Correspondence: Didier Blanot, Laboratoire des Enveloppes Bactériennes et Antibiotiques, IBBMC, UMR 8619 CNRS, Bâtiment 430, Univ Paris-Sud, 91405 Orsay, France. Tel.: +33 1 69 15 81 65; fax: +33 1 69 85 37 15; e-mail: firstname.lastname@example.org
The biosynthesis of bacterial cell wall peptidoglycan is a complex process that involves enzyme reactions that take place in the cytoplasm (synthesis of the nucleotide precursors) and on the inner side (synthesis of lipid-linked intermediates) and outer side (polymerization reactions) of the cytoplasmic membrane. This review deals with the cytoplasmic steps of peptidoglycan biosynthesis, which can be divided into four sets of reactions that lead to the syntheses of (1) UDP-N-acetylglucosamine from fructose 6-phosphate, (2) UDP-N-acetylmuramic acid from UDP-N-acetylglucosamine, (3) UDP-N-acetylmuramyl-pentapeptide from UDP-N-acetylmuramic acid and (4) d-glutamic acid and dipeptide d-alanyl-d-alanine. Recent data concerning the different enzymes involved are presented. Moreover, special attention is given to (1) the chemical and enzymatic synthesis of the nucleotide precursor substrates that are not commercially available and (2) the search for specific inhibitors that could act as antibacterial compounds.
Peptidoglycan (or murein) is a major component of the cell wall of almost all eubacteria. It is a complex heteropolymer that is composed of long glycan chains that are cross-linked by short peptides (Rogers et al., 1980). The glycan chains are made up of alternating N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) residues linked by β1→4 bonds. The d-lactoyl group of each MurNAc residue is substituted by a peptide stem with a composition most often seen as l-Ala-γ-d-Glu-meso-A2pm (or l-Lys)-d-Ala-d-Ala (A2pm, 2,6-diaminopimelic acid) in nascent peptidoglycan, the last d-Ala residue being removed in the mature macromolecule. Cross-linking of the glycan chains generally occurs between the carboxyl group of d-Ala at position 4 and the amino group of the diaminoacid at position 3, either directly or through a short peptide bridge. Minor variations in the glycan chain, the peptide stem or the peptide bridge are present in the bacterial world and are detailed in the accompanying review by Vollmer et al. (2008).
The biosynthesis of peptidoglycan is a complex process that involves c. 20 reactions that take place in the cytoplasm (synthesis of the nucleotide precursors) and on the inner side (synthesis of lipid-linked intermediates) and outer side (polymerization reactions) of the cytoplasmic membrane. The present review deals with the cytoplasmic steps of peptidoglycan synthesis; the subsequent steps are reviewed by Bouhss et al. (2008) and Sauvage et al. (2008) in this issue.
The cytoplasmic steps (Fig. 1) can be divided into four sets of reactions: (1) formation of UDP-GlcNAc from fructose-6-phosphate, (2) formation of UDP-MurNAc from UDP-GlcNAc, (3) assembly of the peptide stem leading to UDP-MurNAc-pentapeptide and (4) ‘side’ or ‘annex’ pathways of synthesis of d-glutamic acid and the dipeptide d-alanyl-d-alanine. The discovery of the enzyme activities involved in these processes was the subject of the pioneering works by J. T. Park, J. L. Strominger, E. Ito, E. J. J. Lugtenberg, F. C. Neuhaus and others during the 1960s and 1970s. Their results have been summarized in previous reviews to which we refer the reader (Rogers et al., 1980; Neuhaus & Hammes, 1981; Ward, 1984; Walsh, 1989; Bugg & Walsh, 1992; van Heijenoort, 2001; El Zoeiby et al., 2003a; Katz & Caufield, 2003; Kotnik et al., 2007a). In the present review, we will focus on recent data concerning the specificities, the kinetic and catalytic mechanisms and the three-dimensional structures of these enzymes. Moreover, because the nucleotide precursors that are substrates (or products) of several of these activities are not commercially available, we will give special attention to their chemical or enzymatic synthesis. Finally, as these enzymes are targets for antibacterial compounds, we present an overview of the existing inhibitors and of the current search for new specific inhibitors.
Biosynthesis of UDP-N-acetylglucosamine
In bacteria, UDP-GlcNAc biosynthesis from fructose-6-phosphate requires four successive enzyme activities: glucosamine-6-phosphate synthase (GlmS), phosphoglucosamine mutase (GlmM), glucosamine-1-phosphate acetyltransferase and N-acetylglucosamine-1-phosphate uridyltransferase (the latter two activities are carried by the GlmU bifunctional enzyme) (Fig. 1). UDP-GlcNAc is also present in eukaryotes because GlcNAc is an important building block for major biomolecules such as chitin and glycoproteins. However, as the eukaryotic pathway of UDP-GlcNAc biosynthesis is different from the prokaryotic pathway, the latter can be considered to be a target for specific antibacterial compounds.
GlmS is an amidotransferase that takes part in the first committed step of hexosamine metabolism. It has been purified from Escherichia coli (Badet et al., 1987) and Thermus thermophilus (Badet-Denisot et al., 1997), and characterized. It is an essential and dimeric enzyme that catalyses the conversion of d-fructose-6-phosphate into d-glucosamine-6-phosphate, using l-glutamine as the nitrogen source. It follows an ordered bi–bi mechanism (Badet et al., 1988).
The GlmS monomer is composed of two structurally and functionally distinct domains (Fig. 2). The N-terminal 30-kDa glutaminase domain promotes glutamine hydrolysis into glutamate and ammonia, whereas the C-terminal 40-kDa isomerase domain binds the nitrogen acceptor and uses the ammonia that is produced for the conversion of fructose-6-phosphate into glucosamine-6-phosphate. The glutamine hydrolysis reaction uses the N-terminal cysteine thiol, which forms a γ-glutamyl thioester intermediate. The ketose/aldose isomerase activity proceeds by abstraction of the C-1 pro-R hydrogen of a putative fructosimine-6-phosphate intermediate, to form a transient cis-enolamine that, upon reprotonation at the re face of the C-2 sp2 carbon, yields glucosamine-6-phosphate (Fig. 3) [see references in Badet-Denisot et al. (1993)].
The individual crystal structures of these two domains as complexes with their respective reaction products have been solved (Obmolova et al., 1994; Isupov et al., 1996; Teplyakov et al., 1998). As shown by its conserved N-terminal catalytic cysteine, GlmS belongs to the N-terminal nucleophile (Ntn) family and shares common structural and catalytic mechanisms with other Ntn family members. Two other invariant residues appear to have key roles in substrate recognition by anchoring the α-COO− and α-NH3+ functions of glutamine. However, no sequence similarity has been detected between other isomerases and GlmS. Moreover, with respect to other isomerases, the catalytic residues of GlmS belong to different polypeptide chains, whereas they are usually located on only one subunit in other isomerases. The intact Escherichia coli GlmS crystal structure was solved (Teplyakov et al., 2001), thus allowing an understanding as to how the ammonia produced at the glutamine site is delivered to the sugar phosphate site: an intramolecular nitrogen channelling between the two active sites that are separated by 18 Å is involved. Both sites contribute to the formation of this channel due to several residues that are strictly conserved in the GlmS sequence (Fig. 2). Moreover, the dimerization of GlmS appears to be crucial for the closure of the sugar-binding site and the formation of the channel. Recently, a study combining molecular dynamics simulations and site-directed mutagenesis allowed to define amino acid residues involved in the channelling process (Floquet et al., 2007a).
The GlmS activity is inhibited by glutamine analogues that may be either naturally occurring or synthetic. Some of these possess an electrophilic functionality at the γ-position that can establish a covalent bond with the N-terminal cysteine residue located in the glutamine-binding site. Among the naturally occurring compounds, there is anticapsin, an epoxyamino acid that is liberated after hydrolysis of the dipeptide antibiotic bacilysin (tetaine) (Kenig et al., 1976) and that inactivates the enzyme; glutamine protects against inactivation (Chmara et al., 1984). Other natural inhibitors are known, like azaserine and albizziin (Winterburn & Phelps, 1971). On the basis of these naturally occurring compounds, synthetic inhibitors have been studied, like 6-diazo-5-oxo-l-norleucine (DON) (Badet et al., 1987) and N3-fumaroyl l-2,3-diaminopropionic acid derivatives (Chmara et al., 1985; Badet et al., 1988). Other glutamine analogues with low micromolar Ki values have been synthesized, such as the γ-dimethylsulfonium derivative (Walker et al., 2000). Some carbohydrate compounds, like 2-amino-2-deoxy-d-glucitol-6-phosphate, have been shown to inhibit the enzyme competitively with respect to fructose-6-phosphate (Badet-Denisot et al., 1995); recently, other carbohydrate-based inhibitors have been designed as analogues of the reaction intermediates (Bearne & Blouin, 2000). N-iodoacetylglucosamine-6-phosphate is an active site-directed irreversible inactivator of GlmS from Escherichia coli, and it interacts with both the sugar- and the glucosamine-binding sites (Bearne, 1996). Very recently, new inhibitors of bacterial GlmS have been discovered through virtual screening; interestingly, these compounds were predicted to interact with the protein region that forms a pocket at the interface between the two enzyme monomers, which opens the way to new molecules that alter the dimerization process (Floquet et al., 2007b).
GlmM is the second enzyme involved in the biosynthesis of UDP-GlcNAc, and it catalyses the interconversion of the glucosamine-6-phosphate and glucosamine-1-phosphate isomers. GlmM was first characterized and purified to near homogeneity in Escherichia coli (Mengin-Lecreulx & van Heijenoort, 1996). Its amino acid sequence contains the characteristic signature of hexosephosphate mutases, including the serine residue (Ser102) for which phosphorylation is required for enzyme activity. Accordingly, GlmM is active only in its phosphorylated form, although the dephosphorylated form also exists in vivo and both forms of the enzyme can be separated by HPLC (Jolly et al., 1999). The GlmM enzymes from Staphylococcus aureus and Pseudomonas aeruginosa have also been purified (Jolly et al., 1997; Tavares et al., 2000).
The reaction catalysed by GlmM follows a ping-pong bi–bi mechanism, where GlcN-1,6-diphosphate appears as an intermediate in the catalytic process, acting as both the first product and the second substrate (Fig. 4). The GlmM enzyme also catalyses the interconversion of the 1-phosphate and 6-phosphate isomers of glucose, although with reduced rate constants (Jolly et al., 1999). The inactive, dephosphorylated form of GlmM undergoes an autophosphorylation reaction when it is incubated with ATP in the presence of divalent cations. The site of phosphorylation has been shown to be the aforementioned Ser102 residue (Jolly et al., 2000).
