OUP user menu

The biosynthesis of peptidoglycan lipid-linked intermediates

Ahmed Bouhss, Amy E. Trunkfield, Timothy D.H. Bugg, Dominique Mengin-Lecreulx
DOI: http://dx.doi.org/10.1111/j.1574-6976.2007.00089.x 208-233 First published online: 1 March 2008

Abstract

The biosynthesis of bacterial cell wall peptidoglycan is a complex process involving many different steps taking place in the cytoplasm (synthesis of the nucleotide precursors) and on the inner and outer sides of the cytoplasmic membrane (assembly and polymerization of the disaccharide-peptide monomer unit, respectively). This review summarizes the current knowledge on the membrane steps leading to the formation of the lipid II intermediate, i.e. the substrate of the polymerization reactions. It makes the point on past and recent data that have significantly contributed to the understanding of the biosynthesis of undecaprenyl phosphate, the carrier lipid required for the anchoring of the peptidoglycan hydrophilic units in the membrane, and to the characterization of the MraY and MurG enzymes which catalyze the successive transfers of the N-acetylmuramoyl-peptide and N-acetylglucosamine moieties onto the carrier lipid, respectively. Enzyme inhibitors and antibacterial compounds interfering with these essential metabolic steps and interesting targets are presented.

Keywords
  • peptidoglycan
  • undecaprenyl phosphate
  • UppS synthase
  • UppP phosphatases
  • MraY translocase
  • MurG transferase

Introduction

Peptidoglycan (murein) is a major heteropolymer of bacterial cell walls that consists in long glycan chains made of alternating units of N-acetylmuramoyl-peptides (MurNAc-peptides) and N-acetylglucosamine (GlcNAc) that are cross-linked together via the short peptide chains (Rogers et al., 1980; Park et al., 1996; Vollmer et al., 2008). The main essential function of this giant cell-sized macromolecule is to protect cells against the deleterious effects of the internal osmotic pressure. It also contributes to the maintenance of the characteristic cell shape and serves as a platform for the anchoring of other cell envelope components including proteins (Braun & Sieglin, 1970; Marraffini et al., 2006; Dramsi et al., 2008) and polysaccharides (Neuhaus & Baddiley, 2003).

The biosynthesis of peptidoglycan is a complex process involving many different cytoplasmic and membranes steps (van Heijenoort, 2001b). The first stage consists in the formation of the soluble nucleotide precursors, from UDP-GlcNAc to UDP-MurNAc-pentapeptide. In particular, the synthesis of the peptide moiety is performed by a series of enzymes designated as the Mur ligases (MurC, MurD, MurE and MurF) which are responsible for the respective additions of l-alanine, d-glutamic acid, meso-diaminopimelic acid (A2pm) or l-lysine, and d-alanyl-d-alanine to UDP-MurNAc (Barreteau et al., 2008). As reported by Schleifer & Kandler (1972) and exemplified in the accompanying reviews (Barreteau et al., 2008; Vollmer et al., 2008), the structure of this peptide and the substrate specificity of these enzymes exhibit some variations in the bacterial world. The membrane steps then begin with the transfer of the phospho-MurNAc-pentapeptide moiety from the cytoplasmic precursor to the membrane acceptor undecaprenyl phosphate (C55-P), a reaction catalyzed by the transferase MraY (also named translocase) yielding undecaprenyl-pyrophosphoryl-MurNAc-pentapeptide (lipid I) (Fig. 1). Thereafter, the transferase MurG catalyzes the transfer of the GlcNAc moiety from UDP-GlcNAc to lipid I yielding undecaprenyl-pyrophosphoryl-MurNAc-(pentapeptide)-GlcNAc (lipid II) which, after its passage through the membrane by a yet unknown mechanism, will be used as the substrate for the polymerization reactions (van Heijenoort, 2001a, b) (Fig. 1). The C55-P carrier lipid plays a central role in these steps and is also required for the synthesis of other cell-wall polymers. This lipid as well as all the enzymes participating in peptidoglycan synthesis are essential, specific for the bacterial world and therefore constitute interesting potential targets to be exploited for the discovery of new antibacterials.

1

Membrane steps of peptidoglycan biosynthesis. M, G and the five colored beads linked to M represent MurNAc, GlcNAc and the pentapeptide, respectively. C55-PP and C55-P are for undecaprenyl pyrophosphate and undecaprenyl phosphate, respectively.

This review summarizes recent advances in the understanding of the membrane steps of peptidoglycan synthesis. In particular, it will focus on recent significant contributions to the knowledge of the metabolism of the carrier lipid and on progress made toward the biochemical and structural characterization of the MraY and MurG activities and the search of inhibitors of these enzymes.

Undecaprenyl phosphate metabolism

Undecaprenyl phosphate (C55-P), also referred to as bactoprenol, is a key lipid involved in the biosynthesis of peptidoglycan and a variety of other cell-wall polysaccharide components such as lipopolysaccharides, the enterobacterial common antigen, capsule polysaccharides, and teichoic acids (Wright et al., 1967; Scher et al., 1968; Troy et al., 1971, 1975; Watkinson et al., 1971; Johnson & Wilson, 1977; Rohr et al., 1977; Reeves et al., 1996; Rick et al., 1998; van Heijenoort, 2001b; Raetz & Whitfield, 2002; Neuhaus & Baddiley, 2003). C55-P-linked saccharides are also used for N-linked protein glycosylation that occurs in certain prokaryotes (Glover et al., 2005; Szymanski & Wren, 2005). That a single lipid participates in the synthesis of various wall polymers has been earlier considered as a potential site of control that prevents an imbalance in the formation of the cell envelope as a whole (Anderson et al., 1972). Although C55-P remains the classical carrier lipid form encountered in bacterial world, the lower-size homologs decaprenyl phosphate (C50-P) and nonaprenyl phosphate (C45-P) were shown to fulfill this essential function in mycobacterial species (Kaur et al., 2004; Mahapatra et al., 2005) and Paracoccus denitrificans (Ishii et al., 1986), respectively.

In the peptidoglycan pathway, C55-P is needed for the synthesis and transport of hydrophilic GlcNAc-MurNAc-peptide monomeric structures across the hydrophobic environment of the cytoplasmic membrane to the externally located sites of polymerization. Although this function is essential, the knowledge of the metabolism of C55-P was still very limited recently and based on fragmentary data obtained from various bacterial species (Fig. 2). In particular, only few data on the genes and enzymes involved in steps 2, 3 and 4 were available (Higashi, 1970a; Sandermann & Strominger, 1972; Willoughby et al., 1972; Poxton et al., 1974; Kalin & Allen, 1979).

2

Metabolism of undecaprenyl phosphate in bacteria. The site of action of bacitracin, an antibiotic which acts by sequestering of undecaprenyl pyrophosphate, is indicated. Steps 1–4 are catalyzed by undecaprenyl pyrophosphate synthase, undecaprenyl pyrophosphate phosphatase, undecaprenol phosphokinase and undecaprenyl phosphate phosphatase activities, respectively.

Biosynthesis and recycling of undecaprenyl pyrophosphate

The precursor for C55-P, undecaprenyl pyrophosphate (C55-PP), is synthesized by addition of eight C5 isopentenyl units (cis[Z]-configuration) onto C15 (all-trans[E])-farnesyl pyrophosphate (FPP) (Fig. 2). This reaction is catalyzed by the undecaprenyl pyrophosphate synthase UppS (di-trans,poly-cis-decaprenylcistransferase; EC 2.5.1.31), which belongs to the family of cis-prenyltransferases of group IV (Ogura & Koyama, 1998). Farnesyl pyrophosphate itself results from head-to-tail condensation of isopentenyl pyrophosphate with dimethylallyl pyrophosphate, generating C10 geranyl pyrophosphate, followed by a second condensation of isopentenyl pyrophosphate, a reaction catalyzed by the farnesyl pyrophosphate synthase that is a prototype of the trans-prenyltransferase family (Ogura & Koyama, 1998; Liang et al., 2002). The structure and mixed E,Z stereochemistry of the C55-prenyl product (Fig. 3), as deduced from the recent knowledge of the reaction mechanism of the UppS synthase, confirmed earlier data of mass and nuclear magnetic resonance spectrometry of the undecaprenol isolated from bacterial membranes (Scher et al., 1968; Gough et al., 1970).

3

Structure of the polyprenyl carrier lipid involved in cell wall biosynthesis.

