruitment of the different genes as well as their origins remain unknown, except for vanHAX that may originate from glycopeptide producers.

Micro-evolution of van gene clusters

The vanB gene cluster does not confer resistance to teicoplanin because this glycopeptide does not trigger induction of the resistance genes (Evers & Courvalin, 1996). Mutations leading to high-level cross-resistance to vancomycin and teicoplanin have been selected under laboratory conditions (Baptista et al., 1997), animal models of experimental endocarditis (Aslangul et al., 1997; Lefort et al., 2004), and in patients treated with glycopeptides (Abelson et al., 1998). Most mutations mapped in the gene encoding VanSB and affected different functions of the sensor kinase (Baptista et al., 1997). Loss of the phosphatase activity due to amino acid substitutions near the phosphorylation site (His233) led to high-level constitutive expression of the resistance genes. Alteration of the specificity of VanSB due to amino acid substitutions in the sensor domain led to induction by both vancomycin and teicoplanin. Finally, null mutations led to a complex phenotype referred to as ‘heterogeneous’ because the resistance genes were expressed in a minor subpopulation in the absence of glycopeptides (Baptista et al., 1999; Arthur & Quintiliani, 2001). Expression of the resistance genes was inducible by vancomycin and teicoplanin in the other subpopulation probably due to cross-regulation between the VanRB response regulator and a host kinase.

In addition to mutations in the gene encoding VanSB, glycopeptides select null mutations in the host ligase gene (ddl) (Baptista et al., 1997). These mutations increase the level of vancomycin resistance by preventing synthesis of precursors ending in d-Ala. The ddl null mutants are dependent on vancomycin for growth because the d-Ala-d-Lac ligase is the only functional ligase and this enzyme is only produced under inducing conditions. Selection for growth in the absence of glycopeptides releases the ‘addiction’ to vancomycin by selecting for mutations in vanSB that inactivate the phosphatase activity of the sensor kinase. The resulting mutants constitutively confer high-level resistance to vancomycin and teicoplanin. Because the mutants have lost the capacity to produce precursors ending in d-Ala due to mutations in ddl, reversion toward susceptibility to glycopeptides or regulated expression of resistance is very unlikely. A combination of mutations in the host ligase and the sensor kinase abolishes the futile cycles associated with synthesis/hydrolysis d-Ala-d-Ala and phosphorylation/dephosphorylation of the response regulator.

Together, these observations suggest that serial selection for high-level resistance to vancomycin and vancomycin independence is a likely scenario for the microevolution of glycopeptide-resistant enterococci that may ultimately result in the replacement of susceptible clones by resistant ones in natural populations. Accordingly, combinations of null mutations in the host ligase and the sensor kinase genes have been detected in clinical isolates harboring various van gene clusters (Depardieu et al., 2007).

Intrinsic resistance to glycopeptides

Several bacterial species used in the fermentation process, including lactococci and lactobacilli, are intrinsically resistant to glycopeptide antibiotics by production of d-Lac-ending precursors (Deghorain et al., 2007). This property has remained largely unrecognized because the fifth residue of stem peptides is generally absent from mature peptidoglycan (Schleifer & Kandler, 1972). In addition, the connection with glycopeptide resistance was not established as these organisms are not pathogenic and are therefore of no interest in the field of medical bacteriology. Glycopeptide resistance is significantly associated with the production d-Lac as the end product of glycolysis. Thus, glycopeptide resistance could be an indirect consequence of a metabolic strategy presenting the advantage of sparing d-Ala by recycling a waste product. However, a d,d-dipeptidase homologous to VanX has been detected recently in Lactobacillus plantarum (Deghorain et al., 2007), indicating that production of a d-Lac-ending precursors may also be selectively advantageous because of resistance to glycopeptide antibiotics produced by competitors of L. plantarum.

The genes involved in intrinsic glycopeptide resistance are constitutively expressed and scattered in the chromosome in contrast to their counterparts involved in acquired resistance in the enterococci or self-defense in the antibiotic producers, which are organized in clusters and regulated by environmental signals. Although intrinsic resistance is not expected to be mediated by mobile genes, production of d-Lac-ending precursor is a variable character in the lactobacilli and lactococci (Deghorain et al., 2007).

Production of precursors ending in d-serine

Glycopeptide resistance can also result from substitution of d-Ala5 by d-Ser5. The substitution leads only to a moderate (eightfold) decrease in the affinity of vancomycin for its target and accordingly to low-level resistance to the antibiotic (Billot-Klein et al., 1994). Incorporation of d-Ser is mediated by ligases distantly related to VanA and the classical d-Ala-d-Ala ligases (Evers et al., 1996). The substrate of the ligase is produced by a serine racemase, a multimodular enzyme that also comprises a membrane-associated domain probably involved in serine transport (Arias et al., 1999). A peptidase of the d,d specificity eliminates precursors ending in d-Ala (Reynolds et al., 1999). The enzyme combines the specificities of VanX and VanY as it hydrolyzes both the dipeptide d-Ala-d-Ala and precursors containing a stem pentapeptide ending in d-Ala. This mode of resistance concerns all members of four bacterial species (Enterococcus gallinarum, Enterococcus casseliflavus, and Enterococcus flavescens and Clostridium inocuum) and was therefore classified as intrinsic although it shares several features with acquired resistance including the gene organization in clusters (Arias et al., 2000), regulation mediated by two-component regulatory systems (Arias et al., 2000), and, in one instance, acquisition of resistance by a susceptible species (Abadia Patino, 2002).

By-pass of PBPs

Analyses of the muropeptide composition of peptidoglycan from Escherichia coli revealed that a small proportion of the dimers were unlikely to be generated by PBPs because the cross-links involved two mDAP residues instead of the classical d-Ala4-mDAP3 cross-links (Glauner et al., 1988; Goffin & Ghuysen, 2002). PBPs are transpeptidases of the ‘d,d’ specificity as these enzymes cleave the terminal d-Ala4-d-Ala5 peptide bond of a pentapeptide acyl donor and link the carbonyl of d-Ala4 to the side chain amino group of the acyl acceptor (Frere & Joris, 1985). Note that the reacting amino group of mDAP in the acyl acceptor has coincidentally a d configuration although this is irrelevant to the ‘d,d’ designation of the transpeptidase. By analogy, mDAP3-mDAP3 cross-links detected in Escherichia coli were thought to be generated by a transpeptidase of the ‘l,d’ specificity because their formation was expected to involve cleavage of the mDAP3-d-Ala4 peptide bond before linkage of the carbonyl of d-Ala4 to the side chain amino group of the acceptor. However, the enzyme remained unidentified as its relation to the classical PBPs (Goffin & Ghuysen, 2002). The contribution of l,d-transpeptidases to β-lactam resistance has been anticipated (Holtje et al., 1998) but this has only been established in mutants of Enterococcus faecium selected in vitro as described in the following section.