No crystal structure of the GlmM enzyme is available to date. However, the crystal structure of another member of the hexosephosphate mutase family, rabbit muscle phosphoglucomutase, has been reported (Dai et al., 1992). With this phosphoglucomutase, the estimated volume of the active site cleft involving the catalytic serine residue has been described as being large enough to accommodate an ATP molecule.
First identified in Bacillus subtilis, the GlmU protein was initially thought to be a GlcNAc-1-phosphate uridyltransferase that catalyses the formation of UDP-GlcNAc from GlcNAc-1-phosphate and UTP (Hove-Jensen, 1992; Mengin-Lecreulx & van Heijenoort, 1993). However, in Escherichia coli, GlmU was shown to be a bifunctional enzyme, catalysing both acetyltransfer and uridyltransfer during the transformation of GlcN-1-phosphate to UDP-GlcNAc (Mengin-Lecreulx & van Heijenoort, 1994). Moreover, the order of the chemical reaction is not random but imposed by the enzyme: GlmU first catalyses acetyltransfer from AcCoA to GlcN-1-P with the release of GlcNAc-1-P, then uridyltransfer from UTP to GlcNAc-1-P in the presence of Mg2+, yielding inorganic pyrophosphate and UDP-GlcNAc (Mengin-Lecreulx & van Heijenoort, 1994; Gehring et al., 1996). However, it should be mentioned that, under certain conditions, GlmU is capable of catalysing the two reactions in the reverse order (i.e. uridyltransfer precedes acetyltransfer), but with greatly reduced kinetic parameters (Pompeo et al., 2001).
The construction of truncated forms of GlmU has shown that this bifunctional enzyme is organized in two domains that operate without substrate channelling and that are individually active (Gehring et al., 1996; Brown et al., 1999; Pompeo et al., 2001). However, both these activities are essential for cell viability (Pompeo et al., 2001). The sizes of the two domains appear to be roughly equivalent, each one representing about half of the 49-kDa protein (Fig. 5).
The crystal structure of a truncated form of GlmU has shown that the two distinct domains are connected by a long α-helical arm (Brown et al., 1999). The C-terminal domain catalyses the first reaction, which is responsible for the CoA-dependent acetylation of GlcN-1-phosphate, and it shows sequence similarities with a number of acetyltransferases (Mengin-Lecreulx & van Heijenoort, 1994; Raetz & Roderick, 1995; Gehring et al., 1996). It is characterized by an imperfect, tandem hexapeptide repeat sequence motif [LIV]-[GAED]-X2-[STAV]-X (Vaara, 1992), which folds into a left-handed β-helix (LβH) (Raetz & Roderick, 1995) (Fig. 5). As with most enzymes that contain an LβH structure, the GlmU acetyltransferase domain has a trimeric organization that is absolutely required for the activity. Moreover, the catalytic site is formed by complementary regions of contact between the three adjacent monomers, as confirmed by the crystal structures of the truncated and the entire GlmU (Olsen & Roderick, 2001; Pompeo et al., 2001). In the C-terminal domain, there are also four cysteine residues, which are apparently located near the active site. These do not appear to be directly involved in the catalytic process, but nevertheless two of them, Cys307 and Cys324, have important roles in the acetyltransferase activity (Pompeo et al., 1998). Recently, the crystal structure of the Escherichia coli GlmU acetyltransferase active site was determined in complexes with AcCoA and with CoA/GlcN-1-P (Olsen et al., 2007).
The second reaction that is catalysed by GlmU involves uridyltransfer from UDP to GlcNAc-1-phosphate and occurs at the N-terminal domain. This domain shares sequence homology with a variety of nucleotidyltransferases over residues Met1-Ala120, which are also known as nucleotide diphosphate sugar pyrophosphorylases, and has strict conservation of the L-X2-G-X-G-T-X-M-(X)4-P-K motif (Mio et al., 1998). Its crystal structure shows that the uridine-binding site is a large open pocket, bounded by two lobes. The first lobe comprises residues that interact with the nucleotide (Asn3-Val111 and His216-Asn227), and the second lobe comprises those residues that interact with the sugar moiety (Glu112-Val215) (Brown et al., 1999; Olsen & Roderick, 2001) (Fig. 5). Contrary to the acetyltransferase domain, trimerization is not essential for expression of the uridyltransferase activity of GlmU; however, some of the interactions between the two domains appear to participate in the folding and stability of the N-terminal domain (Pompeo et al., 2001).
The crystal structure of the GlmU protein from Streptococcus pneumoniae has been solved both in its apo form and in its complex with UDP-GlcNAc and Mg2+ (Kostrewa et al., 2001; Sulzenbacher et al., 2001). Pneumococcal GlmU contains the common LβH motif with high-sequence conservation, suggesting that similar LβH motifs exist in all the GlmU structures and have important roles in the acetyltransferase activity.
Comparison of the prokaryotic and eukaryotic systems
In eukaryotes, the pathway for UDP-GlcNAc biosynthesis appears to be significantly different: acetyltransfer occurs on GlcN-6-P and not on GlcN-1-P and, most importantly, acetyltransferase and uridyltransferase activities are carried by two distinct monofunctional enzymes (Fig. 6).
In mammalian cells, the GlmS equivalent, known as glutamine : fructose-6-P amidotransferase (GFAT), is an insulin-regulated enzyme that controls the flux of glucose into the hexosamine pathway (Traxinger & Marshall, 1991). GFAT is about 280 kDa in size, and it is composed of four subunits; it belongs to the Ntn-amidotransferase family and possesses a Cys1 residue, as does the Escherichia coli GlmS (Huynh et al., 2000). However, unlike its bacterial equivalent, GFAT is subject to allosteric regulation by UDP-GlcNAc (Traxinger & Marshall, 1991) and can also be regulated by its product GlcN-6-P (Broschat et al., 2002). In mammals, the pyrophosphorylase that condenses UTP and GlcNAc-1-P has also been identified; purified to near homogeneity from pig liver extracts, it appears to be a homodimer that is composed of two 64-kDa subunits. It requires a divalent cation (Mn2+) for activity and has an unusual specifity: at high concentrations, it uses UDP-GalNAc as a substrate as well as UDP-GlcNAc in the reverse direction and GalNAc-1-P as well as GlcNAc-1-P in the forward direction (Szumilo et al., 1996). Two isoforms of this enzyme, AGX1 and AGX2, have been identified in humans (Peneff et al., 2001b).
In yeast such as Saccharomyces cerevisiae and Candida albicans, the hexosamine metabolism has also been well studied (Milewski et al., 2006) and the four different enzymes involved in this pathway have been shown to be essential for cell viability. The first reaction, which sees the formation of GlcN-6-P from fructose-6-P, is catalysed by GFA1 (Watzele & Tanner, 1989). Like its mammalian homologue, this enzyme is a homotetramer of 80 kDa subunits that has been crystallized (Raczynska et al., 2007). Then, GlcN-6-P is N-acetylated by the GNA1 acetyltransferase to yield GlcNAc-6-P (Mio et al., 1999). The three-dimensional structure of GNA1 has been shown to be a dimer of two identical subunits (Peneff et al., 2001a). GlcNAc-6-P is further isomerized into GlcNAc-1-P by the GlcNAc-phosphate mutase AGM1 (Hofmann et al., 1994). The crystal structures of this 60-kDa protein from Candida albicans have been reported, both for the apoform and for complexes with substrates and products (Nishitani et al., 2006). Finally, UDP-GlcNAc is produced from GlcNAc-1-P by a monofunctional UDP-GlcNAc pyrophosphorylase, known as UAP1, whose sequence is well conserved in the human homologue (Mio et al., 1998; Peneff et al., 2001b). The crystal structure of the Candida albicans enzyme has been solved recently. Unlike prokaryotic GlmU, which has a metal ion that acts as a cofactor, no metal ion has been seen in this candidal UAP1; instead, the terminal amino function of a conserved lysine residue occupies the virtual metal ion-binding site (Maruyama et al., 2007).
Biosynthesis of UDP-N-acetylmuramic acid
The first committed stage towards the creation of the peptidoglycan polymer involves the formation of UDP-MurNAc from UDP-GlcNAc by two enzymes: MurA and MurB (Fig. 1). MurA (formerly known as MurZ) catalyses the first step of this transformation by transferring the enolpyruvate moiety of phosphoenolpyruvate (PEP) to the 3′-hydroxyl of UDP-GlcNAc with the release of inorganic phosphate (Pi). The resulting product, UDP-GlcNAc-enolpyruvate, undergoes a reduction catalysed by MurB using one equivalent of NADPH and a solvent-derived proton. This two-electron reduction creates the lactyl ether of UDP-MurNAc.
The MurA reaction constitutes a rare biochemical process. The only other known enolpyruvyl transfer from PEP to an OH group with the concomitant release of Pi occurs during the shikimic acid pathway, in the reaction catalysed by 5-enolpyruvylshikimate-3-phosphate synthase AroA (Walsh et al., 1996; Byczynski et al., 2003). Interestingly, although MurA and AroA share only c. 25% sequence identity, they exhibit the same protein architecture. The MurA enzymes from Enterobacter cloacae and Escherichia coli have been overproduced and purified (Marquardt et al., 1992; Wanke et al., 1992), thereafter serving for extensive mechanistic and structural studies. The MurA reaction pathway (Fig. 7) follows an addition–elimination mechanism, with the formation of a noncovalently bound phospholactoyl-UDP-GlcNAc tetrahedral intermediate (Marquardt et al., 1993). Two other intermediates have been characterized: a phospholactoyl-enzyme adduct and an O-phosphothioketal intermediate, both of which are linked to the Cys115 residue (Wanke & Amrhein, 1993; Brown et al., 1994; Ramilo et al., 1994). However, it was demonstrated that they are not essential for catalysis (Kim et al., 1996).
The X-ray structures of several forms of the protein have been solved, both unliganded and in a complex (Schönbrunn et al., 1996; Skarzynski et al., 1996, 1998; Eschenburg & Schönbrunn, 2000; Schönbrunn et al., 2000a, b) (Fig. 8). Unliganded MurA appears as a two-domain protein with an unusual fold (inside-out α/β barrel) that is built up from the sixfold repetition of the same βαβαββ motif (Fig. 8a). Near the hinge region, there is an isoaspartyl residue that is the product of a posttranslational modification of the Asn67–Gly68 dipeptide moiety. Upon UDP-GlcNAc binding, the reaction follows an induced-fit mechanism in which the two-domain structure undergoes large conformational changes that lead to a closed form. This has been confirmed by fluorescence spectroscopy and small-angle X-ray scattering (Schönbrunn et al., 1998). The Pro112–Pro121 loop containing Cys115 is flexible; it is solvent-exposed in the open conformation, but forms a lid around the interdomain section in the closed conformation (Fig. 8b).