The UppS synthase had been partially purified and characterized from several bacteria including Salmonella newington, Micrococcus luteus, Lactobacillus plantarum, Bacillus subtilis and Escherichia coli (Christenson et al., 1969; Baba & Allen, 1978, 1980; Takahashi & Ogura, 1982; Baba et al., 1985; Fujisaki et al., 1986). Like the enzymes involved in fatty acid synthesis, UppS is a soluble cytoplasmic enzyme that generates a membrane embedded product (undecaprenyl pyrophosphate). The first uppS gene was identified much more recently, in 1998 by Ogura's group, using a genomic DNA library of M. luteus B-P 26 constructed in E. coli and a screening of recombinant clones for overexpression of the synthase activity (Shimizu et al., 1998). Orthologs were subsequently identified in various Gram-positive and Gram-negative bacterial species (Apfel et al., 1999) and this gene was demonstrated to be essential in E. coli (Kato et al., 1999) and Streptococcus pneumoniae (Apfel et al., 1999). Interestingly, these newly identified proteins did not exhibit significant sequence similarity with members of the trans-prenyltransferase family and in particular they did not carry the characteristic aspartate-rich DDXXD motif that is involved in substrate binding via a Mg2+ bridge in the latter enzyme family (Chen et al., 1994). The construction of expression vectors allowed the purification of mg quantities of various UppS in either wild-type or histidine-tagged form (Shimizu et al., 1998; Apfel et al., 1999; Pan et al., 2000). The enzyme activity showed both a detergent (Triton X-100) and MgCl2 dependency (Apfel et al., 1999), the latter cation being required for the binding and subsequent condensation of isopentenyl pyrophosphate (Chen et al., 2002b). UppS has now been characterized in great detail, both biochemically and structurally, by different groups. The crystal structure of M. luteus and E. coli enzymes have been solved, either in apo form or in complex with Mg2+, isopentenyl phosphate, FPP, and product analogues (Fujihashi et al., 2001; Ko et al., 2001; Chang et al., 2004; Guo et al., 2005). The data showed that not only the primary but also the three-dimensional (3-D) structure of cis-prenyltransferases was totally different from that of trans-prenyltransferases (Takahashi & Koyama, 2006). The enzyme is a homodimer of 29-kDa subunits and each monomer is composed of six parallel β-strands forming a central β-sheet core, which is surrounded by five of the seven α-helices (Fig. 4). A largely hydrophobic 30 Å depth cleft was observed at the protein surface whose entrance carries several positively charged residues as well as a ‘structural P loop’ motif that is characteristic of phosphate recognition enzymes. A flexible domain located in the vicinity of the H3 α-helix and cleft entrance was identified as an important region for the catalytic process and the two subunits were shown to alternate between two different ‘closed’ and ‘open’ conformations (Ko et al., 2001; Chen et al., 2002b), the ‘closed’ form being catalytically active (Chang et al., 2003). Based on the 3-D structure and site-directed mutagenesis experiments, a model for substrate binding and a catalytic mechanism were proposed (Fujikura et al., 2000, 2003; Pan et al., 2000; Kharel et al., 2001; Takahashi & Koyama, 2006). How the ultimate product chain length of cis-prenyltransferases is determined was also investigated in some detail (Ko et al., 2001; Kharel et al., 2006; Takahashi & Koyama, 2006). Site-directed mutagenesis of E. coli and M. luteus UppS and a sequence comparison with the C70−120 product synthesizing eukaryotic cis-prenyltransferases highlighted some residues that play a critical role in the determination of the product chain length. It was proposed that charged residues present at the hinge region of the H3 α-helix might control the bending direction of the growing hydrophobic prenyl chain along the hydrophobic interior of the H3 helix so that the hydrophobic cleft could accommodate the bulk of the prenyl chain to fit a suitable size during enzymatic elongation. As the substrate specificity and catalytic properties of the FPP synthase and cis-prenyltransferases are varying to some extent in the bacterial world, the size and stereochemistry of the ultimate polyprenyl product, i.e. the carrier lipid, are bacterial species specific: undecaprenyl phosphate in most cases, rarely nonaprenyl phosphate (C45) (Ishii et al., 1986), both with the classical ω,di-trans,poly-cis conformation, or decaprenyl phosphate (C50) with an unusual ω,trans,octa-cis conformation (Wolucka et al., 1994; Kaur et al., 2004; Mahapatra et al., 2005) (Fig. 3). Trace amounts of nona-, deca-, and dodeca-prenyl alcohol derivatives were always detected together with the predominant C55-P lipid in membrane extracts, confirming the high but not absolute specificity of the corresponding UppS synthase activities observed during in vitro assays (Thorne & Kodicek, 1966; Higashi et al., 1967, 1970b; Scher et al., 1968; Gough et al., 1970; Umbreit et al., 1972; Umbreit & Strominger, 1972b; Apfel et al., 1999).

4

Crystal structure of undecaprenyl pyrophosphate synthase dimer (UppS) from Escherichia coli. The figure was prepared using PyMol and the atomic coordinates (1JP3) deposited by Ko (2001). The seven α-helices and six β-strands are shown in red and green for subunit A and pink and blue for subunit B, respectively. The flexible loop without observable electron densities (residues 72–83) is represented as a dotted line in one subunit. The hydrophobic cleft at the molecular surface of each subunit is indicated by an arrow.

The bacitracin antibiotic has been shown to inhibit bacterial cell wall biosynthesis through sequestration of C55-PP, the product of the UppS synthase, thereby provoking the loss of the cell integrity and lysis (Siewert & Strominger, 1967; Stone & Strominger, 1971; Storm & Strominger, 1973). This potent antibiotic produced as a mixture of related cyclic polypeptides by some strains of Bacillus licheniformis and B. subtilis is extensively used for prophylaxis and therapy in food animals. It is clinically used for treatments of surface tissue infections in combination with other antimicrobial drugs. Its oral use had been earlier suggested for the control of vancomycin-resistant enterococci (VRE) (O'Donovan, 1994) although without evident success (Mondy et al., 2001; Hachem & Raad, 2002).

Formation of undecaprenyl phosphate

The dephosphorylation of C55-PP (step 2 in Fig. 2) is required before the lipid carrier becomes available for use in the various biosynthetic pathways. This reaction must also occur after each cycle of polymerization of cell wall components (e.g. of peptidoglycan) and the release of the linked saccharides, because the lipid carrier is in most cases liberated in the pyrophosphate form. However, some exceptions exist: for instance, the transfer of 4-amino-4-deoxy-l-arabinose (l-Ara4N) units to lipid A catalyzed by the ArnT membrane protein uses undecaprenyl-phosphoryl-l-Ara4N as the donor substrate and releases the lipid in the C55-P form (Trent et al., 2001).

The membrane-bound phosphatase catalyzing this reaction had been partially purified from M. luteus by Goldman & Strominger (1972) about 30 years ago and some of its properties were investigated. Its optimal pH for activity was near 7.5, the enzyme did not require any cation, was stimulated by nonionic Triton detergents, and failed to hydrolyze isopentenyl pyrophosphate. The gene for this activity, however, remained to be identified.

In 1993, an E. coli gene whose overexpression resulted in a decreased susceptibility to bacitracin had been identified and the authors hypothesized that this gene should encode an undecaprenol phosphokinase (Cain et al., 1993). This hypothesis was based on the assumption that a significant pool of free C55-OH may exist in E. coli membranes, as earlier demonstrated in some Gram-positive bacteria (Higashi et al., 1970b), that would be directed towards the formation of C55-P by the overproduced kinase, thereby reducing cell requirements for C55-PP molecules and the cell sensitivity to bacitracin. In fact, this question was recently revisited and the bacA gene product was finally unambiguously proved to be a C55-PP phosphatase (El Ghachi, 2004). The overproduction of the BacA protein, which allowed E. coli cells to resist to high concentrations of bacitracin, was correlated with a large (280-fold) increase of C55-PP phosphatase activity in membranes (El Ghachi, 2004). The increased level of C55-PP phosphatase activity likely accelerated the conversion of the pool of C55-PP, the bacitracin target, to C55-P, resulting in an increased cell resistance to the antibiotic. The 30 kDa protein BacA was predicted to be an integral membrane protein with eight transmembrane segments. It was successfully extracted from cell membranes by the n-dodecyl-β-d-maltoside detergent and purified to near homogeneity in the histidine-tagged form (El Ghachi, 2004). The E. coli enzyme exhibited a high C55-PP phosphatase activity of ca. 2200 nmol min−1 mg−1 of protein, a value about 7300-fold higher than the basal activity detected in wild-type cell membranes. It did not show any detectable C55-OH phosphokinase activity. Considering this newly identified function, it was proposed to rename the bacA gene uppP (El Ghachi, 2004), for undecaprenyl pyrophosphate phosphatase (also named undecaprenyl-diphosphatase; EC 3.6.1.27), to follow the nomenclature previously adopted with uppS that encodes the C55-PP synthase (Apfel et al., 1999).

The dephosphorylation of C55-PP was predicted to be an essential metabolic step. The finding that the bacA gene could be deleted from the chromosome of E. coli (El Ghachi, 2004), Mycobacterium smegmatis (Rose et al., 2004), Staphylococcus aureus and Streptococcus pneumoniae (Chalker et al., 2000), without loss of viability or any apparent effect on growth rate or morphology, was therefore quite surprising. Its deletion however resulted in impaired biofilm and smegma formation in M. smegmatis, and attenuated virulence in mouse models of infection in S. aureus and Streptococcus pneumoniae. All the bacA deletion mutants showed enhanced susceptibility to bacitracin (Chalker et al., 2000; El Ghachi, 2004; Rose et al., 2004). The unexpected viability of the bacA deletion mutants suggested the existence of other cell proteins with C55-PP phosphatase activity. The detection of a 25% residual phosphatase activity in the membranes of the E. coli mutant was consistent with this hypothesis (El Ghachi, 2004). No bacA homologue was found in the genomes of the aforementioned species, indicating that these putative phosphatases belonged to a distinct protein family. Although a copy of bacA gene was found in most bacterial genomes sequenced to date (up to three putative bacA orthologs were found in some species, e.g. Bacillus cereus), it was apparently absent in some bacteria such as Helicobacter pylori (unpublished data).

