Morphogenesis of rod-shaped sacculi


  • Editor: Arie van der Ende

Correspondence: Tanneke den Blaauwen, Molecular Cytology, Swammerdam Institute for Life Sciences, University of Amsterdam, Kruislaan 316, 1098 SM Amsterdam, The Netherlands. Tel.: +31 205255196; fax: +31 205257934; e-mail:


For growth and division of rod-shaped bacteria, the cylindrical part of the sacculus has to be elongated and two new cell poles have to be synthesized. The elongation is performed by a protein complex, the elongase that inserts disaccharidepentapeptide units at a limited number of discrete sites while using the cytoskeletal MreB helix as a tracking device. Upon initiation of cell division by positioning of the cytoskeletal Z-ring at mid cell, a switch from dispersed to concentrated local peptidoglycan-synthesis occurs. From this point on, peptidoglycan synthesis is for a large part redirected from elongating activity to synthesis of new cell poles by the divisome. The divisome might be envisioned as an extended elongase because apart from its basic peptidoglycan synthesizing activity, specific functions have to be added. These are conversion from a cylinder to a sphere, invagination of the outer membrane and addition of hydrolases that allow separation of the daughter cells. The elongase and the divisome are dynamic hyperstructures that probably share part of their proteins. Although this multifunctionality and flexibility form a barrier to the functional elucidation of its individual subunits, it helps the cells to survive a variety of emergency situations and to proliferate securely.


The shape of bacteria is maintained by the shape of their peptidoglycan layer, as the isolated peptidoglycan bag or sacculus retains the morphology of the cell in its isolated form (Fig. 1). An understanding of the mechanism that allows bacteria to differentiate their shape into cell poles and straight cylinders, banana-shaped cylinders, spirals and American footballs, to mention a few variations (Young, 2003), is therefore a prerequisite for the understanding of bacterial morphology. To discover these mechanisms, it is not sufficient to understand all the enzymatic reactions involved in the biosynthetic pathway of the sacculus (Barreteau et al., 2008; Bouhss et al., 2007).

Figure 1.

 The sacculus determines the shape of the cell. Electron microscopy image of an isolated sacculus of the stalk containing Caulobacter crescentus cell on the left and a sacculus of a dividing Escherichia coli on the right. The sacculi were isolated and processed for electron microscopy. They were labeled on a grid with antibodies against murein and secondary antibodies conjugated to 6-nm gold particles and subsequently stained by uranyl-acetate. The scalebar equals 200 nm. Courtesy of Waldemar Vollmer and Heinz Schwarz.

For the synthesis of a three-dimensional (3D) structure, the coordinates of synthesis and hydrolysis or its topography also have to be included to answer the question of where the peptidoglycan is synthesized. Topological information has to be added because the sacculus is embedded in the bacterial envelope that is separated from its metabolic compartment: the cytoplasm. Lastly, the dimension, time, has to be implicated because of the time-dependent nature of the bacterial cell cycle. This review will deal with the questions where, when and with the involvement of which proteins the peptidoglycan layer synthesis takes place in rod-shaped bacteria.

Owing to the differential affinity of a variety of β-lactams for penicillin-binding proteins (PBPs), one of the earliest observations has been that rod-shaped bacteria could grow temporarily either as spheres or as filaments. As a consequence, construction of the cylindrical part of the cell, which might be termed elongation, and of the cell poles, termed division, have become separate events in most studies. Therefore, the two processes will be discussed in this order, although this separation might not be completely justified as will follow from this review. By default, the Gram-negative Escherichia coli will be the model organism for rod-shaped bacteria and when necessary illustrative data from the Gram-positive Bacillus subtilis or the Gram-negative Caulobacter crescentus will be used.