The role of the amino acid residues involved in the reaction pathway (Fig. 7) has been recently questioned. Cys115 was initially suggested to act as an acid–base catalyst in the addition–elimination reaction (Skarzynski et al., 1998; Krekel et al., 2000). However, it seems that it is essential for product release only (Eschenburg et al., 2005a). Similarly, Asp305, which was considered to be the base abstracting a proton from the 3′-hydroxyl group of UDP-GlcNAc (Skarzynski et al., 1996; Samland et al., 2001), would be responsible for the final proton abstraction from the C-3 atom of the tetrahedral intermediate (Eschenburg et al., 2003). The Lys22 residue, which is strictly conserved in the MurA enzymes, was shown to be involved in the binding of PEP and to participate in the conformational change that leads to the formation of the catalytically competent enzyme complex (Samland et al., 1999, 2001); it has been presumed recently to also accomplish the two-proton transfer required for the addition of the substrates (Eschenburg et al., 2003).
The stereochemical course of the reaction has been revised as well. Initial studies using PEP analogues had led to the conclusion that the proton addition step at C-3 of PEP proceeded at the 2-si face and that the stereochemistry of the pair of addition and elimination steps was anti/syn (Kim et al., 1995; Lees & Walsh, 1995; Skarzynski et al., 1998). However, recent examination of the crystal structure of the D305A mutant complexed with the tetrahedral intermediate favours an addition at the re face of PEP and an anti/syn stereochemistry (Eschenburg et al., 2003).
MurA is the target of the naturally occurring broad-spectrum antibiotic fosfomycin, which forms a covalent adduct with the reactive Cys115 residue (Marquardt et al., 1994). In the MurA enzymes of species that are naturally resistant to fosfomycin, such as Mycobacterium tuberculosis (De Smet et al., 1999) and Chlamydia trachomatis (McCoy et al., 2003), the corresponding Cys residue is changed into Asp. Escherichia coli MurA that contains the C115D mutation is enzymatically active and resistant to inactivation by fosfomycin (Kim et al., 1996).
Gram-negative bacteria have one copy of the murA gene (Brown et al., 1995), while Gram-positive bacteria have two (murA1 and murA2), which have probably arisen from gene duplication (Du et al., 2000). The MurA1 and MurA2 enzymes from Streptococcus pneumoniae have been purified: their catalytic parameters are similar, and they are both inhibited by fosfomycin (Du et al., 2000).
It has been reported recently that UDP-MurNAc tightly binds to and inhibits Escherichia coli MurA; a possible role of the nucleotide in the regulation of peptidoglycan biosynthesis has been inferred (Mizyed et al., 2005).
MurB from Escherichia coli has been overproduced and purified. It is a 38-kDa protein that contains a stoichiometric amount of bound FAD (Benson et al., 1993; Tayeh et al., 1995). It follows a ping-pong bi–bi mechanism, with weak and strong substrate inhibition by NADPH and UDP-GlcNAc-enolpyruvate, respectively. It is activated by cations, such as K+, NH4+ and Rb+ (Dhalla et al., 1995).
The reaction catalysed by MurB involves two half-reactions in which FAD serves as the redox intermediate (Fig. 9). The first half-reaction is the reduction of FAD to FADH2 by NADPH. This starts with the binding of NADPH to MurB and the transfer of the 4-pro-S hydrogen of NADPH to N-5 of the enzyme-bound flavin. The release of NADP+ is followed by the binding of UDP-GlcNAc-enolpyruvate. The second half-reaction is the reduction of the vinylic enol ether by FADH2. Hydride transfer from the reduced flavin to C-3 of the enolpyruvyl moiety of the nucleotide substrate generates a carbanionic intermediate that is then protonated at C-2 by a solvent-equilibrated proton (Benson et al., 1993, 1997b). The stereochemistry of the reduction has been studied through the use of UDP-GlcNAc-(E)-enolbutyrate as a mechanistic probe (Lees et al., 1996).
The crystal structures of unliganded MurB and of its dead-end complex with UDP-GlcNAc-enolpyruvate have been solved (Benson et al., 1995, 1996, 1997a) (Fig. 10). The protein is composed of three domains: domains 1 and 2 mediate FAD binding, and domain 3 mediates the binding of its substrates. The binding of UDP-GlcNAc-enolpyruvate induces a substantial movement of domain 3. With the perdeuterated, 13C/15N-labelled MurB studied by NMR spectroscopy (Farmer et al., 1996; Constantine et al., 1997), it was deduced that NADP+ binds in the same pocket as UDP-GlcNAc-enolpyruvate, inducing structural changes, and that NADPH transfers a hydride to the si face of the FAD isoalloxazine ring.
Comparison of the X-ray structures of Escherichia coli and Staphylococcus aureus MurB enzymes has revealed notable distinctions with respect to their structural elements. Escherichia coli MurB, classified as type I, contains a Tyr loop and a split βαββ fold (Fig. 10a), whereas Staphylococcus aureus MurB, classified as type II, lacks these secondary elements (Fig. 10b). This has consequences on the mode of substrate binding (Benson et al., 2001). Recently, a type II MurB enzyme (Thermus caldophilus) was crystallized in the presence of UDP-GlcNAc-enolpyruvate. X-ray data and sequence alignments allowed the definition of two subtypes, type II-a (e.g. Staphylococcus aureus and Bacillus subtilis) and type II-b (e.g. Thermus caldophilus and Chlamydia pneumoniae), that contain a serine or a cysteine residue, respectively, as a proton donor to quench the carbanionic intermediate (Kim MK et al., 2007). Six other conserved amino acids were also shown by site-directed mutagenesis to be essential for MurB activity (Nishida et al., 2006).
Inhibitors of MurA and MurB
As already mentioned, the most well-known inhibitor of MurA is fosfomycin 1 (Fig. 11), an epoxide compound that reacts with the Cys115 residue. A number of novel inhibitors of MurA have been discovered recently through various high-throughput screening efforts in the pharmaceutical industry (Dai et al., 2002; DeVito et al., 2002; Labaudiniere et al., 2005). Three noncovalent inhibitors show submicromolar IC50 values (IC50=0.2–0.9 μM); however, they also show nonspecific inhibition of DNA, RNA and protein biosynthesis (Baum et al., 2001). In a whole-cell peptidoglycan synthesis assay, two new inhibitors were identified: a derivative of diarylmethane and a substituted imidazole (Barbosa et al., 2002). Sesquiterpene lactones [e.g. 2: IC50=10.3 μM (Pseudomonas aeruginosa) and 16.7 μM (Escherichia coli)] were shown to alkylate the thiol group of Cys115 of MurA from Pseudomonas aeruginosa and Escherichia coli and thereby to act as irreversible inhibitors (Bachelier et al., 2006). A dodecapeptide inhibitor of MurA from Pseudomonas aeruginosa (IC50=200 μM) was selected by phage display and appeared to be competitive with respect to UDP-GlcNAc (Molina-López et al., 2006).
The fluorescent dye 8-anilino-1-naphthalene sulfonate 3 binds to the solvent-exposed region of MurA from Enterobacter cloacae, and its X-ray cocrystal structure provides the basis for an alternative approach in the design of new inhibitors (Schönbrunn et al., 2000a). Thus, 5-sulfonoxy-anthranilic acid derivatives that were obtained by high-throughput screening bind to the same region of MurA and obstruct its transition from the open to the closed forms (Eschenburg et al., 2005b).
Tri-substituted thiazolidinones [e.g. 4: IC50 (Escherichia coli)=7.7 μM] were the first small-molecule inhibitors of MurB; they were designed to mimic the diphosphate moiety of UDP-GlcNAc-enolpyruvate and prepared by a parallel synthesis approach (Andres et al., 2000). As their bioisosteric replacement, a series of imidazolinone analogues were synthesized and found to possess potent MurB inhibitory activity (the best IC50 values for Escherichia coli ranged from 12 to 40 μM) as well as promising antibacterial activity against Staphylococcus aureus (MIC values: 2–4 μg mL−1) (Bronson et al., 2003). Two inhibitors of Staphylococcus aureus MurB, with Kd values in the submicromolar range (0.19 and 0.14 μM), were discovered by high-throughput screening (Sarver et al., 2002). 4-Alkyl and 4,4′-dialkyl 1,2-bis(4-chlorophenyl)pyrazolidine-3,5-dione derivatives were found to inhibit MurA (the best IC50 values for Escherichia coli ranged from 9.8 to 50 μM) and MurB (the best IC50 values for Escherichia coli and Staphylococcus aureus ranged from 5.1 to 50 μM) (Kutterer et al., 2005). In addition, four structurally related 3,5-dioxopyrazolidines inhibited MurB with IC50 values between 4 and 35 μM, and also MurA and MurC to weaker extents. The crystal structure of a complex with compound 5 indicated that the 3,5-dioxopyrazolidine core occupies the same region of the MurB active site as the N-acetyl group of the substrate (Yang et al., 2006). Antibacterial activity against Gram-positive bacteria was seen for many inhibitors from the pyrazolidine-3,5-dione series; however, when tested in the presence of bovine serum albumin (BSA), the activity was lost, indicating high protein-binding properties of these compounds (Kutterer et al., 2005; Yang et al., 2006). A similar decline in antibacterial activity due to the presence of BSA was also seen for thiazolyl urea and carbamate derivatives [e.g. 6: IC50 (Staphylococcus aureus)=19 μg mL−1], which were good inhibitors of MurA and MurB as well as of the growth of some Gram-positive bacteria (Francisco et al., 2004).