A search in databases for putative membrane phosphatases activities identified three proteins of unknown function that formed the so-called BcrR family: BcrC from B. licheniformis, YbjG from E. coli, and YwoA from B. subtilis (El Ghachi, 2005). Interestingly, the bcrC gene product was described as one of the three components of the ABC transporter system responsible for the protection of B. licheniformis against the antibiotic it produces, bacitracin (Podlesek et al., 1995, 2000), and the overexpression of the E. coli ybjG and B. subtilis ywoA genes were reported to increase bacitracin resistance in the corresponding bacterial species (Harel et al., 1999; Cao & Helmann, 2002; Bernard et al., 2003). Two other E. coli genes displaying similarity with ybjG that encoded members of the PAP2 phosphatidic acid-phosphatase family were also identified: yeiU, of unknown function, and pgpB, encoding one of the two phosphatidylglycerolphosphate phosphatases (Icho & Raetz, 1983). All of these candidate proteins were predicted to be integral membrane proteins and contained a sequence identical or quite similar to the characteristic phosphatase signature KX6RP-(X12−54)-PSGH-(X31−54)-SRX5HX3D (El Ghachi, 2005) that had been identified previously by Stukey & Carman (1997) and Neuwald (1997). Interestingly, this conserved phosphatase motif was not detected in the sequence of BacA and the absence of significant sequence homology between BacA and the above mentioned proteins clearly indicated that they belonged to two distinct protein families. The chromosomal ybjG, yeiU and pgpB genes could be disrupted individually without apparent effect on cell growth but the coinactivation of the three genes bacA, ybjG and pgpB was shown to be lethal (El Ghachi, 2005). A thermosensitive conditional triple mutant strain was generated which lysed at the restrictive temperature due to the depletion of C55-PP phosphatase activity and arrest of peptidoglycan synthesis (El Ghachi, 2005). It confirmed the implication of at least the three proteins BacA, YbjG and PgpB in the formation of the C55-P carrier lipid in vivo. As observed with bacA, the overexpression of the individual ybjG, yeiU and pgpB genes was correlated to an increased resistance to bacitracin and an increased level of C55-PP phosphatase activity in membranes. The B. subtilis ywoA gene product was also subsequently purified and proved to be a C55-PP phosphatase (Bernard et al., 2005).

Therefore, two different classes of integral membrane proteins which belong to the BacA and PAP2 phosphatase families, respectively, could catalyze the dephosphorylation of C55-PP into C55-P in bacteria. The number of these proteins could apparently vary from one bacterial species to another. The situation in E. coli is one BacA protein and at least two members of the PAP2 family. Most bacterial species have only one copy of bacA in their genome but some species seem to express several bacA orthologs (e.g. Bacillus anthracis, B. cereus) and some others apparently do not express any (e.g. H. pylori). The situation is similar for members of the PAP2 family but the precise number of those proteins that effectively exhibit C55-PP phosphatase activity remains to be determined for each species.

Interestingly, genes belonging to either of these two classes were also found in gene clusters conferring antibiotic resistance in the bacitracin-producing and bacitracin-resistant species. For instance, bcrC and ywoA genes that encode phosphatases from the PAP2 family were located within clusters expressing bacitracin efflux systems in B. licheniformis and B. subtilis strains, respectively (Podlesek et al., 1995; Cao & Helmann, 2002; Bernard et al., 2003), and one bacA ortholog was recently found to play a similar essential role in acquired bacitracin resistance in Enterococcus faecalis (Manson et al., 2004). C55-PP phosphatase encoding genes are thus used by bacteria either for generating the essential carrier lipid C55-P or for depleting cells of the pool of C55-PP, the bacitracin target, as a mechanism of resistance to this antibiotic.

The purification and further biochemical characterization of these different proteins is now required to discern phosphatase activities specifically involved in C55-P metabolism from non- or less-specific phosphatases, of distinct metabolic function, that can also use C55-PP as a substrate. C55-PP is synthesized at the inner side of the cytoplasmic membrane but is released at the outer side of the membrane by the peptidoglycan polymerization machinery. Whether the dephosphorylation of C55-PP occurs on both membrane sides or only on one side is at present unknown. A topological analysis of these different membrane proteins will help to answer this question. The involvement of either different phosphatases with catalytic sites orientated towards the cytoplasm and the periplasm or a single phosphatase could be envisaged, but in both cases a trans-bilayer movement of the carrier lipid should occur, suggesting here also the involvement of a putative flippase (see ‘The possible existence of a flippase’).

The reaction of dephosphorylation of C55-PP was considered as an interesting potential target in a search for new antibiotics. The recent discovery that two classes of enzymes and multiple orthologs in each class could participate in this process could render this search more problematic. However, one efficient way to inhibit this step remains the sequestration of the C55-PP substrate, as demonstrated with bacitracin (Siewert & Strominger, 1967; Stone & Strominger, 1971; Storm & Strominger, 1973). An attack of the pyrophosphate moiety of C55-PP was also suggested to be part of the mechanism of action of nisin, a lantibiotic whose primary target is lipid II (Bonev et al., 2004). It was earlier hypothesized that the reaction of dephosphorylation of C55-PP could be the site of action of colicin M, a bacteriolytic toxin produced by some E. coli strains that kills sensitive E. coli strains and related species (Harkness & Braun, 1989a, b). This question was recently revisited and colicin M was in fact identified as an enzyme catalyzing the specific degradation of lipids I and II peptidoglycan intermediates (El Ghachi, 2006).

Undecaprenol: a storage form of lipid carrier?

The intriguing presence of free undecaprenol (C55-OH) in bacterial membranes had been earlier reported in Gram-positive species. More than 90% of the endogenous C55-isoprenyl lipid of S. aureus was found in this nonfunctional alcohol form (Higashi et al., 1970b) and a similar situation was observed in Enterococcus faecalis (Umbreit et al., 1972), Listeria plantarum (Thorne & Kodicek, 1966; Gough et al., 1970), and Listeria monocytogenes (Vilim et al., 1973). It could represent a reserve pool for the regulation of the C55-P pool, an hypothesis that seemed to be corroborated by the detection of two membrane-associated enzyme activities, undecaprenol phosphokinase and undecaprenyl phosphate phosphatase, catalyzing the interconversion of C55-OH and C55-P (steps 3 and 4 in Fig. 2), in some of the latter species (Higashi, 1970a; Sandermann & Strominger, 1972; Willoughby et al., 1972; Poxton et al., 1974; Kalin & Allen, 1979). The corresponding genes however remained to be identified.

Only one report on the characterization of a C55-P phosphatase had been published to date (Willoughby et al., 1972) but a potential involvement of nonspecific phosphatases that could explain the generation of the large pool of C55-OH was also suggested (Bohnenberger & Sandermann, 1976). The C55-P phosphatase activity detected in particulate fractions from S. aureus did not require any cation, was optimal at pH 5, and had an apparent Km for C55-P of 1.5 μM (Willoughby et al., 1972). All attempts to extract it from membranes were unsuccessful and the enzyme was consequently not purified nor further characterized. In the same report, the authors mentioned the failure to detect a similar activity in particulate fractions from B. subtilis, Enterococcus faecalis, M. luteus and E. coli. This activity observed in S. aureus probably accounted for the stimulation by ATP of the overall rate of peptidoglycan synthesis observed in vitro with enzyme preparations from this organism (Anderson et al., 1966).

A C55-OH phosphokinase activity was shown to be extractable by butanol from S. aureus and Klebsiella aerogenes membranes (Higashi, 1970a; Poxton et al., 1974). Its activity was optimal at pH around 8.5 and was dependent on the presence of Mg2+ (Higashi et al., 1970a). The enzyme from S. aureus was purified to near homogeneity and its molecular weight (MW) was estimated at 14 kDa (Sandermann & Strominger, 1972). The estimated Km value was 57 μM for both C55-OH and the nucleotide cosubstrate ATP, and ADP was confirmed as a reaction product. The phosphokinase from L. plantarum was partially solubilized by a variety of methods utilizing Triton X-100 and was characterized in some detail (Kalin & Allen, 1979). Its apparent Km values for C55-OH and ATP were estimated at 14 μM and 2 mM, respectively. No other nucleoside triphosphate was shown to substitute for ATP. Interestingly, it was very recently suggested that the diacylglycerol kinase (DGK) from Streptococcus mutans could also use C55-OH as an alternative substrate (Lis & Kuramitsu, 2003). Although this was not unambiguously established, the physiological significance of this putative undecaprenol kinase activity of DGK was further supported by an increased susceptibility to bacitracin of the dgk mutant strain as compared with that of the parental strain. It could thus be hypothesized that the C55-OH phosphokinase activity that had been purified from S. aureus membranes about 30 years ago was due, at least in part, to the DGK enzyme. The S. aureus dgk gene codes for a 114-residues protein with a MW of 12.97 kDa, a value close to that (14 kDa) earlier estimated for the C55-OH phosphokinase from this organism (Sandermann & Strominger, 1972), but no evidence that this protein also exhibits C55-OH phosphokinase activity has been provided. It should be noted that the DGK from E. coli was proved to be inactive on C55-OH (Bohnenberger & Sandermann, 1979; Lis & Kuramitsu, 2003). Moreover, C55-OH has never been detected in E. coli cells membranes, except in nonphysiological conditions, e.g. following enzymatic degradation of lipid intermediates by colicin M (El Ghachi, 2006). The existence of a pool of C55-OH and expression of the couple of kinase and phosphatase that catalyze its interconversion with C55-P have thus only been demonstrated in a restricted number of bacterial species. Their effective implication in the regulation of the pool of C55-P and of its use for the synthesis of the different cell-wall polymers remain to be demonstrated.

Lipid I biosynthesis

The lipid I is an essential intermediate molecule in the peptidoglycan biosynthesis pathway (Fig. 1). Its existence was reported for the first time by Chatterjee & Park (1964) and Struve & Neuhaus (1965). Afterwards, it was identified as a lipid intermediate produced by transfer of the phospho-MurNAc-pentapeptide moiety from UDP-MurNAc-pentapeptide onto a lipid fraction (Anderson et al., 1965) with concomitant release of UMP. The elucidation of the lipid structure had been achieved by Higashi (1967), who found that the acceptor was a C55 isoprenoid alcohol phosphate, undecaprenyl phosphate (C55-P). However, as detailed in ‘Undecaprenyl phosphate metabolism’, the length and stereochemistry of this carrier lipid appeared to be slightly different in certain bacterial species, such as M. smegmatis (Mahapatra et al., 2005) and P. denitrificans (Ishii et al., 1986). In E. coli, the pool of the lipid I was estimated at about 700 molecules cell−1 (van Heijenoort, 1992). Such an extremely low pool level was explained by the fact that this compound is an intermediate in the pathway whose synthesis and utilization reactions are efficiently coupled.

