OUP user menu

Peptidoglycan structure and architecture

Waldemar Vollmer, Didier Blanot, Miguel A. De Pedro
DOI: http://dx.doi.org/10.1111/j.1574-6976.2007.00094.x 149-167 First published online: 1 March 2008


The peptidoglycan (murein) sacculus is a unique and essential structural element in the cell wall of most bacteria. Made of glycan strands cross-linked by short peptides, the sacculus forms a closed, bag-shaped structure surrounding the cytoplasmic membrane. There is a high diversity in the composition and sequence of the peptides in the peptidoglycan from different species. Furthermore, in several species examined, the fine structure of the peptidoglycan significantly varies with the growth conditions. Limited number of biophysical data on the thickness, elasticity and porosity of peptidoglycan are available. The different models for the architecture of peptidoglycan are discussed with respect to structural and physical parameters.

  • peptidoglycan
  • murein
  • bacterial cell wall


Peptidoglycan (murein) is an essential and specific component of the bacterial cell wall found on the outside of the cytoplasmic membrane of almost all bacteria (Rogers et al., 1980; Park, 1996; Nanninga, 1998; Mengin-Lecreulx & Lemaitre, 2005). Its main function is to preserve cell integrity by withstanding the turgor. Indeed, any inhibition of its biosynthesis (mutation, antibiotic) or its specific degradation (e.g. by lysozyme) during cell growth will result in cell lysis. Peptidoglycan also contributes to the maintenance of a defined cell shape and serves as a scaffold for anchoring other cell envelope components such as proteins (Dramsi et al., 2008) and teichoic acids (Neuhaus & Baddiley, 2003). It is intimately involved in the processes of cell growth and cell division. Peptidoglycan (and the genetic arsenal necessary to its biosynthesis) is absent in Mycoplasmas, Planctomyces and the scrub typhus agent Orientia (Rickettsia) tsutsugamushi (Moulder et al., 1993; Tamura et al., 1995). It has never been detected in Chlamidiae although most biosynthetic genes exist (Chopra et al., 1998; Ghuysen & Goffin, 1999). Conversely, peptidoglycan is present in the photosynthetic organelles (‘cyanelles’) of the glaucocystophyte algae (Aitken & Stanier, 1979). A few biosynthetic genes (but no peptidoglycan) have been found in the green plant Arabidopsis thaliana (five genes) and the moss Physcomitrella patens (nine genes); these genes would be involved in chloroplast division (Machida et al., 2006).

This review provides a brief overview on the diversity and variability of the chemical structure of peptidoglycan in different bacteria, and summarizes the available data on the biophysical properties of the cell wall. Finally, structural aspects required for modelling the architecture of the peptidoglycan sacculus are discussed.

Chemical structure of peptidoglycan

General structure and characteristic features

The main structural features of peptidoglycan are linear glycan strands cross-linked by short peptides (Rogers et al., 1980) (Fig. 1). The glycan strands are made up of alternating N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) residues linked by β-1→4 bonds. The d-lactoyl group of each MurNAc residue is substituted by a peptide stem whose composition is most often 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 lost in the mature macromolecule. Cross-linking of the glycan strands generally occurs between the carboxyl group of d-Ala at position 4 and the amino group of the diamino acid at position 3, either directly or through a short peptide bridge. Therefore, the chemical traits of this heteropolymer involve the presence of an unusual sugar (MurNAc), of γ-bonded d-Glu, of ld (and even dd) bonds and of nonprotein amino acids (e.g. A2pm).

Figure 1

Structure of the peptidoglycan of Escherichia coli. The glycan strands consist of alternating, β-1→4-linked GlcNAc and MurNAc residues, and are terminated by a 1,6-anhydroMurNAc residue. The yellowish labelled part represents the basic disaccharide tetrapeptide subunit (monomer), which is also written with the conventional amino acid and hexosamine abbreviations on the left-hand side. The middle part shows a cross-linked peptide, with the amide group connecting both peptide stems drawn in red. [Reproduced with permission from Mengin-Lecreulx D & Lemaitre B (2005), Copyright (© SAGE Publications, 2005), by permission of Sage Publications Ltd].

The structural features outlined in the preceding paragraph are retrieved in all bacterial species known to date. However, a certain degree of variation exists either in the peptide stem, in the glycan strands or in the position or composition of the interpeptide bridge. Interspecies variation is the general case; it has been the subject of the monumental work of Schleifer & Kandler (1972) 35 years ago and has served in the establishment of the tri-digital system of peptidoglycan classification used universally. However, in the same species, there can be variations in the fine structure according to the growth conditions (growth phase, medium composition, intra/extracellular growth, presence of antibiotics). In the next sections, this study will present an overview of the different types of variations encountered.

The glycan strands in peptidoglycan

The glycan strands are formed by oligomerization of monomeric disaccharide peptide units (lipid II) by transglycosylation reactions. Secondary modifications in the glycan strands such as N-deacetylation, O-acetylation and N-glycolylation are frequently found and are the topic of another review in this issue (Vollmer, 2008). In the Gram-positive Staphylococcus aureus, the glycan strands may contain either a MurNAc or a GlcNAc residue at the reducing end; the latter residue indicates that cleavage of the strand by an N-acetylglucosaminidase had occurred (Boneca et al., 2000). In all Gram-negative and some Gram-positive species (e.g. Bacillus subtilis), the glycan strands do not have a reducing (MurNAc or GlcNAc) end but terminate with a 1,6-anhydroMurNAc residue, which has an intramolecular ring from C1 to C6. It is not known whether the 1,6-anhydroMurNAc residues present in the sacculi have been formed during termination of glycan strand synthesis, or whether they are the result of degradation by lytic transglycosylases, or both. In species with high activity of glycan strand-cleaving enzymes (glucosaminidases and muramidases), the peptidoglycan may contain glycan strands with all possible combinations of GlcNAc and MurNAc residues at the ends. These hydrolytic enzymes must be inactivated rapidly and removed quantitatively when peptidoglycan is prepared for glycan strand length analysis to avoid cleavage of the strands after peptidoglycan isolation (Ward, 1973). Different methods have been applied to determine the average length of the glycan strands and the length distribution: (1) quantification of the fraction of the reducing hexosamine residues after chemical reduction (Rogers, 1970; Ward, 1973), (2) enzymatic addition of galactosamine residues to the GlcNAc end and their quantification (Schindler et al., 1976), (3) quantification of the fraction of the 1,6-anhydroMurNAc-containing disaccharide peptide subunits that are the hallmark of the glycan strand ends in some species (Glauner et al., 1988; Quintela et al., 1995a) and (4) release of the glycan strands by an amidase and their purification by an anion exchange column, followed by their separation by reverse-phase HPLC (Harz et al., 1990). The latter method is restricted to the separation of glycans from 1 to 30 disaccharide units. Glycans that are longer than 30 disaccharide units elute together in one peak.

There are only limited data on the average chain length and the chain length distribution of the glycan strands in different species. Remarkably, the average chain length of the glycan strands does not correlate with the thickness of the peptidoglycan layer, as there are Gram-positive species with a thick cell wall with either short (S. aureus) or long (B. subtilis) glycan strands. Similarly, there are Gram-negative species with either short (Helicobacter pylori) or long (Proteus morganii) glycan strands.

The glycan strands in the peptidoglycan of Bacilli (B. subtilis, Bacillus licheniformis and Bacillus cereus) have an average chain length between 50 and 250 disaccharide units (Hughes, 1971; Warth & Strominger, 1971; Ward, 1973). In contrast, the glycan strands of S. aureus are much shorter, with an average chain length of about 18 disaccharide units (Tipper et al., 1967; Ward, 1973). Separation of the staphylococcal glycan strands by HPLC revealed that the predominant chain length was between 3 and 10 disaccharide units. Longer glycan strands with more than 26 disaccharide units represented about 10–15% of the total glycan material (Boneca et al., 2000). The l-ornithine-containing peptidoglycan of Deinococcus radiodurans Sark, a Gram-positive bacterium that is extremely resistant to ionizing radiations, had glycan strands terminated by 1,6-anhydroMurNAc residues with an average chain length of about 20 disaccharide units (Quintela et al., 1999a).