Proteins involved in elongation and their interaction

The biosynthetic reaction pathway to peptidoglycan occurs in three stages. The cytoplasmic stage takes care of the synthesis of the activated nucleotide cytoplasmic precursors UDP-N-acetyl-glucosamine (UDP-GlcNAc) and UDP-N-acetylmuramyl-pentapeptide (UDP-MurNAc-pp) via an enzymatic reaction cascade described elsewhere (Barreteau et al., 2008). This is followed by the membrane-linked lipid cycle of reactions that translocate the lipid-linked β-(1,4) disaccharidepentapeptide building-unit known as lipid II across the cytoplasmic membrane (Bouhss et al., 2007). In the periplasmic stage, these precursors are inserted into the peptidoglycan layer by the PBPs (Sauvage et al., 2008). The cytoplasmic stage is thought to be identical for elongation and division of the bacterium. It is in the membrane that the first signs of discrimination of the two steps seem to occur. Proteins that cause the cell to grow as spheres when they are defective, such as MreB, MreC, MreD, RodA and PBP2, are either associated with or integrated in the membrane.

RodA is a 370 amino acid residue integral membrane protein of 40 kDa (Matsuzawa et al., 1989) that was discovered as a morphological mutant that caused the cells to grow as spheres (Matsuzawa et al., 1973). Conditional mutants or deletions of homologues of RodA in B. subtilis, Salmonella thyphimurium and even in the ovoid coccus Streptococcus thermophiles result in spherical growth (Henriques et al., 1998; Costa & Anton, 1999; Thibessard et al., 2002). Because RodA in association with PBP2 seems to be essential for peptidoglycan synthesis in PBP1B-free membrane preparations of E. coli (Ishino et al., 1986), it has been postulated that RodA could be involved in the translocation of the lipid II peptidoglycan precursors across the cytoplasmic membrane. Although it has been established that proteins must be involved in the lipid II translocation and that translocation seems to be coupled to the transglycosylation reaction of the PBPs or monofunctional transglycosylases (van Dam et al., 2007), the identity of the protein(s) responsible for the translocation is still a mystery.

Mecillinam or amdinocillin (FL1060; Lund & Tybring, 1972) specifically inhibits the peptidoglycan transpeptidase PBP2 (Spratt, 1975; Spratt & Pardee, 1975) causing spherical growth, a substantial increase in the diameter of the cells and subsequently cell death. The latter can be prevented by overexpression of the cell division proteins FtsQAZ (see ‘Proteins involved in cell division and their interactions’) or by increasing the stringent response transcription regulator ppGpp (Vinella et al., 1992, 1993; Joseleau-Petit et al., 1994; Costa & Anton, 1999). The interpretation is that PBP2 is essential for the maintenance of the diameter of the cells (see also ‘Is the cellular localization of PBPs related to their function?’) and in its absence, the diameter and therefore the cell volume increases uncontrolled. Consequently, the concentration of FtsZ becomes too low to initiate cell division and new polar peptidoglycan synthesis. The other rod shape-determining RodA and Mre proteins are also not essential, provided that the strains contain mutations that prevent uncontrolled increase of their diameter. Such strains may be mecillinam resistant (Costa & Anton, 1993), overproduce ppGpp (Costa & Anton, 1999), overproduce cell division proteins (Wachi & Matsuhashi, 1989; Kruse et al., 2005), or may have to be osmotically protected (Formstone & Errington, 2005). Because PBP2 has transpeptidase activity only, it will have to associate with at least one class A PBP with glycosyltransferase and transpeptidase activity and perhaps also with a monofunctional glycosyltransferase. Based on in vivo crosslinking experiments, some evidence exists that the majority of the PBPs are part of two different protein complexes (Alaedini & Day, 1999).