Translocase I reaction and identification of the mraY gene

In 1965, Strominger and Neuhaus laboratories demonstrated for the first time the transfer of the phospho-MurNAc-pentapeptide moiety from the soluble UDP nucleotide precursor onto the C55-P carrier lipid, using membrane preparations from S. aureus and M. luteus (Anderson et al., 1965; Struve & Neuhaus, 1965). Embedded Image

This reaction, which is generally referred to as the ‘transfer reaction’, does not lead to any modification of the basal MurNAc-pentapeptide structure synthesized by the Mur synthetases in the cytoplasm. It essentially consists in its translocation onto the C55 carrier lipid present in the cytoplasmic membrane. This anchoring is required before subsequent steps could occur, i.e. the addition of the GlcNAc residues by the MurG transferase, the passage of the peptidoglycan monomeric structures through the membrane and finally their polymerization on the outer side of the cytoplasmic membrane. The enzyme catalyzing this first transfer/translocation reaction, the phospho-MurNAc-pentapeptide translocase or MraY (E.C. 2.7.8.13), therefore insures the link between the cytoplasmic and periplasmic steps of peptidoglycan biosynthesis (van Heijenoort, 2001b; Bugg et al., 2006).

MraY was also found to be able to exchange radio-labelled UMP for the unlabelled UMP moiety of UDP-MurNAc-pentapeptide, consistent with the overall reaction referred to as the ‘exchange reaction’: Embedded Image

The mraY gene had been identified by Ikeda (1991) within a large cluster of genes, named mra for ‘murein region A’, located at 2 min on the chromosome map of E. coli. In this region the genes are tightly packed and appear in the order: pbpB-murE-murF-mraY-murD-ftsW-murG-murC-ddlB-ftsQ-ftsA-ftsZ-envA. They all code for proteins involved in peptidoglycan biosynthesis and cell division (Mur synthetases MurC/D/E/F, penicillin-binding protein PBP3; division proteins FtsW/Q/A/Z, MurG transferase, and d-Ala-d-Ala ligase DdlB). A quite similar organization of this cluster was found in other bacterial species. Ikeda (1991) showed that the overexpression of the E. coli mraY gene resulted in an increase of the UDP-N-acetylmuramoyl-pentapeptide: undecaprenyl-phosphate phospho-N-acetylmuramoyl-pentapeptide transferase activity in membranes, demonstrating that this gene encoded the latter activity. Expression of the mraY gene was shown to be dependent on the Pmra promoter (Hara et al., 1997; Mengin-Lecreulx et al., 1998). This gene is essential for the bacterial viability (Boyle & Donachie, 1998; Thanassi et al., 2002) and a conditional mraY mutant strain was shown to accumulate peptidoglycan nucleotide precursors under restrictive growth conditions (Lara et al., 2005). One copy of the mraY gene was found in all bacterial genomes sequenced to date but was not detectable in eukaryotic organisms and archaebacteria which both lack peptidoglycan. Interestingly, however, the presence of an mraY gene orthologue was recently demonstrated in the plant Arabidopsis thaliana (Mondego et al., 2003). These authors postulated that this MraY-like product could participate in the biosynthesis of specific proteoglycan arabinogalactan proteins that reach peak expression during late flower bud development. It was also reported that the mraY gene product from A. thaliana had putative plastid-targeting signals (Machida et al., 2006).

MraY protein, a member of the polyprenyl-phosphate N-acetyl hexosamine 1-phosphate transferase superfamily

The alignment of MraY orthologue sequences from both Gram-negative and Gram-positive species available in databases allowed Bouhss (1999) to identify a set of five well-conserved hydrophilic sequences (I–V) containing 34 invariant amino acid residues (Fig. 5). In the E. coli MraY sequence these domains were defined as H65-L80 (I), I111-K133 (II), N189-L200 (III), L251-G275 (IV) and V296-F342 (V). The size of the MraY protein appears to be fairly well conserved in the bacterial world, the sequences from Gram-negative species being generally longer than those found in Gram-positive species (∼360 vs. 320 amino acid residues), owing mainly to an extension at the N-terminal extremity. Unusually large MraY proteins (∼420 residues) are found, however, in some bacteria such as Bacteroides (Price & Momany, 2005). The presence of alternating hydrophobic and hydrophilic segments in its primary structure clearly suggested that MraY was an integral membrane protein spanning the cytoplasmic membrane several times (Ikeda et al., 1991). Moreover, a lipid microenvironment was shown to be required for the MraY activity (Heydanek & Neuhaus, 1969; Umbreit & Strominger, 1972a; Geis & Plapp, 1978; Weppner & Neuhaus, 1979). More recently, Bouhss (1999) determined the two-dimensional membrane topology of both the E. coli and S. aureus MraY translocases, using β-lactamase fusion experiments. A common topological model was proposed which contained ten transmembrane segments joining four periplasmic loops and five cytoplasmic sequences corresponding to the conserved hydrophilic patterns I–V (Fig. 5). Both the N- and C- terminal extremities were located in the periplasmic space. This model was in agreement with many structural features predicted from a sequence comparison of MraY orthologues, strongly suggesting its validity for all eubacterial MraY proteins (Bouhss et al., 1999).

5

Membrane topology of the MraY translocase (Escherichia coli). The topological model consists in ten transmembrane segments, four periplasmic loops and five cytoplasmic sequences corresponding to the conserved hydrophilic patterns I–V (Bouhss et al., 1999). Conserved residues by identity (red) and by similarity (blue) are indicated.

The cytoplasmic sequences II, III and IV of MraY were also found in other prokaryotic and eukaryotic proteins catalyzing the same kind of reaction, namely the transfer of a lipid phosphate to the β-phosphate of an UDP-linked hexosamine. These three conserved patterns thus defined a superfamily of enzymes termed polyprenyl-phosphate N-acetylhexosamine-1-phosphate transferases or UDP-HexNAc: polyprenyl-P HexNAc-1-P transferases. In addition to MraY, this superfamily contained the prokaryotic enzymes WecA, TagO, WbcO, WbpL and RgpG involved in the biosynthesis of different cell envelope polymers (enterobacterial common antigen, lipopolysaccharide O-antigen, teichoic acids or rhamnose-glucose polysaccharide) (Soldo et al., 2002; Lehrer et al., 2007) and a eukaryotic paralogue, GPT, involved in protein N-glycosylation (Lehrman, 1994; Dal Nogare, 1998; Burda & Aebi, 1999). These enzymes share a common membrane-bound acceptor substrate, undecaprenyl phosphate in bacteria or dolichyl phosphate in eukaryotes, but they differ in their selectivity for the soluble UDP-N-acetyl-hexosamine substrate. The sugar nucleotide donors are UDP-MurNAc-pentapeptide and UDP-GlcNAc for MraY and WecA, TagO and GPT, respectively, the MraY substrate carrying an additional 3-O-lactoyl-pentapeptide group.

The cytoplasmic sequences I and V are highly specific for all MraY orthologues and are also present in some bacterial paralogues with some differences. However, they are not found in the eukaryotic paralogue sequences (GPTs). Sequence and membrane topology analyses revealed that all of the invariant or highly-conserved residues identified within MraY, and/or WecA and the eukaryotic GPTs, were located on the cytoplasmic side of the membrane, consistent with the active site being orientated towards the cytoplasm (Bouhss et al., 1999; Lloyd et al., 2004; Lehrer et al., 2007). The patterns II, III and IV would be involved in the substrate binding and/or the catalytic process that are common features in the enzyme superfamily, while the patterns I and V would be involved in the substrate specificity and in particular the recognition of the sugar nucleotide substrate.

Purification and biochemical characterization of MraY

Solubilization, expression, and purification of MraY

The availability of a pure, stable, soluble and active preparation of the integral membrane MraY protein was a prerequisite to the development of detailed biochemical investigations. Since 1969, many soluble and active preparations of MraY were described, extracted from membranes of various bacteria such as micrococci, S. aureus and E. coli using detergents and in particular Triton X-100 or CHAPS (Heydanek & Neuhaus, 1969; Umbreit & Strominger, 1972a; Pless & Neuhaus, 1973; Brandish, 1996a; Breukink et al., 2003; Lloyd et al., 2004; Stachyra et al., 2004). Upon detergent extraction, the S. aureus MraY was shown to require the presence of a phospholipid, either phosphatidylcholine, dioleoyl phosphatidyl choline or phosphatidylglycerol (Pless & Neuhaus, 1973). Brandish (1996a) reported that overexpressed E. coli MraY protein was preferentially activated by phosphatidylglycerol. All previous attempts to overexpress significantly and purify any MraY protein had been unsuccessful. Thus, only partially purified enzyme preparations were generally used for enzymatic assays (Brandish, 1996a; Zawadzke et al., 2003; Lloyd et al., 2004; Stachyra et al., 2004). Recently, a comparative study performed with recombinant MraY proteins from E. coli, S. aureus, B. subtilis and Thermotoga maritima, expressed in the E. coli C43(DE3) host strain, identified n-dodecyl-β-d-maltoside and N-lauroyl-sarcosine as the most efficient detergents for the extraction of this protein from cell membranes (Bouhss et al., 2004). In the same report, conditions allowing the high-level overexpression of a MraY protein and, for the first time, its purification to homogeneity in milligram quantities were described. The specific activity of the pure B. subtilis MraY protein was estimated at 1900 nmol min−1 mg−1 (Bouhss et al., 2004).