Biosynthesis of the UDP-MurNAc-peptides
Generality of the Mur ligases
The stepwise assembly of the peptide stem of peptidoglycan is ensured by a series of four essential enzymes, known as the Mur ligases (MurC, D, E and F). These provide for additions of l-alanine (MurC), d-glutamic acid (MurD), a diamino acid, generally meso-diaminopimelic acid or l-lysine (MurE) and dipeptide d-Ala-d-Ala (MurF) onto the d-lactoyl group of UDP-MurNAc (Fig. 1). A fifth enzyme, Mpl, adds the tripeptide l-Ala-γ-d-Glu-meso-A2pm directly onto UDP-MurNAc during peptidoglycan recycling. The Mur ligases catalyse the formation of an amide or a peptide bond with simultaneous formation of ADP and Pi from ATP. A divalent cation, Mg2+ or Mn2+, is essential for the reaction. Their mechanism of action has been studied through biochemical experiments, amino acid sequence examination, site-directed mutagenesis, assays of transition-state analogue inhibitors and X-ray structure determination. These studies have shown that the Mur ligases share three characteristics:
1They have the same reaction mechanism, which consists first in the activation of the carboxyl group of the UDP-precursor by ATP, generating an acyl phosphate intermediate and ADP; the acyl phosphate then undergoes the nucleophilic attack of the amino group of the condensing amino acid (or dipeptide), leading to the formation of a high-energy tetrahedral intermediate, which eventually breaks down into amide or peptide and Pi (Fig. 12) [see references in Bouhss et al. (2002)].
2They have a series of six invariant residues in addition to an ATP-binding consensus sequence. This finding led to the definition of the Mur ligases as a new family of enzymes (Bouhss et al., 1997, 1999b; Eveland et al., 1997). Three other enzymes that are not related to peptidoglycan biosynthesis also belong to this family: the folylpoly-γ-l-glutamate synthetase FolC (Sheng et al., 2000), the C-terminal region of cyanophycin synthetase CphA (Ziegler et al., 1998) and the poly-γ-glutamate synthetase CapB from Bacilli (Candela & Fouet, 2006).
3They have the same three-dimensional structures in three domains, as seen by crystallographic studies (Fig. 13). The N-terminal domain is involved in the binding of the UDP-precursor, the central domain in the binding of ATP, and the C-terminal domain in the binding of the amino acid or dipeptide. Whereas the topologies of the central and C-terminal domains are similar among the Mur ligases, that of the N-terminal domain shows differences, with MurC and MurD on the one hand, and MurE and MurF on the other hand. These differences are related to the lengths of the UDP-precursor substrates. These enzymes exist in ‘closed’ and ‘open’ conformations. The closure of the conformation is thought to be provoked by ligand binding. The description and the comparison of the three-dimensional structures of the Mur ligases has been reviewed excellently by Smith (2006).
The MurC ligase adds the first amino acid of the peptide stem. In most bacterial species, this amino acid is l-alanine; in rare cases, glycine or l-serine is added instead (Schleifer & Kandler, 1972). MurC from Escherichia coli (Liger et al., 1995; Falk et al., 1996; Gubler et al., 1996), Mycobacterium tuberculosis and Mycobacterium leprae (Mahapatra et al., 2000), Pseudomonas aeruginosa (El Zoeiby et al., 2000) and Chlamydia trachomatis (Hesse et al., 2003) have been purified and characterized. The enzymatic properties of MurC from Escherichia coli have been studied extensively. Its preferred substrate is l-Ala, although Gly and l-Ser, as well as several compounds that are structurally related to l-Ala, can be added with lower efficiencies (Liger et al., 1991, 1995; Emanuele et al., 1996). The stereospecificity is strict: d-Ala is not a substrate (Liger et al., 1995). A sequential, ordered kinetic mechanism has been demonstrated, with ATP binding first, UDP-MurNAc second and l-Ala third (Emanuele et al., 1997). The reaction is reversible and the exchange reaction is phosphate- and ADP-dependent (Liger et al., 1996). The existence of the acyl phosphate intermediate has been inferred from several studies, namely isotope exchange (Falk et al., 1996), rapid kinetics (Emanuele et al., 1997), radioactive labelling (Liger et al., 1996) and chemical trapping (Bouhss et al., 2002). Although MurC activity is dependent on the presence of a reducing agent, site-directed mutagenesis of the two cysteine residues present in the active site of MurC has shown that these two are not essential (Nosal et al., 1998). In solution, the Escherichia coli enzyme is present as an equilibrium between monomeric and dimeric forms; however, this has no influence on its specific activity (Jin et al., 1996).
The specificities of the MurC enzymes from species other than Escherichia coli have also been studied. Interestingly, MurC from Mycobacterium tuberculosis and Mycobacterium leprae have the same in vitro specificity patterns towards l-Ala and Gly; however, the amino acid found in the first position of the peptide stem is different (l-Ala for the former and Gly for the latter). This appears to be due to growth conditions (Mahapatra et al., 2000). Another interesting case is that of Chlamydia trachomatis, which contains a bifunctional protein with a MurC domain and a Ddl domain (Chopra et al., 1998). The MurC domain has been purified alone (Hesse et al., 2003): it adds l-Ala, l-Ser and Gly with similar efficiencies, thereby preventing the deduction of the nature of the first amino acid of the putative chlamydial peptidoglycan (Moulder, 1993). The entire MurC-Ddl fusion protein has been purified recently (McCoy & Maurelli, 2005) (see ‘Formation of d-Ala-d-Ala’).
The crystal structures of the MurC enzymes from Haemophilus influenzae (Mol et al., 2003) (Fig. 13a), Thermotoga maritima (Spraggon et al., 2004) and Escherichia coli (Deva et al., 2006) have been solved. Unexpectedly, in the apo structures of these last two enzymes, the N-terminal and central domains were seen to be in their closed conformation, suggesting that in its unliganded form MurC may exist in different conformations in solution (Smith, 2006).
The second amino-acid residue of the peptide stem is in most species d-glutamic acid (Schleifer & Kandler, 1972). The few variations encountered (d-isoglutamine, threo-3-hydroxyglutamic acid) are due to modifications at a latter stage in the biosynthesis, and thus d-Glu can be said to be the amino acid substrate of MurD in all species (Vollmer et al., 2008). The enzymes from Escherichia coli (Pratviel-Sosa et al., 1991; Auger et al., 1998), Staphylococcus aureus, Haemophilus influenzae and Enterococcus faecalis (Walsh et al., 1999) have been purified and characterized. Here again, the specificity of the enzyme from Escherichia coli has been studied in detail (Pratviel-Sosa et al., 1994). An almost exclusive preference for d-Glu was seen. Only a few closely related derivatives (homocysteic acid, 3- or 4-methyl-d-Glu, cyclopentane or cyclohexane analogues of d-Glu) were fairly good substrates. l-Glu is not a substrate (Pratviel-Sosa et al., 1994); however, an N-sulfonyl derivative of l-Glu was shown to be a competitive inhibitor towards d-Glu (Kotnik et al., 2007b). Among the four Mur ligases, MurD differs by its low specificity towards the UDP-precursor: whereas the UMP moiety is essential for the three other ligases, 1-phospho-MurNAc-l-Ala is a substrate for MurD (Michaud et al., 1987). The reaction is reversible; however, in contrast to MurC, the exchange reaction is not ADP-dependent (Vaganay et al., 1996). Chemical trapping experiments have established the existence of the acyl phosphate intermediate (Bouhss et al., 2002), and the tight binding of phosphinate analogues of the high-energy tetrahedral intermediate has strongly suggested its occurrence (Tanner et al., 1996; Gegnas et al., 1998). MurD is capable of producing adenosine 5′-tetraphosphate, which originates from a reaction of the acyl phosphate with ATP (Bouhss et al., 1999a); among the Mur ligase family, only FolC appears to share this property (Dementin, 2001; Sun et al., 2001). MurD from other species have been characterized. Differences between Gram-negative and Gram-positive bacteria regarding substrate inhibition by UDP-MurNAc-l-Ala and effects of monovalent ions have been seen. These differences have been interpreted in terms of regulation of peptidoglycan synthesis (Walsh et al., 1999).
The first crystal structure of a Mur ligase that was solved was that of Escherichia coli MurD complexed with UDP-MurNAc-l-Ala (Bertrand et al., 1997). Other structures (complexes with products, metal ions or inhibitors, ‘open’ forms) were then reported (Bertrand et al., 1999, 2001; Kotnik et al., 2007b), with strong structural similarities between MurD and FolC being seen (Sheng et al., 2000; Bertrand et al., 2001). The MurD molecule contains two magnesium ions (Fig. 13b): the ‘classical’ one (Mg1) involved in ATP binding, and a second one (Mg2) involved in acyl phosphate formation. Mg2 is coordinated with two water molecules that are hydrogen bonded to a carbamoylated lysine residue (Bertrand et al., 1999). The importance of this carbamoyl group, which is also present in MurE and MurF, has been investigated by chemical rescue experiments (Dementin et al., 2001). Recently, a targeted molecular dynamics study has increased our understanding of the substrate binding and domain closure processes (Perdih et al., 2007).
The third amino acid of the peptide stem is generally either meso-A2pm (most Gram-negative bacteria and Bacilli) or l-lysine (most Gram-positive bacteria), although in certain species, other amino acids are encountered (l-ornithine, ll-A2pm, meso-lanthionine, l-diaminobutyric acid, l-homoserine, for example) (Schleifer & Kandler, 1972). In most cases, the MurE enzyme is highly specific for the relevant amino acid, incorporation of a ‘wrong’ amino acid (e.g. l-Lys in Escherichia coli) leading to cell lysis (Mengin-Lecreulx et al., 1999). Bacillus sphaericus possesses two MurE enzymes: one that adds l-Lys that is active during vegetative growth, and one that adds meso-A2pm that is active during spore cortex formation (Anwar & Vlaovic, 1986). As for the second amino acid, certain variants of the third amino acid (amidated meso-A2pm, acetylated diaminobutyric acid) necessitate a subsequent enzymatic activity (Vollmer et al., 2008).
The l-Lys-adding enzyme from Bacillus sphaericus and the meso-A2pm-adding enzyme from Escherichia coli are strongly activated by phosphate, a product of the reaction (Anwar & Vlaovic, 1986; Michaud et al., 1990). Previous studies on MurE from Escherichia coli have shown that a few analogues of meso-A2pm (ll-A2pm, lanthionine and cystathionine) can be accepted as substrates either in vitro or in genetically engineered cells (Mengin-Lecreulx et al., 1988, 1994; Michaud et al., 1990; Richaud et al., 1993; Auger et al., 1996). The specificities of the MurE enzymes from a Gram-negative (Escherichia coli) and a Gram-positive (Staphylococcus aureus) species towards the amino acid substrate have been compared recently (Boniface, 2007). While the Escherichia coli enzyme has a very weak l-Lys-adding activity, the Staphylococcus aureus MurE is totally unable to add meso-A2pm. Furthermore, Escherichia coli MurE does not accept l-Orn as a substrate, contrary to the staphylococcal enzyme, which has a weak l-Orn-adding activity.
