Correspondence: Eric Sauvage, Centre d'Ingénierie des Protéines, Institut de Physique B5a, University of Liège, B-4000 Sart Tilman, Belgium. Tel.: +32 43663620; fax: +32 43663748; e-mail: firstname.lastname@example.org
Penicillin-binding proteins (PBPs) have been scrutinized for over 40 years. Recent structural information on PBPs together with the ongoing long-term biochemical experimental investigations, and results from more recent techniques such as protein localization by green fluorescent protein-fusion immunofluorescence or double-hybrid assay, have brought our understanding of the last stages of the peptidoglycan biosynthesis to an outstanding level that allows a broad outlook on the properties of these enzymes. Details are emerging regarding the interaction between the peptidoglycan-synthesizing PBPs and the peptidoglycan, their mesh net-like product that surrounds and protects bacteria. This review focuses on the detailed structure of PBPs and their implication in peptidoglycan synthesis, maturation and recycling. An overview of the content in PBPs of some bacteria is provided with an emphasis on comparing the biochemical properties of homologous PBPs (orthologues) belonging to different bacteria.
Bacterial peptidoglycan not only enables the bacteria to resist the intracellular pressure of several atmospheres that exists in the cell but also provides the bacterium with a well defined cell-shape that is reproduced from generation to generation. The peptidoglycan is made of glycan chains of alternating N-acetylglucosamine and N-acetylmuramic acid cross-linked by short stem peptides attached to the N-acetylmuramic acid (Ghuysen, 1968; Schleifer & Kandler, 1972; Vollmer et al., 2008a). Penicillin-binding proteins (PBPs) catalyze the polymerization of the glycan strand (transglycosylation) and the cross-linking between glycan chains (transpeptidation). Some PBPs can hydrolyze the last d-alanine of stem pentapeptides (dd-carboxypeptidation) or hydrolyze the peptide bond connecting two glycan strands (endopeptidation). Endopeptidation and transpeptidation are reverse activities. Because of the structural resemblance between their natural substrate, the d-Ala-d-Ala end of the stem pentapeptide precursors, and penicillin, the late stage peptidoglycan synthesizing enzymes are sensitive to penicillin with which they form a long-lived acyl-enzyme that impairs their peptidoglycan cross-linking capability (Tipper & Strominger, 1965).
Classification and overview of the content in PBPs of selected bacteria
Bacteria possess a variable number of PBPs (Fig. 1). Because the sequencing of many bacterial genomes has been achieved, the number of PBPs of each bacterium is well determined. PBPs have been divided into two main categories, the high molecular mass (HMM) PBPs and the low molecular mass (LMM) PBPs. HMM PBPs are multimodular PBPs responsible for peptidoglycan polymerization and insertion into pre-existing cell wall (Goffin & Ghuysen, 1998; Born et al., 2006). Their topology consists of a cytoplasmic tail, a transmembrane anchor, and two domains joined by a β-rich linker located on the outer surface of the cytoplasmic membrane where peptidoglycan synthesis takes place (Goffin & Ghuysen, 1998; Macheboeuf et al., 2006; Lovering et al., 2007). Depending on the structure and the catalytic activity of their N-terminal domain, they belong either to class A or class B PBPs. The C-terminal penicillin-binding (PB) domain of both classes has a transpeptidase activity, catalyzing peptide cross-linking between two adjacent glycan chains. In class A, the N-terminal domain is responsible for their glycosyltransferase activity, catalyzing the elongation of uncross-linked glycan chains. In class B, the N-terminal domain is believed to play a role in cell morphogenesis by interacting with other proteins involved in the cell cycle (Höltje, 1998; den Blaauwen et al., 2008; Zapun et al., 2008a). Monofunctional enzymes (MGTs) similar to the glycosyltransferase domain of class A PBPs also exist in some bacteria but their exact role is still unknown (Spratt et al., 1996).
LMM PBPs are sometimes referred to as LMM PBPs of class A, B and C. As a whole LMM PBPs are frequently described with the general term of class C PBPs, sometimes with C1, C2 and C3 as subdivisions. We will rather use four subcategories and will refer to their main PBP representative in Escherichia coli, i.e. type-4 for PBPs similar to Escherichia coli PBP4, type-5 for enzymes similar to Escherichia coli PBP5, type-7 for PBPs similar to Escherichia coli PBP7 and type-AmpH for PBPs similar to Escherichia coli AmpH.
The numbering of PBPs is historically based on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) migration and this may lead to some confusion. For example, Staphylococcus aureus PBP2 is a class A PBP similar to Escherichia coli PBP1a, Staphylococcus aureus PBP3 is similar to Escherichia coli PBP2 and Staphylococcus aureus PBP1 is similar to Escherichia coli PBP3. The complete set of PBPs of 10 important and widely studied bacteria, with their numbering and grouping in subclasses on the basis of their sequence is provided in Fig. 1. The classification used herein is based on amino acid sequence alignment aided by the knowledge of structural features.
The 12 PBPs of Escherichia coli have been the subject of numerous investigations. Escherichia coli possesses three class A PBPs (PBP1a, PBP1b and PBP1c) and two class B PBPs (PBP2 and PBP3). PBP1a and PBP1b are the major transpeptidases-transglycosylases. Deletion of one of them is not lethal for the bacteria (Suzuki et al., 1978; Denome et al., 1999; Meberg et al., 2001). The role of PBP1c is not well understood. PBP1c is unaffected by most β-lactams and overexpression of PBP1c does not suppress the autolysis phenotype of a mutant lacking both PBP1a and PBP1b (Schiffer & Holtje, 1999). The two class B PBPs of Escherichia coli are monofunctional transpeptidases. PBP2 is involved in the ‘elongase’, a dynamic protein complex specific to cell elongation while PBP3 is a major protein of the divisome, the cell division complex (these complexes are dealt with in den Blaauwen et al., 2008).
The seven LMM PBPs of Escherichia coli are involved in cell separation, peptidoglycan maturation or recycling. PBP4 and PBP7 are two endopeptidases that cleave cross-bridges between two glycan strands. They probably should be considered as hydrolase members of the autolysins pool of Escherichia coli (Vollmer & Holtje, 2004; Vollmer et al., 2008b). PBP5 is the major carboxypeptidase, i.e., the most abundant, and it only cleaves the terminal d-Ala-d-Ala bond, making the stem peptide unavailable for transpeptidation (Spratt & Strominger, 1976). PBP6 and PBP6b both have sequences homologous to PBP5 and their activity is, like PBP5, that of a strict carboxypeptidase. Both PBP4b and AmpH have a sequence close to the paradigmatic Streptomyces R61 dd-peptidase (Henderson et al., 1997; Vega & Ayala, 2006). The role of PBP4b is undetermined whereas AmpH is associated with peptidoglycan recycling.
Bacillus subtilis is the model organism for sporulating Gram-positive bacteria. Most of its 16 PBPs have been extensively studied for their role in vegetative peptidoglycan synthesis and in sporulation. Bacillus subtilis has four class A PBPs. PBP1 is part of the cell division machinery and is required for the efficient formation of the asymmetric sporulation septum (Scheffers & Errington, 2004). Among the six class B PBPs, PBP2b is the cell division specific class B transpeptidase (Daniel et al., 2000). PBP4a is the equivalent of Escherichia coli PBP4 and PBP5 is the major carboxypeptidase. Two other carboxypeptidases similar to PBP5 are present in B. subtilis, PBP5* and dacF. PBP5* is required for proper spore cortex synthesis (Popham et al., 1995) and dacF regulates the degree of cross-linking of spore peptidoglycan (Popham et al., 1999). PBP4* and PBPX, two class C PBPs of AmpH-type, are supposed to be involved somehow in sporulation (Scheffers, 2005). Of note is the absence of type-7 PBP in B. subtilis.
Staphylococcus aureus is a Gram-positive coccus. It incorporates peptidoglycan at the division site and has no elongase complex (Pinho & Errington, 2003; Zapun et al., 2008a). Consistently, its unique class A PBP localizes to the septum. Strains of Staphylococcus aureus susceptible to β-lactam antibiotics have two class B PBPs (Pinho et al., 2000; Pereira et al., 2007), but resistant strains have acquired an additional PBP, PBP2a, with a low sensitivity to β-lactams (Zapun et al., 2008b). Staphylococcus aureus has only one LMM PBP, which is of type-5, but unlike Escherichia coli PBP5, Staphylococcus aureus PBP4 has a transpeptidase activity necessary to achieve the high degree of cross-linking observed in the peptidoglycan of staphylococci (Wyke et al., 1981).