MreB (originally EnvB) mutants that grew as spheres were discovered in 1969 by Normark (Normark, 1969) and in combination with MreC and MreD by Wachi et al. (1987, 1989). The fusion of GFP to the B. subtilis MreB protein in 2001 [(Jones et al., 2001) and in other species (Shih et al., 2003; Figge et al., 2004)], which showed that MreB (and its paralogs in B. subtilis, Mbl and MreBH) localizes as a helix underneath the cytoplasmic membrane and the subsequent publication of its actin-like crystal structure (van den Ent et al., 2001), focused the attention of the scientific community on these proteins. MreB is a soluble cytoplasmic protein of 37 kDa that polymerizes in an ATP-dependent fashion (van den Ent et al., 2001; Esue et al., 2005). MreC is a 40-kDa bitopic membrane protein with a large periplasmic domain and MreD is an integral membrane protein of 19 kDa. Bacterial two-hybrid analysis (Karimova et al., 1998) showed that MreC interacts with itself, presumably forming a dimer (van den Ent et al., 2006), and it interacts with MreD and with MreB (Kruse et al., 2005). MreB interacts with MreC but not with MreD (Fig. 2). The association of the three proteins is essential for lateral growth as depletion of each of these proteins is sufficient for spherical growth (Kruse et al., 2005). Interestingly, MreB is not able to localize in its helical pattern in spherical cells that are depleted of RodA, MreC, or MreD (Kruse et al., 2005), but it localizes normally in cells that are spherical due to inhibition of the function of PBP2 by mecillinam (Karczmarek et al., 2007). Inhibition of MreB by its specific inhibitor A22 (Iwai et al., 2002, 2004) or inhibition of PBP2 by mecillinam results in a change in shape from a rod to a sphere in the same time window (Karczmarek et al., 2007) with identical changes in the muropeptide composition of the sacculus (Varma et al., 2007). These changes are a significant shortening of the glycan chain length, a 60% decrease in pentapeptides, a 20% increase in crosslinking, and an increase in the percentage of lipoprotein-containing muropeptides (Varma et al., 2007). This clearly points to a relationship between MreB and PBP2. PBP2 of C. crescentus localizes in a band-like pattern that is dependent on the presence of MreB and MreC (Campo et al., 2004; Figge et al., 2004; Divakaruni et al., 2005). MreC itself was also shown to localize in a helical pattern in the periplasm (Divakaruni et al., 2005; Dye et al., 2005; Leaver & Errington, 2005). Ramoplanin binds to the reducing ends of the nascent glycan chains (Walker et al., 2005; Fang et al., 2006). Application of a subinhibitory concentration of fluorescent ramoplanin or vancomycin to growing cells revealed a helical staining pattern along the side-walls of B. subtilis (Tiyanont et al., 2006) and of E. coli (Varma et al., 2007). Together, these data suggest the presence of a protein complex consisting of MreBCD, RodA, and PBP2 that directs lateral peptidoglycan synthesis (Fig. 2). The helical track provided by the MreBCD proteins most likely ensures the disperse insertion of new peptidoglycan precursors into the existing lateral cell wall. Like actin, MreB monomers treadmill through MreB filaments by preferential polymerization at one filament end and depolymerization at the other filament's end (Defeu Soufo & Graumann, 2006; Kim et al., 2006). It seems likely that this dynamic movement is passed on to the precursor-inserting protein complex (see also ‘Topology of insertion sites in the lateral wall sacculi’).

Figure 2.

 Schematic representation of the elongase. Proteins are shown that are supposed to interact with each other and form a complex responsible for the elongation of the cylindrical part of the cell (a). Because the majority of these interactions have been found by bacterial two-hybrid systems or by coimmunoprecipitation, the precise interactions are not established. Therefore, alternative interactions cannot be excluded. Instead of a mysterious protein or RodA, the translocase could also be formed by a combination of integral membrane proteins (e.g. MraY and RodA as shown in b or RodA and MreD, not shown).