The average chain length of the glycan strands in Gram-negative bacteria can be calculated from the fraction of 1,6-anhydroMurNAc residues containing disaccharide peptide subunits. Different species differ in the average chain length of the glycan strands but the normal range lies between 20 and 40 disaccharide units (Tuomanen et al., 1989; Quintela et al., 1995a, b). As shown for Escherichia coli, the average chain length of the glycan strands also varies to some extent with the strain and growth conditions (Glauner, 1988). Escherichia coli glycan strands of up to 30 disaccharide units have been separated by HPLC (Harz et al., 1990). The average chain length of the glycan strands from 1 to 30 disaccharide units was 8.9 disaccharide units in strain W7. Glycan strands longer than 30 disaccharides represented about 25–30% of the total material, and their average chain length was 45.1 disaccharide units. The average chain length of all glycan strands was estimated as 21 disaccharide units, which is slightly less than the average chain length of 25–35 disaccharide units calculated from the fraction of 1,6-anhydroMurNAc residues (Gmeiner et al., 1982; Glauner, 1988). The average glycan chain length is greater (ca. 50 disaccharide units) in the peptidoglycan from P. morganii, as calculated from the fraction of 1,6-anhydroMurNAc containing muropeptides (Quintela et al., 1995a). The glycan strands from H. pylori are particularly short, with an average chain length of <10 disaccharide units (Costa et al., 1999), and they become even shorter during transition from a spiral to a coccoid cell shape in the stationary growth phase (Costa et al., 1999; Chaput et al., 2007). It might be important for the integrity of the peptidoglycan net that the glycan strand ends are hyper-cross-linked in H. pylori (Costa et al., 1999).

Variation in the peptide stem

The variations of the peptide stem can be divided into two categories: (1) those due to the specificity of the Mur ligases, the enzymes responsible for its biosynthesis, and (2) those occurring at a later step of the biosynthesis [see accompanying chapters in this issue (Barreteau et al., 2008; Bouhss et al., 2008)]. These variations are enumerated in Table 1.

View this table:
Table 1

Amino acid variations in the peptide stem

PositionResidue encounteredExamples
1l-AlaMost species
GlyMycobacterium leprae, Brevibacterium imperiale
l-SerButyribacterium rettgeri
2d-IsoglutamateMost Gram-negative species
d-IsoglutamineMost Gram-positive species, Mycobacteria
threo-3-HydroxyglutamateMicrobacterium lacticum
3meso-A2pmMost Gram-negative species, Bacilli, Mycobacteria
l-LysMost Gram-positive species
l-OrnSpirochetes, Thermus thermophilus
l-Lys/l-OrnBifidobacterium globosum
l-Lys/d-LysThermotoga maritima
ll-A2pmStreptomyces albus, Propionibacterium petersonii
meso-LanthionineFusobacterium nucleatum
l-2,4-DiaminobutyrateCorynebacterium aquaticum
l-HomoserineCorynebacterium poinsettiae
l-AlaErysipelothrix rhusiopathiae
l-GluArthrobacter J. 39
Amidated meso-A2pmBacillus subtilis
2,6-Diamino-3-hydroxypimelateAmpuraliella regularis
l-5-HydroxylysineStreptococcus pyogenes
N γ-Acetyl-l-2,4-diaminobutyrateCorynebacterium insidiosum
4d-AlaAll bacteria
5d-AlaMost bacteria
d-SerEnterococcus gallinarum
d-LacLactobacillus casei, Enterococci with acquired resistance to vancomycin
  • * These residues result from reactions occurring posterior to the action of Mur ligases.

  • The process of formation of these residues (direct incorporation by MurE or subsequent hydroxylation of the nonhydroxylated residue) is unclear (Muñoz, 1966; Perkins et al., 1969).

  • In this organism, a 10: 1 ratio of lysine to hydroxylysine was found Muñoz (1966).

The first amino acid of the peptide stem is added by the MurC ligase. In most bacterial species, this amino acid is l-Ala; in rare cases, Gly or l-Ser is added instead (Table 1). Two interesting cases deserve to be mentioned. First, the enzymes from Mycobacterium tuberculosis and Mycobacterium leprae have the same in vitro specificity pattern towards l-Ala and Gly; however, the amino acid found at 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 the growth conditions (Mahapatra et al., 2000). The second case is that of Chlamydia trachomatis: the MurC activity adds l-Ala, l-Ser and Gly with similar in vitro efficiencies. This absence of specificity prevents from deducing the nature of the first amino acid of the putative chlamydial peptidoglycan (Hesse et al., 2003).

The amino acid at the second position is added by the MurD ligase. In all species this enzyme adds d-Glu, the modifications encountered (Table 1) occurring at a later step.

The greatest variation is found at position 3. The addition of the third amino acid is catalyzed by the MurE ligase. This amino acid is generally a diamino acid, either meso-A2pm (most Gram-negative bacteria, Mycobacteria, Bacilli) or l-Lys (most Gram-positive bacteria); in certain species, other diamino acids (l-Orn, ll-A2pm, meso-lanthionine, l-2,4-diaminobutyric acid, d-Lys) or monoamino acids (l-homoserine, l-Ala, l-Glu) are encountered (Table 1). As for the second position, further modifications of the third amino acid occur posterior to MurE action (Table 1).

In most cases, the MurE enzyme is highly specific for the relevant amino acid; this has been demonstrated for the meso-A2pm-adding and l-Lys-adding enzymes from E. coli and S. aureus, respectively (Boniface et al., 2007). In Corynebacterium pointsettiae, where l-homoserine (l-Hse) is found at position 3, MurE is specific both for the amino acid and the nucleotide precursor UDP-MurNAc-Gly-d-Glu, the ‘wrong’ nucleotide with l-Ala at position 1 being a poor substrate; this ensures a correct synthesis of the ‘right’ peptide stem Gly-γ-d-Glu-l-Hse-d-Ala-d-Ala (Wyke & Perkins, 1975). However, the MurE enzyme sometimes seems to be devoid of strict specificity, and this affects the final composition of peptidoglycan. In certain species of the genus Bifidobacterium, two diamino acids, l-Lys and l-Orn, are retrieved at the third position of the peptide stem (Schleifer & Kandler, 1972). As a matter of fact, it has been shown that MurE from Bifidobacterium globosum can incorporate both diamino acids indifferently (Hammes et al., 1977). MurE from Thermotoga maritima, a Gram-negative species whose peptidoglycan contains similar proportions of both enantiomers of lysine, but no meso-A2pm (Huber et al., 1986), is capable of adding in vitrol-Lys, d-Lys and meso-A2pm with comparable efficiencies (Boniface et al., 2006). In the l-Lys-containing UDP-MurNAc-tripeptide product, the d-Glu-l-Lys bond has the conventional γ→α arrangement; however, in the d-Lys-containing product, d-Lys is acylated by the γ-carboxylate of d-Glu on its ɛ-amino function. This leads to the synthesis of two peptide stems: the conventional l-Ala-d-Glu (γ→α) l-Lys-d-Ala-d-Ala and the unusual l-Ala-d-Glu (γ→ɛ) d-Lys. The absence of meso-A2pm in T. maritima peptidoglycan is explained by its very low intracellular pool with respect to those of l- and d-Lys (Boniface et al., 2006).