Streptococcus pneumoniae has three class A PBPs (Hoskins et al., 1999), two class B PBPs and one type-5 class C PBP (Morlot et al., 2005). Unlike Staphylococcus aureus, Enterococcus faecalis or L. monocytogenes, Streptococcus pneumoniae does not have a penicillin-resistant PBP resembling Staphylococcus aureus PBP2a, and the resistance to β-lactams in Streptococcus pneumoniae arises from the alteration of the sequence and structure of its PBPs (Hakenbeck, 2000; Hotomi et al., 2006; Zapun et al., 2008b).
Probably because of their complex life cycle and their ability to produce β-lactam molecules [e.g. Streptomyces clavuligerus produces cephamycin and clavulanic acid (Ward & Hodgson, 1993)], members of the Streptomyces genus have a great number of PBPs that can be expressed at the different stages of the bacterial development or when the bacterium effectively produces β-lactam metabolites that can interfere with peptidoglycan biosynthesis. Streptococcus coelicolor has 21 PBPs: three class A, nine class B, and nine class C PBPs. The time schedule of their expression is unknown.
Mycobacterium tuberculosis produces two class A PBPs, two class B PBPs and a lipoprotein sharing some motifs with the class B PBPs (Goffin & Ghuysen, 2002). Six class C PBPs, one type-4, one type-5, one type-7 and three putative type-AmpH, complete the set of PBPs of M. tuberculosis.
The cyanobacterium Anabaena sp. PCC7120 has 12 PBPs. Six class A PBPs, two class B PBPs, two type-4 PBPs and two type-AmpH PBPs (Lazaro et al., 2001; Leganes et al., 2005). Strikingly, Anabaena sp. is devoid of type-5 (and type-7) class C PBP, a property shared by all cyanobacteria (Leganes et al., 2005)
The PB domain and the dd-peptidase activity
PBPs share a common dd-peptidase activity, whether a dd-transpeptidase, a dd-carboxypeptidase or a dd-endopeptidase activity (Ghuysen, 1991; Goffin & Ghuysen, 1998). The carboxypeptidation and transpeptidation reactions catalyzed by PBPs follow a three-step mechanism: the rapid, reversible formation of a noncovalent Henri–Michaelis complex between the enzyme and a peptidoglycan stem pentapeptide [l-Ala-γ-d-Glu-A2pm (or l-Lys)-d-Ala-d-Ala], called the donor strand, is followed by the attack of the active serine on the carbonyl carbon atom of the C-terminal d-Ala-d-Ala peptide bond, leading to the formation of an acyl-enzyme intermediate and the concomitant release of the C-terminal d-Ala (acylation). The final step (deacylation) consists of either hydrolysis, with release of the shortened peptide (carboxypeptidation), or cross-link formation with a second peptidoglycan stem peptide called the acceptor strand (transpeptidation). The dd-endopeptidase activity of PBPs consists of the hydrolysis of the cross-bridge resulting from the dd-transpeptidase activity.
The dd-peptidase activity of PBPs is catalyzed by a common PB domain, which binds β-lactam antibiotics (Fig. 2). The latter are aimed at inhibiting the enzymatic dd-peptidase activity of the PB domain, by playing on the structural similarity between penicillin and the d-Ala-d-Ala dipeptide that ends the natural substrate of PBPs, the pentapeptide precursors of the peptidoglycan (Tipper & Strominger, 1965; Ghuysen, 1991). The PB domain is made of two subdomains, a five stranded β-sheet covered by three α-helices and an all-helical domain. The active site lies at the interface of these two subdomains (Fig. 3a). There may be some flexibility between these two subdomains, and this can influence the binding capability of some PBPs towards various ligands (Lim & Strynadka, 2002). The active site encompasses nine residues broadly conserved in PBPs. The active serine is positioned at the beginning of helix α2 and is followed by a lysine to form a S*xxK motif. A second motif, SxN, is situated in a loop between helices α4 and α5. Four conserved residues form the third motif KTG(T/S), a ninth residue, a glycine situated in the rear of the active site, are also strictly conserved (Fig. 3a).
Structural information regarding the interaction of PBPs with their substrate mainly comes from structures of LMM PBPs in complex with substrates that mimic the natural substrate. McDonough et al. (2002) obtained two high resolution structures with the dd-peptidase of Streptomyces R61, one showing the noncovalent binding of the two products of the carboxypeptidase reaction and the other showing the enzyme–substrate Henri–Michaelis complex resulting from the utilization of an inactivated enzyme. Nicola et al. reported a 1.6 Å resolution structure of Escherichia coli PBP5 in complex with a substrate-like peptide boronic acid. The boronyl group mimics the transition state of a PBP-catalyzed deacylation reaction (Nicola et al., 2005). We recently reported on the structure of B. subtilis PBP4a with α-aminopimelyl-ɛ-d-alanyl-d-alanine, a peptide that mimics the peptidoglycan ending tripeptide of Bacillus (Sauvage et al., 2007). This structure shows the acyl-enzyme α-aminopimelyl-ɛ-d-alanyl-PBP4a and an unbound d-alanine, the result of the PBP-catalyzed acylation reaction.
From these structures, it emerges that the penultimate d-alanine of the donor stem peptide fits tightly in the active site with its amide group wedged between the side chain of the asparagine of the second motif and the backbone of the β3 strand that lines the active site (Fig. 3b). The carbonyl oxygen of the penultimate d-alanine lies in the oxyanion hole and its methyl group is inserted into a hydrophobic pocket, underlining the importance of the conserved glycine at the rear of the active site. The carboxylate of the leaving d-alanine is oriented towards the two hydroxyl groups of the third motif. In brief, from the nine conserved residues, the asparagine of the second motif and both hydroxyl group of the third motif are important for the correct positioning of the substrate. The glycine of the third motif is needed to avoid steric hindrance as a bulkier residue would block the entry of an incoming substrate to the active site. The glycine in the rear of the active site is important for the binding specifity of PBPs towards the penultimate d-alanine of the peptidoglycan stem peptide (Adediran et al., 2005).
The acylation of PBPs by the penultimate d-alanine of the donor strand, with the concomitant release of the last d-alanine, requires the abstraction of a proton from the active serine. The lysine of the first motif, assumed deprotonated, can perform the withdrawal of the proton (Gordon et al., 2000; Lim & Strynadka, 2002; Nicola et al., 2005; Macheboeuf et al., 2006; Sauvage et al., 2007). The proton can subsequently be back-donated to the leaving d-alanine amine group via the serine of the second motif. Alternatively, the direct transfer of a proton from the active serine to the serine of the second motif concomitant to the transfer of a proton from the latter to the amine group of the leaving d-alanine in a one step process may be considered (Dive & Dehareng, 1999).
In the deacylation step, a proton of the acceptor (an amino group in transpeptidation, water in carboxypeptidation) has to be abstracted to allow the activated group to attack the carbonyl carbon atom of the ester bond of the acyl-enzyme (Ghuysen, 1991; Goffin & Ghuysen, 1998). The proton can subsequently be back-donated to the active serine. Withdrawal of a proton from the acceptor group can be performed by the lysine of the third motif, which needs to be in a deprotonated form (Nicola et al., 2005). Extraction of the proton can also be carried out by the serine of the second motif with the help of the lysine of the third motif. In the dd-peptidase of Streptomyces R61, the tyrosine of the second motif, the hydroxyl group of which is positioned in the active site, the same way that the hydroxyl group of the serine belonging to the SxN motif of other PBPs, is believed to be responsible for deacylation of this enzyme (Silvaggi et al., 2003).