In at least two species, the MurE enzyme appears to be devoid of strict specificity. MurE from Bifidobacterium globosum can incorporate two amino acids indifferently, l-Lys and l-Orn, which are both retrieved in peptidoglycan (Hammes et al., 1977). MurE from Thermotoga maritima, a Gram-negative species for which peptidoglycan contains similar proportions of both enantiomers of lysine, but no meso-A2pm (Huber et al., 1986), can add l-Lys, d-Lys and meso-A2pm in vitro with comparable efficiencies (Boniface et al., 2006). In the UDP-MurNAc-tripeptide products, the d-Glu-l-Lys bond has the conventional γ→α arrangement; however, d-Lys is acylated on its ɛ-amino group, leading to the synthesis of a new nucleotide, UDP-MurNAc-l-Ala-d-Glu(γ→ɛ)d-Lys. The absence of meso-A2pm in Thermotoga maritima peptidoglycan is explained by its very low intracellular pool (Boniface et al., 2006).
The X-ray structure of Escherichia coli MurE in complex with its product, UDP-MurNAc-l-Ala-γ-d-Glu-meso-A2pm, has been solved (Gordon et al., 2001) (Fig. 13c). A binding pocket for the distal (nonreacting) site of meso-A2pm is seen (Fig. 14a). Sequence alignments have revealed consensus sequences in the binding pockets of meso-A2pm and l-Lys: DNPR and D(D,N)P(N,A), respectively (Dementin, 2001; Gordon et al., 2001) (Fig. 14b). The main difference between these consensus sequences is the arginine residue, which is H-bonded with the carboxyl group of the distal site of meso-A2pm (Gordon et al., 2001) (Fig. 14a). In the Thermotoga maritima enzyme, the consensus sequence (DDPR) is undoubtedly a meso-A2pm-adding one (Fig. 14b), thereby explaining the ‘upside-down’ binding of d-lysine. Interestingly, MurE from Escherichia coli produces UDP-MurNAc-l-Ala-d-Glu(γ→ɛ)d-Lys in vitro, although at a low rate (Mengin-Lecreulx et al., 1994; Boniface et al., 2006). The determination of the three-dimensional structure of an l-Lys-binding pocket would be of great interest; however, despite attempts with MurE from Streptococcus pneumoniae (Blewett et al., 2004) and Staphylococcus aureus (Boniface, 2007), no data are currently available.
The residues in positions 4 and 5 of the peptide stem are added as a dipeptide by MurF. d-Ala-d-Ala is the most usual dipeptide. d-Ala-d-Ser and d-Ala-d-Lac are found in vancomycin-resistant strains (Healy et al., 2000a). The enzyme from Escherichia coli has been purified and studied (Duncan et al., 1990; Anderson et al., 1996), and as for MurC, it follows a sequential, ordered kinetic mechanism. The two forms of UDP-MurNAc-tripeptide (meso-A2pm and l-Lys) are equally effective as substrates, and a strong inhibition by excess of UDP-MurNAc-tripeptide has been seen; this effect was suppressed by the addition of 0.5 M NaCl (Anderson et al., 1996). The specificity profile for the dipeptide substrate has been the subject of many studies; however, because most of them used in vivo systems [see references in van Heijenoort (2001)], the results obtained are indirect. When the pure enzyme has been available, it was firmly established that it has a high degree of specificity for the C-terminal amino acid (Duncan et al., 1990; Bugg et al., 1991). This is complementary to the specificity of d-Ala : d-Ala ligase (Ddl), which resides mainly on the N-terminal amino acid, and this constitutes a ‘double sieving’ mechanism that ensures the synthesis of UDP-MurNAc-pentapeptide ending mainly in d-Ala-d-Ala (Neuhaus & Struve, 1965; Duncan et al., 1990). An interesting observation is the ability of MurF to incorporate the dipeptide 3-fluoro-d-Ala-3-fluoro-d-Ala that is synthesized by Ddl in vivo, thereby explaining the autoantagonistic effect of high concentrations of the antibiotic 3-fluoro-d-Ala (Kollonitsch et al., 1973; Duncan et al., 1990). The MurF enzyme from Thermotoga maritima has been isolated; not unexpectedly, it adds d-Ala-d-Ala to the l-Lys-containing UDP-MurNAc-tripeptide, but not to the d-Lys-containing nucleotide. However, the fact that the latter is a good substrate for MraY explains the incorporation of d-Lys into Thermotoga maritima peptidoglycan (Boniface et al., 2006).
The crystal structure of the MurF apoenzyme from Escherichia coli has been determined; it is an open structure that is expected to undergo domain closure upon substrate binding (Yan et al., 2000). Recently, three structures of the enzyme from Streptococcus pneumoniae were cocrystallized with sulfonamide inhibitors (Fig. 13d) and showed interdomain closure (Longenecker et al., 2005; Stamper et al., 2006).
Murein peptide ligase (Mpl) is a nonessential enzyme that is found in some Gram-negative species. The mpl gene was identified by a search of databases for proteins with significant homology with MurC. Mpl participates in the recycling of peptidoglycan by adding the tripeptide l-Ala-γ-d-Glu-meso-A2pm onto UDP-MurNAc (Mengin-Lecreulx et al., 1996). The enzyme from Escherichia coli has been purified recently and its in vitro substrate specificity has been studied thoroughly (Hervéet al., 2007). The meso-A2pm-containing tri-, tetra- and pentapeptides were accepted as substrates with high, similar catalytic efficiencies. Their l-Lys-containing counterparts were accepted with a lower (c. 500-fold), albeit still significant, efficiency. Weak additions of l-Ala (a MurC-type activity) and l-Ala-d-Glu were seen.
Although not essential, the Mpl enzyme may be interesting as a potential target for antibacterial compounds. Indeed, its broad specificity raises the possibility of incorporating toxic peptides into peptidoglycan. In this regard, Escherichia coli cells were grown in the presence of large concentrations of the synthetic tripeptide l-Ala-γ-d-Glu-l-Lys; however, no effects on cell growth or morphology were seen (Hervéet al., 2007). This disappointing result was explained by the poor uptake of the lysine-containing peptides by Escherichia coli (Le Roux et al., 1991). Nevertheless, further work is in progress to identify more permeant Mpl peptide substrates that are endowed with antibacterial activity.
Synthesis of UDP-MurNAc and the UDP-MurNAc-peptides
The UDP-MurNAc and the UDP-MurNAc-peptides were initially isolated from Staphylococcus aureus by Park (1952). Over the next four decades, these compounds were mainly prepared from bacterial extracts [see references in Flouret et al. (1981)]. These procedures were tedious and time-consuming; moreover, for some nucleotides (e.g. UDP-MurNAc-l-Ala), the yields were low (Michaud et al., 1987). Two advances contributed to major improvements in this field: (1) the development of methods of synthesis of UDP-MurNAc and (2) the availability of pure Mur ligases in large quantities.
The first chemical synthesis of UDP-MurNAc was published 40 years ago, although the description was incomplete and the authors could not separate the α and β anomers of the product (Heymann et al., 1968). Blanot and coworkers described a synthesis starting from commercially available benzyl N-acetyl-4,6-O-benzylidene muramic acid and separated the two anomers of the product by HPLC (Blanot et al., 1994). This synthesis was improved and scaled up by Dini et al. (2000). Recently, the synthesis was further modified and optimized using different protecting and coupling strategies (Babič & Pečar, 2007; Kurosu et al., 2007). The key steps of all these synthetic schemes are the introduction of the phosphate group at the anomeric centre of a suitably protected MurNAc derivative, and the coupling of MurNAc-1-phosphate with uridine 5′-phosphomorpholidate.
The in vitro enzymatic synthesis of UDP-MurNAc from UDP-GlcNAc using MurA and MurB was described for the first time by Benson et al. (1993). The sensitivity of MurB to substrate and product inhibition and the intrinsic NADPH oxidase activity of MurB complicate this procedure. Improvements have been described, though, such as operating under an argon atmosphere (Reddy et al., 1999) or using an in situ NADPH regeneration system (Liu et al., 2001). An enzymatic synthesis of UDP-[14C]MurNAc has been published (Bouhss et al., 2002), while using MurA alone, UDP-GlcNAc-enolpyruvate, the substrate of MurB, can be prepared (Benson et al., 1993b).
The analysis and purification of the UDP-MurNAc-peptides have been aided considerably by the introduction of reverse-phase HPLC (Flouret et al., 1981). As well as for synthesis, this technique has found applications in the measurement of the enzymatic activity of the Mur ligases (Liger et al., 1991; Auger et al., 1995) and in the analysis of the intracellular pools of these nucleotide precursors (Mengin-Lecreulx et al., 1982).
For the setting up of in vitro inhibitory assays, studies have used the in situ production of UDP-MurNAc-peptides from UDP-GlcNAc by MurA, MurB and the Mur ligases (Wong et al., 1998; El Zoeiby et al., 2001). Similarly, a MurF inhibitory assay in which the meso-A2pm-containing or l-Lys-containing UDP-MurNAc-tripeptide is produced by Mpl in situ has been described (Baum et al., 2006). In both cases, HPLC was used to analyse the nucleotides formed.
Inhibitors of the Mur ligases
Several simple compounds that are structurally related to l-Ala have been reported to be moderate inhibitors of MurC from Escherichia coli (Liger et al., 1991, 1995). A series of phosphinate transition-state analogues of MurC were prepared and compound 7 (Fig. 15) was identified as its most potent inhibitor, with an IC50 value of 49 nM. Biochemical characterization revealed that it has a mixed-type inhibition with respect to all three substrates. Any structural modification of this inhibitor significantly reduced the inhibitory activity (Marmor et al., 2001; Reck et al., 2001). Benzylidene rhodanines inhibited MurC with micromolar IC50 values (12–27 μM): in a whole-cell assay, they were active against methicillin-resistant Staphylococcus aureus (MRSA), but not against Escherichia coli (Sim et al., 2002). 2-Phenyl-5,6-dihydro-2H-thieno[3,2-c]pyrazol-3-ol derivatives with the general formula 8 showed good inhibitory activities against Staphylococcus aureus MurB (the best IC50 values ranged from 3.6 to 24 μg mL−1), MurC (IC50 values between 7 and 25 μg mL−1) and MurD (IC50 values between 8.3 and 25 μg mL−1) enzymes and promising antimicrobial activities against Gram-positive bacteria, including resistant strains. Also here, the MIC values increased to above 128 μg mL−1 when these compounds were tested in the presence of BSA (Li et al., 2003). A benzofuran acyl-sulfonamide derivative was discovered by AstraZeneca and shown to act competitively with ATP and UDP-MurNAc (IC50=2.3 μM), but unfortunately it also has high-affinity binding to BSA (Ehmann et al., 2004). More than 180 pulvinones [e.g. 9: IC50 (MurC)=8 μg mL−1)] were synthesized and evaluated as inhibitors of MurA-D. They consistently inhibited MurC (IC50 values in the 1–10 μg mL−1 range), and also MurA and MurB to lesser extents, while demonstrating antibacterial activities against Gram-positive bacteria, including resistant strains (Antane et al., 2006). Two peptide inhibitors of MurC from Pseudomonas aeruginosa, with IC50 values of 1.5 and 0.9 mM, were selected by phage display (El Zoeiby et al., 2003b).