Amino acids at positions 4 and 5 are added as a dipeptide, in most cases d-Ala-d-Ala. The synthesis of the dipeptide is carried out by the Ddl enzyme and its incorporation into the peptide stem by the MurF ligase. It has been established that the latter 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 the former, which resides mainly on the N-terminal amino acid, and this constitutes a ‘double-sieving’ mechanism ensuring the synthesis of a pentapeptide stem ending mainly in d-Ala-d-Ala (Neuhaus & Struve, 1965; Duncan et al., 1990). d-Ala is predominantly found at position 4 in all species. d-Lactate (d-Lac) or d-Ser is found at position 5 in strains endowed with natural or acquired resistance to vancomycin (Table 1), the affinity of the d-Ala-d-Lac and d-Ala-d-Ser moieties for the antibiotic being much lower than that of the conventional d-Ala-d-Ala moiety (Healy et al., 2000). A certain proportion of Gly, presumably escaping from the double-sieving mechanism, is often found at position 4 or 5 in lieu of d-Ala. The proportion is low (ca. 1%) in E. coli, but reaches 19% in Caulobacter crescentus (Markiewicz et al., 1983; Glauner et al., 1988).

Other variations in peptidoglycan composition (amidation, hydroxylation, acetylation, attachment of amino acids or other groups, attachment of proteins) occur after the action of the Mur ligases, often at the level of lipid II. These modifications concern essentially positions 2 and 3. It should be mentioned that most enzymes responsible for these modifications are still unknown.

Amidation of the α-carboxyl of glutamic acid (yielding d-isoglutamine) and the ɛ-carboxyl group of meso-A2pm is very frequent. It has been shown in Mycobacterium smegmatis that lipid II is the substrate for amidation reactions; in fact, lipid II in this species appears as a mixture of non-, mono- and di-amidated molecules (Mahapatra et al., 2005). The hydrocarbon chain of d-Glu, meso-A2pm or l-Lys is hydroxylated in some species. In the case of d-Glu, it was demonstrated that hydroxylation occurs after the cytoplasmic steps (Schleifer et al., 1967); it depends on the oxygen supply during growth, cells grown under microaerophilic conditions displaying almost no hydroxylated glutamic acid (Schleifer et al., 1968). In Corynebacterium insidiosum, where unsubstituted l-2,4-diaminobutyric acid is the substrate for MurE, the γ-amino group of the diamino acid is acetylated in the UDP-MurNAc-pentapeptide precursor and in peptidoglycan (Perkins, 1971; Wyke & Perkins, 1975). In certain organisms, an amino acid or another amine-containing moiety, such as glycine (Micrococcus luteus, Arthrobacter tumescens), glycine amide (Arthrobacter athrocyaneus), d-alanine amide (Arthrobacter sp. NCIB9423), cadaverine (Selenomonas ruminantium) or N-acetylputrescine (Cyanophora paradoxa), is added to the α-carboxyl group at position 2 (Schleifer & Kandler, 1972; Kamio et al., 1981; Pfanzagl et al., 1996).

The peptide stem constitutes the point of covalent anchoring of cell envelope proteins to peptidoglycan (Dramsi et al., 2008). In E. coli and other related Gram-negative bacteria, Braun's lipoprotein is the only protein covalently linked to peptidoglycan known to date (Braun & Rehn, 1969; Braun & Sieglin, 1970). It is a 58-amino acid protein whose N-terminal glyceryl-cysteine residue is modified by the addition of three fatty acids. It is attached to the α-carboxyl group of meso-A2pm through the ɛ-amino function of its C-terminal lysine residue: in this transpeptidation reaction, the tetrapeptide moiety l-Ala-γ-d-Glu-meso-A2pm-d-Ala is the acyl donor while the side chain amine of the C-terminal residue of the lipoprotein is the acyl acceptor. Lately, it has been demonstrated that in E. coli, enzymes homologous to the l,d-transpeptidase of Enterococcus faecium are responsible for the attachment of Braun's lipoprotein to the peptide stem (Magnet et al., 2007b).

Gram-positive bacteria contain many surface proteins (e.g. protein A, fibrinonectin-binding proteins, collagen adhesin) anchored to peptidoglycan and involved in pathogenic processes. The anchoring reaction is catalyzed by a membrane protein called sortase A (Marraffini et al., 2006). Sortase A from S. aureus cleaves the Thr–Gly peptide bond of the sorting signal (consensus sequence Leu–Pro–Xaa–Thr–Gly, where Xaa is any amino acid) present in the surface protein and links the α-carboxyl group of Thr to the side-chain amine at the third position of a peptide stem. In this case, the surface protein is the acyl donor whereas the peptide moiety l-Ala-γ-(α-d-Glu-NH2)-l-Lys(Nɛ-Gly5)-d-Ala is the acyl acceptor. It has been demonstrated that the acceptor substrate of sortase A is lipid II.

Variation in cross-linking

Most variations of the peptide moiety of peptidoglycan occur in its mode of cross-linkage and in the composition of the interpeptide bridge (Fig. 2 and Table 2). There are two main groups of cross-linkage (Schleifer & Kandler, 1972). In the first group (3–4 cross-linkage), the cross-linkage extends from the amino group of the side-chain of the residue at position 3 of one peptide subunit (acyl acceptor) to the carboxyl group of d-Ala at position 4 of another (acyl donor). As mentioned above, this is the most common kind of cross-linkage. It can be either direct (most Gram-negative bacteria) or through an interpeptide bridge (most Gram-positive bacteria). In the second group (2–4 cross-linkage), found only among coryneform bacteria, especially the phytopathogenic corynebacteria, the cross-linkage extends between the α-carboxyl group of d-Glu at position 2 of one peptide subunit (acyl acceptor) and the carboxyl group of d-Ala at position 4 of another (acyl donor). In this case, for the first subunit to be an acceptor, an interpeptide bridge containing a diamino acid must be present. The cross-linking reactions are catalyzed by the transpeptidase domain of penicillin-binding proteins, enzymes that have been studied extensively, in particular in human pathogenic bacteria (Sauvage et al., 2008).

Figure 2

Examples of cross-linkage and interpeptide bridge. (a), Escherichia coli (direct 3–4 cross-link); (b), Staphylococcus aureus (3–4 cross-link with a pentaglycine bridge); (c), Corynebacterium pointsettiae (2–4 cross-link with a d-ornithine bridge); (d), Micrococcus luteus (3–4 cross-link with a bridge consisting of a peptide stem). G, N-acetylglucosamine; M, N-acetylmuramic acid.

View this table:
Table 2

Characterized branching enzymes, and nature of the interpeptide bridges synthesized

FmhB, FemA, FemBFem transferase(Gly)5Staphylococcus aureusSchneider et al. (2004)
BppA1, BppA2Fem transferasel-Ala-l-AlaEnterococcus faecalisBouhss et al. (2002)
FemXFem transferasel-Ala-l-Ser or l-Ala-l-Ser-l-AlaWeissella viridescensBiarrotte-Sorin et al. (2004)
MurM, MurNFem transferasel-Ser-l-Ala or l-Ala-l-AlaStreptococcus pneumoniaeFiser et al. (2003)
AslfmATP-graspd-AsxEnterococcus faeciumBellais et al. (2006)
  • * FemX catalyses the addition of the first amino acid residue (l-Ala) of the side chain; the second (l-Ser) and third (l-Ala) residues are added by unknown Fem transfereases (Villet et al., 2007).

  • The substrate of Aslfm is d-Asp, which is engaged through its β-carboxyl group; in mature peptidoglycan, the α-carboxyl group is partially amidated.