The active site of PBPs can be seen as a double lysine–serine system, one for acylation and one for deacylation, although both systems are not independent. Simple catalytic hydroxyl/amine dyads (Lys–Ser) are frequently encountered in proteins (Paetzel & Dalbey, 1997; Lee et al., 2007). Withdrawal of a proton from the serine by the deprotonated lysine might be a too simplistic view. The proton must lie somewhere between the hydroxyl oxygen atom of the serine and the nitrogen atom of the lysine, and the resulting orientation of the lone pairs of the oxygen should be crucial for the catalytic mechanism. Quantum level calculation on lysine–serine dyads is worthy of consideration to analyze the mechanisms of acylation and deacylation of PBPs.
Crystallographers have been successful in trapping acyl-enzymes of PBPs and β-lactam antibiotics. Most of the structures of PBPs have been solved in the apo form and in complex with β-lactams (Kuzin et al., 1995; Gordon et al., 2000; Lim & Strynadka, 2002; Sauvage et al., 2002, 2005; Macheboeuf et al., 2005; Nicola et al., 2005; Silvaggi et al., 2005; Kishida et al., 2006). X-ray structures show that many antibiotics covalently linked to the active site serine adopt a common standard positioning that shares some characteristics with the acyl-enzyme PBP4a-α-aminopimelyl-ɛ-d-alanyl: (1) the amide group of the side chain is inserted between the asparagine of the second motif and the backbone of β3 strand, (2) the carboxylate associated to the thiazolidine or to the dihydrothiazine ring of the antibiotic is hydrogen bonded to one or both hydroxyl groups of the KTGT motif, and (3) the carbonyl oxygen lies in the oxyanion hole. The acyl enzyme PBP2a-benzylpenicillin (Lim & Strynadka, 2002) is a typical example of this ‘canonical’ conformation (Fig. 3c). The catalytic mechanism can follow the scheme described for PBPs with the d-Ala-d-Ala end of the stem pentapeptide although a different mechanism could be at work.
Class A PBPs and MGTs
From sequence alignment, class A PBPs can be grouped in at least seven subclasses (Fig. 1). As defined by Goffin & Ghuysen (1998), subclasses A1 and A2 group Gram-negative PBPs, whereas subclasses A3, A4, and A5 form three clusters of Gram-positive PBPs. Subclass A6 contains unusual PBPs such as Escherichia coli PBP1c and Anabaena sp. PBP2. The three Streptococcus coelicolor class A PBPs are grouped with the penicillin-resistant M. tuberculosis PBP1a in subclass A7.
The presence of a class A PBP is necessary for cell growth in most but not all bacteria. Loss of Escherichia coli PBP1a or PBP1b is tolerated but loss of both proteins is lethal even in the presence of PBP1c (Denome et al., 1999). Deletion of PBP1, the only class A PBP in N. gonorrhoeae, results in the loss of viability (Ropp et al., 2002). In Streptococcus pneumoniae, the three class A PBPs are dispensable individually but PBP1a or PBP2a is required for growth in vitro (Hoskins et al., 1999).
The absence of all its class A PBPs is not lethal to B. subtilis. A strain lacking all four class A PBPs is still viable and produces a peptidoglycan with only small structural differences from that of the wild type. The growth rate of the quadruple mutant is much lower than that of a strain lacking only three of the class A PBPs and wall abnormalities are more frequent (McPherson & Popham, 2003). A situation similar to the one observed in B. subtilis prevails in Enterococcus faecalis. Deletion of the three class A PBPs is not lethal although this leads to an increase in generation time and a decrease in peptidoglycan cross-linking. As there is no monofunctional glycosyltransferase in Enterococcus faecalis, the glycan chain polymerization in the triple mutant must be performed by a novel type of glycosyltransferase, that is not inhibited by moenomycin (Arbeloa et al., 2004).
The role of Escherichia coli PBP1c is not well understood. It is a penicillin-insensitive class A PBP that binds only specific β-lactams. Its transpeptidase activity could not be measured, which suggests that it may function in vivo as a glycosyltransferase only. Interestingly, the pbpB gene coding for Anabaena sp. strain PCC7120 PBP2, a class A PBP with a sequence similar to Escherichia coli PBP1c, is required for anaerobic nitrogen fixation (Lazaro et al., 2001). A PBP similar to PBP1c exists only in heterocyst-forming filamentous cyanobacteria (Leganes et al., 2005).
In their attempt to determine the structure of a class A PBP, Dessen and colleagues crystallized both PBP1a and PBP1b of Streptococcus pneumoniae (Macheboeuf et al., 2005; Contreras-Martel et al., 2006). At the same time, the group of Strynadka also solved the structure of PBP1b of Streptococcus pneumoniae (Lovering et al., 2006). The structure of the transpeptidase domain associated with a linker domain could be determined but the structure of the transglycosylase domain remained undeciphered. The determination of an X-ray structure of a class A PBP (Staphylococcus aureus PBP2) with its transglycosylase domain was achieved recently by the group of Strynadka (Lovering et al., 2007). The general fold of class A PBPs is made of an N-terminal domain coupled to the C-terminal PB domain. The interdomain linker is composed of a small β-sheet, and one α helix (Fig. 4a). A small number of class A PBPs, e.g. the (penicillin-resistant) PBP1 of M. tuberculosis, contain an additional C-terminal domain made of one or two repeating units known as Penicillin-binding protein And Serine/Threonine kinase Associated domains (PASTA), because this domain is also found in the C-termini of serine/threonine kinases (Yeats et al., 2002). The PASTA domain is a small globular domain consisting of three β strands and one α-helix (Young et al., 2003). An Fn3 (fibronectin type III) domain can also be found at the C-terminus of some class A PBPs whereas a FHA (forkhead-associated) domain, consisting of a typical 11-stranded β sandwich fold, may be found at the N-terminus. This observation suggests a more complex organization and other functions acquired by these PBPs. The significance and role of these domains are still unknown, but most often Fn3 domains are involved in some manner in cell surface binding and FHA domains are protein–protein interaction domains.
The global fold of the transpeptidase domain, as observed in PBP1a and PBP1b of Streptococcus pneumoniae and PBP2 of Staphylococcus aureus, is identical to the general structure of the PB domain described previously. However, both structures described for Streptococcus pneumoniae PBP1b show important conformational differences in the active site conformation of the apoenzyme. Macheboeuf et al. (2005) observed a ‘closed’ active site in the absence of substrate or antibiotic. Residues following the KTGT motif on strand β3 move away from strand β4 and an asparagine situated in the loop connecting these two strands makes hydrogen bond contacts with the backbone of a loop situated on the other side of the active site, thus blocking the entry of the active site. When crystals of PBP1b are grown in the presence of a substrate, then washed and back-soaked in the presence of nitrocefin or cefotaxime, the active site retains a ‘canonical’ configuration and the antibiotic forms with the PBP1b a classical acyl-enzyme. Analysis of the structure solved by Strynadka et al. reveals an ‘open’ active site which resembles the acyl-enzyme active-site topology and is more similar to the one observed in the PB domain of other PBPs (Lovering et al., 2006). It has been suggested that both the observed apoenzyme structures may illustrate a conformational sampling of the ‘open’ and ‘closed’ forms that may play a regulatory role in PBP1b catalytic activity.
Despite the difficulties of finding well-behaved glycosyltransferases, the efforts deployed by several laboratories recently led to the determination of two X-ray structures: the bifunctional Staphylococcus aureus PBP2 (lacking the cytoplasmic and transmembrane anchor) in apo form and in complex with moenomycin (Fig. 4b) (Lovering et al., 2007), and the isolated glycosyltransferase domain of Aquifex aeolicus PBP1a (Yuan et al., 2007). The two structures revealed that the glycosyltransferase domain contains almost only α-helices and its fold resembles that of phage λ-lysozyme but differs from all known glycosyltransferase structures. Like lysozyme, the glycosyltransferase domain consists of a large and a small lobe separated by an extended cleft that contains the active site (Fig. 4a). An important feature of the glycosyltransferase domain is the presence of a hydrophobic region that probably mediates its interaction with the membrane and the lipid II substrate.