The lipid II peptidoglycan precursor that is synthesized by the membrane-associated glycosyltransferase MurG (37.7 kDa) (Mengin-Lecreulx et al., 1991; Ha et al., 2000; Chen et al., 2002b) is supposed to be the substrate of this elongase. It is, therefore, not surprising that in E. coli MurG could be coimmunoprecipitated with MreB and that the lateral localization pattern of MurG was lost in an MreCD deletion background (Mohammadi et al., 2007). Similarly, the integral membrane protein MraY (40 kDa) (Ikeda et al., 1991; Boyle & Donachie, 1998; Bouhss et al., 1999) that catalyzes the transfer of the phospho-MurNAc-pp motif of UDP-MurNAc-pp to the lipid carrier undecaprenyl phosphate, to form the so-called lipid I peptidoglycan precursor to be used by MurG (Brandish et al., 1996; Stachyra et al., 2004), was shown to coimmunoprecipitate with MreB and MurG (Mohammadi et al., 2007). Most likely, MraY and MurG are part of the protein complex that inserts the peptidoglycan precursors into the peptidoglycan layer during elongation of the cells, and that therefore could be termed elongase (Fig. 2). In addition, the elongase should contain the lipid II translocase if the RodA protein does not provide this function. Alternatively, a combination of the integral membrane proteins might form the translocase (Fig. 2b).

Proteins involved in cell division and their interactions

MurG and MraY are both essential proteins with no alternatives available in E. coli. Therefore, both proteins are the most likely candidates to provide lipid I and II precursors for the synthesis of the new cell poles. In agreement with this notion, MurG was found to localize as well-separated foci in the cylindrical part of the cell and as concentrated foci at the site of division in E. coli (Mohammadi et al., 2007) and in C. crescentus (Aaron et al., 2007). The presence of the mid cell localization of MurG was only visible after assemblage of all cell division proteins at mid cell in E. coli (Mohammadi et al., 2007) but coincided with the assemblage of the Z-ring in C. cresentus (Aaron et al., 2007). The observation in E. coli does not exclude a somewhat less abundant activity of MurG at the initiation of cell division that coincides with the assembly of the Z-ring.

Assembly of the Z-ring

The tubulin homologue FtsZ (Oliva et al., 2004) is a GTPase that is conserved in most bacteria. This abundant protein [∼5000 copies per cell in E. coli (Pla et al., 1991; T. den Blaauwen, unpublished data)] polymerizes in a GTP-dependent fashion. It is a highly dynamic structure. Subunits continuously exchange with a half-life of 8–9 s (Anderson et al., 2004). The Z-ring is built from overlapping segments of protofilaments of on average about 30 subunits (∼120 nm in length) (Chen & Erickson, 2005; Li et al., 2007). FtsZ outside the ring localizes in a helix-like pattern and moves rapidly within this pattern independently from the MreB helix or MinD oscillation (see for a review on the min system, Harry et al., 2006). Thus, FtsZ not only forms the Z-ring but is also part of a highly dynamic, potentially helical cytoskeleton (Thanedar & Margolin, 2004; Peters et al., 2007).

The formation of the Z-ring, which is the earliest step in cell division, requires the presence of FtsA or ZipA that tethers FtsZ filaments to the membrane. They interact with a short conserved sequence of the C-terminal end of FtsZ (Addinall & Lutkenhaus, 1996; Hale & de Boer, 1999; Pichoff & Lutkenhaus, 2007). ZipA binds to the membrane through an essential N-terminal transmembrane segment (TMS) that is linked to the FtsZ-interacting C-terminal globular domain by an extended linker region (Hale & de Boer, 1997).

FtsA is an actin structural homolog (van den Ent & Lowe, 2000). It binds to the membrane through a conserved C-terminal amphipathic helix that does not appear to be specific as it can substitute for that of MinD. FtsA is proposed to bind to the membrane before interacting with FtsZ; this interaction plays an essential role in cell division (Pichoff & Lutkenhaus, 2005). Cardiolipin-rich domains are present in the septal and in the polar membrane regions of E. coli (Mileykovskaya & Dowhan, 2000; Koppelman et al., 2001). The amphipathic helix of FtsA has basic residues on its hydrophilic side and might interact with acidic cardiolipin in the septal region analogous to the polar assembly of MinD, which is promoted by cardiolipin (Mileykovskaya & Dowhan, 2005). Interestingly, overexpression of MurG increases the amount of cardiolipin per cell sevenfold (van den Brink-van der Laan et al., 2003). This could imply that the MurG attracts cardiolipin at mid cell.