The effects of various analogues of d-Glu on MurD from Escherichia coli have been studied, and have yielded some moderate inhibitors (Pratviel-Sosa et al., 1994). Many phosphonic acids and phosphinates (e.g. 10: IC50=20 nM) have been developed so far as substrate analogues and tetrahedral transition-state analogue inhibitors of MurD, respectively (Auger et al., 1995; Tanner et al., 1996; Gegnas et al., 1998; Snyder et al., 1999; Victor et al., 1999; Gobec et al., 2001; Štrancar et al., 2006), and quantitative structure–activity relationship (QSAR) studies have been performed for some of these (Kotnik et al., 2004). From a library of N-acyl-d-Glu derivatives, several inhibitors have been identified; those containing an indole moiety appeared to be of special interest (Victor et al., 1999). Recently, the high-resolution crystal structures of MurD in complexes with N-sulfonyl-d-Glu (IC50=280 μM) and N-sulfonyl-l-Glu (IC50=710 μM) (11) have been solved (Fig. 16). The binding modes of these inhibitors have also been characterized kinetically (Kotnik et al., 2007b). Using a de novo structure-based molecular design, a series of macrocyclic inhibitors 12 were developed, which showed IC50 values between 0.7 and 5.1 μM (Horton et al., 2003). In addition, peptide inhibitors of MurD from Pseudomonas aeruginosa (best IC50 value: 4 μM) have been obtained by the screening of phage display libraries using competitive biopanning approaches (Paradis-Bleau et al., 2006).
The first inhibitors of MurF were pseudo-tripeptide and pseudo-tetrapeptide aminoalkylphosphinic acids of the general structure X-Lys-Ψ(PO2H)-Gly-Ala that had been synthesized as transition-state analogues. They acted as reversible competitive inhibitors, with Ki values in the range of 200–700 μM (Miller et al., 1998). With the affinity selection screening, two small-molecule MurF leads (e.g. 14: IC50=1 μM) were discovered by Abbott Laboratories and cocrystallized with the enzyme (Gu et al., 2004; Longenecker et al., 2005; Comess et al., 2006) (Fig. 13d). Both these were bound in the substrate-binding region and induced domain closure. After the structure-based lead optimization, a series of potent inhibitors were obtained, culminating in compound 15 (IC50=22 nM), although none of these exhibited significant antibacterial activities even in the presence of bacterial cell permeabilizers (Gu et al., 2004; Stamper et al., 2006). Recently, three QSAR models were constructed to further facilitate the search for new inhibitors with extensive physicochemical properties (Kong et al., 2007). Using an Mpl-based assay, a thiazolylaminopyrimidine series of MurF inhibitors with IC50 values as low as 2.5 μM have also been identified (Baum et al., 2006).
Besides the hexoses and nucleotides described in the previous sections, the cytoplasmic steps of peptidoglycan biosynthesis involve a certain number of substrates (Fig. 1). Most participate in many metabolic pathways (acetyl-CoA, UTP, PEP, NADPH, ATP and l-amino acids), but some are more specific (meso-A2pm, d-Ala, d-Ala-d-Ala and d-Glu). Because meso-A2pm is the last intermediate in the l-Lys biosynthesis pathway, which has been the subject of several reviews (Patte, 1983; Cox, 1996; Born & Blanchard, 1999; Cox et al., 2000; Hutton et al., 2007), it will not be considered here. Therefore, only the formation of d-Ala-d-Ala (and its precursor d-Ala) and d-Glu will be described.
Formation of d-Ala-d-Ala
Essentially found in peptidoglycan, d-Ala is also present in lipoteichoic acids of Gram-positive organisms (Volkman et al., 2001). It is produced from l-Ala through the action of a pyridoxal 5′-phosphate (PLP)-dependent enzyme, alanine racemase. Some organisms contain only one alanine racemase, while others have two, encoded by the alr and dadX genes. The latter group includes Escherichia coli (Wild et al., 1985), Salmonella typhimurium (Wasserman et al., 1984; Esaki & Walsh, 1986) and Pseudomonas aeruginosa (Strych et al., 2000). Expression of the alr gene is constitutive and provides the d-Ala that is necessary to maintain cell growth, while the dadX-encoded racemase is inducible and required only when l-Ala is used as a carbon and energy source. While Alr from Enterococcus faecalis (Badet & Walsh, 1985), Salmonella typhimurium (Esaki & Walsh, 1986), Thermus thermophilus (Seow et al., 2000), Shigella sp. (Yokoigawa et al., 2001) and Helicobacter pylori (Saito et al., 2007) have been described as monomers, those from Escherichia coli, Pseudomonas aeruginosa (Strych & Benedik, 2002), Bifidobacterium bifidum (Yamashita et al., 2003), Corynebacterium glutamicum (Oikawa et al., 2006), Acidophilium organovorum (Seow et al., 1998), Mycobacterium sp. (Strych et al., 2001) and Bacillus stearothermophilus (Inagaki et al., 1986) are dimers that are formed between identical polypeptide chains of c. 40 kDa. Structural studies on Alr from Bacillus stearothermophilus (Shaw et al., 1997; Morollo et al., 1999) and Mycobacterium tuberculosis (LeMagueres et al., 2005) have established that each monomer consists of two different domains: an N-terminal domain made up of an α/β-barrel, and a C-terminal one primarily made up of β-strands. Each subunit contains one PLP molecule (Fig. 17). Kinetic analyses (Sawada et al., 1994) and X-ray crystallographic studies (Shaw et al., 1997; Watanabe et al., 2002) have revealed that the Alr reaction proceeds via a two-base mechanism (Fig. 18). First, the PLP bound to the active-site residue Lys39 (Bacillus stearothermophilus numbering) reacts with l-Ala to form an external Schiff base through transaldimination. The reaction continues with the abstraction of the Cα proton of the external aldimine (l-Ala-PLP adduct) by Tyr265 from the second monomer. In the next step, the planar carbanion intermediate is reprotonated by Lys39 from the opposite side to yield the d-enantiomer. Finally, the internal aldimine between Lys39 and PLP is formed again, displacing d-Ala from its covalent linkage to the cofactor. Conversion of d- into l-enantiomer proceeds in the opposite manner. The importance of the catalytic bases Lys39 and Tyr265 has been revealed by several studies using mutant enzymes (Watanabe et al., 1999a, b). Replacing Tyr265 by Ala in Alr from Bacillus stearothermophilus resulted in a novel aldolase activity (Seebeck & Hilvert, 2003). Another interaction of great importance for the racemization process is the direct hydrogen bond between the unprotonated pyridine nitrogen of PLP and Arg219 (Shaw et al., 1997; Morollo et al., 1999). In other PLP-dependent enzymes, but not with Alr, the pyridine nitrogen of the cofactor is in the protonated form and acts as an effective electron sink, stabilizing the carbanion intermediate by forming a quinonoid species. The unusual interaction between the positively charged guanidino group of Arg219 and the pyridine nitrogen of PLP in Alr was explored by analysis of the R219K, R219A and R219E mutants (Sun & Toney, 1999). It was thus demonstrated that for efficient catalysis, a positively charged residue is required in this position. Molecular dynamics simulations have revealed that the enhancement of the carbon acidity of the α-amino acid by PLP via unprotonated pyridine is mainly due to solvation effects, in contrast to the intrinsic electron-withdrawing stabilization by the pyridinium ion to form a quinonoid intermediate (Major & Gao, 2006).
Cycloserine is a naturally occurring suicide substrate of many PLP-dependent enzymes; in this respect, d-cycloserine (Fig. 19) inactivates Alr in a time-dependent manner (Wang & Walsh, 1978). After the initial formation of an external aldimine between cycloserine and PLP, a proton is transferred from the C-2 of the substrate to the C-4′ of the cofactor, resulting in the formation of a ketimine species. This ketimine proceeds through a second prototropic shift, forming a stable isoxazole (Fig. 19). Together with kinetic studies, the crystal structures of l- and d-cycloserine-inactivated Alr (Fenn et al., 2003) and the Y265F mutant enzyme (Fenn et al., 2005) from Bacillus stearothermophilus have emphasized the importance of the catalytic bases Tyr265′ and Lys39 for the inactivation process.
A series of N(2)-substituted derivatives of compound 16 (Kim et al., 2003b) (Fig. 20) and five- and six-membered heterocycles (Kim et al., 2003a) were prepared and evaluated for inhibitory activities against Alr from various bacterial species, as well as in growth inhibition assays. Some of the heterocycles had moderate inhibitory activities against Alr from Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa and Mycobacterium tuberculosis. Besides d-cycloserine, several other compounds are irreversible time-dependent mechanism-based inactivators (suicide substrates) of Alr, like O-carbamyl-d-serine (Wang & Walsh, 1978), β-chloro- and β-fluoro-d-alanine (Wang & Walsh, 1978; Badet et al., 1984), β,β-difluoro- (Wang & Walsh, 1981) and β,β,β-trifluoroalanine 17 (Faraci & Walsh, 1989) and 3-halovinylglycines 18 (Thornberry et al., 1987, 1991). Other alanine analogues that are inhibitors of Alr have been reviewed by Neuhaus & Hammes (1981). To improve in vivo antimicrobial activities, several di- and tripeptides containing β-chloro-l-Ala (Cheung et al., 1983, 1986; Boisvert et al., 1986; Le Roux et al., 1991) and halovinylglycine (Patchett et al., 1988) were synthesized as transport systems for the intracellular delivery of potentially bactericidal amino acids. l-Norvalyl-l-chlorovinylglycine showed good activity against Gram-positive organisms, including methicillin-resistant Staphylococcus species (Patchett et al., 1988). Furthermore, (β-chloro-, (β-β-dichloro- and (β-β-β-trichloro-α-aminoethy) phosphonic acids (Vo-Quang et al., 1986a) and the phosphonic analogue of vinylglycine, dl-(1-amino-2-propenyl)phosphonic acid (Vo-Quang et al., 1986b), have been evaluated against Alr and Ddl from Pseudomonas aeruginosa and Enterococcus faecalis. The monochloro, dichloro and vinylglycine derivatives of Ala-P exhibited strong inhibition of Alr from both species tested, whereas only the Ddl from Enterococcus faecalis was inhibited by these compounds.