Activation of the l,d-transpeptidation pathway

Among the d,d-transpeptidases of Enterococcus faecium, low-affinity PBP5 is responsible for intrinsic low-level β-lactam resistance with minimal inhibitory concentrations (MICs) of ampicillin comprised between 0.5 and 4 μg mL−1 (Rice et al., 2004). In clinical isolates, acquisition of higher levels of resistance to ampicillin is generally associated with increased production of PBP5 (MICs ≤16 μg mL−1) or with amino acid substitutions that further decrease the interaction of the d,d-transpeptidase module of PBP5 with β-lactams (up to 256 μg mL−1) (Rice et al., 2004). A search for alternate mechanisms of β-lactam resistance was carried out in a mutant of Enterococcus faecium hypersusceptible to ampicillin (MIC of 0.06 μg mL−1) due to spontaneous loss of the pbp5 gene (Mainardi et al., 2000). Five selection steps on increasing concentrations of ampicillin led to a mutant insensitive to the drug (MIC >2000 μg mL−1). Analysis of the peptidoglycan structure indicated that ampicillin resistance was due to target substitution because the d-Ala4-d-iAsx-l-Lys3 cross-links generated by the d,d-transpeptidase activity of PBPs were replaced by l-Lys3-d-iAsx-l-Lys3 cross-links (Fig. 1e) (Mainardi et al., 2000). By-pass of the d,d-transpeptidases was therefore due to an l,d-transpeptidase that cleaved the l-Lys3-d-Ala4 peptide bond of an acyl donor and linked the carbonyl of l-Lys3 to the side chain d-iso-asparagynil or d-iso-aspartyl (d-iAsx) residue of the acceptor. Of note, by-pass of the glycosyltransferase activity of PBPs has also been reported in Enterococcus faecalis (Arbeloa et al., 2004b).

The gene encoding the l,d-transpeptidase has been identified in the genome of Enterococcus faecium based on partial purification of the enzyme and N-terminal sequencing (Mainardi et al., 2005). The enzyme, designated Ldtfm, was produced in Escherichia coli and its activity was studied using synthetic substrates and disaccharide-peptide fragments isolated from the peptidoglycan. The enzyme was shown to be stereospecific and to catalyze formation of peptide as well as ester bonds. As expected, the enzyme was not inhibited by ampicillin. Incubation of the enzyme with the Enterococcus faecium disaccharide-tetrapeptide ending in d-Ala4 led to the formation of dimers containing the l-Lys3-d-iAsx-l-Lys3 cross-links observed in vivo. Because muropeptides containing a pentapeptide stem ending in d-Ala4-d-Ala5 are not used as acyl donors by Ldtfm, activation of the l,d-transpeptidation pathway requires production of a tetrapeptide from the pentapeptide synthesized in the cytoplasm (Fig. 1e). The corresponding d,d-carboxypeptidase activity was detected in membrane fractions of the mutant resistant to ampicillin (Mainardi et al., 2002). Similar to Ldtfm, this d,d-carboxypeptidase was not inhibited by ampicillin. The enzyme was active in the cytoplasm because large amounts of UDP-MurNAc-tetrapeptide were detected (60% of the pool of cytoplasmic precursors).

The five mutations required for full activation of the l,d-transpeptidation pathway remain unknown. The l,d-transpeptidase activity was detected in the parental strain susceptible to ampicillin (Mainardi et al., 2002) and, 33 years ago, in a wild-type strain belonging to a species closely related to Enterococcus faecium (Enterococcus hirae formerly Streptococcus faecalis) (Coyette et al., 1974). In the parental strain, Ldtfm participated in peptidoglycan synthesis because a minority of the cross-links (0.7%) was of the l-Lys3-d-iAsx-l-Lys3 type (Mainardi et al., 2002). In the mutant highly resistant to ampicillin, DNA sequencing and Western blot analysis indicated that neither the structure of Ldtfm nor the level of production of the enzyme had been altered during the selection process. In contrast, the β-lactam-insensitive d,d-carboxypeptidase was cryptic in the parental strain and production of the enzyme was activated at the fourth selection step. Thus, the l,d-transpeptidase is produced and functional in the parental strain but this enzyme cannot by-pass the PBPs because its tetrapeptide acyl donor substrate is not present in sufficient amounts. Production of the tetrapeptide substrate of Ldtfm accounts for one of the five selection steps, presumably one of five mutations, required for activation of the l,d-transpeptidation pathway and acquisition of β-lactam resistance. The other mutations do not appear to affect any enzyme in the peptidoglycan assembly pathway because the structures of the peptidoglycan and its precursors are very similar in the parental strain and in the resistant mutant, except for the presence of the d-Ala4-d-iAsx-l-Lys3 cross-links and the production of cytoplasmic precursors containing a tetrapeptide stem. The mutations may therefore affect components of the peptidoglycan polymerization complexes that co-operate for peptidoglycan assembly, including the glycosyltransferases.

Structure of the l,d-transpeptidase of Enterococcus faecium

Analysis of the sequence of Ldtfm revealed the presence of an N-terminal cluster of hydrophobic amino acids that may act as a membrane anchor for the protein (Fig. 3) (Mainardi et al., 2005). By analogy with the topology of classical d,d-transpeptidases belonging to the PBP family, Ldtfm may comprise a short cytoplasmic stretch of amino acids (residues 1–12), a transmembrane segment (13–28), whereas the bulk of the protein (29–466) is located in the periplasm where nascent peptidoglycan is expected to be assembled following translocation of lipid intermediate II. The primary sequence of Ldtfm is unrelated to proteins of known function in databases. The C-terminal domain of Ldtfm (314–466) is the first functionally characterized member of a conserved family of proteins designated ErfK-Ybis-YhnG or Pfam 03734 in protein family databases. This domain contains an invariant Cys residue at position 442 that has been shown to be essential for l,d-transpeptidase activity by site-directed mutagenesis and chemical modification. This residue most probably corresponds to the catalytic residue of the protein. Sequences related to the C-terminal domain of Ldtfm are sporadically distributed among taxonomically distant bacteria including both Gram-positive and Gram-negative bacteria. Close homologues of Ldtfm have been detected in Gram-positive bacteria including Bacillus anthracis and Enterococcus faecalis but not in Streptococcus pneumoniae and Staphylococcus aureus.

3

(a) Sequence and structural organization of the Enterococcus faeciuml,d-transpeptidase (Ldtfm). The crystallized fragment of Ldtfm comprises two domains colored in pink and in blue. The gray box represents an additional domain of unknown structure, which contains the putative membrane anchor (indicated in yellow). (b) Three views of the molecular surface of Ldtfm with the same color coding as in (a). The top view shows the two paths, indicated by arrows, to the catalytic site. The bottom and right views show the first and second path, respectively. Cys442, His421, and Asp422 of the catalytic site are represented in pink, cyan, and green, respectively. (c) Zoom of the catalytic site, with Cys442, His421, and Asp422 represented as sticks.