A conserved region in subdomain 2B of FtsA responsible for interaction with FtsZ has been identified by isolation of FtsA mutants that are able to bind to the membrane but fail to interact with FtsZ (Pichoff & Lutkenhaus, 2007). FtsA* contains a modification of R286 into W in subdomain 2B and can bypass the requirement for ZipA in the assembly of the Z-ring and its functioning in cell division by stabilizing the FtsZ ring (Geissler & Margolin, 2005; Geissler et al., 2007). Subdomain 1C of FtsA is involved in the recruitment of the other components of the divisome (Rico et al., 2004).

Two nonrelated intracellular nonessential proteins [ZapA in E. coli (Low et al., 2004; Small et al., 2007) and EzrA in B. subtilis (Levin et al., 1999; Chung et al., 2007)] appear to affect the stability of the Z-ring. A proper ratio between FtsZ to FtsA or ZipA is required for assembly and maintenance of the Z-ring. FtsE/X is an ABC transporter combination of an integral membrane protein FtsX and a cytosolic membrane-associated protein FtsE (Schmidt et al., 2004). The complex interacts with FtsZ (Corbin et al., 2007) and allows the cell to assemble the Z-ring under conditions of high osmolarity (Reddy, 2007).

Maturation of the divisome

The Z-ring is required for the localization of other cell division proteins, FtsK, FtsQ, FtsL, FtsB, FtsW, PBP3 (also called FtsI) and FtsN to form the divisome (also called septosome) at mid cell. All these proteins localize in a sequential and interdependent manner at the division site (Figs 3 and 4). Proteins FtsK-FtsN are recruited almost simultaneously at the division site c. 17 min after the assembly of the Z-ring (Aarsman et al., 2005; Wang et al., 2005b). A mutant that underexpresses S-adenosylmethionine synthase forms filaments without visible septa in which FtsQ, FtsW, PBP3 or FtsN fail to assemble to the Z-ring. This result suggests that some methylation is required before the complete divisome can be assembled (Wang et al., 2005a). MraW, also called YabC, might be the methylase. Interestingly, the gene mraW with an unknown function is located in the dcw cell division gene cluster upstream from ftsL (Hara et al., 1997).

Figure 3.

 Schematic representation of the assembly of the components of the divisome. The boxed proteins represent subcomplexes: (a) the Z-ring with FtsZ-FtsA-ZipA-ZapA, which interacts with FtsE/X; (b) the subcomplex FtsQLB, which contains a heterodimer of FtsL and FtsB; (c) the subcomplex FtsW-PBP3; (d) PBP1B and FtsN interact with PBP3 and could be part of the subcomplex FtsW-PBP3. FtsK appears at the division site at the same time as FtsQLB and FtsW-PBP3, is back recruited by the complex QLB and interacts with most of these proteins. MurG, PBP2 and PBP5, AmiC and EnvC are located at the division and are part of the cell division machinery. Dashed lines: interaction detected using a two-hybrid system (Di Lallo et al., 2003; Karimova et al., 2005; D'Ulisse et al., 2007 and unpublished results). Solid lines: interactions detected using different techniques. Courtesy of Benoît Wolff.

Figure 4.

 Schematic representation of the divisome. The membrane topology and compartmentalization of cell divisome proteins that have been localized at mid cell either by GFP fusions or by immunofluorescence microscopy are shown. Transmembrane helices are represented by cylinders and the cytosolic or periplasmic (above the membrane) domains are sized approximately according to the molecular weight of the protein domains. MraY has been added in analogy to its presence in the elongase but has not yet been localized. EnvA, PBP2 and PBP5 are not shown because, presently, information is lacking on their possible interactions with the divisomal proteins. Courtesy of Benoît Wolff.