The phosphonic analogue of alanine, l-1-aminoethylphosphonic acid 19 (l-AlaP) (Atherton et al., 1979b), is a time-dependent irreversible inactivator of Gram-positive (Bacillus, Staphylococcus and Streptococcus), but not Gram-negative (Escherichia coli, Salmonella and Pseudomonas), racemases (Badet & Walsh, 1985; Badet et al., 1986). It was used as the phosphonodipeptide alaphosphin (l-Ala-l-AlaP), which can be transported by bacterial cell wall permeases and then hydrolysed to l-Ala and l-AlaP (Allen et al., 1978; Atherton et al., 1979a). The formation of an AlaP-PLP Schiff base linkage was shown by solid-state 15N-NMR of the [15N]AlaP-racemase complex (Copiéet al., 1988) and the cocrystal structure of l-AlaP with Alr (Stamper et al., 1998). The latter showed an inappropriate orientation of the external aldimine for efficient Cα proton abstraction and revealed interactions of the phosphonate group with putative catalytic residues, thereby rendering them unavailable for catalysis. Several phosphonodipeptides containing 1-amino-1-methylethanephosphonic (Zboinska et al., 1990), 1-aminomethylphosphonic (Atherton et al., 1982) and 1-aminocyclopropanephosphonic acid 20 (Erion & Walsh, 1987) were generally less potent inhibitors of the Alr and Ddl enzymes than phosphonodipeptides based on AlaP.
The replacement of the AlaP phosphonate with a boronic acid group led to the alanine analogue (1-aminoethyl)boronic acid 21, which is a slow-binding, time-dependent inhibitor of Alr (Bacillus stearothermophilus) and Ddl (Salmonella typhimurium). The inhibition is most probably due to the formation of a tetrahedral boronate anion that acts as a transition-state analogue (Duncan et al., 1989). On the other hand, compounds mimicking the PLP-Ala complex were inactive (Leung et al., 1985).
A structure-based strategy for the identification of novel Bacillus stearothermophilus Alr inhibitors with a dynamic receptor-based pharmacophore model was developed using the LigBuilder programme (Mustata & Briggs, 2002); however, these compounds have not yet been evaluated for their inhibitory activities.
The condensation of two molecules of d-Ala that were formed previously by alanine racemase is catalysed by a specific ATP-dependent D-Ala:D-Ala ligase (Ddl). Early kinetic and specificity studies of the reaction were carried out with the ligase purified from Enterococcus faecalis (Neuhaus, 1962a): they provided evidence for two d-Ala-binding sites that have different specificity patterns and Michaelis constants (Neuhaus, 1962b). The binding site for the N-terminal d-Ala, designated as the donor site, is highly specific for d-Ala whereas the C-terminal d-Ala-binding site, the acceptor site, is less specific and accepts a variety of d-amino acids. Ddl enzymes from Oceanobacillus iheyensis, Synechocystis sp. (Sato et al., 2005) and Thermotoga maritima (Sato et al., 2006) were also isolated and characterized for their substrate specificities. Recently, the entire MurC-Ddl fusion protein from Chlamydia trachomatis (see ‘Biosynthesis of the UDP-MurNAc-peptides’) was purified; the presence of the MurC domain appeared to be required for Ddl activity (McCoy & Maurelli, 2005). The existence of the two distinct genes ddlA and ddlB in Escherichia coli (Zawadzke et al., 1991; Al-Bar et al., 1992) and Salmonella typhimurium (Daub et al., 1988) was demonstrated. Both genes were cloned and overexpressed, and their products have been purified. Despite a difference in size, both enzymes show very similar kinetic characteristics and substrate specificities.
The kinetic mechanism, which has been well studied with the enzyme from Salmonella typhimurium, has been shown to be an ordered ter–ter reaction with ATP as the first substrate to bind, and ADP the last product off. The reaction is reversible; however, as the exchange reaction is not strictly ADP-dependent, this can argue in favour of some randomness in the kinetic mechanism (Mullins et al., 1990).
The reaction starts with the attack on the first d-Ala by the γ-phosphate of ATP, to give an acylphosphate. This is followed by an attack by the amino group of the second d-Ala to yield a tetrahedral intermediate, which collapses into d-Ala-d-Ala and Pi (Mullins et al., 1990; Healy et al., 2000b). In this respect, the mechanism is similar to that of the Mur ligases (see ‘Biosynthesis of the UDP-MurNAc-peptides’). Ddl is strongly inhibited by its reaction product d-Ala-d-Ala (Mullins et al., 1990).
Despite the similarity of the Ddl and Mur ligase reactions, Ddl does not belong to the Mur ligase family. Sequence alignments and crystal structures have revealed that it belongs to the ATP-grasp family, which is composed of highly diverse enzymes that catalyse the ATP-dependent ligation of a carboxyl group to an amino or imino nitrogen, a hydroxyl oxygen or a thiol sulphur (Galperin & Koonin, 1997; Kobayashi & Go, 1997) (Fig. 21).
The most important inhibitor of Ddl is undoubtedly a structural analogue of d-Ala, d-cycloserine (formula in Fig. 19) (Strominger et al., 1960; Neuhaus & Lynch, 1964). A series of phosphinates [e.g. 22: IC50 (Streptococcus faecalis)=35 μM] (Fig. 20), phosphonates and phosphonamidates have been developed as transition-state analogue inhibitors or as analogues of d-alanyl phosphate (Lacoste et al., 1979; Parsons et al., 1988; Chakravarty et al., 1989; Lacoste et al., 1991; Ellsworth et al., 1996). It was shown that transition-state mimetics can be phosphorylated by Ddl and inhibit the reaction by their tight binding to the enzyme after this phosphorylation (Duncan & Walsh, 1988; McDermott et al., 1990). Although their antibacterial activities are low, they enabled the crystallographic determination of complexes of Escherichia coli DdlB with ADP/phosphorylated phosphinate (Fan et al., 1994) (Fig. 21) and the Escherichia coli Y216F DdlB mutant with ADP/phosphorylated phosphonate (Fan et al., 1997). Recently, two patent applications have been filed describing structure-based drug discovery methods for identifying inhibitors that target the ATP-binding site as well as new heterocyclic inhibitors (Moe et al., 2003; Navia et al., 2003). Using a de novo structure-based molecular design, cyclopropane derivative 23 [Ki (Escherichia coli)=12.5 μM] was developed as an inhibitor of DdlB (Besong et al., 2005). An allosteric inhibitor of Ddl from Staphylococcus aureus (24: Ki=4 μM) was discovered by high-throughput screening, and it was cocrystallized with the enzyme (Liu et al., 2006). Recently, diazenedicarboxamides [e.g. 25: IC50 (Escherichia coli)=15 μM] were described as inhibitors of DdlB from Escherichia coli (Kovačet al., 2007). In addition, the crystal structure of the apo form of Ddl from Thermus caldophilus has been solved, providing an insight into the substrate-induced conformational changes, which could be important for inhibitor design (Lee et al., 2006a).
In vancomycin-resistant strains, the d-Ala-d-Ala termini are replaced by either d-Ala-d-Lac or by d-Ala-d-Ser. The affinities of these alternate termini for the antibiotic vancomycin are three and one orders of magnitude, respectively, lower than that of the conventional d-Ala-d-Ala termini. Depsipeptide d-Ala-d-Lac and dipeptide d-Ala-d-Ser are synthesized by enzymes known as VanA-G according to the vancomycin-resistance phenotype considered. Together with Ddl, the Van enzymes constitute a superfamily of d-Ala : d-X ligases (X=d-Ala, d-Lac or d-Ser) that share a common reaction mechanism and the ATP-grasp three-dimensional structure. Differences in the specificity of the acceptor site have been explained through their pH profiles and active-site residues [see Healy et al. (2000a) for a review]. The crystal structures of d-Ala:d-Lac ligases from Leuconostoc mesenteroides and Enterococcus faecium are also available (Kuzin et al., 2000; Roper et al., 2000).
Formation of d-glutamic acid
In addition to its occurrence in peptidoglycan (Vollmer et al., 2008), d-glutamic acid is a key component of capsular poly-γ-glutamate of some bacterial genera (Ashiuchi & Misono, 2002; Candela & Fouet, 2006). Two distinct enzymes have been identified for the formation of d-glutamate: glutamate racemase (MurI) and d-amino acid aminotransferase (d-AAT).
Glutamate racemase catalyses the interconversion of d- and l-enantiomers of glutamate (Doublet et al., 1993). The murI genes (also called racE) from various bacteria have been cloned. Contrary to the situation in most organisms (including Escherichia coli) that contain only one glutamate racemase gene [see references in Dodd et al. (2007)], the Bacillus subtilis and Bacillus anthracis genomes have two genes that have been named racE1 and racE2 for the former species, and racE (also called glr) and yrpC for the latter (Ashiuchi et al., 1999; Kimura et al., 2004; Shatalin & Neyfakh, 2005).
The MurI racemases from c. 10 species, including Bacillus pumilus, Pediococcus pentosaceus, Lactobacillus fermenti and Escherichia coli, have been purified [see references in Dodd et al. (2007)], as well as the RacE1 and RacE2 enzymes from Bacillus anthracis (Dodd et al., 2007; May et al., 2007) and the RacE and YrpC enzymes from Bacillus subtilis (Ashiuchi et al., 1998, 1999). Thorough biochemical studies have been performed for some of them. The Escherichia coli MurI racemase requires UDP-MurNAc-l-Ala for its activation, and it has been shown that this nucleotide precursor exerts its effects by increasing both the substrate-binding affinity and the turnover rate of the enzyme (Doublet et al., 1994; Ho et al., 1995). Therefore, the formation of d-glutamic acid is regulated by UDP-MurNAc-l-Ala and thus adjusted to the requirement of peptidoglycan synthesis. Moreover, an excessive racemization of the large intracellular pool of l-glutamic acid is avoided (Doublet et al., 1993, 1994). This activating mechanism appears to be unique for Escherichia coli because glutamate racemases from other bacteria are not activated by UDP-MurNAc-l-Ala (Yoshimura & Esaki, 2003; Lundqvist et al., 2007). It was hypothesized that a 21-amino-acid extension at the N-terminus of the Escherichia coli enzyme, which is the major difference between Escherichia coli and other bacteria, was responsible for the nucleotide-mediated activation (Doublet et al., 1994; Ho et al., 1995); however, the N-terminal-truncated Escherichia coli enzyme retained the activation, although to lesser extents (Ho et al., 1995; Doublet et al., 1996). The glutamate racemase from Pediococcus pentosaceus is inhibited by haemin due to the formation of a stoichiometric complex (Choi et al., 1994). Recently, it was shown that glutamate racemases from Escherichia coli, Bacillus subtilis (YrpC, but not RacE) and Mycobacterium tuberculosis inhibit DNA gyrase activity; in the case of Escherichia coli, the presence of the nucleotide precursor was required (Ashiuchi et al., 2002, 2003; Sengupta et al., 2006).