The size of the interpeptide bridge ranges from one to seven amino acid residues. Various amino acids are encountered: Gly, l-Ala, l- or d-Ser, d-Asx, l- or d-Glu, etc. As already mentioned, the interpeptide bridges of 2–4 cross-links contain necessarily (but not exclusively) a diamino acid (l- or d-Lys, d-Orn, d-2,4-diaminobutyrate). For many years, the important diversity of composition of the interpeptide bridges has contrasted with the poor knowledge of the enzymes responsible for their biosynthesis (‘branching enzymes’). In fact, it is only recently that some branching enzymes have been purified and characterized (Table 2). They can be divided into two groups according to the nature of the amino acid incorporated: (1) Glycine and l-amino acids are activated as aminoacyl-tRNAs and transferred to the precursors by a family of nonribosomal peptide bond-forming enzymes called Fem transferases (Mainardi et al., 2008). (2) d-Amino acids are activated as acyl phosphates by proteins belonging to the ATP-grasp family, which is composed of highly diverse enzymes that catalyze the ATP-dependent ligation of a carboxyl group to an amino or imino nitrogen, a hydroxyl oxygen or a thiol sulpfur (Galperin & Koonin, 1997). The precursor substrate of the branching enzymes varies among species: lipid II for S. aureus enzymes, UDP-MurNAc-pentapeptide for Weissella viridescens FemX or both for Enterococcus faecalis enzymes.

An interesting case deserves to be mentioned. In M. luteus (lysodeikticus), a large proportion of the peptide stems [l-Ala-γ-(α-d-Glu-Gly)-l-Lys-d-Ala] becomes detached from the MurNAc residues by an amidase reaction and forms peptide cross-bridges between positions 3 and 4 (Fig. 2). These bridges often contain several pentapeptides joined ‘head-to-tail’ (Ghuysen et al., 1968). The nature of the enzyme catalyzing the unusual transpeptidation reaction between d-Ala (acyl donor) and the N-terminal l-Ala (acyl acceptor) is unknown.

d-Ala at position 4 is not the only possible acyl donor: the carboxyl group of the amino acid at position 3 can also play this role. This gives rise to the appearance of 3–3 cross-links, which were originally discovered in Mycobacteria (Wietzerbin et al., 1974). In M. smegmatis, 3–3 cross-links represents one third of all cross-links. In other bacteria, their proportion is low but increases during the stationary phase or in β-lactam-resistant strains (Pisabarro et al., 1985; Mainardi et al., 2000). Their formation is catalyzed by penicillin-insensitive l,d-transpeptidases (Mainardi et al., 2005).

As for the main peptide chain, the interpeptide bridge can be further modified after its assembly. In Enterococcus faecium, where the Aslfm enzyme ligates the β-carboxyl group of d-Asp to the UDP-MurNAc-pentapeptide (Table 2), part of the α-carboxyl groups are subsequently amidated, thereby leading to a peptidoglycan composed of β-d-aspartyl and d-isoasparaginyl residues in similar amounts (Bellais et al., 2006). In Thermus thermophilus, where the amino acid at position 3 is l-Orn and the bridge consists of a diglycyl residue between positions 3 and 4, a significant proportion of glycyl residues not engaged in the bridge with the donor peptide stem are acylated with phenylacetic acid (Quintela et al., 1995b).

Besides the diversity in the nature of cross-linkage, there is a considerable variation in the degree of cross-linkage, which varies from ca. 20% in E. coli to over 93% in S. aureus (Rogers et al., 1980). Translated in terms of muropeptide content, these figures mean that in E. coli, most peptidoglycan units appear as monomers (ca. 50%) or dimers (ca. 40%), higher oligomers being a minority (Glauner et al., 1988); conversely, in S. aureus, the percentage of monomers is low (<10%), most peptidoglycan units being present as oligomers (lengths up to nonamers were detected by HPLC and up to eicosamers deduced from calculation) (Snowden & Perkins, 1990).

Classification of peptidoglycans according to their structure

Considering that the variations in peptidoglycan structure have taxonomic implications, Schleifer and Kandler established a tri-digital system of classification of peptidoglycans. The first digit, a Roman capital letter, represents the mode of cross-linkage (A and B for the 3–4 and 2–4 cross-linkages, respectively). The second digit, a number, refers to the type of interpeptide bridge, or lack of it, involved in the cross-linkage. The third digit, a Greek letter, indicates the amino acid found at position 3 of the peptide stem. As a consequence, the examples of peptidoglycan depicted in Fig. 2 have the following types: A1γ (a), A3α (b), B2β (c) and A2α (d). For further details about this classification, the reader is invited to refer to the review of Schleifer & Kandler (1972) or to the book of Rogers (1980).

Variations in peptidoglycan fine structure

The fine structure of the bacterial sacculi is reflected in the detailed muropeptide composition of peptidoglycan as determined by means of high-resolution techniques. Information on the abundance and peculiarities of families of muropeptides with specific structural functions is crucial to understand the architecture and physiology of the sacculus itself. In a first approximation, the most relevant muropeptide families are: cross-linked muropeptides, as their abundance is related to the mesh size and strength of the sacculus; 1,6-anhydroMurNAc muropeptides (if present), which relate to the average length of peptidoglycan strands, and muropeptides acting as links for other molecules such as Braun's lipoprotein in Gram-negative and wall teichoic acids or polysaccharides in Gram-positive species (Glauner & Höltje, 1990; Höltje & Glauner, 1990; Höltje, 1998; de Pedro, 2004).

Systematic studies in the model bacterium E. coli revealed that sacculi are endowed with a large metabolic plasticity, which is indicated by the variability of peptidoglycan muropeptide composition and surface density (amount of peptidoglycan per unit of cell surface area) in response to the growth and environmental conditions (Mengin-Lecreulx & van Heijenoort, 1985; Prats & de Pedro, 1989; Höltje, 1998).

Investigations of the mechanisms of precursor incorporation into the sacculus indicated that newly inserted and old peptidoglycans had different muropeptide compositions, and that the macromolecule suffers a maturation, or aging, before new and old materials become indistinguishable (de Pedro & Schwarz, 1981; Burman & Park, 1983; Pisabarro et al., 1985; Glauner & Höltje, 1990). In E. coli, new peptidoglycan is less cross-linked, and has little covalently bound lipoprotein, but is richer in muropeptides with a pentapeptide side chain (potential donors in dd-transpeptidation reactions) and is made up of longer glycan strands. Aging brings with it a progressive variation of the indicated parameters in a process that apparently requires about one mass doubling time to complete. Interestingly, most of the difference in cross-linkage between new and old peptidoglycans can be accounted for by de novo synthesis of ld-A2pm-A2pm (3–3) cross-linked muropeptides (de Pedro & Schwarz, 1981; Glauner & Höltje, 1990).

Peptidoglycan fine structure is also subjected to global variations when the state of growth changes (Pisabarro et al., 1985; Glauner et al., 1988). The transition of E. coli from exponential growth into a resting phase brings with it a drastic modification in the composition, and likely structure, of sacculi. Most remarkable alterations affect the degree of cross-linkage, which increases by 30–40% (from 27–30 up to 36–42% of cross-linked muropeptides), the mean glycan chain length, which declines to about one half (from 30–35 down to 15–18 disaccharides chain−1), and the lipoprotein content, which increases by about 50% (from 8–10 up to 13–15% of lipoprotein-bound muropeptides) (Pisabarro et al., 1985; Blasco et al., 1988; Glauner et al., 1988). As for peptidoglycan aging, most of the variation in cross-linkage could be accounted for by a dramatic increase in the relative abundance of ld-A2pm-A2pm cross-linked muropeptides, from 5–6% to 11–13% of the total muropeptides (Pisabarro et al., 1985). Of course, the inverse transition also takes place when cells resume active growth from a resting condition. Although data are quite limited, recovery of the muropeptide composition characteristic for actively growing cells might involve active modification of total peptidoglycan in addition to the expected variation due to mixing of old (resting phase) and new (growth phase) peptidoglycans (de la Rosa, 1985; Pisabarro et al., 1985).