FtsK is a multifunctional and multidomain protein. The cytoplasmic C-terminal domain forms the translocation machine, which is involved in chromosome segregation (Bigot et al., 2007). The N-terminal domain is formed of four TMSs and is required for cell division (Dorazi & Dewar, 2000). Expression of ftsA(R286W) or overexpression of ftsQ allows the cell with a complete deletion of ftsK to survive and divide, although many of these ftsK null cells formed multiseptate chains (Geissler & Margolin, 2005). The cytoplasmic and transmembrane domains of FtsQ are sufficient to confer viability to ftsK null cells. These data suggest that FtsK is involved in stability of the cell division machine and has a role in the closure of the pole. Deletion of dacA encoding PBP5 dd-carboxypeptidase can suppress the ftsK44(ts) phenotype, which is due to a modification of G80 into A in a TMS. This modification might affect the conformation of a periplasmic loop of FtsK in contact with peptidoglycan (Begg et al., 1995).

FtsQ, FtsL and FtsB are bitopic membrane proteins with a small N-terminal intracellular region, a TMS, and a larger periplasmic domain. The crystal structure of the extracellular region of FtsQ (van den Ent et al., 2008) reveals two domains; the N-terminal α-domain has a striking similarity to POTRA domains (Sanchez-Pulido et al., 2003; Robson & King, 2006) and the C-terminal β-domian forms an extended β-sheet overlaid by two, slightly curved α-helices. Both domains seem to act co-operatively to bring FtsQ to mid cell (Scheffers et al., 2007). Mutagenesis experiments demonstrate that two functions of FtsQ, localization and recruitment, occur in two separate domains. Proteins that localise FtsQ need the second β-strand of the POTRA domain and those that are recruited by FtsQ, like FtsL/FtsB require the surface formed by the tip of the last α-helix and the two, C-terminal, β-strands (Chen et al., 2002a; D'Ulisse et al., 2007; van den Ent et al., 2008). FtsQ interacts with FtsL, FtsB, PBP3, FtsW, and FtsN as shown by a two-hybrid system and coimmunoprecipitations (Di Lallo et al., 2003; Karimova et al., 2005; D'Ulisse et al., 2007). Overproduction of FtsQ inhibits division of cells producing FtsZ, FtsA, or PBP3 mutants. These data suggest that FtsQ could play a role in regulating new pole synthesis (Dai & Lutkenhaus, 1992; Piette et al., 2004).

FtsL and FtsB seem to interact in a coiled-coil structure through their periplasmic domain (Buddelmeijer et al., 2002). They are dependent on each other for proper localization and depend on FtsQ for their localization at the division site. A fusion of FtsQ to ZapA is recruited to the Z-ring independently of FtsA and directs the recruitment of FtsL, FtsB, and PBP3. It can also back recruit FtsK, indicating that both proteins can interact directly (Goehring et al., 2005). A complex of FtsL, FtsB and FtsQ can be coimmunoprecipitated in the absence of FtsK, FtsW and PBP3, which indicates that these proteins probably form a complex before assembling into the divisome (Buddelmeijer & Beckwith, 2004).

Expression of ZapA-FtsL or ZapA-FtsB allows the recruitment of FtsW and PBP3 in the absence of FtsQ, showing that the complex L-B is sufficient to recruit FtsW and PBP3. Expression of ZapA-FtsW results in efficient localization of PBP3 in cells depleted of FtsA or FtsQ. Targeting of the complex FtsW-PBP3 restores localization of the FtsQ-L-B complex in cells depleted of FtsA. Thus, the late proteins appear capable of associating into preassembled complexes (FtsQ-FtsB-FtsL and FtsW-PBP3) within the cell (Goehring et al., 2006). As the average number of FtsQ proteins in the cell is c. 20–40 and that of PBP3 c. 100, a limited number of protein subcomplexes can be expected along the Z-ring.