Sequence alignments have revealed that all glycosyltransferase domains shared five conserved motifs. The first three motifs are found in the catalytic cleft. Motif 1 (EDxxFxxHxG) and motif 3 (RKxxE) contain the two conserved catalytic glutamic acids (Fig. 4b). The essential glutamic acid in motif 1 of the glycosyltransferase domain and the catalytic Glu19 of phage λ-lysozyme are located at the same position. Motif 2 which divides the cleft in two pockets is more likely to be involved in substrate recognition. Motif 4 forms the back wall of the cleft and motif 5 is located in the large lobe farther from the active site. They were both suggested to play mainly a structural role (Lovering et al., 2007). Comparison of the ligand-free and moenomycin bound PBP2 structures show that moenomycin induces conformational change in the small subdomain and that some flexibility occurs around the linker region connecting the glycosyltransferase and the transpeptidase domains.
Monofunctional enzymes (MGTs) similar to the glycosyltransferase domain of class A PBPs also exist in some bacteria but their exact role is still unknown (Spratt et al., 1996). Both the glycosyltransferase domain of class A PBPs and MGTs belong to GT51 family, which is characterized by the five conserved motifs and uses lipid II [undecaprenyl-P-P-MurNAc-(pentapeptide)-GlcNAc] as substrate (for a description of lipid II synthesis, see Bouhss et al., 2007). The MGTs represent about 20% of the sequences in the GT51 database (http://CAZy.org) showing that the majority of the glycosyltransferase are bifunctional PBPs. This proportion reflects the distribution and the number of class A PBPs and MGT among bacteria.
In class A PBPs, the N-terminal domain is responsible for their glycosyltransferase activity, catalyzing the elongation of uncross-linked glycan chains of the peptidoglycan. In recent years, a substantial progress has been made in the synthesis of lipid II (either radioactive or labeled with different fluorescent probes) using enzymatic route (van Heijenoort et al., 1992; Breukink et al., 2003), chemoenzymatic synthesis (Ye et al., 2001; Schouten et al., 2006) or complete chemical synthesis (Schwartz et al., 2001, 2002; VanNieuwenhze et al., 2002). The availability of a sufficient amount of substrate allowed significant progress in the characterization of glycosyltransferase enzymes. About a dozen glycosyltransferase proteins have now been successfully purified and characterized in vitro with lipid II (Table 1).
Table 1. Catalytic efficiency of glycosyltransferases
Properties of class A PBPs and glycosyltransferase domain derivatives
Escherichia coli PBP1a and 1b use monomeric-, tri-, tetra-, and pentapeptide as an acceptor in the in vitro assay transpeptidase reaction with lipid II as substrate (Bertsche et al., 2005; Born et al., 2006). The glycan strands polymerized by PBP1a have an average length of 20 disaccharide units with 18–26% of the peptides involved in cross-linking (Born et al., 2006). In addition, PBP1a catalyzes the attachment of nascent peptidoglycan to mature cell wall by transpeptidation (Born et al., 2006). The catalytic efficiency of the PBP1b glycosyltransferase is about 10 times higher than that of PBP1a (Table 1) (Schwartz et al., 2002), and the length of glycan strands obtained by polymerization is >25 disaccharide units with almost 50% of the peptides engaged in cross-linking (Bertsche et al., 2005). The activity of the two PBPs is inhibited by moenomycin in the nanomolar range (Terrak et al., 1999; Chen et al., 2003; Welzel, 2005). Different glycosyltransferase domain derivatives of PBP1b were produced. They show similar moenomycin sensitivity and their catalytic activity is lower or comparable to that of the full-length protein (Table 1). The C-terminal boundary of the glycosyltransferase domain was identified between residues 385 and 478 (Barrett et al., 2004).
The full length PBP2 of Staphylococcus aureus was purified and characterized with heptaprenyl-lipid II (C35) substrate (Barrett et al., 2005). It has a lower catalytic efficiency than Escherichia coli PBP1b and distinct properties from those of PBP1b in terms of pH and metal ions requirement.
The PBP4 of L. monocytogenes EGD was overexpressed in Escherichia coli and purified (Zawadzka-Skomial et al., 2006). Its catalytic efficiency is slightly lower than that of Escherichia coli PBP1a and Staphylococcus aureus PBP2 (Table 1). Disruption of the lmo2229 gene encoding PBP4 increased the resistance of L. monocytogenes EGD to moenomycin.
Streptococcus pneumoniae PBP1a, PBP1b and PBP2a were overexpressed without transmembrane segment (Δtm) as GST fusions in Escherichia coli, purified and characterized (Di Guilmi et al., 1998; Di Guilmi et al., 1999, 2003a, b). GST-PBP2a Δtm fusion required a detergent for solubility but not GST-PBP1b Δtm. A glycosyltransferase domain derivative of PBP1b (82–300) was also produced in the absence of the detergent. These three proteins catalyze the polymerization of peptidoglycan from dansyl lipid II substrate and the glycosyltransferase activity is inhibited by moenomycin (Table 1).
Because the expression of the full length PBP1a of A. aeolicus was poor (Yuan et al., 2007), three derivatives of its glycosyltransferase domain (29/51/67–243) were realized and expressed in Escherichia coli (Yuan et al., 2007). The 67–243 derivative catalyzes peptidoglycan polymerization from heptaprenyl lipid II with an efficiency similar to that of the glycosyltransferase domain of full size PBPs. The glycan chains have a length of at least 40 disaccharides. The 51–243 derivative leads to the X-ray structure solution. The role of residues Glu83, Asp84, His90 and Lys153 of the PBP1a glycosyltransferase domain were analyzed by mutagenesis. The results confirmed the role of the glutamic acid of motif 1 as the catalytic residue of the GT51 family (Terrak et al., 1999; Terrak & Nguyen-Disteche, 2006).
The glycosyltransferase domain of Thermotoga maritima PBP1a (Glu34–Thr244) was expressed and purified (Offant et al., 2006). The use of detergents and/or moenomycin helped in reducing the aggregation state and increasing the protein homogeneity. The truncated glycosyltransferase domain of PBP1a had polymerase activity with fluorescent lipid II which was inhibited by moenomycin (IC50 value of 4.8 μM).
Owing to their intrinsic hydrophobic properties, most class A PBPs have a tendency to aggregate even in the presence of detergents. The complex assay buffer mixture, often with dimethyl sulfoxide and a detergent, helps solubilize the lipid substrate and the protein but may not be the better condition for enzyme stability and activity. In addition, the assay systems and lipid II substrate used to measure the activity of the proteins shown in Table 1 were not all the same and these may account for some of the differences observed in catalytic rates between the proteins. Therefore, the in vitro measurement of the kinetic parameters of the glycosyltransferase is only helpful to perform biochemical in vitro studies but do not reflect their real activities in vivo.
Properties of monofunctional glycosyltransferases
A soluble and active form of Staphylococcus aureus MGT was overexpressed in Escherichia coli and purified (Wang et al., 2001; Liu & Wong, 2006; Terrak & Nguyen-Disteche, 2006). The enzyme catalyzes glycan chain polymerization from lipid II with an efficiency of 5800 M−1 s−1 and was sensitive to moenomycin. The properties of Staphylococcus aureus MGT in term of pH and ions requirement were similar to those of Escherichia coli PBP1b (Terrak & Nguyen-Disteche, 2006). A MGT able to polymerize uncross-linked PG in vitro from lipid II substrate was isolated from Escherichia coli (Hara & Suzuki, 1984). In contrast to Staphylococcus aureus MGT, Escherichia coli MGT was insensitive to moenomycin (Di Berardino et al., 1996).
The crystal structures of PBP2 and its complex with moenomycin, brought direct evidence that the lipid II is the acceptor and the growing chain is the donor (Lovering et al., 2007). From the kinetic characterization of Escherichia coli PBP1b (Schwartz et al., 2002) and its interaction with moenomycin (Welzel, 2005), mutagenesis studies (Terrak et al., 1999) and the recently determined crystal structure of Staphylococcus aureus PBP2 (Lovering et al., 2007), it was concluded that the mechanism of glycan chain elongation by the glycosyltransferase proceeds by successive attacks of the growing chain (donor) at the reducing end by lipid II (acceptor) catalyzed by deprotonation of the 4-OH nucleophile of GlcNAc by the active site base (Glu of motif 1), concomitant with stabilization of the leaving diphospho-undecaprenyl group by the second glutamate (motif 3) probably via a divalent metal (Schwartz et al., 2002). The exact role of metal ion in the activity of the glycosyltransferase needs to be clarified. It should be noted that apart from Staphylococcus aureus PBP2 (Barrett et al., 2005), the activity of most characterized glycosyltransferases was dependent on metal ions (Schwartz et al., 2002; Terrak & Nguyen-Disteche, 2006; Zawadzka-Skomial et al., 2006).