The crystal structure of a fragment of Ldtfm (residue 217–466; PDB 1ZAT) has been determined at 2.4 Å resolution, revealing two structural domains displaying novel folds (Fig. 3) (Biarrotte-Sorin et al., 2006). Domain I (residues 217–338) is elongated (65 × 20 Å) and composed of nine β strands and four α helices. This domain could act as a pedestal (Biarrotte-Sorin et al., 2006) placing the C-terminal catalytic domain in the proximity of the nascent peptidoglycan as proposed for PBP3 of Streptococcus pneumoniae (Morlot et al., 2005). Domain II (residues 339–466) is globular and composed of 10 β strands and two α helices (Biarrotte-Sorin et al., 2006). This structural domain corresponds to the ErfK-Ybis-YhnG/Pfam 03734 domain defined by primary sequence comparisons. The β-strands of domain II fold into two five-stranded mixed polarity β-sheets sharing two bent strands. The catalytic residue, Cys442, is located in a buried cavity accessible by two paths that may correspond to the binding sites for the acyl donor and acceptor substrates of the transpeptidation reaction. The hydrogen bond network connecting Cys442, His421, and Asp422 suggests that these residues define the catalytic triad of the l,d-transpeptidase. By analogy with cysteine and serine proteases (Dodson & Wlodawer, 1998), the Nɛ2 of His421 could capture the Sγ hydrogen released by Cys442 to assist nucleophilic attack of the carbonyl of the l-Lys3-d-Ala4 peptide bond (Biarrotte-Sorin et al., 2006). Thus, Ldtfm and the PBPs are structurally unrelated, comprise different active site architectures, and nucleophiles (Cys and Ser, respectively).

Coresistance to β-lactams and glycopeptides

Peptidoglycan cross-linking by Ldtfm cannot be inhibited by glycopeptides because this l,d-transpeptidase uses a tetrapeptide donor lacking d-Ala5, a residue essential for drug binding (Nieto & Perkins, 1971). However, the mutant of Enterococcus faecium selected in vitro in five steps, as described above, remained susceptible to glycopeptides (Cremniter et al., 2006). Thus, by-pass of the PBPs by Ldtfm was in itself not sufficient for glycopeptide resistance. Four additional selection steps on increasing concentrations of glycopeptides were necessary to obtain high-level resistance to β-lactams, vancomycin, and teicoplanin. Cross-resistance to these antibiotics was also obtained in a total of nine steps using exclusively glycopeptides for selection. In the corresponding mutants, the UDP-MurNAc-pentapeptide pool was almost completely converted to UDP-MurNAc-tetrapeptide by the β-lactam-insensitive d,d-carboxypeptidase. UDP-MurNAc-pentapeptide accounted for 92% of the cytoplasmic precursors in the parental strain susceptible to ampicillin and glycopeptides, 40% in the mutant resistant to ampicillin only, and 8% or <1% in the mutants selected by glycopeptides or serially by β-lactams and glycopeptides, respectively. Thus, coproduction of UDP-MurNAc-pentapeptide and UDP-MurNAc-tetrapeptide in similar amounts leads to resistance to ampicillin because sufficient tetrapeptide-containing precursors are present to sustain peptidoglycan cross-linking by Ldtfm. Because these glycopeptides can still bind to pentapeptide-containing precursors and block the transglycosylation reaction, coproduction of the two types of precursors does not lead to glycopeptide resistance. Coresistance to ampicillin and glycopeptides is only obtained for an almost complete elimination of precursors containing the peptidyl-d-Ala-d-Ala target of glycopeptides. As described above, the ‘dominant’ effect of precursors containing a pentapeptide stem may be due to sequestration of the lipid carrier due to formation of complexes between lipid intermediate II and glycopeptides.

Production of precursors containing a tetrapeptide stem or a pentadepsipeptide stem ending in d-Lac constitutes two alternate mechanisms of resistance appearing under the selective pressure of glycopeptides. Because the l,d-transpeptidase uses the energy of the l-Lys3-d-Ala4 peptide bond for cross-link formation, full elimination of d-Ala5 is only possible if Ldtfm by-passes the d,d-transpeptidase activity of the PBPs. In contrast, PBPs use the energy of the d-Ala4-d-Ala5 or the d-Ala4-d-Lac5 bond. Resistance to glycopeptides by production of precursors containing a tetrapeptide stem is therefore dependent on production of a transpeptidase of the ‘l,d’ specificity.

Inhibition of Ldtfm by carbapenems

β-Lactams are molecular mimics of the peptidyl-d-Ala4-d-Ala5 extremity of peptidoglycan precursors and act as ‘suicide’ substrates of the d,d-transpeptidase module of PBPs (Fig. 1) (Tipper & Strominger, 1965). Because the tetrapeptide donor of Ldtfm lacks the d-Ala4-d-Ala5 motif, by-pass of the PBPs was expected to confer high-level cross-resistance to all β-lactams (Mainardi et al., 2000, 2005). However, a specific class of β-lactams, the carbapenems, has recently been shown to remain active against mutants highly resistant to ampicillin (Mainardi et al., 2007). Imipenem totally inhibited the l,d-transpeptidase activity of Ldtfm at a concentration equal to the MIC of the drug (0.5 μg mL−1). Peptidoglycan cross-linking was also inhibited, indicating that Ldtfm is the in vivo target of imipenem. Stoichiometric and covalent modification of Ldtfm by imipenem was detected by mass spectrometry. The mass increment matched the mass of imipenem, indicating that inactivation of Ldtfm is likely to involve rupture of the β-lactam ring. The same type of modification was observed for other carbapenems (ertapenem and meropenem). Before the covalent modification was mapped into the trypsin fragment of Ldtfm containing Cys442, the catalytic residue of Ldtfm is most probably the site of acylation. In addition, the Cys442Ala substitution prevented imipenem binding. Thus, Ldtfm‘commits suicide’ by catalyzing formation of a thioester bond between its catalytic Cys442 residue and the β-lactam ring of carbapenems. Inhibition of Ldtfm by carbapenems unexpectedly extends the spectrum of activity of β-lactams to include transpeptidases of the ‘l,d’ specificity. Ldtfm is the first representative of a novel class of active site cysteine peptidases that are inhibited by a specific class of β-lactams. Synergy between ampicillin and imipenem was observed in a strain of Enterococcus faecium endowed with the dual capacity to manufacture peptidoglycan with transpeptidases of the ‘l,d’ and ‘d,d’ specificities. This indicates that combined therapy with two β-lactams could be useful against strains coproducing two types of low-affinity transpeptidases (PBP5 and Ldtfm in this case).