FtsW has 10 TMSs and shares 30% identity with RodA and SpoVE, which are involved in elongation and in Bacillus sporulation, respectively. These proteins appear to work in coordination with one class B PBP to catalyze peptidoglycan polymerization during the cell cycle (see also ‘Proteins involved in elongation and their interaction’). The periplasmic loop of FtsW from residue P368 to P375 plays an important role in the septal recruitment of PBP3, whereas the E240-A249 periplasmic amphiphilic sequence appears to be a key element of the functioning of FtsW in the septal peptidoglycan assembly machineries (Peters et al., 2007). The first 75 amino acid residues of FtsW are sufficient to interact with FtsQ (D'Ulisse et al., 2007).

The multimodular class B PBP3 that is specifically involved in septal peptidoglycan synthesis consists of a short intracellular M1-R23 peptide fused to an F24-L39 membrane anchor that is linked via a G40-S70 peptide to the noncatalytic module itself linked to the penicillin-binding module with transpeptidase activity (Begg et al., 1990; Adam et al., 1997). The first 56 amino acid residues of PBP3 possess the structural determinants required to target the protein to the cell division site, and none of the putative protein-interacting peptides present in the N-terminal noncatalytic module are essential for the positioning of the protein at the division site (Piette et al., 2004). Using an E. coli two-hybrid system, the first 56 amino acid residues of PBP3 were shown to interact with FtsW (M. Nguyen-Distèche, unpublished data). Residues 1–70 of PBP3 are sufficient to interact with FtsQ, whereas the membrane-bound noncatalytic module appears to interact with FtsL (Karimova et al., 2005).

PBP1B localizes at the division site and in the cylindrical part of the cell and interacts directly with PBP3 as shown by affinity chromatography, surface plasma resonance, two-hybrid system and immunoprecipitation (Bertsche et al., 2006). Its localization depends on the physical presence of PBP3 but not on its activity. A monofunctional glycosyl transferase (Mgt) is able to interact with three constituents of the divisome, PBP3, FtsW and FtsN, in a bacterial two-hybrid assay (Derouaux et al., 2007). Finally, PBP1C interacts with PBP3 and PBP1B (Schiffer & Holtje, 1999) and the structural protein MipA interacts with PBP1B as well as with the hydrolase MltA (Vollmer et al., 1999, 2008a). The localization of these proteins is not yet known.

FtsN is a bitopic protein with an N-terminal intracellular domain, a TMS and a periplasmic domain. This region is composed of three short helices (residues 62–123), a long glutamine-rich, unstructured region (residues 124–242) and a C-terminal domain that binds peptidoglycan, but that is not essential for cell division (Yang et al., 2004). FtsN interacts with PBP3 (Karimova et al., 2005) and with PBP1B (Müller et al., 2007). Despite not being essential, the N-terminal region (residues 1–62) of FtsN can compensate for a complete lack of FtsK (Goehring et al., 2007).

A fusion of the subdomain 1C of FtsA to the B. subtilis DivIVA protein is targeted to the cell pole and is able to recruit FtsN and PBP3 to the poles independently from FtsZ (Corbin et al., 2004). FtsN is not recruited by prematurely targeted protein complexes FtsQLB or FtsW-PBP3 and required the presence of FtsA at the division site (Goehring et al., 2006). FtsA mutants that increase the integrity of the Z-ring can compensate for the loss of FtsN and can partially compensate for a deletion of FtsK or ZipA (Bernard et al., 2007). FtsN could play a role in regulating conformational changes within the divisome subassemblies, which in turn regulate the activity of the Z-ring. FtsA appears to be conformationally flexible and could be a key modulator of the divisome function at all stages (Bernard et al., 2007).

AmiC is a periplasmic amidase that cleaves the septal peptidoglycan and promotes cell separation. It depends on the presence of FtsN for its localization and it contains an N-terminal domain that is necessary and sufficient to target the protein at the division site (Heidrich et al., 2001; Bernhardt & de Boer, 2003) (see ‘Topology of insertion sites in the lateral wall sacculi’). EnvC, a periplasmic metallo-endopeptidase, also participates in the splitting of the peptidoglycan septum and is located at the division site (Bernhardt & de Boer, 2004).