A rather surprising ability of E. coli is its aptitude to modify peptidoglycan surface density. Studies conducted under conditions limiting supply of precursors showed that E. coli can reduce the amount of peptidoglycan per surface unit down to one half the normal value, maintaining a typical morphology and growth parameters (Prats & de Pedro, 1989; Caparros et al., 1992). Cells with reduced peptidoglycan content were nevertheless more sensitive to penicillin and other damaging agents. It has been proposed that the ability to reduce the content of peptidoglycan could help cells to deal with transient inhibitions of cell wall biosynthesis and therefore increase their chances of survival (Prats & de Pedro, 1989).

A comprehensive survey of peptidoglycan fine structure in different bacterial species is simply nonexistent at present. Only a few Gram-positive bacteria have lent themselves to analysis by HPLC and in most cases their composition could be only partially solved (Garcia-Bustos et al., 1987; Laitinen & Tomasz, 1990; de Jonge, 1992; Atrih et al., 1996, 1999a, b; Quintela et al., 1999a, b; Severin, 2004; Boneca et al., 2007). A large variability in fine structure is evident, as expected from their heterogeneity in chemical composition and cross-linking. Even among the more homogeneous Gram-negative organisms, large differences in fine structure have been clearly shown in spite of the limited number of well-studied organisms (Folkening et al., 1987; Höltje & Glauner, 1990; Quintela et al., 1995a, b, 1999b; Costa et al., 1999). Therefore, it seems that there is no optimal or standard value for parameters as cross-linkage or glycan strand length, but rather each species selects the values (or range of values) appropriate for its particular life conditions.

Variations in peptidoglycan fine structure have also been associated with bacterial pathogenesis in a number of cases. Well-documented examples are the changes exhibited by peptidoglycan in Neisseria gonorrhoeae peptidoglycan related to secretion of cytotoxin (Cloud & Dillard, 2002); in H. pylori associated with morphological transitions (Costa et al., 1999); and in Salmonella typhimurium associated with intracellular colonization of epithelial cells (Quintela et al., 1997).

To conclude, it seems that bacteria are remarkably able to ‘ fine tune’ the structure of the cell wall to best suit their individual needs, and to better adapt to changing, and often challenging, environmental conditions. The nature of the profits bacteria derived from these adaptations is still unknown, but is likely relevant for their survival.

Variation upon amino acid supplementation or in genetically engineered bacteria

When present at a high concentration in the growth medium, analogues of peptidoglycan amino acids can be incorporated into the macromolecule and modify its composition. The most-known example is that of glycine, which can replace alanine at position 1, 4 or 5 in several bacterial species (Hammes et al., 1973). Similarly, d-amino acids such as d-Met, d-Trp or d-Phe can replace d-Ala at position 4 in E. coli (Caparros et al., 1992). The fact that several A2pm analogues are able to complement A2pm auxotrophy in E. coli dap mutants suggests their incorporation at position 3 in peptidoglycan (see references in Mengin-Lecreulx 1994). Interestingly, the presence of hydroxylysine, which is often considered to be a natural constituent of the peptidoglycan of certain species, is in most cases the result of particular growth conditions, namely lysine deprivation and hydroxylysine supplementation (see e.g. Smith & Henderson, 1964; Shockman et al., 1965).

Peptidoglycan composition varies in mutants or genetically engineered cells with respect to wild type. This is well documented in E. coli for the amino acid at position 3 of the peptide stem. A dapF mutant lacking A2pm epimerase was shown to contain a huge pool of ll-A2pm that was incorporated into peptidoglycan (Mengin-Lecreulx et al., 1988). Similarly, in a dapA strain in which genes of the methionine pathway are either deleted or overexpressed, A2pm was totally replaced by meso-lanthionine and/or l-allo-cystathionine (Richaud et al., 1993; Mengin-Lecreulx et al., 1994). Recenty, the peptidoglycan of E. coli cells overexpressing the murE gene of S. aureus or T. maritima was shown to contain high amounts of lysine (Mengin-Lecreulx et al., 1999; Boniface et al., 2006). The replacement of meso-A2pm at position 3 by an analogue resulted in a decrease of the proportion of dimer. For cells overexpressing the S. aureus murE gene, this defect in the degree of cross-linking even led to a lytic phenotype.

There kinds of experiments were also applied to genes coding for Fem transferases. The heterospecific expression of the femhB, femA and femB genes of S. aureus in Enterococcus faecalis led to the production of peptide stems substituted by mosaic side chains that were effectively engaged in cross-bridges (Arbeloa et al., 2004). Similar results were obtained when the bppA1 gene of Enterococcus faecalis was expressed in Enterococcus faecium (Magnet et al., 2007a).

Peptidoglycan composition and structure can also evolve under the selective pressure of antibiotics. This aspect is developed in the present issue by Mainardi (2008).

Biophysical properties of peptidoglycan

Peptidoglycan sacculi are bag-shaped molecules with unique biophysical properties. On the one hand, sacculi have the strength to withstand the cell's turgor of up to 25 atmospheres present in Gram-positive bacteria. On the other hand, the sacculi are not rigid walls but are flexible, allowing reversible expansion under pressure, and they have relatively wide pores, enabling diffusion of large molecules such as proteins. Because the peptidoglycan completely surrounds the cytoplasmic membrane, the sacculus has a similar size and shape as the bacterial cells from which it was isolated.

Thickness of the cell wall peptidoglycan

Upon staining with a heavy metal, the thin sacculi from Gram-negative bacteria appear in electron microscopy (EM) pictures as flat, empty cell envelopes (Fig. 3). The thickness of peptidoglycan has been determined from electron micrographs of thin sections of E. coli cells. However, the results should be considered with caution when chemical fixation and dehydration are used in combination with certain staining techniques. Depending on the procedure applied, the peptidoglycan layer has a thickness between 1.5 and 15 nm (Murray et al., 1965; de Petris, 1967; Hobot et al., 1984; Leduc et al., 1985, 1989; Beveridge, 1999). The main concerns with these methods are (1) that the dehydration and fixation might change the thickness of the peptidoglycan and (2) that the measurements determine the thickness of the line formed by the contrasting metal, which is not necessarily identical to the thickness of the peptidoglycan layer (de Petris, 1967; Wientjes et al., 1991). The introduction of cryo-transmission electron microscopy (cryo-TEM) of frozen hydrated sections proved to be a major improvement because this technique omits the chemical fixation and staining procedures. With this method, the peptidoglycan can be seen as a thin line beneath the outer membrane in thin sections of Gram-negative bacteria. In E. coli and Pseudomonas aeruginosa, the peptidoglycan layer is 6.35±0.53 and 2.41±0.54 nm thick, respectively (Matias et al., 2003). These data are in agreement with measurements of the thickness of isolated sacculi from E. coli with two different methods. Small-angle neutron scattering of hydrated, isolated sacculi (without bound lipoprotein) yielded a thickness of 2.5 nm of 75–80% of the surface, corresponding to the thickness of a single layer (Labischinski et al., 1991). The remaining 20–25% of the surface had a maximum thickness of 7 nm, which would correspond to a triple layer. By atomic force microscopy (AFM), a thickness of 3.0±0.5 nm of air-dried sacculi from E. coli was determined, and this value increased to 6.0±0.5 nm upon rehydration of the sacculi (Yao et al., 1999). AFM showed, too, that sacculi from P. aeruginosa were thinner, 1.5±0.5 nm in the air-dried form and 3.0±0.5 nm when rehydrated, as compared with sacculi from E. coli.

Figure 3

EM of purified sacculi. Sacculi purified from Escherichia coli (a, b) and from the extreme thermophile Thermus thermophilus (c) by repeated incubations in boiling 4% sodium dodecyl sulfate were spread onto copper grids, washed, stained with 1% uranyl acetate and observed under a transmission electron microscope. Sample A was subjected to carbon platinum shadowing at an angle of 15° to emphasize the surface features and thickness. Scale bars represent 0.5 μm. Note the folds at the polar regions as the cylindrical structure collapses onto a flat surface.