Substrate specificity and kinetic properties

MraY has two substrates: C55-P and UDP-MurNAc-pentapeptide. The structure of the lipid substrate is expected to be conserved in most bacterial species but a few exceptions exist, as mentioned in ‘Undecaprenyl phosphate metabolism’. The structure of the sugar nucleotide substrate shows important variations in the bacterial world, particularly in the peptide moiety (Schleifer & Kandler, 1972; Barreteau et al., 2008; Vollmer et al., 2008). This peptide is a pentapeptide that generally contains at the third position either a meso-A2pm residue in Gram-negative bacteria (as E. coli) and Bacillus species or a lysine residue (more rarely an ornithine) in most Gram-positive bacteria (as S. aureus). In vivo complementation experiments have shown that the S. aureus MraY was functional in E. coli and restored growth of a mraY thermosensitive mutant, indicating that it accepts the A2pm-containing sugar nucleotide (Bouhss et al., 1999). Similarly, the overexpression of the S. aureus MurE synthetase in E. coli (which introduces lysine instead of A2pm at the third position of the nucleotide substrate) resulted in a massive incorporation of lysine into the peptidoglycan, demonstrating the efficient utilization of the lysine-containing UDP-MurNAc-pentapeptide by the E. coli MraY enzyme (Mengin-Lecreulx et al., 1999). The relatively low specificity of the MraY translocase and its tolerance toward the variability of the peptide chain was a well-known characteristic (Hammes & Neuhaus, 1974) that has been confirmed in various other circumstances both in vivo and in vitro. Shorter or longer peptides as well as modified peptides were shown to be accepted: dipeptides (Ornelas-Soares et al., 1994), tripeptides (Hammes & Neuhaus, 1974; Pisabarro et al., 1986; van Heijenoort, 1992), tetrapeptides (Hammes & Neuhaus, 1974), acetylated and dansylated pentapeptides (Ward & Perkins, 1974; Weppner & Neuhaus, 1977; Brandish, 1996a; Stachyra et al., 2004), as well as hexa- and heptapeptides (Billot-Klein et al., 1997). More recently, the MraY enzyme from T. maritima was shown to act on two different substrates in vivo: a d-lysine-containing UDP-MurNAc-tripeptide and an l-lysine-containing UDP-MurNAc-pentapeptide, with similar efficiencies (Boniface et al., 2006).

The replacement of the l-Ala and d-Ala residues at positions 1 and 4 of the pentapeptide chain, respectively, by glycine reduced the S. aureus MraY catalytic efficiency by 135-fold (Hammes & Neuhaus, 1974). The use of UDP-MurNAc-tetrapeptide (lacking one d-Ala) and UDP-MurNAc-tripeptide (lacking two d-Ala) as substrates reduced the MraY catalytic efficiency by fourfold and 80-fold, respectively, as compared with the nucleotide pentapeptide. Stickgold & Neuhaus (1967) determined that 5-fluorouracil-substituted UDP-MurNAc-pentapeptide was utilized at <2% of the rate of the native substrate and was a competitive inhibitor (Ki=0.12 mM) of the transfer reaction, just as 5-fluoro-UMP is a competitive inhibitor (Ki=50 μM) of the exchange reaction. However, 2′-deoxy UMP was accepted as a good substrate in the exchange reaction (Neuhaus, 1971).

Various MraY assays using radiolabeled or fluorescent substrates, resolution of reaction mixtures by paper or thin-layer chromatography, or direct measurement using microplates, have been developed for analyzing the kinetic properties of this enzyme and for the screening of inhibitors (Weppner & Neuhaus, 1977; Brandish et al., 1996a, b; Bouhss, 2004; Stachyra et al., 2004). Typical Michaelis–Menten kinetics were observed with both purified and partially purified enzyme preparations (Anderson et al., 1966; Struve et al., 1966; Stickgold & Neuhaus, 1967; Hammes & Neuhaus, 1974; Brandish, 1996a; Bouhss et al., 2004; Stachyra et al., 2004). Km values observed for UDP-MurNAc-pentapeptide were generally in the 5–30 μM range with nonpurified enzyme preparations. However, a much higher value of 1 mM was determined with the purified MraY enzyme from B. subtilis (Bouhss et al., 2004).

The structure and size of the C55-P lipid acceptor substrate present in the membranes is essentially determined by the substrate specificity and catalytic properties of the undecaprenyl pyrophosphate synthase UppS. However, the E. coli MraY translocase was shown to accept in vitro heptaprenyl (C35) as well as dodecaprenyl (C60) phosphate as alternative substrates, with Km values (10–20 μM) similar to that of the natural substrate (Brandish et al., 1996a). Breukink (2003) have reported that the Micrococcus flavus MraY was active in vitro on polyprenyl substrates containing from two to 25 isoprenyl units, with a maximal efficiency observed for substrates bigger than C35. The purified B. subtilis MraY exhibited a Km for C55-P about 10-fold higher than that of the nonpurified MraYs from other bacterial species (Bouhss et al., 2004). It could thus be assumed that the membrane environment affects the recognition of the lipid substrate by the enzyme.

The MraY activity is dependent on the presence of both mono- and divalent metal ions. The activity of the E. coli and B. subtilis enzymes was stimulated (two- to fourfold) in presence of 10–100 mM of potassium or sodium (Brandish, 1996a; Bouhss et al., 2004). Heydanek (1970) also observed that the S. aureus MraY was stimulated by many monovalent metal ions such as K+, Rb+, Cs+ and NH4+. The MraY enzyme has an absolute requirement for a divalent metal ion, particularly Mg2+, at 5–40 mM concentrations (Heydanek et al., 1970; Bouhss et al., 2004). Mn2+ can replace Mg2+ but the activity is decreased by two orders of magnitude (Bouhss et al., 2004). Lloyd (2004) proposed that the D115 or D116 aspartate residues of E. coli MraY, which are located in the conserved hydrophilic sequence II (Fig. 5) and whose replacement by Asn leads to loss of catalytic activity, are involved in Mg2+ chelation.

Catalytic mechanism of MraY

In 1969, on the basis of kinetic evidences, Heydanek (1969) proposed a two-step catalytic mechanism for the MraY reaction (Fig. 6). It consisted in an attack by a nonidentified nucleophile residue of MraY on the β phosphate of UDP-MurNAc-pentapeptide, generating a covalent enzyme-phospho-MurNAc-pentapeptide intermediate with concomitant release of UMP. The second step corresponded to the attack by an oxyanion from C55-P on the phosphate of the covalent intermediate, resulting in the formation of lipid I and regeneration of the native enzyme form. Recently, Lloyd (2004) proposed that the D267 aspartate residue of E. coli MraY, which is located in the conserved hydrophilic sequence III, plays the role of the catalytic nucleophile. Three observations supported the two-step mechanism: (1) phospho-MurNAc-pentapeptide was formed during the transfer reaction that likely resulted from the hydrolysis of the enzyme-linked phospho-MurNAc-pentapeptide intermediate; (2) the enzyme could catalyze the exchange of UMP with the UMP moiety of UDP-MurNAc-pentapeptide in the presence or absence of the cosubstrate C55-P (Pless & Neuhaus, 1973); (3) dodecylamine inhibited the synthesis of lipid I and caused the release of the phospho-MurNAc-pentapeptide.

6

Proposed MraY translocase reaction mechanisms. The two-step mechanism involves a covalent enzyme-phospho-MurNAc-pentapeptide intermediate and the single-step mechanism consists in a direct attack of the phosphate oxyanion of C55-P onto the β-phosphate of the nucleotide substrate.

A major criticism of the experiments performed by Heydanek et al. that could have resulted in a misinterpretation of the results was the use of a nonpurified MraY enzyme. The different observations they made could effectively be attributed to the presence of contaminating enzymes (phosphatases) and C55-P lipid acceptor in the preparation used. An alternative mechanism consists in a direct attack of the phosphate oxyanion of C55-P onto the β phosphate of UDP-MurNAc-pentapeptide. This would lead to the formation of lipid I and UMP in only one step (Fig. 6). The recent availability of pure MraY protein (Bouhss et al., 2004) will allow to revisit this question and elucidate the catalytic mechanism.

Recent progress on inhibitors of the MraY-catalysed reaction

MraY is the target for several classes of natural product antibiotics, which in some cases have been studied in detail through structure/function studies and/or detailed kinetic or mechanistic studies. The structures of the MraY inhibitors have been reviewed in detail (Kimura & Bugg, 2003; Dini et al., 2005; Bugg et al., 2006). This section will briefly survey the different inhibitor classes, and small molecule inhibitors arising from structure/function studies.

There are five different classes of uridine-containing nucleoside natural antibiotic products that target MraY, illustrated in Fig. 7. The tunicamycins (related to the streptovirudins and corynetoxins) contain a uridine disaccharide, attached to a fatty-acyl chain. Tunicamycin is a reversible, competitive inhibitor of MraY, with Ki 0.6 μM (Brandish et al., 1996b), but is not useful as an antibacterial agent, since it is toxic to mammals, through potent inhibition of GlcNAc-1-P transferase in the dolichol cycle of N-linked glycoprotein biosynthesis (Heifetz et al., 1979). The mureidomycins (related to the pacidamycins and napsamycins) are peptidyl nucleoside natural products, containing a 3′-deoxyuridine sugar attached via an enamide linkage to an unusual peptide chain. The peptide chain contains an N-methyl 2,3-diaminobutyric acid residue, and a urea linkage to a C-terminal aromatic amino acid, which can be meta-tyrosine, Trp, or Phe. Mureidomycin A is a slow-binding inhibitor of E. coli MraY, with Embedded Image=2.2 nM (Brandish et al., 1996a). The enamide functional group is not essential for inhibition, and a range of dihydro-pacidamycin analogues have been synthesized and tested for biological activity (Boojamra et al., 2001). The mechanism of action of mureidomycin A has been studied, and it was proposed that the amino-terminus binds to the Mg2+ cofactor-binding site, and is positioned by an N-methyl amide cis-amide bond (Howard & Bugg, 2003). A uridinyl dipeptide analogue of mureidomycin A retained biological activity against Pseudomonas putida (Howard & Bugg, 2003).