Diversity of l,d-transpeptidases

Homologues of Ldtfm from Bacillus subtilis (LdtBs) and Enterococcus faecalis (Ldtfs) have been produced in Escherichia coli and purified in order to compare the in vitro activity of enzymes isolated from bacteria producing peptidoglycan of different chemotypes (Magnet et al., 2007a). Bacillus subtilis produced peptidoglycan precursors containing mDAP at the third position of the pentapeptide stem (Schleifer & Kandler, 1972). In Enterococcus faecalis, this position is occupied by an l-Lys substituted by two l-Ala instead of d-iAsx in Enterococcus faecium (Schleifer & Kandler, 1972). The three enzymes, LdtBs, Ldtfs, and Ldtfm, were functional in vitro, with their cognate muropeptides containing tetrapeptide stems, and catalyzed formation of dimers containing mDAP3-mDAP3, l-Lys3-l-Ala2-l-Lys3, and l-Lys3-d-iAsx-l-Lys3 cross-links, respectively (Magnet et al., 2007a). Ldtfs displayed additional activities including the hydrolysis of the l-Lys3-d-Ala4 peptide bond (l,d-carboxypeptidase activity) and the use of both tetrapeptide and pentapeptide as acyl donor substrates. LdtBs, Ldtfs, and Ldtfm used exclusively their cognate acyl acceptors in the cross-linking reaction. Ldtfs was also specific for its cognate acyl donor in contrast to Ldtfm and LdtBs, which tolerated substitutions of l-Lys3-d-iAsx by l-Lys3-l-Ala2 and of mDAP3 by l-Lys3-d-iAsx and l-Lys3-l-Ala2, respectively. Thus, diversification of the structure of peptidoglycan precursors associated with speciation has led to a parallel evolution of the substrate specificity of the l,d-transpeptidases affecting mainly the recognition of the acyl acceptor. Peptidoglycan cross-linking is not the only function performed by members of the Ldtfm family because homologues from Escherichia coli were recently found to be responsible for anchoring the Braun lipoprotein to peptidoglycan (Magnet et al., 2007b).

In spite of its wide distribution (Mainardi et al., 2005), the physiological role of peptidoglycan l,d-transpeptidases remains unknown. Cross-links generated by l,d-transpeptidation have been detected in various bacteria including Clostridium perfringens (Leyh-Bouille et al., 1970), Streptomyces albus G (Leyh-Bouille et al., 1970), Mycobacterium smegmatis (Wietzerbin et al., 1974), Mycobacterium bovis strain BCG (Wietzerbin et al., 1974), and various members of the Enterobacteriaceae family (Glauner et al., 1988; Quintela et al., 1995, 1997). In all these species, the cross-links generated by d,d-transpeptidases remain predominant and by-pass of the PBPs leading to β-lactam resistance has not been reported. The l,d-transpeptidases may have a role under conditions where pentapeptide donors, which could drive the d,d-transpeptidase reaction, are in short supply (Holtje, 1998). In agreement, a significant increase in the number of the mDAP3-mDAP3 cross-links has been observed in the stationary phase of growth or during bacteriolysis caused by antibiotics in Escherichia coli (Pisabarro et al., 1985; Driehuis & Wouters, 1987; Tuomanen & Cozens, 1987; Blasco et al., 1988; Kohlrausch & Holtje, 1991). An increase in the abundance of cross-links generated by l,d-transpeptidation has been associated with an increase of the covalently bound form of the Brown lipoprotein (Pisabarro et al., 1985) and a reduction of the length of glycan strands (Pisabarro et al., 1985; Tuomanen & Cozens, 1987; Glauner et al., 1988). The mature peptidoglycan of most bacterial species contains small amounts of pentapeptide stems due to the combined action of d,d-carboxypeptidases and d,d-transpeptidases. Thus, the l,d-transpeptidases could be important to remodel or maintain the peptidoglycan network under conditions where de novo synthesis of precursors containing a stem pentapeptide is spatially or temporally restricted (Cremniter et al., 2006). Although antibiotic resistance is certainly not the primary role of l,d-transpeptidases in eubacteria, there is a risk of activation of the l,d-transpeptidation pathway in response to the selective pressure of glycopeptides and β-lactams that may lead to cross-resistance to the two major classes of antibiotics available to treat severe infections due to Gram-positive bacteria. Fortunately, carbapenems retain activity against l,d-transpeptidases (Mainardi et al., 2007).

Aminoacyl transferases of the Fem family

Most Gram-positive bacteria produce branched peptidoglycan precursors due to the addition of a side chain to the diamino acid at the third position of peptidoglycan precursors (Schleifer & Kandler, 1972). The extremity of the side chain carries the amino group of the acyl acceptor used by the transpeptidases in the cross-linking reaction (Fig. 1a) (Frere & Joris, 1985). Following transpeptidation, the amino acids of the side chain are referred to as cross-bridges. In contrast to the stem pentapeptide of peptidoglycan precursors, the side chain is highly variable with respect to its mode of synthesis, its length, and its sequence (Schleifer & Kandler, 1972). The dicarboxylic amino acid d-Asp is incorporated into the side chain by aspartate ligases that belong to the ATPGrasp superfamily and activate their substrate by ATP-dependent formation of a β-aspartyl-phosphate intermediary (Staudenbauer & Strominger, 1972; Bellais et al., 2006). d-Glu is probably incorporated by the same mechanism. In contrast, l-amino acids and glycine are activated as aminoacyl-tRNAs and incorporated into the side chain by Fem transferases that belong to the GCN5-related N-acetyltransferase (GNAT) protein superfamily (Kopp et al., 1996; Berger-Bachi & Tschierske, 1998). The length of the side chain varies from one to seven amino acids and contains in most cases combinations of Gly, l-Ala, l-Ser, and l-Thr (Schleifer & Kandler, 1972). The aminoacyl transferases responsible for incorporation of these amino acids have been characterized biochemically in the late 1960s and early 1970s (Matsuhashi et al., 1965; Plapp & Strominger, 1970). The enzymes were shown to catalyze transfer of aminoacyl residues from aminoacyl-tRNAs to the peptidoglycan precursors independently from the ribosome. The aminoacyl-transferases were genetically identified ca. 15 years later using a totally distinct approach based on random insertion mutagenesis of the chromosome of methicillin-resistant Staphylococcus aureus and screening for impaired expression of β-lactam resistance (Berger-Bachi et al., 1989). This led to the identification of factors essential for methicillin resistance (fem), including a locus composed of two genes (femA and femB) that was essential for incorporation of glycine at positions 2–5 of the pentaglycine side chain. More recently, recombinant aminoacyl-transferases have been produced in Escherichia coli and purified, confirming that the genetically identified genes of the femAB family were actually encoding the aminoacyl-transferases for side chain synthesis (Bouhss et al., 2001; Hegde & Shrader, 2001). In this section, the authors shall first review recent investigations on the catalytic mechanism and the structure of Fem aminoacyl transferases. Analysis of the in vitro specificity of the enzymes for their peptidoglycan precursor and the amino-acyl-tRNA substrates will introduce two sections concerning the design of Fem inhibitors and the physiological role of Fem transferases in peptidoglycan synthesis and β-lactam resistance.

Catalytic mechanism

The reaction catalyzed by the Fem transferases (Fig. 4a) is an acyl transfer reaction: the transfer of an aminoacyl group from the 3′ end of the aminoacyl-tRNA substrate to the ɛ-amino group of l-Lys or mDAP at position 3 of the pentapeptide side chain of lipid intermediate II or UDP-MurNAc-pentapeptide (Hegde & Shrader, 2001; Hegde & Blanchard, 2003). The Fem-catalyzed reaction is likely to proceed via nucleophilic attack of the l-Lys or mDAP ɛ-amino group, perhaps assisted by base catalysis, on the aminoacyl carbonyl group of the aminoacyl-tRNA, to form a tetrahedral oxyanion intermediate, which is presumably stabilized by hydrogen bonding or electrostatic interactions at the enzyme active site (Fig. 4a). The oxyanion would then break down, with protonation of the 3′-hydroxyl group of the terminal adenosine base, to complete the acyl transfer reaction.