Application of the cryo-TEM technique to Gram-positive bacteria revealed a bipartite organization of the cell wall (Matias & Beveridge, 2005, 2006, 2007; Zuber et al., 2006). In all species analyzed, there is a zone of low density presumably lacking polymeric wall structure next to the plasma membrane. This ‘inner wall zone (IWZ)’ or ‘periplasmic space’ has a thickness between 16 nm (in S. aureus) and 22 nm (in B. subtilis). The ‘outer wall zone’ (OWZ) of higher density is likely to represent the polymeric peptidoglycan–teichoic acid complex (with attached surface proteins). The thickness of the OWZ most likely varies with the species, growth phase of the cells and growth conditions, and was determined to be in the range of 15–30 nm (in S. aureus, B. subtilis, Streptococcus gordonii and Enterococcus gallinarum). Interestingly, the septal cell wall region in S. aureus has five layers, with two layers of high density (OWZs) sandwiched between three layers of lower density (IWZs) (Matias & Beveridge, 2007). AFM allowed visualization of concentric rings surrounding a central depression on a new hemisphere after daughter cell separation, as well as a network of fibers and large holes with diameters of 50–60 nm at the older regions of the cell surface (Touhami et al., 2004). Upon mechanical removal of patches of outer wall, AFM produced images of regularly arranged, about 26 nm thick strands running perpendicular to the long axis of rod-shaped cells of Lactobacillus helveticus. Although the nature of these structures is unknown, it has been proposed that they are made of bundles of twisted peptidoglycan strands (Firtel et al., 2004).

Elasticity of sacculi

Low-angle laser light scattering was used to determine the change of the mean surface of E. coli sacculi following an alteration of the net charge, either by changing to a low or a high pH, or by chemical modification. From these studies, it was concluded that the sacculi are elastic and can reversibly expand and shrink threefold without rupture (Koch & Woeste, 1992). Osmotic challenge of growing E. coli cells results in a maximum shrinkage in the surface area of 33%, or a maximum swelling of 23% (Koch, 1984; Baldwin et al., 1988). Similarly, living cells as well as isolated cell walls of Bacillus megaterium contract when they are transferred from water to salt solution (Marquis, 1968). When the cytoplasmic membrane of filamentous grown E. coli cells was destroyed with a detergent, there was a sudden decrease in the cell surface area of about 45% due to the relaxation of the peptidoglycan (Koch et al., 1987). Apparently, the peptidoglycan is under dynamic stress in the living cell due to the cell's turgor and there is a limit to which the peptidoglycan can be maximally stretched.

Interestingly, isolated E. coli sacculi are significantly more deformable in the direction of the long axis of the cell (elastic modulus, 1.5 × 107–3 × 107 N m−2; average, 2.5 × 107 N m−2) than in the direction perpendicular to the long axis (elastic modulus, 3.5 × 107–6 × 107 N m−2; average, 4.5 × 107 N m−2) (Yao et al., 1999). The elastic modulus is lower for material with greater elasticity. This is consistent with the observation that the changes in the volume of osmotically shocked E. coli cells are mainly due to changes in the cell length, whereas the cell diameter is virtually constant (van den Bogaart, 2007). It was suggested that the anisotropy in elasticity is the consequence of the predominant alignment of the (more flexible) peptides in the direction of the long axis of the cell and of the (more rigid) glycan strands perpendicular to the direction of the long axis. Such a network has been modelled, and the theoretical calculations of the elastic moduli are in good agreement with the measured values (Boulbitch et al., 2000).

Porosity of sacculi

Fluorescence-labelled dextrans of different sizes have been used to determine the diameters of holes in the peptidoglycan network in E. coli and B. subtilis (Demchick & Koch, 1996). Interestingly, the pores were of similar average sizes in peptidoglycans from Gram-negative and Gram-positive species and they were relatively homogenous in size: the mean radius of the pores was 2.06 nm for E. coli peptidoglycan and 2.12 nm for B. subtilis peptidoglycan, which is a similar value as the pore radius of at maximum 2.5 nm for B. subtilis and B. megaterium cell walls described in a previous study (Hughes et al., 1975). From these values it was calculated that globular, uncharged proteins with molecular weights of up 22–24 kDa should be able to penetrate the isolated, relaxed peptidoglycan. Globular proteins of up to 50 kDa or more should be able to diffuse through stretched peptidoglycan layer in the cell. Indeed, disruption of the outer membrane of E. coli in combination with a hyper-osmotic shock released a subset of proteins from the cell which were similar to those proteins which were able to pass through a 100-kDa cut-off filter. Perhaps this value is determined by the molecular sieving properties of the stretched peptidoglycan layer (Vazquez-Laslop et al., 2001).

Modelling the structure of the bacterial sacculus

To understand how a biological structure grows, a detailed knowledge of how the individual components are organized and arrayed with respect to each other is of prime relevance. Unravelling the molecular architecture of the bacterial sacculus has been a constant aspiration for many microbiologists, but it is proving to be a frustrating topic. In particular, the architecture of the cell wall of Gram-positive bacteria is far from being understood. Gram-positive species not only have a thick, multi-layered peptidoglycan but other major surface polymers linked to it (Vollmer, 2008). The anionic poly(ribitol-phosphate) or poly(glycerol-phosphate) teichoic acids may account for up to 50% of the mass of the cell wall. Many species also have capsular polysaccharides which are often covalently linked to the peptidoglycan. In addition, there are many surface proteins either linked covalently to peptidoglycan or bound noncovalently to cell wall polymers. To decipher the architecture of this three-dimensional assembly of different polymeric components and its enlargement during bacterial growth will be a major challenge for the future. With respect to the architecture of the sacculus of Gram-negative bacteria, E. coli has been the best-studied subject. Therefore, the structural aspects of sacculi from this organism will be concentrated on. The idea of the following comments is not so much to criticize existing models, but to point out aspects of cell wall biology and biochemistry which are overlooked, but must be accounted for, by present day models to help improve future developments.

As commented above the sacculus is a covalently closed structure built up from glycan strands that are cross-linked to each other through peptide bridges. In E. coli, and in general in Gram-negative bacteria, the sacculus is very thin, in particular compared with the thick, dense appearance of cell walls from Gram-positive species. These basic properties, defined long ago, naturally lead to the concept of the sacculus as static, regular and planar net-like polymeric macromolecule, a concept which can still be traced down to present-day textbooks. However, this somewhat simplistic vision seems to be far from reality and the bacterial sacculus is proving itself to be a particularly intractable subject for structural studies. Application of even the more powerful tools in structural analysis, as X-ray diffraction, EM, AFM, low angle neutron scattering, and others have provided only limited information (Formanek et al., 1976; Labischinski et al., 1979, 1985, 1991; Yao et al., 1999; Matias et al., 2003). Perhaps the most important concept to derive from these studies is that the distribution of subunits in the sacculus is far from the highly regular ‘quasi crystalline’ schemes proposed in the past and so often used to represent the cell wall.

From a structural point of view, the basic problem is to define how individual glycan strands are arranged relative to each other and to the cell axes. The interactions among neighbouring glycan strands are in turn conditioned by three parameters; thickness, cross-linkage and length distribution of the glycan strands. These three parameters determine the number of chemical bonds per unit of surface area opposing the turgor, and define the basic constrains in model making.

Are sacculi from E. coli mono- or multilayered?