7

Structure of MraY translocase inhibitors.

The liposidomycins are fatty acyl nucleosides, whose structures contain a sulfated aminoglycoside residue. Liposidomycin B is a slow-binding inhibitor (Embedded Image=80 nM) of solubilized E. coli MraY (Brandish et al., 1996b). A synthetic analogue containing the aminoribofuranose monosaccharide attached to the 5′ position of uridine showed moderate levels of inhibition (IC50=50 μM) against translocase I when assayed in toluene-permeabilized E. coli cells (Dini et al., 2000; Stachyra et al., 2004). Further structure–function studies in this series of compounds (riburamycins) gave more active inhibitors, and a synthetic analogue containing a C12-fatty-acyl chain showed antibacterial activity against S. aureus (Dini et al., 2002; Stachyra et al., 2004).

The muraymycins, reported by Wyeth Research in 2002, contain an aminoribofuranoside monosaccharide, attached to a short peptide chain, containing a urea linkage to a C-terminal amino acid (McDonald et al., 2002). Members of this family were potent inhibitors of MraY in vitro (IC50=0.027 μg mL−1), showed antibacterial activity against S. aureus (MIC=2–16 μg mL−1) and enterococci (MIC=16–64 μg mL−1) and were able to protect mice against S. aureus infection (ED50=1.1 mg kg−1) (McDonald et al., 2002). Synthetic analogues lacking the aminoribofuranoside showed reduced activity in vitro, but retained some antibacterial activity (Yamashita et al., 2003).

Capuramycin, a further uridine nucleoside antibiotic, is a potent inhibitor of MraY in vitro (IC50=0.017 μg mL−1), and a methylated derivative showed antibacterial activity against M. smegmatis (MIC=2–16 μg mL−1) (Muramatsu et al., 2003). Acylation derivatives of capuramycin showed very potent activity (MIC=0.06 μg mL−1) against several Mycobacterium strains (Hotoda et al., 2003).

Genetic studies by Young and coworkers showed that MraY was also the target for the bacteriolytic E protein from bacteriophage φX174 (Bernhardt et al., 2000). The E protein is a 91-amino acid protein, containing a transmembrane domain. The killing action of E also required a host protein SlyD, a peptidyl-prolyl isomerase. Recent studies on the mechanism of action of E showed that a 37-amino acid peptide containing the transmembrane domain of E was a potent inhibitor of membrane-bound MraY (IC50=0.8 μM), but did not inhibit solubilized MraY, unlike the small molecule inhibitors (Mendel et al., 2006). It has been proposed that E inhibits MraY via a protein–protein interaction, blocking the formation of protein–protein interactions between MraY and other cell wall assembly proteins in the cytoplasmic membrane at cell division (Mendel et al., 2006).

Interestingly, it was recently demonstrated that colicin M was an enzyme acting by specifically targeting and destroying the peptidoglycan lipid intermediates, thereby provoking the arrest of peptidoglycan synthesis and cell lysis (El Ghachi, 2006). The cleavage site was located between the C55 and pyrophosphoryl groups, as demonstrated in vitro by appropriate assays and in vivo by the observation that colicin M-treated cells accumulated C55-OH, a lipid form that is normally not detectable in E. coli cell membranes (El Ghachi, 2006). To the authors' knowledge this is the first example of a mechanism of inhibition of peptidoglycan biosynthesis occurring via enzymatic degradation of its precursors, in this case the MraY and MurG lipid reaction products.

Lipid II biosynthesis

The translocase II reaction and identification of the murG gene

The translocase II (MurG) catalyses the second membrane associated step of peptidoglycan synthesis (Fig. 1). This glycosyl transferase of the GT-B superfamily transfers the GlcNAc moiety from UDP-GlcNAc to the C4 hydroxyl group of lipid I to form a β-linked disaccharide (lipid II). It was originally identified in E. coli but has since been recognized in all other bacteria that make peptidoglycan.

The murG gene was first discovered by Salmond (1980) who identified, cloned and mapped the gene within the mra region of E. coli. The nucleotide sequence of the E. coli murG coding region (Mengin-Lecreulx et al., 1990) revealed an ORF of 1065 nucleotides theoretically coding for a moderately hydrophobic 37.8 kDa protein. When cultures of a thermosensitive murG mutant strain (GS58) growing exponentially at 30 °C were shifted to the nonpermissive temperature of 42 °C, cells progressively lost their rod shape, became ovoid with a greatly increased volume and finally lysed (Mengin-Lecreulx et al., 1991). This mutant accumulated significant amounts of UDP-MurNAc-pentapeptide, UDP-GlcNAc and to a lesser extent a lipid compound as labelled A2pm-containing precursors, suggesting that the mutational block was in a membrane step. The fact that the ratio of the two lipid intermediates I and II, which had previously been reported to be between 0.3: 1 and 0.6: 1 in E. coli strains (Ramey & Ishiguro, 1978), was reversed and reached the considerably high value of 8.2: 1 in this mutant clearly indicated that it was the second membrane step that was altered. It was consequently deduced that murG encoded the UDP-GlcNAc: MurNAc-(pentapeptide)-pyrophosphoryl-undecaprenol GlcNAc transferase (EC 2.4.1.227) which catalyses the formation of lipid II from lipid I. This protein was recently shown to interact with the MraY and MreB proteins and was suggested to participate in two multi-protein complexes involved in cell elongation and cell division, respectively (Aaron et al., 2007; Mohammadi et al., 2007; den Blaauwen, 2008).

Purification and biochemical characterization of MurG

Crouvoisier (1999) first reported the overproduction, solubilization and purification of the protein. MurG was solubilized using salt or detergent from the cell membranes and purified using anion exchange for the wild-type protein or Ni2+ affinity for the His-tagged protein. Ha (2000) also reported a similar procedure where the protein was solubilized with detergent directly from cell pellets of E. coli and purified using cation exchange followed by gel filtration. van den Brink-van der Laan (2003) showed that the cardiolipin content of E. coli cell membranes was increased following overexpression of MurG and that lipid vesicles copurified with MurG. The activity of MurG was increased in the presence of cardiolipin, suggesting specific interactions of the protein with phospholipids and in particular with cardiolipin.

Using optimal conditions in HEPES buffer, pH 7.9, supplemented with 5 mM MgCl2 or MnCl2, Ha (1999) determined the kinetic parameters for E. coli MurG. Because lipid I is almost impossible to isolate from bacterial cells and difficult to handle, a synthetic lipid I analogue was used. The assay used radiolabelled UDP-[14C]GlcNAc and a biotin-labelled lipid I analogue. A biotin capture technique was used to sequester products from the reaction which were then counted for bound radioactivity. The parameters were KUDP-GlcNAc=58±30 μM, Klipid I analogue=37±4 μM, kcat=16±2 min−1 in the presence of MgCl2. Using a different (C35) lipid I analogue and a new assay based on HPLC, Auger (2003) showed that a high concentration (35%) of dimethylsulfoxide was necessary for maximal enzyme activity. The kinetic constants they determined in these conditions were: Km UDP-GlcNAc=150±20 μM, Km lipid I analogue=2.8±1 μM, kcat=56±5 min−1. Ha (1999) also investigated the acceptor and donor substrate specificity for MurG. The results showed that a biotin-labelled UDP-MurNAc-pentapeptide was also a substrate for the enzyme with a relative rate of c. 20% when compared with the biotin-labelled lipid I analogue. The results also revealed that the enzyme was sensitive to the identity of the nucleotide, and required the presence of the diphosphate linkage. The enzyme was inhibited by UDP, but not by UMP, nor any other nucleotide diphosphate. The enzyme also showed high specificity for the equatorial stereochemistry at the C4 position of the donor. Therefore, UDP-N-acetylgalactosamine failed to show inhibitory activity even at millimolar concentrations and showed very little donor activity. Liu (2003) reported the development of a continuous fluorescence coupled enzymatic assay in which the formation of UDP was coupled to a pyruvate kinase-lactate dehydrogenase assay and the fluorescence signal of NADH monitored. This group tested a variety of lipid I analogues, with various substituents replacing the undecaprenyl moiety, as substrates. MurG accepted all of them with Km values of around 20–50 μM but there were, however, large differences in the specific activity (kcat). Acceptor substrates with long saturated linear alkyl chains were better substrates than the natural lipid I demonstrating an increase in kcat from 11±2 min−1 for the natural substrate to up to 180±13 min−1 for an analogue with a C14H29 chain.

Using synthetic substrate analogues and products containing different length lipid chains, as well as a synthetic dead-end acceptor analogue, Chen (2002a) showed that MurG follows an ordered Bi-Bi mechanism in which UDP-GlcNAc binds first (Fig. 8). The enzyme likely utilizes a mechanism that involves partial participation of the lone pair on the sugar ring oxygen and therefore an oxocarbenium-ion-like transition state. Evidence exists that supports the procedure of glycosyltransferase-mediated reactions through an oxocarbenium-ion-like transition state, similar to that proposed for similar enzymes (Singh et al., 1987; Kim et al., 1988; Takayama et al., 1999). Using a synthetic, radiolabelled-analogue of lipid II, Auger (2003) reported for the first time that the reaction catalyzed by MurG was reversible.