Relation transglycosylation/transpeptidation in peptidoglycan synthesis
For the bifunctional PBPs, the transglycosylation can proceed while the transpeptidase domain is either inhibited by penicillin, mutation or is completely deleted (Terrak et al., 1999; Born et al., 2006). In contrast, inactivation of the glycosyltransferase domain by mutation (Terrak et al., 1999; Born et al., 2006) or by moenomycin completely blocks the peptidoglycan polymerization. Study with Escherichia coli PBP1a showed that, for the transpeptidation to occur, it must be coupled to the transglycosylase activity. The two reactions are carried out by the same PBP molecule because the mixture of PBP1a mutated in the glycosyltransferase active site and the PBP1a mutated in the transpeptidase active site were inactive (Barrett et al., 2005). Time-course of PG synthesis in vitro by PBP1a and PBP1b showed that peptide cross-linking catalyzed by PBP1b occurs simultaneously with glycosyltransferase of lipid II (Bertsche et al., 2005). With PBP1a the transpeptidation is delayed until oligomerized glycan strands are formed indicating that PBP1a transpeptidase prefers the peptide of elongating glycan strands (Born et al., 2006).
Class B PBPs
Streptococcus pneumoniae PBP2x was the first HMM PBP whose crystal structure was solved (Pares et al., 1996). More recently, the structures of Staphylococcus aureus PBP2a (Lim & Strynadka, 2002) and Enterococcus faecium PBP5 (Sauvage et al., 2002) were solved at the same time. Staphylococcus aureus PBP2a and Enterococcus faecium PBP5 are orthologous proteins responsible for the high level of resistance encountered in Staphylococcus aureus and Enterococcus faecium strains. Sequence alignment allowed Goffin & Ghuysen (1998) to define five subclasses (Fig. 1). Staphylococcus aureus PBP2a and Enterococcus faecium PBP5 are in the subclass B1 and all the PBPs in this subclass are supposed to show a low affinity for penicillin (Zapun et al., 2008b). The subclass B2 contains an elongase complex specific to PBP2 type of Gram-negative bacteria (e.g. Escherichia coli PBP2), whereas the subclass B3 contains divisome specific to PBP3 type of Gram-negative bacteria (e.g. Escherichia coli PBP3) (den Blaauwen et al., 2008). Streptococcus pneumoniae PBP2x is part of subclass B4 together with other enzymes involved in division of Gram-positive bacteria (Zapun et al., 2008a), and subclass B5 contains enzymes from Gram-positive not directly involved in septation. Subclasses B-like I, II and III are defined by Goffin & Ghuysen (2002) and contains PBPs from mycobacteria, streptomycetes and related bacteria.
The overall fold of class B PBPs is made of a complex N-terminal domain coupled to the C-terminal transpeptidase domain. In the transpeptidase domain, the residue following motif 3 [KTG(T/S)] is, with very few exceptions, an alanine in class B PBPs [KTG(T/S)A] while it is a threonine or a serine in class A PBPs [KTG(T/S)(T/S)], and this might be related to their specific role as class A or class B PBPs. Gram-negative PBP2 types are also distinguished in their active site by the presence of an aspartic acid at the third position of motif 2 (SxD). The other subclasses have an asparagine as in the canonical active site of PBPs (SxN).
The N-terminal domain was hypothesized to interact with other proteins (Höltje, 1998) or to serve as a pedestal for PBPs to reach their target (Macheboeuf et al., 2006). Small loops or domains may interact with other proteins as shown by a complementation analysis of the PBP3 deficiency in Enterococcus hirae (Leimanis et al., 2006). The interdomain linker is made of four to six β-strands. Five conserved motifs (I–V) in or close to the linker constitute the signature of class B PBPs (Goffin & Ghuysen, 1998). The N-terminal domain should be seen as small subdomains or loops that are connected to the interdomain linker (Fig. 5). The number of residues in each loop or subdomain between the conserved motifs is characteristic of each subclass. In the absence of a B2 PBP structure, the features distinguishing Escherichia coli PBP2 from Escherichia coli PBP3 are unknown. Interestingly, in Gram-negative bacteria, length of the loop between motifs II and III is similar in all B2 PBPs and in all B3 PBPs but differs between type-2 and -3. This holds as well for the loop between motifs IV and V, but the significance of such characteristics awaits further structural information.
The length of the loop between motifs III and IV (Fig. 5) is similar in subclasses B1 and B2. As well, the loop between motifs IV and V and the loop between motifs V and I of the transpeptidase domain are of same length in subclasses B1 and B2 whereas they have different length in the other subclasses.
Role of class B PBPs in peptidoglycan biosynthesis
Escherichia coli PBP2 is essential for cell elongation and shape maintenance. PBP2 localizes in the lateral wall and its inhibition with mecillinam results in spherical cells which are unable to form colonies (Spratt & Pardee, 1975). PBP2 also localizes at the division site, as shown by using a functional green fluorescent protein (GFP)-PBP2 (Den Blaauwen et al., 2003). Nevertheless PBP2 is not a stable partner of the cell division machinery of Escherichia coli because its localization at the constriction site disappears just before the cell separation. It was suggested that PBP2 might play a transient role in the initiation of the cell division (Den Blaauwen et al., 2003).
Sequence alignment suggests that enzymes of subclass B1 are the equivalent in Gram-positive bacteria of Escherichia coli PBP2 but experimental evidence with B. subtilis PBP2a and PBPH suggests rather that the equivalent of Gram-negative B2 PBPs are enzymes of subclass B5 (Murray et al., 1998; Wei et al., 2003). Bacillus subtilis PBP2a is required for normal outgrowth of spores (Murray et al., 1998). PBP2a and PBPH play redundant roles in determining the rod cell shape and the activity of one of these proteins is required for viability (Wei et al., 2003). In the double mutant, septa are formed extremely irregularly and cells have a greater diameter (Wei et al., 2003). PBPH is expressed most strongly in late log phase and during the transition to stationary phase and may play a unique role during that part of the life cycle.
In Gram-positive cocci, the situation is different. There is no apparent phenotype associated with the inactivation of pbp3 (PBP3) in Staphylococcus aureus regarding growth rate, cell wall muropeptide profile, and impact on methicillin resistance in a methicillin-resistant Staphylococcus aureus strain. The only observable effect was a decrease in autolysis rate (Pinho et al., 2000). The problem of the presence of two or three class B PBPs in Gram-positive cocci and the role of class B PBPs not involved in bacterial division is addressed in Zapun et al. (2008a)
Gram-positive division specific PBPs belong to subclass B4. They all possess a supplementary domain made of two repeating units known as PASTA domains (cf class A PBPs). An interesting example is the serine/threonine kinase PknB of M. tuberculosis that phosphorylates PBPA, a class B PBP required for cell division (Dasgupta et al., 2006). The PASTA domain is believed to bind unlinked peptidoglycan because a cefuroxime was observed associated to the first PASTA domain in the X-ray structure of Streptococcus pneumoniae PBP2x (Gordon et al., 2000) but further experimental evidence is needed to prove this assertion.
Mycobacterium tuberculosis has one B3 PBP (pbpB or mtu3), one in subclass B-like I (PBPA) and one in subclass B-like II, which is annotated as PB lipoprotein. Streptococcus coelicolor has one PBP (PBP2, sco2608) classified in subclass B2, but with a weak similarity to other members, and one in subclass B3 (PBP3, sco2090). Streptococcus coelicolor has also four B-like I and three B-like III PBPs. Anabaena sp. PCC7120 has one B2 PBP (PBP5, alr5045) and one type-3 PBP (PBP6, alr0718). Their sequences are close to those of their respective Escherichia coli orthologues. The roles of these proteins are largely unknown.
Resistance to β-lactams
Class B PBPs play an important role in the resistance to β-lactams of many bacteria. One resistance mechanism found in some Gram-positive bacteria is the presence of an endogenous or acquired penicillin-resistant PBP that can take over the transpeptidase function of all other PBPs, like Staphylococcus aureus PBP2a and Enterococcus faecium PBP5. Strikingly a similar PBP in L. monocytogenes (lmo0441) is responsible for the resistance to monobactams and some cephalosporins but not penicillin (Guinane et al., 2006).