Whereas the multilayered structure of sacculi from Gram-positive bacteria is generally accepted from EM and biochemical data (Beveridge & Matias, 2006; Matias & Beveridge, 2007), the nature of the Gram-negative sacculi is still uncertain. The extreme thinness (3–4 nm) assigned to the E. coli sacculus in early EM work (Murray et al., 1965; de Petris, 1967) was consistent with the sacculus being a planar monolayer of peptidoglycan. However, later evidence from different fields suggests a more complex situation. Neutron scattering studies on purified cell walls indicated that sacculi consist of a mixture of mono (75%) and trilayered (25%) regions (Labischinski et al., 1991), although no data is available about the distribution of the mono and trilayered areas. That the sacculus is ‘more than a monolayer’ finds additional support on the experimental determination of the amount of peptidoglycan per cell which was close to theoretical estimations for a single layered sacculus, but could still accommodate sizeable areas with a trilayered structure (Wientjes et al., 1991), on the ability of E. coli to keep proper shape and growth rate with a severely reduced amount of peptidoglycan per unit of cell surface area (Prats & de Pedro, 1989; Caparros et al., 1992), and on the identification of cross-linked trimers and tetramers in sacculi (Glauner et al., 1988) which also favours a (partially) multi-layered structure. Application of AFM (Yao et al., 1999) and advanced sample preparation techniques for cryo-TEM (Matias et al., 2003) indicated that E. coli sacculi are considerably thicker (6.0 nm by AFM, 6.3 nm by cryo-TEM) than first thought. Perhaps more convincing than absolute thickness measurements is the fact that sacculi from another typical Gram-negative organism, P. aeruginosa, had only one half the thickness of those from E. coli (3.0 and 2.5 nm for AFM and cryo-TEM, respectively). As sacculi from both species are made of identical subunits (Quintela et al., 1995a), the more likely explanation is that sacculi from E. coli, and probably other Gram-negative bacteria, indeed include sizable regions of multilayered peptidoglycan.

The length distribution of glycan strands

The sacculus is made up of glycan strands cross-linked to each other through peptide bridges. As the physico chemical properties of the glycan and peptide moieties are very different, in particular the ability of each to change conformation under stress (Barnickel et al., 1979; Labischinski et al., 1985; Labischinski & Maidhof, 1994; Koch et al., 2000b), an important parameter for the understanding of the macromolecular structure of the sacculus is the length distribution of the glycan strands. The peptide stems are of a fixed length, and in principle distributed regularly along the glycan backbone. Therefore the number of possible interstrand connections is also a direct function of glycan chain length (GCL). For a long time the only way to determine GCL was based on quantification of glycan strand terminal muropeptides (Schindler et al., 1976; Beachey et al., 1981; Glauner et al., 1988). This permits calculation of a mean value for GCL (mGCL). Reported results (Pisabarro et al., 1985, 1987; Glauner et al., 1988; Glauner & Höltje, 1990) show four structurally relevant features: (1) wild-type E. coli strains in exponential growth have mGCL from 25 to 35 disaccharide units; (2) the mGCL changes widely in response to the state of growth of the cells: in E. coli mGCL drops to ca. 15 disaccharides in resting cells; (3) more than 80% of 1,6-anhydroMurNAc terminal muropeptides are cross-linked and there are indications of a similar situation for the GlcNAc termini as deduced from peptidoglycan labelled with galactosyl transferase (Schindler et al., 1976; Beachey et al., 1981; C. Altmutter, unpublished data).

Application of a new method based on enzymatic clipping of peptide stems with human serum amidase followed by HPLC separation of the resulting linear polysaccharides permitted an accurate analysis of the size distribution of glycan strands (Harz et al., 1990). However the method still suffers from a key limitation in that only glycan strands between 1 and 30 disaccharide units can be individually resolved. Longer strands cannot be separated from each other but at least the proportion of muropeptides in strands longer than 30 disaccharides can be evaluated, and an average value can be calculated from the proportion of 1,6-anhydroMurNAc-containing muropeptides. Information gathered by this method (Harz et al., 1990; Obermann & Höltje, 1994; Höltje, 1998; Ishidate et al., 1998), showed some unexpected, but crucial features of sacculi, that can be summarized as: (1) about two thirds of total muropeptides are assembled in strands shorter than 30 disaccharide units; (2) the distribution is continuous with strands of every length between 1 and 30 subunits; (3) the mean and modal values for the distribution of short strands (<30 disaccharides chain−1) are about 7–8 and 9–10 disaccharide units per strand, respectively, with a large positive skewness; (4) the ca. 40% of total material in long strands (>30 disaccharides chain−1) is recovered as a pool with a mean GCL of 45 disaccharides chain−1.

Therefore, sacculi of E. coli contain relatively short glycan strands with a broad length distribution and with a strong tendency to cross-link to each other through the glycan strand terminating muropeptides.

Cross-linking peptidoglycan strands

In Gram-negative bacteria peptidoglycan strands are bound (cross-linked) to each other through the peptide side chains of the monomeric disaccharide penta- (tetra)- peptide subunits (Schleifer & Kandler, 1972). Cross-linking in Gram-negative species, in particular in E. coli, is relatively low and mediated by peptide bridges consisting of the heptapeptide l-Ala→d-Glu→meso-A2pm→d-Ala→meso-A2pm←d-Glu←l-Ala (dd-Ala-meso-A2pm, or 3–4, bridges), the result of dd-transpeptidation reactions responsible for incorporation of new precursors into the cell wall (Schleifer & Kandler, 1972; Quintela, 1995a; Höltje, 1998).

Application of HPLC methods demonstrated the presence of a second kind of cross-linking in E. coli (Glauner et al., 1988) and other Gram-negative bacteria (Quintela et al., 1995a), the so-called (ld)-A2pm-A2pm (or 3–3) bridges. These bridges are shorter by one amino acid (a hexapeptide instead of the normal heptapeptide) and are, apparently, the product of penicillin-insensitive ld-transpeptidase activity (Höltje, 1998). Sacculi from growing, wild type E. coli strains usually exhibit a degree of cross-linkage (molar proportion of cross-linked muropeptides relative to total muropeptides) of about 25–35%. That means that on average every third to second disaccharide in a peptidoglycan strand would be cross-linked to another adjacent strand. Similar values seem to be common among Gram-negative species, although data are still limited (Quintela, 1995a; Costa et al., 1999). ld-A2pm-A2pm bridges account for 7–16% of total cross-linked muropeptides (Pisabarro et al., 1985; Glauner, 1988; Glauner & Höltje, 1990), a minor but certainly significant fraction. However the function and distribution of A2pm-A2pm bridges in the sacculus remains unknown. The positive identification of cross-linked trimers and tetramers by means of HPLC techniques in E. coli sacculi demonstrated that a particular subunit in one glycan strand could be linked to two, or even three, other glycan strands (Glauner, 1988; Glauner & Höltje, 1990). Although trimers account for only about 3–4% of total muropeptides (dimers add up to 30–40%), the fact that they represent connection nodes for multiple glycan strands suggest a relevant structural role for this minor class of muropeptides (Glauner et al., 1988; Höltje, 1998). Tetramers are barely detectable with reported abundances about 0.2% of total muropeptides, a proportionally low value, which nevertheless corresponds to some thousands of muropeptides per sacculus. Following the same argument as above scarce tetramers could still be structurally significant (Glauner et al., 1988). As commented above the degree of cross-linkage is a variable parameter influenced by the state of growth of the cell, and aging of newly inserted muropeptides (de Pedro & Schwarz, 1981; Pisabarro et al., 1985; Glauner & Höltje, 1990).