4

(a) Catalytic mechanism for Fem aminoacyl transferases. (b) Transition state analogue. The phosphonate inhibits the Streptococcus pneumoniae MurM with an IC50 of 100 μM. (c) Stable substrate analogue of Ala-tRNAAla containing 1,2,4 oxadiazole as a mime of the 3′ amino acyl ester. FemXWv from Weissella viridescens was inhibited with an IC50 of 1.4 μM.

Kinetic studies have been performed with the FemXWv aminoacyl transferase from Weissella viridescens (Hegde & Blanchard, 2003), which catalyzes in vivo the transfer of the first residue of an l-Ala-l-Ser or l-Ala-l-Ser-l-Ala side chain onto the cytoplasmic precursor UDP-MurNAc-pentadepsipeptide (Villet et al., 2007). FemXWv catalysis proceeds by an ordered bi-bi mechanism with sequential fixation of the UDP-MurNAc-pentapeptide and Ala-tRNAAla substrates and sequential release of the tRNAAla and UDP-MurNAc-hexapeptide products (Hegde & Blanchard, 2003). Site-directed mutagenesis of FemXWv based on the alignment of the primary sequence of Fem proteins surprisingly revealed that substitutions at invariant or highly conserved positions had a rather moderate (<20-fold) impact on activity, except for Asp108 (230-fold) (Hegde & Shrader, 2001; Hegde & Blanchard, 2003). This residue was therefore proposed to be the catalytic base although the subsequent resolution of the crystal structure of a FemXWv:UDP-MurNAc-pentapeptide complex revealed that Asp108 is not located in the vicinity of the substrate (Biarrotte-Sorin et al., 2004).

Structure of Fem transferases

The Fem transferases constitute a family of proteins of 330–450 amino acids that have been divided into two classes depending on the presence or absence of a coiled-coil region of about 60 amino acids (Fig. 5). Structures of one member of each class have been solved: FemA of Staphylococcus aureus [PDB code 1LRZ (Benson et al., 2002)] and FemXWv of W. viridescens [PDB code 1NE9 (Biarrotte-Sorin et al., 2004)]. FemA, possessing a coiled-coil region, catalyzes the transfer of the second and third residues of the pentaglycine side chain to lipid intermediate II. FemXWv, lacking the coiled-coil region, catalyzes the transfer of the first l-Ala residue to UDP-MurNAc-pentadepsipeptide. Excluding the coiled-coil region, FemA and FemXWv display 23% sequence identity and similar structures (rms deviation of 2.8 Å for 302 Cα atoms in common), suggesting that this fold can be generalized to all members of the family.

5

(a) Sequence and structural organization of Fem transferases. Domain 1 is represented in blue, domain 2 in green, and the coiled-coil region in red. The sequence numbering refers to the structure of FemA of Staphylococcus aureus. In FemXWv, the coiled coil is absent (corresponding numbering is 1–145 and 317–335 for domain 1, and 146–316 for domain 2). (b) Structure of domain 1 (top left), domain 2 (top right) of Weissella viridescens FemXWv, histone acetyltransferase of Tetrahymena thermophilia (bottom left), and serotonin acetyltransferase of Ovis ovaries (bottom right). The common secondary structures are represented in blue, absent or less frequent in yellow, and always present but different in green.

Fem proteins are organized into two structurally equivalent domains of about 170 amino acids separated by a cleft containing the binding site of the peptidoglycan precursor as revealed by the structure of the binary complex of FemXWv and UDP-MurNAc-pentapeptide [PDB code 1P4N (Biarrotte-Sorin et al., 2004)]. When present, the antiparallel two-helix coiled-coil moiety is inserted into nearly the middle of domain 2. Each domain consists of a mixed α/β structure centered on a twisted seven-stranded mixed polarity β sheet. The first N-terminal strand can be absent, as this is the case for domain 1 of FemXWv. Two helices flank the concave side of the β sheet, and three helices are localized on the opposite side. Domain 2 possesses a sixth α helix, inserted between the first N-terminal β strand and the first α helix of the core domain fold. The particularity of this two-domain organization is that domain 2 is inserted into domain 1, formed consequently by discontinuous sequence segments. From a structural point of view, this means that the last 20 C-terminal residues of the protein correspond to the N-terminal part of domain 1, providing an additional example of circular permutation, when comparing the structures of the two domains. The additional helix of domain 2 is structurally equivalent to the part where domain 1 is disrupted in sequence. To summarize, the fold of the Fem transferases is the result of a double insertion: the coiled-coil region, when present, is inserted into domain 2, which is itself inserted into domain 1 (Fig. 5a).

The fold of each domain is reminiscent of a large N-acetyltransferases superfamily, the GNAT, universally distributed in nature, which catalyze the transfer of an acetyl group from acetyl coenzyme A (acetylCoA) to a primary amine (Neuwald & Landsman, 1997). This family includes aminoglycoside, serotonin, glucosamine-6-phosphate, and the histone N-acetyltransferases (Fig. 5b). To date, about 25 members of the family have been structurally characterized (Vetting et al., 2005), revealing a conserved core fold composed of a six-stranded mixed polarity β sheet surrounded by four α helices, despite no detectable primary sequence homology. Differences between GNAT structures are generally confined to the immediate N-terminus, with much greater variation at the C-terminus, which can be extended significantly.

In addition to Fem aminoacyl transferases, duplication of the GNAT fold has been detected in mycothiol synthase, which transfers an acetyl group from acetylCoA to a mycothiol precursor in Mycobacterium tuberculosis, and in N-myristoyltransferase, which transfers a myristoyl group from myristoylCoA to the N-terminal glycyl residue of nascent eukaryotic polypeptides. The structure of mycothiol synthase in complex with acetylCoA revealed one cofactor molecule bound to each domain although only the C-terminal domain seems to be catalytically active (Vetting et al., 2003). In the case of N-myristoyltransferases, only one myristoylCoA is found in the structure of the complex, and it is located in the C-terminal GNAT domain, suggesting that when two GNAT folds are present in a structure, the C-terminal domain is the donor-binding site in the transfer reaction (transfer of an acetyl group in most cases or of a myristoyl group for N-myristoyltransferases).

Recently, the GNAT superfamily has been extended to include aminoacyl-transferases implicated in the N-end rule pathway of protein degradation. Based on sensitive sequence comparison, arginyltransferases have been proposed to be evolutionarily related to Fem transferases, and to consist of two GNAT domains (Rai et al., 2006). The C-terminal catalytic domain was predicted to bind the Arg-tRNAArg substrate. Two crystal structures of the Escherichia coli leucyl/phenylalanyl-tRNA-protein transferase (L/F-transferase) revealed that the C-terminal domain of the protein is topologically similar to the C-terminal GNAT domain of Fem transferases whereas the N-terminal domain is unrelated to GNAT (Suto et al., 2006; Dong et al., 2007). The C-terminal domain contains the aminoacyl-tRNA-binding site according to the structure of the complex of the L/F-transferase with puromycin, which mimics the 3′-termini of aminoacyl-tRNAs. In FemXWv, the peptidoglycan substrate is located at the cleft between the two domains and interacts mostly with residues in the N-terminal GNAT domain (Biarrotte-Sorin et al., 2004). The long channel running across the C-terminal GNAT domain of FemXWv has been proposed to be the tRNA-binding site.