8

Proposed MurG reaction mechanism. The model predicts a deprotonation of the C4 hydroxyl group of the MurNAc moiety of lipid I by a residue of MurG (probably His18). The oxyanion thus generated then attacks the C1 of GlcNAc of the nucleotide substrate to form the oxocarbenium-ion-like transition state. Finally, UDP is released generating lipid II. R=−C(CH3)CO-l-Ala-γ-d-Glu-meso-A2pm-d-Ala-d-Ala, Un=undecaprenyl.

3-D structure of the MurG protein

The 1.9 Å crystal structure of E. coli MurG (Ha et al., 2000) revealed that the free enzyme consists of two protein molecules in an asymmetric unit. Each protein chain has two domains separated by a cleft which is c. 20 Å deep and 18 Å across at its widest point. Each domain adopts a α/β open sheet motif which is characteristic of domains that bind nucleotides. The N-terminal domain contains seven parallel β-strands and six α-helices. The C-terminal domain contains six parallel β-strands and eight α-helices including one irregular bipartite helix that connects the N-domain to the first β-strand of the C-domain. The structural homology between the two domains is high, despite low sequence homology. A subsequent, 2.5 Å crystal structure of E. coli MurG complexed with the donor substrate, UDP-GlcNAc, was reported in 2003 (Hu et al., 2003a) (Fig. 9). Before this, no X-ray crystal structures containing intact substrates had been obtained for any of the NDP-glycosyltransferase superfamily. The substrate bound enzyme was reported to have only a 16 Å wide cleft, c. 2 Å narrower than the free enzyme. UDP-GlcNAc binds in both protein subunits along with four glycerol molecules and 121 water molecules. Only one of the UDP-GlcNAc molecules is in the correct orientation for catalytic activity and is presumed to be the Michaelis complex where the UDP is displaced by the incoming nucleophile. MurG contains a sequence motif that is found in most members of the GT-B superfamily. This α/β/α subunit has two α-helices located near to the cleft between the domains. The UDP-GlcNAc substrate makes several contacts to these helices and also to the loops that connect them to the adjacent β-strands. The GlcNAc moiety of the donor substrate makes contacts with the invariant residues of MurG through hydrogen bonding interactions between the backbone amide of the A263 residue, the side chain amides of N291 and N127 and the C4 hydroxyl group of GlcNAc. Interactions also exist between the side chain amide of Q287 to both the C3 and C4 hydroxyl groups of GlcNAc. The catalytic base appears to be histidine H18 situated 9.52 Å across the domain cleft in line with the anomeric bond. The contacts between the enzyme and the diphosphate are purely hydrogen bonding. S191 has been shown to be an important residue whose mutation to alanine affects all kinetic parameters including an increase in the Km for lipid I binding from 0.053 to 0.179 mM. This serine residue is located on a GGS loop that is conserved in all MurG homologs which moves up towards the donor substrate as it binds. Because MurG is believed to utilise a sequential mechanism in which lipid I binds following UDP-GlcNAc (Chen et al., 2002a), it is thought that conformational changes in the GGS loop may play a role in the adjustments required for lipid I binding, as well as directly contributing to UDP-GlcNAc binding. The uracil nucleotide is anchored in a pocket by hydrogen bonds from the backbone amide of I244 to the N3H and O4 atoms. There is also a possible interaction between the uracil O2 atom and R163. The F243 residue also rotates, as the substrate binds, to cap the binding pocket. Contacts between the enzyme and the ribose 2′ and 3′ hydroxyl groups exist through hydrogen bonds from E268. To evaluate their role in the activity of the E. coli MurG enzyme, 13 residues that are invariant or highly conserved in the MurG enzyme family (T15, H18, Y105, H124, E125, N127, N134, S191, N198, R260, E268, Q288, N291) and located within or very close to the active site were recently submitted to site-directed mutagenesis (Hu, 2003a; Crouvoisier et al., 2007). Most of these mutations resulted in a great loss of enzyme activity, consistent with the interactions observed in the crystal structures of MurG alone or in complex with UDP-GlcNAc (Ha et al., 2000; Hu et al., 2003a). The important role of these residues either in substrate binding/catalysis or maintenance of the protein's conformation was confirmed (Hu, 2003a; Crouvoisier et al., 2007). Interestingly, a hydrophobic patch has been identified in the MurG structure, consisting of I74, L78, F81, W84 and W115 residues, which appears to be the site of interaction between this extrinsic membrane protein and the phospholipid bilayer (Ha et al., 2000).

9

Structure of Escherichia coli MurG complexed with UDP-GlcNAc. This structure, drawn with Swiss PDB Viewer, shows the interactions between UDP-GlcNAc and the enzyme active site. Binding of the lipid I substrate is presumed to take place in the cleft between the two domains of the MurG structure. From Hu (2003a).

Recent progress on inhibitors of the MurG reaction

The cyclic lipoglycodepsipeptide antibiotic ramoplanin has excellent activity against a wide range of Gram-positive bacteria. It was demonstrated that ramoplanin forms complexes with both lipids I and II (Reynolds & Somner, 1990; Lo et al., 2000; Helm et al., 2002; Hu et al., 2003b), however, ramoplanin was also shown to directly interact with MurG although the exact mechanism of inhibition is not known (Fang et al., 2006).

Using the most active lipid I analogue substrate they had identified, Liu (2003) developed a screening for MurG inhibitors. The potential inhibitors screened were vancomycin, moenomycin and two chloro-biphenyl derivatives of vancomycin. All of these compounds were found to inhibit the enzyme with one of the vancomycin derivatives being particularly potent. These known inhibitors of the bacterial cell wall biosynthesis pathway, however, cannot penetrate the bacterial cell membrane and so they do not encounter MurG in vivo because the enzyme localizes on the cytoplasmic side of the cell membrane (Bupp & van Heijenoort, 1993).

A recent publication by Helm (2003) detailed the high throughput screening of almost 50 000 compounds from a commercial library as inhibitors of MurG. The assay employed in the screening used an N-acyl fluoresceinated UDP-GlcNAc analogue instead of the natural substrate. As MurG was added to the labelled substrate, an increase in polarization occurred indicating that the substrate had bound. The compounds were then screened at 25 μg mL−1 for in vitro activity as competitive inhibitors by looking for a decrease in polarization. Eleven compounds were found to reproducibly inhibit the enzyme by 50% or more. Seven of these compounds contained a five-membered, nitrogen containing heterocyclic core with an alkyl or aryl substituent at N-1 and an arylidine substituent at the 3 position. A later publication by Hu (2004) utilized the same assay to screen a further 64 000 compounds from a variety of different compound libraries, including commercial and diversity-orientated synthesis (DOS) libraries. Fifty-five compounds were found to exhibit >40% inhibition of MurG at a concentration of 2.5 μg mL−1. Interestingly, 31 of the 55 compounds were found to contain one of four common cores (Fig. 10). It is not known exactly how these compounds bind to MurG and antibacterial activity has not yet been reported.

10

General structures of MurG inhibitors. From Hu (2004). X=S, O.

A direct continuous fluorescence assay for MurG based on fluorescence resonance energy transfer (FRET) was recently described (Li & Bugg, 2004). It used a dansyl-labelled lipid I (λex=500 nm, λem=550 nm) and a fluorescent UDP-GlcNAc analogue (λex=290 nm, λem=500 nm) as substrates for MurG. A linear relationship between enzyme concentration and the rate of increase in fluorescence was observed, indicating that an efficient energy transfer occurred as the substrates were converted to products by the enzyme. This assay is believed to be the first direct continuous fluorescence assay for MurG and may be very useful for determining the activity of potential inhibitors.

Enzymatic modification of lipid II in the bacterial world

The peptidoglycan structure of many bacteria, especially Gram-positive species, contains an additional peptide cross-link between the residue (l-Lys or meso-A2pm in most cases) at position 3 of the pentapeptide chain and the d-alanine residue at position 4 of the cross-linked strand. There is significant variation in the type and structure of these cross-links across the bacterial kingdom (Schleifer & Kandler, 1972; Vollmer et al., 2008). Some examples are shown in Fig. 11.

11

Structure of lipid II and its main modifications in bacterial world. R=H or COOH for Lys and A2pm, respectively; d-Asx=d-Asp or d-Asn.

The addition of the cross-link amino acids occurs most commonly at the level of the lipid II intermediate (Fig. 11), although in Weisselia viridescens it has been shown that it occurs at the level of the nucleotide precursor (Maillard et al., 2005). Kamiryo & Matsuhashi (1972) showed that the formation of the penta-glycine cross-link in S. aureus required glycyl tRNA and lipid II. More recently, the femABX genes have been implicated in the addition of the penta-glycine bridge (Rohrer & Berger-Bächi, 2003): FemX is responsible for addition of the first glycine; FemA the next two glycines; and FemB the final two glycines (Rohrer & Berger-Bächi, 2003). The FemABX reactions have recently been reconstituted using purified recombinant proteins, with lipid II and glycyl tRNA as substrates (Schneider et al., 2004). There is a crystal structure of S. aureus FemA, which contains a pair of extended α-helices, similar in structure to a motif found in seryl tRNA synthetase that interacts with the tRNA substrate (Benson et al., 2002). A crystal structure of W. viridescens FemX complexed with its substrate UDP-MurNAc-pentapeptide has been determined, which reveals enzyme–substrate binding interactions (Biarrotte-Sorin et al., 2004). Because aminoacyl tRNAs are found only in the cytoplasm, the lipid II modification reactions must occur on the cytoplasmic side of the membrane, after the MurG-catalysed reaction, but before flipping to the cell surface.