Mutations or mosaic gene transfers are also frequently observed with naturally competent bacteria, mainly involving the division specific class B PBP (B3 PBP). Alterations of the N. gonorrhoeae or meningitidis PBP2 (a B3 PBP) is associated with reduced susceptibility to penicillin G (Dougherty et al., 1980; Barbour, 1981; Spratt, 1988; Antignac et al., 2001), a situation that also exists in Pseudomonas aeruginosa, Streptococcus pneumoniae and Bacteroides fragilis (Ayala et al., 2005). This situation also occurs in class A PBPs. For example, alterations in PBP 1A confer resistance to β-lactam antibiotics in Helicobacter pylori (Kwon et al., 2003). Details are given in the accompanying paper (Zapun et al., 2008b).
Class C PBPs of type-4
There is generally one or zero type-4 class C PBP in bacteria. Of note is the absence of this type of protein in Gram-positive cocci like staphylococci, enterococci, streptococci or Listeria. Streptococcus coelicolor and some cyanobacteria are the only presumably known bacteria possessing two type-4 PBPs.
Three X-ray structures of type-4 PBPs are available to date: the dd-peptidase of Actinomadura R39 (Sauvage et al., 2005), Escherichia coli PBP4 (Kishida et al., 2006) and B. subtilis PBP4a (Sauvage et al., 2007) (Fig. 6). They show a great similarity in their overall folding. The transpeptidase domain is associated with two other domains, which are not in N- or C-terminal positions but are inserted in the transpeptidase domain in the way of matryoschka dolls (Sauvage et al., 2005), i.e. the third domain is inserted in the second domain, which itself is inserted in the PB domain between the conserved motifs 1 and 2 (of the PB domain). The role of these domains is unknown. Domain II has the topology of half a Rossmann fold, which is also the case of the N-terminal domain of MinC and the 1A region of FtsA, two proteins interacting with FtsZ and involved in the regulation of the septum formation in cell division, but the resemblance of these proteins to the domain II of type-4 PBPs remains limited (Sauvage et al., 2005). Interestingly a positively charged surface appears in domain II of the dd-peptidase of Actinomadura R39 and B. subtilis PBP4a (Fig. 6), and it has been suggested that this positive surface could interact with the teichoic acid present in Bacillus and Actinomycetales (Sauvage et al., 2007).
Domain III consists of two three-stranded β-sheets facing each other. Strikingly, the number of residues between motifs 1 and 2 is reduced by 50 for type-4 PBPs of some bacteria (Streptomyces avermitilis, Streptococcus coelicolor, Corynebacterium diphtheriae, M. tuberculosis, Anabaena sp., …), which corresponds approximatively to the length of the sequence of domain III. Sequence alignment supports the idea that this domain is absent from those bacteria.
The surroundings of the active site are specific to type-4 PBPs. The residue following motif 3 is invariably a leucine or a methionine, a hydrophobic residue turned inside the cavities that pushes the C-terminus of β3 strand slightly to the forefront of the active site. Together with the residues pertaining to domain II, the residues of the C-terminus of β3 strand form a pocket that can accommodate the terminal H3N+–CH–COO− group of the diaminopimelic acid, the antepenultimate amino-acid of the peptidoglycan stem peptide (Sauvage et al., 2007). The residues involved in the binding pocket are widely conserved with some variations (Sauvage et al., 2005).
Type-4 PBPs are not anchored into the cytoplasmic membrane via a transmembrane helix (Gittins et al., 1994; Harris et al., 2002; Stefanova et al., 2003). Nearly all of Escherichia coli PBP4 in overproducing cells is in a soluble form (Korat et al., 1991; Gittins et al., 1994). Neisseria gonorrhoeae PBP3 is a putative outer membrane protein that interacts with peptidoglycan (Judd et al., 1991; Shafer & Judd, 1991). The dd-peptidase of Actinomadura R39 is found in the culture medium, and B. subtilis PBP4a is easily washed away from the cells with a buffer containing 1 M KCl (C. Duez, pers. commun.). These results suggest that type-4 PBPs are very loosely associated with the cytoplasmic membrane. They could be associated directly with the peptidoglycan as suggested for PBP4a. Gram-positive type-4 PBPs could interact with anionic polymers that decorate the peptidoglycan whereas Gram-negative type-4 PBPs could be associated with the outer membrane or directly with the peptidoglycan.
Localization studies by green fluorescent protein fusions have shown that B. subtilis PBP4a is recruited to the lateral wall and is absent from the septum (Scheffers et al., 2004). Still the fact remains that a type-4 PBP is present in N. gonorrhoeae, a coccoïd bacterium without lateral wall.
The absence of a membrane anchoring the helix and the in vivo endopeptidase activity are two arguments in favour of type-4 PBPs being envisioned as members of the pool of autolysins (Höltje, 1995). Type-4 PBPs could be inserted into the peptidoglycan or exposed on the outer face of peptidoglycan where it would exert its endopeptidase activity. It could thereby be indirectly involved in cell morphology (Meberg et al., 2004), in daughter cells separation (Priyadarshini et al., 2006), and could be implicated in biofilm formation (Gallant et al., 2005).
Class C PBPs of type-5
We distinguish the groups of type-5 and type-7 class C PBPs on the basis of the presence in type-5 PBPs of a C-terminal domain that was shown in Escherichia coli to be essential for the correct functioning of the PBP5 (Nelson & Young, 2001). The PBP5 C-terminal domain ends with an amphipathic helix that associates PBP5 to the cytoplasmic membrane. The truncated PBP5, lacking the C-terminal domain, is soluble (like PBP7) and its production leads to cell lysis (Nelson & Young, 2001).
The crystal structures of Escherichia coli PBP5 (Davies et al., 2001; Nicholas et al., 2003) (Fig. 7a), Streptococcus pneumoniae PBP3 (Morlot et al., 2005) and Staphylococcus aureus PBP4 (K.R. Rajashankar, unpublished data) have been determined. The overall fold of type-5 PBPs is biglobular. The PB domain is associated with a β-strand rich C-terminal domain that ends with an associated membrane amphipathic helix. The PB domain lacks the N-terminal helix and a small loop extends on the top of the active site (green in Fig. 7a).
Role of type-5 PBPs in peptidoglycan biosynthesis
With the exception of Staphylococcus aureus PBP4, type-5 PBPs are strict dd-carboxypeptidases unable to catalyze a transpeptidation reaction (Matsuhashi et al., 1979). They play a major role in the control of cell diameter and correct septum formation. Cells of an Escherichia coli mutant lacking only PBP5 had aberrant morphology (Meberg et al., 2004), which was exacerbated by the deletion of other dd-carboxypeptidases (Nelson & Young, 2000, 2001). In L. monocytogenes, cells lacking PBP5 displayed an irregular morphology and shape (Guinane et al., 2006), and a thicker cell wall (Korsak et al., 2005b). Moreover the ratio of pentapeptides to tripeptides was increased in cells lacking PBP5 (Korsak et al., 2005b), as in a dacA (PBP5) mutant of B. subtilis (Atrih et al., 1999). In Streptococcus pneumoniae, PBP3 plays a major role in organizing the growth and division process (Morlot et al., 2004). PBP3 is present on the entire surface of the cell except at the septum, where the pentapeptides are left intact as substrates of the division specific PBPs.
In contrast to previous results, Staphylococcus aureus PBP4 was shown to have a secondary transpeptidase activity (Wyke et al., 1981). Disruption of pbpD, the structural gene of Staphylococcus aureus PBP4, caused a large decrease in the highly cross-linked muropeptide components of the cell wall (Leski & Tomasz, 2005). The striking difference between the activity, function and physiological role of Staphylococcus aureus PBP4 compared with the concordant activity, function and physiological role of its structural equivalent in other bacteria remains to be explained. One possibility is that Staphylococcus aureus PBP4 plays a similar role in restricting the availability of pentapeptides to its division specific PBPs, but instead of simply removing the ultimate d-alanine of the stem peptide, PBP4 utilizes the available d-Ala-d-Ala to further cross-link the glycan chains.