In conclusion, it seems that transferases implicated in peptidoglycan synthesis and protein degradation have evolved from the ancient GNAT fold. The C-terminal GNAT domains of these aminoacyl-transferases have been tailored to use aminoacyl-tRNA instead of acetylCoA as the acyl donor. In addition, the N-terminal domain of FemXWv, while maintaining characteristics of the GNAT-fold, has lost its cofactor-binding role and gained a substrate-binding role.

Specificity for the peptidoglycan precursors

Depending on the specificity of Fem transferases, transfer of the amino acid residue can occur on the nucleotide (UDP-linked) precursor as well as on both lipid intermediates (Billot-Klein et al., 1997). In W. viridescens, UDP-MurNAc-pentadepsipeptide substituted by an l-Ala residue accounts for the substantial majority of the nucleotide precursors, whereas the cross-bridges contain the sequence l-Ala-l-Ser or l-Ala-l-Ser-l-Ala (Villet et al., 2007). Because FemXWv transfers a single residue in vitro, the second (l-Ser) and third (l-Ala) residues are added to the lipid intermediates by additional unknown Fem transferases (Villet et al., 2007). In vitro, FemX (FmhB) of Staphylococcus aureus transfers the first Gly of the pentaglycine side chain only onto the lipid intermediate II (Schneider et al., 2004). Accordingly, the nucleotide precursors extracted from the cytoplasm of Staphylococcus aureus do not contain any side chain amino acid (Billot-Klein et al., 1997). Thus, FemXWv from W. viridescens and FemX from Staphylococcus aureus act at different stages of the assembly of peptidoglycan precursors. Mixed specificities have been detected for other Fem enzymes. For example, the BppA1 and BppA2 transferases for synthesis of the l-Ala2 side chain of Enterococcus faecalis use nucleotide precursors in vitro albeit with a low catalytic efficiency (Bouhss et al., 2001, 2002). The physiological substrate of these enzymes is thought to be the lipid intermediates because only a small fraction of the nucleotide precursors extracted from the cytoplasm of Enterococcus faecalis are substituted by l-Ala. Fem transferases should efficiently interfere with the linear sequence of reactions that leads to formation of lipid I by MraY and of lipid II by MurG in order to assemble the side chain before the translocation of the precursor to the outer surface of the membrane by the unidentified ‘flippase’. The underlying mechanism could involve the substrate specificity of these enzymes and their direct interaction with Fem transferases in order to block processing of incomplete precursors. According to this hypothesis, MraY of W. viridescens would be specific of branched nucleotide precursors whereas the as yet unidentified flippase of Staphylococcus aureus would be specific of branched lipid II.

Structure-based site-directed mutagenesis of the UDP-MurNAc-pentapeptide-binding cavity of FemXWv revealed that a complex hydrogen bond network connecting two residues of the enzyme (Lys36 and Arg211) and two regions of UDP-MurNAc-pentapeptide (both phosphate groups and both d-Ala residues) constrains the substrate in a bent conformation essential for the aminoacyl transferase activity (Maillard et al., 2005). Testing substrate analogues confirmed that removal of the α-phosphate or of d-Ala5 severely impairs enzyme activity, whereas compounds lacking both phosphates or both d-Ala residues are not used as substrates by FemXWv. The mode of recognition of the peptidoglycan precursor may not be conserved because Lys36 and Arg211 of FemXWv are not systematically present in other Fem transferases. In addition, FemXWv catalyzes aminoacyl transfer to nucleotide precursors ending in d-Lac5 and d-Ala5 with similar efficiencies whereas FemX from Streptomyces coelicolor is specific for d-Ala5-ending precursors.

Specificity for the aminoacyl-tRNAs

Extensive analysis of the structure of mature peptidoglycan and of the cytoplasmic precursors of W. viridescens showed that FemXWv is highly specific for incorporation of l-Ala in vivo (Villet et al., 2007). FemXWv was also specific in vitro as l-Ser and Gly are transferred 100- and 4000-fold less efficiently than l-Ala, respectively. Comparison of various natural and in vitro-transcribed tRNAs revealed that the specificity of FemXWv depends mainly on the sequence of the tRNA body of the aminoacyl-tRNA substrate. In addition, unknown posttranscriptional modifications might act as negative (antideterminant) elements for Gly-tRNAGly because incorporation of Gly was 25-fold less efficient with natural tRNAs than with in vitro transcripts. Recognition of the esterified amino acid is most probably an additional identity determinant because Ser is transferred less efficiently than Ala and larger amino acids appear to be excluded from the active site. In combination, these specificity elements account for the specific incorporation of l-Ala into the peptidoglycan of W. viridescens.

The high in vitro specificity of FemXWv was observed with homologous tRNAs extracted from W. viridescens (Villet et al., 2007). FemXWv retained preferential incorporation of alanine with heterologous tRNAs extracted from related Gram-positive species, although the capacity of the enzyme to discriminate between Ala-tRNAAla, Ser-tRNASer, and Gly-tRNAGly was reduced. The use of heterologous Ala-tRNAAla and Ser-tRNASer from Escherichia coli in a previous study (Hegde & Blanchard, 2003) may therefore account for the reported poor discrimination of the two aminoacyl-tRNAs by FemXWv.

Deletion of different portions of the tRNA body of Ala-tRNAAla showed that the acceptor arm is sufficient for FemXWv activity (Hegde & Blanchard, 2003; Villet et al., 2007). Saturation mutagenesis of the acceptor arm revealed that C72 and C71 are required for base-specific recognition of the substrate by FemXWv (Villet et al., 2007). Because double mutations that restored Watson–Crick base pairing at positions 1•72 and 2•71 did not restore the aminoacyl transferase activity of FemXWv, destabilization of the acceptor arm was not responsible for loss of activity associated with substitutions at these positions. Ala-tRNAAla recognition by FemXWv did not involve the G3•U70 wobble base pair (Villet et al., 2007), which is essential for the alanyl-tRNA synthetase (Francklyn & Schimmel, 1989; Giege et al., 1998). Thus, the two enzymes display different modes of recognition of the acceptor stem (Villet et al., 2007). Different modes of recognition of the aminoacyl-tRNA substrate are also predicted for the L/F-transferase (Suto et al., 2006; Dong et al., 2007) and EF-Tu, which delivers aminoacyl-tRNAs to the ribosome during translation (Sanderson & Uhlenbeck, 2007). The L/F-transferase interacts with the tRNA body in a sequence-independent manner and amino acid specificity is solely mediated by a hydrophobic pocket that accommodates Leu and Phe. EF-Tu has roughly equivalent affinities for all elongator aminoacyl-tRNAs because the interactions of the protein with the tRNA body and the amino acid have compensatory roles. The sequence-dependent recognition of the substrate involves hydrogen bonding between Glu390 of EF-Tu and the amino group of a guanine located in the minor groove of the T-stem (base pair 51•63).