Several crystal structures of bacterial glutamate racemases have been reported (Hwang et al., 1999; Ruzheinikov et al., 2005; Kim et al., 2007; Lundqvist et al., 2007; May et al., 2007) (Fig. 23). Along with analytical methods such as size-exclusion chromatography or equilibrium ultracentrifugation, structure examination has led to a classification of glutamate racemases according to their quaternary structure (Lundqvist et al., 2007; May et al., 2007): monomer (Escherichia coli, Lactobacillus fermenti, Bacillus subtilis YrpC) (Fig. 23a), monomer–dimer equilibrium (Aquifex pyrophilus, Bacillus anthracis RacE1 and Bacillus subtilis RacE) or dimer (Helicobacter pylori, Enterococcus faecalis, Enterococcus faecium, Staphylococcus aureus and Bacillus anthracis RacE2). Among dimers, a tail-to-tail orientation (i.e. with the active sites opposed and fully exposed to the solvent) is the general case; however, the Helicobacter pylori enzyme appears as a head-to-head dimer (i.e. with the active sites in close proximity in a face-to-face orientation) (Fig. 23b).
Owing to the essential role of glutamate racemase for the viability of bacteria, this enzyme is an attractive target for the development of antibacterial agents. Several inhibitors of glutamate racemase have been reported to date (Fig. 24). Ashiuchi et al. (1993) showed that l-serine O-sulphate 26 behaves as a suicide substrate of the enzyme from Pediococcus pentosaceus; a mechanism by which an α-aminoacrylate intermediate reacts with a nucleophilic group of the enzyme was proposed. Tanner and coworkers synthesized aziridino glutamate 27, which irreversibly inactivates Lactobacillus fermenti glutamate racemase through the alkylation of a Cys residue in the active site (Tanner & Miao, 1994), and d-N-hydroxyglutamate 28, which was shown to be a competitive inhibitor (Ki=56 μM) (Glavas & Tanner, 1997). The research group at Eli Lilly discovered the first potent competitive inhibitors that showed antibacterial activity: 4-substituted d-Glu analogues with various aryl and heteroaryl substituents (e.g. 29), with IC50 values as low as 30 nM and a good correlation between inhibitory and antibacterial activities against Streptococcus pneumoniae (de Dios et al., 2002). Pyrazolopyrimidinedione derivative 30 was discovered from a high-throughput screen by AstraZeneca; it behaved as an uncompetitive inhibitor of Helicobacter pylor MurI (IC50=1.4 μM) and prevented the growth of Helicobacter pylori cells (MIC=4 μg mL−1) (Lundqvist et al., 2007). Kim et al. (2000) isolated peptide ligands from a random phage display library that inhibit Escherichia coli glutamate racemase activity. Among these, the peptide sequence His-Pro-Trp-His-Lys-Lys-His-Pro-Asp-Arg-Lys-Thr has the highest affinity for the enzyme (IC50=160 μM); however, no bactericidal or antibacterial effects were seen.
The organization of the genes involved in the biosynthetic process reviewed here is well established in Escherichia coli, where it has been facilitated by the isolation of thermosensitive mutants. Most of the genes are clustered in two regions, known as mra and mrb (murein regions A and B, respectively) (Miyakawa et al., 1972). The mra (or dcw, division cell wall) region is located at 2 min on the chromosome and contains genes involved either in peptidoglycan synthesis or in cell division in the order: mraZ-mraW-ftsL-ftsI-murE-murF-mraY-murD-ftsW-murG-murC-ddlB-ftsQ-ftsZ-envA (Mengin-Lecreulx et al., 1989). The first nine genes are under the control of the Pmra promoter (Mengin-Lecreulx et al., 1998). The mrb region is located at 90 min and contains the murI and murB genes. The glmM and murA genes are in the 69 min region, and the glmU and glmS genes in the 83 min region. The ddlA, dadX, alr and mpl genes are located at 8.5, 27, 92 and 96 min, respectively (Berlyn et al., 1996). The mra region has been located in other bacteria such as Bacillus subtilis, Staphylococcus aureus, Enterococcus faecalis and Pseudomonas aeruginosa (Daniel & Errington, 1993; Pucci et al., 1997; El Zoeiby et al., 2000; Azzolina et al., 2001); in the last of these, its organization is exactly the same as in Escherichia coli.
Peptidoglycan integrity being necessary for the bacterial cell's survival, genes involved in its biosynthesis must be essential provided no isogenes are present. Data about the essential character of the genes coding for the cytoplasmic enzymes can be found either in former literature [see references in van Heijenoort (2001)], or inferred from the recent systematic construction of single-gene knock-out mutants of Escherichia coli (Baba et al., 2006). In this organism, the glmS, glmM, glmU, murA, murB, murC, murD, murE, murF and murI genes are essential. Isogenes (ddlA/ddlB, alr/dadX) can be individually inactivated, showing that the remaining isozyme is sufficient to sustain growth (Wild et al., 1985; Zawadzke et al., 1991). As other genes involved in the recycling pathway, the mpl gene is dispensable for Escherichia coli growth (Mengin-Lecreulx et al., 1996).
A systematic gene inactivation was also performed in Bacillus subtilis (Kobayashi et al., 2003). The glmS, ybbT (glmM), gcaD (glmU), murA, murB, murC, murD, murE, alr, ddl and racE genes could not be inactivated. As far as alr and ddl are concerned, this is explained because Bacillus subtilis, contrary to Escherichia coli, contains only one alanine racemase and one Ddl. The racE gene was claimed to be essential as no knockout mutants could be found on solid medium, suggesting that yrpC cannot take over. However, the disruption of the racE gene has been described recently (Ashiuchi et al., 2007); the RacE-less mutants obtained can grow with a reduced rate in liquid medium even in the absence of exogenous d-glutamate. The authors concluded that the RacE activity, although important for ensuring maximum growth rate, is dispensable, and that the YrpC racemase probably operates as an anaplerotic enzyme for RacE (Ashiuchi et al., 2007).
Although deletion of both murA1 and murA2 genes in Streptococcus pneumoniae is lethal, each can be inactivated individually without effect on cell growth (Du et al., 2000). In Bacillus anthracis, while the racE1 knock-out leads to moderate growth defect that can be alleviated fully by d-glutamate, the racE2 knock-out severely inhibits bacterial growth, which is only partially restored by exogenous d-glutamate (Shatalin & Neyfakh, 2005).
It is important to mention that some mutations in genes involved in the cytoplasmic steps in MRSA affect their susceptibility to methicillin; this is the case for glmS (Komatsuzawa et al., 2004), glmM (Jolly et al., 1997; Glanzmann et al., 1999) and murE (Ludovice et al., 1998). In Pseudomonas aeruginosa, a transcriptional repressor of the glmS expression, known as glmR, has been described. The inactivation of the glmR gene dramatically sensitizes the microorganism to a large variety of antibiotics, and in particular the aminoglycosides (Ramos-Aires et al., 2004).
Since the discovery of Park's nucleotide in 1952, the cytoplasmic steps of peptidoglycan biosynthesis have been defined completely. In particular, considerable progress was made during the 1990s, where the overproduction and purification of the enzymes became possible, along with the production of the noncommercially available substrates by chemical and/or enzymatic synthesis. Studies using enzyme kinetics, isotope-based experiments, NMR, sequence alignments, site-directed mutagenesis, etc., have allowed the reaction mechanisms to be established and their substrate specificities to be defined. More recently, the availability of X-ray and NMR structures has improved our knowledge of these mechanisms; furthermore, these investigations have revealed important conformational changes upon ligand binding. Generally, they were performed with a representative enzyme, and often that of Escherichia coli. In the future, it will be important to compare enzymes from different species with respect to ligand binding and reaction mechanism. This is obvious when one of the substrates differs (e.g. MurE from Gram-negative vs. Gram-positive species, d-Ala : d-X ligases); however, even when the substrates are invariant, significant differences may prevail (e.g. MurB types I and II, alanine racemases). Such comparative enzymology of bacterial species needs to be performed.
As peptidoglycan is an essential component of the cell wall of eubacteria, these enzymes represent targets for antibacterial agents. However, they have been underexploited so far, perhaps due to the very few natural antibiotics that are known to inhibit them (bacilysin, fosfomycin, d-cycloserine). The proliferation of resistant bacterial strains renders the search for new antibacterial compounds urgent, and in this respect, these enzymes involved in the cytoplasmic steps need to be considered seriously. Although good inhibitors have been described recently, most are devoid of antibacterial activity due to their inability to cross the cytoplasmic membrane, and thereby to reach their target. The challenge for the next few years will be to design compounds that are endowed with penetration properties while retaining their affinities for their respective targets.
This work was supported by the European Commission through the EUR-INTAFAR project (LSHM-CT-2004-512138), the Centre National de la Recherche Scientifique (UMR 8619 and PICS 3729), the Délégation Générale pour l'Armement (Contrats Jeune Chercheur 036000104 and 056000030 to A.B.), the Ministry of Education, Science and Sports of the Republic of Slovenia, the Franco-Slovene Proteus programme and the Institut Charles Nodier (Ljubljana). The authors thank Samo Turk for help in preparing protein figures, and Chris Berrie for critically reading the manuscript. They apologize to those researchers whose work could not be quoted because of limited space.
Protein figures presented in this review were prepared with PyMol [DeLano WL, The PyMol Molecular Graphics System (2002), DeLano Scientific, Palo Alto, CA, USA; http://www.pymol.org].
Although its essential character had been demonstrated by Mengin-Lecreulx & van Heijenoort (1996), glmM (referred to in the ‘Genetic organization’ section above) was annotated as nonessential in the article of Baba et al. (2006). As a matter of fact, it has been proved that the glmM strain of Baba et al. contains an intact copy of the glmM gene (D. Mengin-Lecreulx, pers. commun.).