Bacteria most often have a major type-5 PBP which is the most abundant LMM PBP they produce. Escherichia coli produces two other type-5 PBPs (Fig. 1). The expression level of Escherichia coli PBP6 considerably rises in stationary-phase cells compared with exponentially growing cells (Buchanan & Sowell, 1982). The level of transcription of dacA (PBP5), dacC (PBP6) and ampC (AmpC) is controlled by the bolA gene (Santos et al., 2002). PBP6 is supposed to play a major role at the onset of stationary-phase (van der Linden et al., 1992). Interestingly, by constructing mosaic proteins in which a 20 amino acids sequence were grafted from PBP5 to PBP6, and vice versa, Ghosh & Young (2003) showed that the shape maintaining characteristics of PBP5 relies on residues following the KTGT motif.
Bacillus subtilis produces three type-5 PBPs (Fig. 1). PBP5 (dacA) is the major carboxypeptidase found in vegetative cells (Lawrence & Strominger, 1970). PBP5 has no role in spore peptidoglycan synthesis but PBP5* (dacB) and DacF (dacF) both function in regulating the degree of cross-linking of spore peptidoglycan (Popham et al., 1999). dacB is expressed only in the mother cell compartment of the developing sporangium whereas dacF is expressed in the forespore compartment and they can act differently on the nascent spore peptidoglycan (Popham et al., 1999).
Class C PBPs of type-7
As we mention for type-5 PBPs, Escherichia coli PBP7 and Escherichia coli PBP5 have homologous sequences. The structure of PBP7 differs from that of PBP5 by the absence of the C-terminal domain and the amphipathic helix that anchors PBP5 to the cytoplasmic membrane.
The crystal structures of M. tuberculosis PBP7 (I. Krieger, unpublished data) and the dd-peptidase of Streptomyces K15 (K15, Fig. 7b) (Fonzéet al., 1999) have been determined. The sequence of K15 is slightly different from the sequences of the other type-7 PBPs (Palomeque-Messia et al., 1991) and close orthologues have been found only in Streptococcus coelicolor and Streptomyces avermitilis. The global fold of M. tuberculosis PBP7 and K15 proteins is similar to the PB domain of Escherichia coli PBP5. Like type-5 PBPs, type-7 PBPs lack the N-terminal helix and exhibit on the top of the active site a β-hairpin protuberance, which in K15 is relatively long and was hypothesized to anchor the protein to the plasma membrane (Fonzéet al., 1999) (green in Fig. 7b).
Role of type-7 PBPs in peptidoglycan biosynthesis
Escherichia coli PBP7 was shown to be loosely associated with the membrane and totally released in the supernatant in the presence of 1 M NaCl (Romeis & Holtje, 1994a). PBP7 is an endopeptidase that hydrolyzes the d-alanyl-ɛ-meso-2,6-diaminopimelyl cross-bridge bond in high-molecular mass sacculi (not with soluble LMM muropeptides) but had no carboxypeptidase activity. Interestingly, a mutant of Escherichia coli lacking PBP7 showed no morphological change but a double mutant ΔPBP5–ΔPBP7 was more abnormal than the single mutant lacking only PBP5 (Meberg et al., 2004). Escherichia coli PBP8, a proteolytic degradation product of PBP7, was also shown to stabilize the soluble lytic transglycosylase Slt70, and indirectly interact with PBP3 (Romeis & Holtje, 1994b). As suggested for type-4 PBPs, Escherichia coli PBP7 should probably be looked at as an autolysin but its precise role remains unclear and the difference between the role of Escherichia coli PBP4 and the role of Escherichia coli PBP7 is undetermined (Vollmer et al., 2008b). Both are endopeptidases implicated in cell morphology (Meberg et al., 2004) and in daughter cells' separation.
Neisseria gonorrhoeae PBP4 is close in sequence to Escherichia coli PBP7. It lacks the characteristic C-terminal domain of type-5 PBPs. PBP4 has a carboxypeptidase activity on l-Lys-d-Ala-d-Ala based substrates, preferentially with Nɛ substituted substrates. It showed no transpeptidase activity in a transpeptidation reaction with Ac2-l-Lys-d-Ala-d-Ala as a donor and glycine as an acceptor. Endopeptidation was not tested. One of the two LMM PBPs (PBP3 or PBP4) can be deleted without severely affecting the cells but the removal of both of them leads to a modest decrease of cell growth and a change in morphology, suggesting that N. gonorrhoeae PBP3 and PBP4 have a redundant function in normal cell wall biosynthesis (Stefanova et al., 2004).
Type-7 PBPs are absent from the Gram-positive bacteria that we have analyzed, with the exception of the vanYDdd-carboxypeptidase (Arthur et al., 1994; Casadewall et al., 2001; Reynolds et al., 2001). The vanYDdd-carboxypeptidase is included in the vanD gene cluster responsible for the resistance to vancomycin of enterococci (Courvalin, 2006). VanYD lacks the C-terminal domain of Escherichia coli PBP5 and hence cannot be associated with the membrane via the C-terminal amphiphatic helix characteristic of type-5 PBPs. The hydrophobicity profile of VanYD rather suggests that there is a single transmembrane segment close to the N-terminus (Reynolds et al., 2001).
The in vivo endopeptidation activity and the carboxypeptidation/transpeptidation activities might result from the location of the protein active site at the outer face (endopeptidase) or inner face (carboxypeptidase/transpeptidase) of the peptidoglycan.
Although lacking transmembrane segments, K15 enzyme is associated with the cytoplasmic membrane. Its overexpression results in the secretion of c. 30% of the synthesized enzyme in the culture medium. K15 can be considered as a strict transpeptidase, for which the rate of transpeptidation depends on the nature of both the scissile bond (peptide, thiolester, or ester) of the carbonyl donor and the acceptor. When assayed on Ac2-l-Lys-d-Ala-d-Ala, the enzyme behaves as a poor hydrolase. It utilizes the small amount of the released d-Ala as an acceptor to catalyze a silent exchange between the tripeptide C-terminal d-Ala and the free amino acid. In the presence of Gly–Gly or Gly-l-Ala, which mimic the structure of the acceptor in the nascent peptidoglycan, virtually no hydrolysis product is formed, and the enzyme behaves as a strict and efficient transpeptidase (Nguyen-Disteche et al., 1982; Fonzéet al., 1999).
Class C PBPs of type-AmpH
The structure of type-AmpH PBPs is mainly based on the structure of the dd-peptidase of Streptomyces R61 (R61) (reviewed in Frère, 2004). R61 folding is similar to the PB domain described for the other classes of PBPs although differences occur in loops or even secondary structures (Fig. 8a). The active site lies at the interface of the two subdomains and the first motif SxxK is similar to other PBPs. The conserved serine of the second motif (SxN) is, in type-AmpH PBPs, replaced by a tyrosine and this is the main feature that distinguishes the enzymes of this family. Nevertheless, when superposing the active site of R61 with that of other PBPs, the tyrosine hydroxyl group of R61 is roughly at the same position as the hydroxyl group of the serine of the second motif of the other PBPs. The third residue of the second motif is generally an asparagine but is an aspartic acid in Escherichia coli PBP4b. The third motif [KTG(T/S)] is not well conserved in this family. A histidine is frequently found at the first position but a lysine can also occur, as in Escherichia coli AmpH. The second position also suffers variations. An arginine and a serine residue are respectively found in the two type-AmpH PBPs (PBP9 and PBP12) of the cyanobacterium Anabaena sp. PCC7120, whereas the type-AmpH PBP of Enterococcus faecalis have a HGG motif.
The exact enzymatic function of these enzymes is generally not determined except in a few cases and we cannot be sure that these proteins have a role in peptidoglycan biosynthesis or that they have another function, for example a β-lactamase function or a d-aminopeptidase function. A d-amino acid amidase from Ochrobactrum anthropi was shown to have a fold similar to that of R61 (Bompard-Gilles et al., 2000; Okazaki et al., 2007). Because a loop impairs its active site entrance, the d-aminopeptidase lacks transpeptidase and carboxypeptidase activities (Fig. 8b).