Synthesis of the side chain of peptidoglycan precursors by Fem transferases in Gram-positive bacteria implies that translation and peptidoglycan synthesis share common aminoacyl-tRNA synthetases and tRNAs. FemXWv is therefore expected to compete with EF-Tu for uptake of aminoacyl-tRNAs before its utilization by the ribosome. FemXWv might overcome this competition using different recognition sites in the aminoacyl-tRNAs as proposed for L/F transferase (Suto et al., 2006). According to this hypothesis, FemXWv would start to bind to its substrate before full dissociation of the EF-Tu:Ala-tRNAAla complex. Alternatively, the bacteria might produce tRNAs dedicated to peptidoglycan synthesis. Evidence for such tRNAs has been provided by sequence analyses that revealed the presence of ‘unusual’ tRNAs genes in staphylococci (Green & Vold, 1993; Thumm & Gotz, 1997). More directly, purification of tRNAGly from Staphylococcus epidermidis identified four isoacceptors that were all used as substrates by Fem transferases whereas two of these four isoacceptors failed to participate in protein synthesis by programmed ribosomes (Roberts et al., 1974; Roberts et al., 1974). Sequencing of the latter tRNAGly, which appear to be dedicated to peptidoglycan synthesis, showed that pseudouridine is replaced by U at position 55 in the T stem. The presence of an additional base pairing in the anticodon stem (C32•G38) reduces the length of the anticodon loop from seven to five bases and could therefore introduce a distortion preventing successful interaction with programmed ribosomes (Roberts et al., 1974). However, neither of the two modifications is expected to affect interaction with EF-Tu.

The in vivo impact of amino acid substitutions on the specificity of Fem transferases has been explored for the MurM enzyme of Streptococcus pneumoniae (Filipe et al., 2001). In this species, preferential incorporation of l-Ala or l-Ser at the first position of the side chain depends on the presence of specific alleles of the murM gene. As found previously for pbp genes, sequence diversity in murM is generated by homologous recombination between the host gene and gene fragments acquired form related streptococcal by natural transformation (Filipe et al., 2000b). Comparison of the sequence of the mosaic genes and site-directed mutagenesis identified the coiled-coil element of MurM as the identity determinant for incorporation of l-Ala or l-Ser (Filipe et al., 2001; Fiser et al., 2003). This analysis cannot be extrapolated to the model for the interaction of FemXWv with Ala-tRNAAla because the latter aminoacyl transferase does not contain any coiled-coil element (Biarrotte-Sorin et al., 2004).

Inhibitors of Fem transferases

Fem transferases are considered to be attractive targets for the development of novel antibiotics active against multi-resistant bacteria. These enzymes have a unique catalytic mechanism and are essential for the synthesis of the appropriate substrate for the d,d-transpeptidases, including low-affinity PBPs responsible for resistance to β-lactam antibiotics (Kopp et al., 1996; Fiser et al., 2003; Rohrer & Berger-Bachi, 2003). Fem transferases have not been studied extensively as drug targets, probably because of the complexity of their nucleotide substrate. Two approaches have been explored recently to obtain transition state (Cressina et al., 2007) and substrate (Chemama et al., 2007) analogues.

Synthetic phosphonate analogues of the putative tetrahedral transition state intermediary have been synthesized recently as inhibitors of Streptococcus pneumoniae MurM (Cressina et al., 2007). Because adenosine is found at the 3′-terminus of all tRNAs, a 3′-substituted adenosine phosphonate is a possible transition state mimic for this family of enzymes. A deoxyadenosine 3′-phosphonate analogue (Fig. 4b) was found to be an effective inhibitor (IC50 100 μM) for Streptococcus pneumoniae MurM (Cressina et al., 2007). The inhibition of MurM by this phosphonate analogue supports the existence of an oxyanion intermediate in the reaction catalyzed by Fem transferases.

A stable analogue of Ala-tRNAAla containing an oxadiazole ring as a mime of the 3′ amino acyl ester (Fig. 4c) inhibits FemXWv with an IC50 of 1.4 μM, a value similar to the Km for the Ala-tRNAAla substrate (1.7 μM) (Chemama et al., 2007). The inhibitor contains a 24-nucleotide micro-helix mimicking the acceptor stem of tRNAAla. Its synthesis involves enzymatic coupling of a 22-nucleotide RNA molecule with a dinucleotide containing the oxadiazole ring using the T4 RNA ligase. Interestingly, the route for synthesis of the modified oligonucleotides could be applicable to inhibitors of other important targets that use aminoacyl-tRNAs as substrates.

Essential role of Fem transferases

The specificity of Fem transferases is essential for the bacterium because incorporation of erroneous amino acids would produce precursors acting as chain terminators, thereby preventing formation of the stress-bearing peptidoglycan network (Frere & Joris, 1985). As described in section 2.5, l,d-transpeptidases are highly specific for their cognate acceptor substrate because substitutions in the side chain abolish in vitro formation of peptidoglycan dimers (Magnet et al., 2007a). The specificity of the classical d,d-transpeptidases from Gram-positive bacteria has not been assessed directly because these enzymes are generally inactive in vitro, except in very special cases involving highly reactive substrates (e.g. thioester) or atypical enzymes [e.g. the soluble R61 d,d-peptidase from Streptomyces spp. (Anderson et al., 2003; Josephine et al., 2006)]. For this reason, the specificity of the d,d-transpeptidases was studied in vivo by evaluating the impact of production of incomplete side chains on peptidoglycan cross-linking following inactivation or conditional expression of fem genes. Heterospecific expression of fem and pbp genes was also used to test the functionality of different combinations of peptidoglycan precursors and cross-linking enzymes. This section reviews the information gained on the in vivo substrate specificity of the d,d-transpeptidases, an approach that relies on sequencing of the cross-bridges by tandem mass spectrometry to establish the participation of specific side chains to the transpeptidation reaction.

In Staphylococcus aureus, conditional expression of the femX (fmhB) gene showed that the aminoacyl transferase for incorporation of the first residue of the pentaglycine side chain is an essential enzyme because the d,d-transpeptidases cannot directly cross-link unsubstituted stem pentapeptide (Tschierske et al., 1997). In contrast, femA and femB null mutants are viable and produced cross-bridges containing 1 and 3 Gly residues. However, inactivation of these genes abolishes expression of methicillin resistance mediated by low-affinity PBP2a. This observation indirectly shows that PBP2a cannot cross-link precursors containing an incomplete side chain. Streptococcus pneumoniae is unusual because direct cross-links coexist with cross-bridges consisting of l-Ala-l-Ala and l-Ser-l-Ala (Filipe & Tomasz, 2000; Filipe et al., 2000a, b; Fiser, 2003). The relative proportion of the direct and indirect cross-links as well as the relative proportion of l-Ala or l-Ser vary between different clinical isolates due to the presence of different mosaic alleles of murM. The peptidoglycan of clinical isolates resistant to penicillin contains a high proportion of indirect cross-links and inactivation of murM leads to susceptibility to this antibiotic. Thus, resistance to penicillin in Streptococcus pneumoniae appears to be mediated by combinations of mosaic murM and pbp genes that encode highly active aminoacyl-transferases and low-affinity PBPs specific for branched peptidoglycan precursors, respectively. In Enterococcus faecalis, deletion of bppA2 leads to production of precursors substituted by a single l-Ala and impaired PBP5-mediated expression of intrinsic β-lactam resistance whereas inactivation of bppA1 has not been obtained, suggesting that the Fem transferase for incorporation of the first l-Ala is an essential enzyme (Bouhss et al., 2002). These observations suggest that a complete side chain is required for the cross-linking activity of low-affinity PBPs in Staphylococcus aureus, Streptococcus pneumoniae, and Enterococcus faecalis, whereas unsubstituted stem peptides are not cross-linked by any PBP in Staphylococcus aureus and Enterococcus faecalis.