Dipeptide Ala-Ala cross-links are found in Enterococcus faecalis (Bouhss et al., 2001), and in strains of highly penicillin-resistant Streptococcus pneumoniae (Garcia-Bustos & Tomasz, 1990), in the latter case linked to the murMN genes (Filipe & Tomasz, 2000). The murMN genes are found in all pneumococci, but specific sequences of murM are found in highly resistant strains (Filipe et al., 2000). Two transferases involved in the formation of the Ala–Ala cross-link in Enterococcus faecalis have also been identified, which are able to add Ala-tRNA to UDP-MurNAc-pentapeptide in vitro (Bouhss et al., 2002). The Streptococcus pneumoniae MurM reaction has recently been reconstituted in vitro using recombinant MurM, lipid II, and alanyl-tRNA (Lloyd et al., 2007). MurM from a penicillin-resistant strain was found to show 80-fold higher kcat/Km than MurM from a sensitive strain, providing an explanation for the higher level of cross-links in the resistant strains (Lloyd et al., 2007). The ATP-dependent enzyme responsible for addition of d-aspartic acid to l-lysine residue in Enterococcus faecium has also been identified (Bellais et al., 2006).

In a number of bacterial strains, the d-glutamic acid residue at position 2 of the pentapeptide chain is amidated, to form d-isoglutamine. In S. aureus, it is known that this amidation is dependent upon ATP and ammonia, and that either lipid I or II can be amidated in vitro (Siewert & Strominger, 1968). A membrane-bound amidating activity has also been observed in Bacillus stearothermophilus that is able to amidate the nucleotide precursor (Linnett & Strominger, 1974). Lipid II isolated from M. smegmatis has been found to contain several modifications: the MurNAc residue is modified as N-glycolylmuramic acid, or as de-acylated muramic acid; the diaminopimelic acid residue is amidated; and the terminal d-alanine residue is methylated (Mahapatra et al., 2005). It appears that a number of modification reactions occur in particular bacterial strains, most commonly at the level of the lipid-linked intermediates (Fig. 11). In S. aureus, the acceptor substrate of sortase A that catalyzes anchoring of surface proteins to peptidoglycan was identified as lipid II (Perry et al., 2002). These different modifications should be taken into account when considering the lipid II as a potential target for new antibacterial agents (Breukink & de Kruijff, 2006).

Chemical and enzymatic synthesis of lipids I and II and analogues

In order to reconstitute in vitro the lipid-linked steps of peptidoglycan biosynthesis, it was necessary to prepare the lipid-linked intermediates. Because they are present naturally in very low abundance, it was not feasible to isolate them from bacterial cells in useful quantities (van Heijenoort, 1992; Guan et al., 2005). Chemical syntheses were published for lipid I (VanNieuwenhze et al., 2001) and for lipid II (Ye et al., 2001; VanNieuwenhze et al., 2002), however the synthetic routes were long and proceeded in fairly low overall yield. Auger, (1997, 2003) first succeeded for the hemi-synthesis of functional analogues of both lipids I and II. Chemical synthesis was used to prepare soluble analogues of lipid I containing 10-carbon (Ha et al., 1999) and 20-carbon chains (Cudic et al., 2001), which were used as soluble substrates for MurG in vitro (Lazar & Walker, 2002).

Alternatively, lipids I and II can be prepared enzymatically. The cytoplasmic precursor UDP-MurNAc-pentapeptide (containing meso-A2pm at position 3) can be accumulated in B. subtilis using a cell wall synthesis medium (Lugtenberg et al., 1972), and isolated in 50 mg quantities. High level expression and purification of enzymes MurA-F from E. coli or Pseudomonas aeruginosa and their use for the preparation of the cytoplasmic intermediates in vitro have been reported (Reddy et al., 1999; El Zoeiby, 2001; Bouhss et al., 2004). An efficient procedure for the conversion of UDP-MurNAc-pentapeptide into lipid I or lipid II was published by Breukink (2003), that used membranes prepared from M. flavus, a species which has elevated levels of MraY and MurG enzyme activities. Using this procedure, they reported the preparation of up to 50 mg quantities of lipids I or II. The availability of purified C55-PP synthase UppS, C55-PP phosphatase BacA, MraY and MurG proteins also facilitated the synthesis of different forms of lipids I and II, with a labelling in either the C55 or the pentapeptide moiety (El Ghachi, 2006).

The ɛ-amino group of meso-A2pm or l-Lys at position 3 of the pentapeptide chain can be chemically modified by fluorophores such as dansyl. The N-dansyl analogue was used to develop a fluorescence assay for MraY, for enzyme kinetic studies (Brandish et al., 1996a), and for high throughput inhibition assays (Stachyra et al., 2004). The N-dansylated lipid II analogue was prepared and was shown to give rise to fluorescence changes upon treatment with E. coli PBP1b (Schwartz et al., 2002). An alternative labelling strategy was recently published, in which the d-alanine residue at position 4 or 5 can be replaced by d-cysteine, by enzymatic synthesis from d-Ala-d-Cys or d-Cys-d-Ala, and the thiol side chain of d-Cys can be labelled by thiol-selective reagents such as pyrene maleimide (Schouten et al., 2006). The S-labelled UDP-MurNAc-pentapeptide analogues prepared in this way were converted into fluorescent lipid I and II analogues, using M. flavus membranes.

Possible existence of a flippase

Lipid II, biosynthesized on the cytoplasmic face of the membrane, must somehow be flipped onto the extracellular face of the membrane (Fig. 1). The possible existence of a ‘flippase’ protein capable of catalysing this process has been discussed for many years (Weppner & Neuhaus, 1978; Ehlert & Höltje, 1996) but no gene has been identified to date. Analysis of fluorescently labelled peptidoglycan precursors in S. aureus membranes by fluorescence energy transfer have earlier concluded that the rate of unassisted flipping was much too low to account for the rate of peptidoglycan synthesis required at cell division (Weppner & Neuhaus, 1978), therefore it was generally assumed that there must be some protein assistance for the flipping of lipid II across the membrane. In fact, the biochemical process was only recently demonstrated experimentally, in particular by analysis of the transmembrane transport of fluorescent lipid II through model and bacterial membranes (Breukink et al., 1999; van Dam, 2007). It was shown that lipid II flop did not occur spontaneously and was not obligatory coupled to lipid II synthesis. This suggested the involvement of one or more integral membrane proteins or membrane-associated proteins in this process. Based on their results, the authors further hypothesized a coupling of the translocation of lipid II with the subsequent reactions of polymerization (transglycosylation) catalyzed by the penicillin-binding proteins (van Dam, 2007).

The existence of ABC transporters that are able to catalyse similar types of membrane transport events suggests that the putative flippase might be similar to this protein family, for which there is now structural information (Chang & Roth, 2001). The bacterial wzx gene was suggested as a candidate flippase gene in lipopolysaccharide biosynthesis, on the basis of gene knockout studies (Liu et al., 1996). A putative flippase gene involved in N-linked protein glycosylation in eukaryotic cells was identified, based upon gene knockout studies (Helenius et al., 2002). The pglK gene encoding an ABC transporter responsible for N-linked protein glycosylation in Campylobacter jejuni was shown to complement a wzx deficiency in E. coli O-antigen biosynthesis, although the corresponding gene products showed no sequence similarity (Alaimo et al., 2006). In contrast, van Dam (2007) found that the flipping of lipid II was not dependent upon ATP, which would argue against the involvement of an ABC transporter. These different studies suggest that there is a flippase protein for peptidoglycan biosynthesis, but its identity and precise mechanism of action are yet to be determined.

Concluding remarks

In summary, the understanding of the biochemistry of the lipid-linked intermediates of bacterial peptidoglycan biosynthesis has increased considerably in the last few years. Several genes encoding the phosphatase and kinase enzymes involved in undecaprenyl phosphate metabolism have now been identified. The corresponding enzymes are potential new targets for antibacterial action. Translocase MraY has been successfully purified, and is the target for a number of natural product and synthetic inhibitors. There is an X-ray crystal structure for glycosyltransferase MurG, which will facilitate the development of novel inhibitors for this enzyme. Both MraY and MurG are attractive targets for antibacterial development, and methods for high-throughput screening of both enzymes have been developed. The putative ‘flippase’ activity responsible for transport of lipid intermediate II across the cytoplasmic membrane remains an interesting target which has not yet been identified. A number of enzymatic modifications of lipid II by the FemABX/MurMN proteins have now been reconstituted; their role in bacterial cell growth and antibiotic resistance mechanisms can now be studied. Further modifications of lipid-linked intermediates take place in certain microorganisms, and the ability to prepare lipids I and II in useful quantities will allow more detailed studies on new modification reactions. Availability of lipid II will also allow more detailed biochemical studies on the transglycosylase and transpeptidase reactions catalysed by the penicillin-binding proteins in future years. How all these different membrane proteins are organized in the membrane and possibly interact together in relation to the cell elongation and cell division processes also remains to be elucidated.

Acknowledgements

The authors thank Didier Blanot and Thierry Touzé for critical reading of the manuscript. This work was supported by the Centre National de la Recherche Scientifique and by the European Community (FP6 projects COBRA, LSHM-CT-2003-503335, and EUR-INTAFAR, LSHM-CT-2004-512138). AET was supported by a PhD studentship from the Engineering and Physical Sciences Research Council.

Footnotes

  • Editor: Jacques Coyette

References

View Abstract