Because disaccharide subunits in peptidoglycan strands are rotated with respect to each other due to the influence of the lactyl group in MurNAc, consecutive peptide stems point out in different directions (Labischinski et al., 1979, 1985). Early estimations indicated rotation angles close to 90° between consecutive disaccharides, and therefore a periodicity of four muropeptides, that is every fourth peptide stem would roughly point in the same direction (Labischinski et al., 1985). However, recent calculations with synthetic peptidoglycan fragments indicate a rotation angle close to 120° (Meroueh et al., 2006), consistent with a periodicity of three muropeptides. The periodicity of peptidoglycan conditions the cross-linking between adjacent strands, as only those peptide stems with the correct relative orientation can be proficient substrates for transpeptidation. An immediate consequence of these facts is that adjacent peptidoglycan strands are unlikely to be cross-linked to each other by consecutive muropeptides (Koch, 1998a, b).

The orientation of the glycan strands

The more controversial aspect related to the structure of the sacculus is the orientation and distribution of the glycan strands (Formanek et al., 1976; Verwer et al., 1978; Höltje, 1998; Koch, 1998a, b, 2000a, b; Dmitriev, 1999, 2003; Yao et al., 1999; Pink et al., 2000; Vollmer & Höltje, 2004; Meroueh et al., 2006). Most available evidence favoured models postulating glycan strands oriented parallel to the cell surface, and in most cases with the glycan backbones transversal to the cell longitudinal axis. More recently an alternative model based on glycan strands oriented perpendicular to the surface of the sacculus has been proposed (Dmitriev et al., 1999, 2003; Meroueh et al., 2006). As indicated above, it is not intended here to enter into a discussion of models so far proposed, but rather point out some aspects, in the authors' opinion, overlooked in those models.

A good starting point is to refer to a recent comment (Young, 2006) on the same topic which emphasizes what a weak point for most models is: the requirement for rather restrictive structural and morphological parameters. All the main models (Höltje, 1996, 1998; Koch, 2000b; Dmitriev et al., 2003) do require a high degree of order in the array of both glycan strands and cross-links, a requirement opposed by experimental evidence (Labischinski et al., 1979, 1985; Labischinski & Maidhof, 1994). As commented in the preceding sections the main structural parameters in the sacculus are subjected to drastic changes on the course of normal growth. This calls for dynamic rather than static models because the sacculus as such is in a continuous state of change. Furthermore, any credible model should have an intrinsic ability to accommodate the size and shape changes (in particular in cell diameter, but also more dramatic ones as round and branched shapes) that cells can exhibit under specific conditions.

Sacculi are made of glycan strands with a very wide distribution of sizes, with a substantial proportion of total peptidoglycan in strands too short to be connected to nearby strands by more than two cross-links. Such strands could structurally be assimilated to a long range cross-link, connecting longer and relatively distant strands (Costa et al., 1999; de Pedro, 2004). Even if it is generally assumed, there is no evidence at all that cross-linkage happens regularly alongside the glycan strands. Indeed, the tendency of glycan strand termini to be highly cross-linked means that cross-linkage in internal positions is lower than the mean value for total peptidoglycan. Therefore, it is likely that long glycan strands might have relatively long uncross-linked stretches. A final comment on the organization of glycan strands comes from the very existence of cross-linked trimers and tetramers. These groups of muropeptides might not be very abundant, but still enough to be significant elements of the sacculus with about (1–3) × 104 and 103 trimers and tetramers per sacculus, respectively. The interesting point about these two families of muropeptides is that they represent connecting hubs for several crossing glycan strands. An attractive idea is that trimers and tetramers represent linking points of short-to-long glycan strands. A normal cross-link bridging two nearby, long glycan strands could act as the attachment point for a short glycan strand acting as a long range connection to another relatively remote long glycan strand. Interestingly, about 30% of cross-linked 1,6-anhydroMurNac muropeptides are trimers, whilst only 10% of cross-linked nonterminal muropeptides are trimers as calculated from data in Glauner (1988). That is, glycan strand termini seem to have a high tendency to cross-link to a dimeric muropeptide to form a trimer. Whether this tendency is more marked in short or long glycan strands cannot be decided at present, and of course trimers could show no preference at all relative to glycan strand length.

As for the remaining, major fraction of cross-linked trimers and tetramers geometrical arguments require that three, or four, peptidoglycan strands cross over (Glauner et al., 1988). There are no data available on the angles between multiple glycan strands linked together at these positions. In the extreme cases, they could run parallel, antiparallel, or perpendicular to each other. Regions of multiple glycan strand crossings could be related with the postulated regions of multilayered peptidoglycan. Whether or not simple crossing of strands could be enough to explain the neutron diffraction results (Labischinski et al., 1991) has not been addressed yet.

Finally, another aspect models should be able to account for is the ability of E. coli cells, and likely others, to accommodate variations in the amount of peptidoglycan per unit of cell surface. Under normal growth conditions results suggest that E. coli has about two times as much peptidoglycan as would be strictly required to preserve physical integrity, normal shape and growth parameters (Prats & de Pedro, 1989; Caparros et al., 1992). It is interesting to note that recent measurements of cell wall thickness in E. coli and P. aeruginosa indicate that the former is twice as thick as the latter (Yao et al., 1999; Matias et al., 2003). If indeed E. coli does have an ‘excess’ of peptidoglycan relative to other species as P. aeruginosa, then Gram-negative bacteria should have to be considered a more structurally heterogeneous group than formerly thought.

Concluding remarks

Many data on the chemical structure and the biophysics of peptidoglycan have been gathered over the last decades. However, knowledge of this fascinating molecule is still very limited. For example, the analysis of peptidoglycan composition with high-resolution techniques has been performed so far only for a few bacterial species. A significantly enlarged data set on peptidoglycan fine structures will be of interest for different research fields including bacterial taxonomy, physiology, and pathogenesis. Moreover, such research can lead to the discovery of peptidoglycans whose structure diverge from the overall structure defined at the beginning of this review. A recent example is represented by T. maritima, in which a d-Lys residue with an ‘upside-down’ arrangement has been identified (Boniface et al., 2006). It is likely that such an unusual motif, which coexists with the conventional l-Lys-containing motif, is at the origin of a particular type of cross-link. Another example is that of Chlamydiae, for which no muramic acid-containing peptidoglycan was detected to date. However, the biosynthetic activities characterized so far (MurA, MurC, Ddl) are functional either in vitro or in a heterologous context (McCoy & Maurelli, 2006). It is difficult to imagine that these enzymes do not participate in the synthesis of a specific macromolecule, the structure of which presumably greatly differs from that of usual peptidoglycan.

In the last two decades, the improvement of analytical methods (HPLC, MS) has allowed to show that, within a particular species, variations in peptidoglycan fine structure occur as a function of aging, medium composition, pathogenesis, or presence of antibiotics. This type of research has implications not only in the field of bacterial physiology, but also in those of innate immunity, pathogenicity, and antibacterial therapy.

One major task for the future is to determine the molecular architecture of peptidoglycan in Gram-positive and Gram-negative species, which is not possible with today's techniques. This includes determination of the orientation of the glycan strands and peptides with respect to the cell's axes and the distribution pattern of particular structures (ld-cross-links, dimeric and trimeric peptides, glycan strand ends, attachment sites for other polymers) on the surface of the cell wall. Knowing the architecture of peptidoglycan is a prerequisite for solving the mechanism(s) of cell wall growth. As outlined in this review, models of peptidoglycan architecture should be based on the length distribution of the glycan strands, the degree of cross-linkage, the thickness of the macromolecule, the number of subunits per cell surface, as well as on biophysical data on the porosity and elasticity of the sacculus.


This work was supported by the European Commission through the EUR-INTAFAR project (LSHM-CT-2004-512138) (to W.V. and D.B.), the Deutsche Forschungsgemeinschaft (DFG) within the Forschergruppe Bakterielle Zellhülle (FOR 449) (to W.V.), the Centre National de la Recherche Scientifique (UMR 8619) (to D.B.), grant BFU2006-04574 from inisterio de Educación y Ciencia, Spain (to M.A.P) and an institutional grant from Fundación Ramón Areces (to M.A.P).


  • Editor: Arie van der Ende


View Abstract