In Enterococcus faecalis, heterospecific expression of the fmhB, femA, and femB genes of Staphylococcus aureus showed that the PBPs cross-link side chains containing from 1 to 5 residues and Gly instead of l-Ala at its N-terminus (Arbeloa et al., 2004a). The substitution of d-iAsx by l-Ala was obtained in the cross-bridges of Enterococcus faecium based on heterospecific expression of bppA1 from Enterococcus faecalis (Magnet et al., 2007a). A similar substitution, d-iAsx by l-Ala-l-Ala and l-Ser-l-Ala, was also found to be tolerated following heterospecific expression of the murMN operon of Streptococcus pneumoniae in Lactococcus lactis (Veiga et al., 2006). Finally, heterospecific of the Enterococcus faeciumd-asapartate ligase in Enterococcus faecalis indicated that the opposite substitution (l-Ala by d-iAsp) was also tolerated (Bellais et al., 2006). The aspartate ligase was shown to be essential in L. lactis, and directly cross-linked stem peptides were detected neither in L. lactis nor in the enterococci (Arbeloa, 2004a; Bellais et al., 2006; Magnet et al., 2007a). These observations indicate that the d,d-transpeptidases cross-link side chains of various lengths ending in amino acids of the d and l configurations as well as glycine whereas unsubstituted stem peptides are not used for cross-bridge formation in vivo. The essential role of the side chain might therefore be to provide a minimum spacing between the stem peptides participating in the cross-linking reaction.

In order to assess the substrate specificity of low-affinity PBPs, the mecA gene encoding PBP2a of Staphylococcus aureus and the pbp5 genes of Enterococcus faecalis and Enterococcus faecium were expressed in enterococcal hosts devoid of their cognate pbp5 genes (Arbeloa et al., 2004a). The PBP5 of Enterococcus faecium did not confer β-lactam resistance in Enterococcus faecalis, whereas the PBP5 of Enterococcus faecalis tolerated substitution of l-Ala-l-Ala by d-iAsx in Enterococcus faecium. PBP2a of Staphylococcus aureus conferred β-lactam resistance in Enterococcus faecalis and Enterococcus faecium, indicating that this low-affinity PBP was functional in spite of substitution of the pentaglycine side chain by l-Ala-l-Ala and d-iAsx. Inactivation of the femA and femB genes (above) led to conflicting results because the complete pentaglycine side chain was judged to be essential for PBP2a-mediated methicillin resistance in Staphylococcus aureus (Henze et al., 1993; Ehlert et al., 1997; Stranden et al., 1997). Thus, the essential role of the femA and femB genes in methicillin resistance may not depend solely on on the specificity of PBP2a, raising the possibility that interaction of the d,d-transpeptidase with the acyl acceptor substrate could involve additional components of the peptidoglycan polymerization complexes.

Resistance to endopeptidases

The pentaglycine cross-bridges of staphylococci are cleaved between the third and fourth glycine residues by lysostaphin and related endopeptidases (Sugai et al., 1997a, b). Staphylococcal strains that produce such lytic enzymes also produce additional Fem aminoacyl transferases that mediate incorporation of Ser instead of Gly at the third and fifth positions of the side chains and thereby act as immunity factors because the endopeptidases do not cleave Ser-Gly and Gly–Ser peptide bonds (Sugai et al., 1997a, b; Thumm & Gotz, 1997; Ehlert et al., 2000). The endopeptidases and the immunity factors generally form gene clusters located on plasmid and are therefore most probably transferable between staphylococci. These observations strongly suggest that resistance to endopeptidases produced by competing bacteria provides the selective pressure for diversification of the side chain of peptidoglycan precursors.

Concluding remarks

In Gram-negative bacteria, the peptidoglycan is polymerized in a periplasmic space protected from the environment by the outer membrane that efficiently excludes the lytic endopeptidases and the glycopeptides. The outer membrane also participates in β-lactam resistance following alteration of the porins, often in association with β-lactamases. In contrast, the peptidoglycan and the polymerization complexes are constantly exposed to these bactericidal agents in Gram-positive bacteria that have developed complex resistance strategies based on diversification of the structure of the peptidoglycan and its mode of polymerization. In the case of glycopeptide resistance, the flexibility of the biosynthesis pathway involves the recruitment of genes from various origins to constitute complex gene clusters able to redirect the assembly of peptidoglycan precursors terminating in d-lactate or d-Ser in various hosts. In the case of β-lactam resistance, the flexibility of the biosynthesis pathway has been revealed by mutagenic activation of a novel transpeptidation pathway that plays only a marginal role in peptidoglycan maintenance in wild-type bacteria. The pathway can also confer cross-resistance to glycopeptides because the l,d-transpeptidase, which by-passes the PBPs, uses as the acyl donor substrate a tetrapeptide that lacks the peptidyl-d-Ala-d-Ala motif essential for drug binding. Finally, in the case of resistance to lytic enzymes, the flexibility of the biosynthesis pathway involves production of highly diverse l-amino acid transferases (Fem) and d-amino acid ligases (Asl) that render the cross-bridges resistant to their endopeptidase activity. The flexibility of the peptidoglycan assembly pathway implies that several enzyme activities previously thought to be essential for viability, such as the d,d-transpeptidase and glycosyltransferase activities of the PBPs, or the d-Ala-d-Ala ligase activity of classical Ddl enzymes, are only conditionally essential because accessory proteins can by-pass their function. Conversely, accessory proteins may become essential for drug resistance, or even irreversibly replace host functions, as proposed for d-Ala-d-lactate ligase from enterococci constitutively resistant to glycopeptides. The accessory proteins are potential targets for the development of novel antibiotics, as discussed for the Fem transferases. However, antibiotics directed against accessory factors will have narrowspectra, limited to a particular pathogen or to a particular resistance mechanism, unlike all therapeutically important antibiotics developed so far. The advantage of such narrow-spectrum antibiotics would be to limit the collateral damage caused by broad-spectrum antibiotics, which destroy commensal flora and select for resistance in opportunistic pathogens.

Acknowledgements

This work was supported by the Fondation pour la Recherche Médicale (Equipe FRM 2006 DEQ200661107918 and fin de thèse to RV) and the European Community (EUR-INTAFAR, Project No. LSHM-CT-2004-512138, 6th PCRD). The authors thank R. Quintiliani for critically reading the manuscript and S. Biarrotte-Sorin for the design of Fig. 5B.

Footnotes

  • Editor: Arie van der Ende

References

View Abstract