Class C β-lactamases
Class C β-lactamases also have a structure close to R61, and although β-lactamases are not included in our analysis of PBPs, we will briefly discuss them because it has been suggested that the AmpC β-lactamase may have an additional cellular function as a peptidoglycan hydrolase (Bishop & Weiner, 1993; Henderson et al., 1997). From the high structural similarity, first observed between R61 and the class C β-lactamase of Enterobacter cloacae P99, has emerged the hypothesis of an evolution from a common ancestral enzyme. The evolution scenario includes structural changes leading in β-lactamases to rejection of the d-methyl substituent of the penultimate d-alanine residue of the dd-peptidase substrate. However molecular modeling studies have shown that the d-alanyl methyl group fits tightly into the space originally occupied by the motif 2 tyrosine side chain and thus allows the bound substrate to adopt a conformation similar to that observed in the R61 active site, which has a hydrophobic pocket for this substituent. Accommodation of the penultimate d-alanyl methyl group seems therefore to be necessary, but not sufficient for an efficient dd-peptidase activity (Adediran et al., 2005).
In Escherichia coli, ampC deletion mutant exhibited a decline in growth rate at mid-exponential phase that could be delayed by the expression of AmpC at the early-exponential phase. Maintenance and expression of the ampC gene represents an unbearable cost for Salmonella enterica in terms of reduction of the growth rate and invasiveness (Morosini et al., 2000). The deleterious AmpC burden could be eliminated by decreasing the production of AmpC when both regulatory ampR and ampC genes were present in Salmonella enterica. Even if AmpC synthesis did not produce major variations in the peptidoglycan composition of Salmonella enterica, data suggested that overexpression may interfere with normal PG synthesis or turnover.
Role of type-AmpH PBPs in peptidoglycan biosynthesis
Although closely related to AmpC and other class C β-lactamases, and despite the fact that they strongly bind penicillin, AmpH and PBP4b show no β-lactamase activity (Henderson et al., 1997; Vega & Ayala, 2006). Mutation of the ampC and/or ampH genes in Escherichia coli lacking PBPs 1a and 5 produced morphologically aberrant cells, particularly in cell filaments induced by aztreonam. Thus, these traits suggest that AmpC and AmpH play roles in the normal course of peptidoglycan synthesis, remodeling or recycling. PBP4b has been shown to have a weak dd-carboxypeptidase activity in vitro (Vega & Ayala, 2006), but extensive assays to identify any enzymatic activity with natural isolated muropeptides have been unsuccessful (J. Ayala, unpublished results).
On the opposite, in B. subtilis, a weak β-lactamase activity is associated with PBP4* (Popham & Setlow, 1993). Dynamic localization of PBP4* and PBPX, in fusion with the GFP, during spore development, indicated that PBP4* would have a function in sporulation. Moreover PBPX seems to change its localization and to accumulate specifically at the prespore. It starts by localizing to the division septum at midcell, and then spirals out in a pattern similar to FtsZ. GFP–PBPX is seen to appear at both poles, as observed for cell division proteins like FtsZ and the sporulation-specific SpoIIE. Despite these observations, PBPX does not seem to be an essential component of the cell division machinery of B. subtilis. Indeed, the inactivation of PBPX does not affect cell growth, or cell shape or sporulation efficiency. A possible role for the PBPX could be the quick removal of PG connecting two cells after vegetative division, or thinning of the sporulation septum before engulfment (Scheffers, 2005).
Resting on the information reviewed in the previous chapters, and using a peptidoglycan model based on glycan chains running parallel to the cytoplasmic membrane, we propose a simplified version of the three-to-one model of peptidoglycan biosynthesis described by Höltje (1998), i.e. a model for the incorporation of a single glycan chain without considering peptidoglycan recycling, which is observed in Gram-negative bacteria. The purpose of this model is to specify for each type of PBPs a role in peptidoglycan biosynthesis and in the incorporation of new glycan chains in the pre-existing peptidoglycan (Fig. 9a). For a review on peptidoglycan structure, see Vollmer et al. (2008a)
Lipid II is synthesized in the cytoplasm by the mur family enzymes (Barreteau et al., 2008), modified in the membrane (Bouhss et al., 2007), and then transferred to the class A PBP glycosyltransferase active site where it is transglycosylated onto a donor strand that may run along the surface of the whole class A PBP (Fig. 9b). The donor strand can adopt a three- or fourfold rotation axis (Höltje, 1998; Meroueh et al., 2006) while running along the class A PBP surface. The stem peptide of a disaccharide unit pentapeptide of the elongating strand (about nine disaccharide units from the glycosyltransferase active site) can be inserted into the groove of the PB domain active site where it may undergo a transpeptidase attack from a pre-existing peptidoglycan transpeptidase-donor strand (Fig. 9b). The orientation of the active site of the transpeptidase domain suggests that class A PBPs can hook the elongating glycan strand beneath the old peptidoglycan.
Although class A PBPs have a transpeptidase domain, the transpeptidase domain of a class B PBP is needed. The precise role of each transpeptidase domain in peptidoglycan biosynthesis is not known and we hypothesize that class B PBPs undertake to cross-link the glycan chain hooked by class A PBPs beneath the old peptidoglycan to a second strand of the peptidoglycan (Fig. 9c). After hydrolysis by an endopeptidase of the bond that cross-links the two old glycan chains, the new strand reaches the surface as a result of the internal pressure. The endopeptidase activity of type-4 or -7 PBPs can serve to cleave the cross-links between the glycan chains (Fig. 10). These enzymes do not need to be directly associated with the elongase complex but they could be associated with other amidases. Moreover, these amidases can break the peptide that cross-links the old strands at different places, making type-4 and type-7 PBPs nonessential enzymes.
Type-5 PBPs remove the ultimate d-Ala of the pentapeptide. They seem not to be directly associated with the proteins of the divisome or the elongase complex. They are dispersed on the entire surface and prevent the septum formation at inappropriate places by removing the pentapeptide substrates. The C-terminal domain can act as a pedestal that positions the PB domain close to the d-Ala stem peptide end of the peptidoglycan (Fig. 11).
Since the resolution about 20 years ago of the S. R61 dd-peptidase structure, a considerable amount of structural data on PBPs have been made available by worldwide laboratories. The domain responsible for the transpeptidation reaction and penicillin binding is now well known, but the role and function of the additional domains are not established yet, except for the transglycosylase domain. Very recently the structure of a membrane class A PBP has been deciphered in its entirety, including both transglycosylase and transpeptidase domains, and further crystallographic experiments regarding the interaction between the transglycosylase domain and its substrates or specific inhibitors can now be envisaged.
Many complexes of PBPs with β-lactam antibiotics have been made available for a long time, and recently the structure of PBPs with peptides mimicking the peptidoglycan has been described. Keeping on with structural studies with fragments of cell wall including peptidic and saccharidic parts would be of interest. It is indeed essential to understand the reactions of transpeptidation and transglycosylation in order to better determine the physiological role of PBPs. Moreover, the experimental data on the structure of the peptidoglycan should bring additional and useful details for the comprehension of the interaction between PBPs and the peptidoglycan.
The future structural studies will have to consider complexes between PBPs and their partners in complex machineries such as the divisome or the elongase complex, in order to better understand the role of each protein in the phenomena of cell division and elongation. This study remains however, a challenge because many partners of these multi-enzymatic machineries are membrane proteins.
This work was supported in part by the Europan Commission Sixth Framework Program grants LSMH-CT-COBRA 2003-503335 and LSMH-CT-EUR-INTAFAR 2004-512138, by the Belgian Program on Interuniversity Poles of Attraction initiated by the Belgian State, Prime Minister's Office, Science Policy programming (IAP no. P6/19), by the Actions de Recherche Concertées (grant 03/08-297), by the Fonds National de la Recherche Scientifique (IISN 4.4505.00, FRFC 9.45/9.99, FRFC 2.4.508.01.F, FRFC 9.4.538.03.F, FRFC 2.4.524.03) and the University of Liège (Fonds spéciaux, Crédit classique, 1999). F.K. is Chargé de Recherche and M.T. is Chercheur Qualifié of the Fonds de la Recherche Scientifique (F.R.S.-FNRS, Brussels, Belgium).
E.S. and F.K. contributed equally to the work described in this paper.