Rate and topology of peptidoglycan synthesis during the cell cycle

The generation and maintenance of proper morphology in rod-shaped bacteria requires a mechanism able to produce a regular cylindrical surface, the lateral peptidoglycan, that grows in length (elongation) while avoiding generation of bending and torsional forces. Otherwise, growth would result in cells of irregular diameter, with curved or twisting morphologies, shapes often associated with mutational impairment of some morphogenes (Young, 2003; Varma & Young, 2004).

At periodical intervals, concomitantly with the onset of cell division, the mode of peptidoglycan synthesis changes from a constant diameter ‘elongative’ mode to a decreasing diameter ‘constrictive’ mode to form the transversal wall at the cell center, which will in turn generate the new cell poles as physical separation of the daughter cells proceeds (Spratt, 1975; Satta et al., 1985; Holtje, 1998; Nanninga, 1998; Errington et al., 2003). Both elongation and constriction ultimately rely on the local activity of a set of complexes where precursors, originating in the cytoplasm, become polymerized and covalently incorporated into the meshwork structure of the sacculus, causing its expansion. However, the mechanisms and proteins involved are different enough as to treat both independently in this discussion.

Topology of insertion sites in the lateral wall of sacculi

Whether incorporation of new precursors into the sacculus occurred at a defined location or in a diffuse way was soon decided in favor of the later alternative. Studies performed in the 1960s already were supportive of diffuse incorporation in a relatively large number of sites (Van Tubergen & Setlow, 1961). Application of high-resolution autoradiography led to some initial confusion, as some early results were interpreted in favor of zonal growth (Ryter et al., 1973). However, further technical and experimental improvements confirmed the disperse nature of precursor insertion into the lateral wall by a series of complementary approaches such as: (1) high-resolution autoradiography following incorporation of radioactive precursors (de Chastellier et al., 1975; Burman et al., 1983a; Woldringh et al., 1987; Wientjes & Nanninga, 1989; Mulder & Woldringh, 1991); (2) evolution of the acceptor–donor radioactivity ratio (a parameter that indicates whether radioactively labeled incoming precursors cross-link preferentially with themselves or with pre-existing material) (Burman et al., 1983b; Cooper et al., 1988); (3) distribution of peptidoglycan hydrolytic (Burman et al., 1983b; de Pedro et al., 2002), biosynthetic (Den Blaauwen et al., 2003; Scheffers et al., 2004) and morphogenetic (Daniel & Errington, 2003; Kruse et al., 2005) activities; (4) analysis of the peptidoglycan segregation pattern by d-amino acid labeling and immunomicroscopy (de Pedro et al., 1997; Varma et al., 2007); and (5) Ramoplanin-affinity labeling of precursors (Tiyanont et al., 2006) and newly incorporated peptidoglycan (Varma et al., 2007). Therefore, that elongation of sacculi occurs by incorporation of precursors in a number of homogeneously distributed sites covering the cylindrical surface of the sacculus seems to be a well-established fact, at least for the model rod-like bacteria E. coli, B. subtilis and related species. The number of insertion sites active at any particular time point is likely to be rather small. Although the earlier estimations suggested ‘200 or more’ (Van Tubergen & Setlow, 1961), later estimates reduced figures to 80–100 sites per cell based on the kinetics of insertion (Burman & Park, 1984). Nonetheless, this figure is likely a considerable overestimation, in view of the scarcity of some critical components of the elongation machinery as PBP2 of E. coli (20–40 molecules per cell) (Spratt, 1975, 1977; Den Blaauwen et al., 2003). The real nature of the material during the course of insertion is still under debate, in particular whether new glycan chains are inserted individually (Cooper & Hsieh, 1988), as pairs (Burman & Park, 1984) or as triplets (Holtje, 1998). Recent evidence supports an intrinsic ability of peptidoglycan synthases to produce pairs of cross-linked glycan strands (Bertsche et al., 2005; Born et al., 2006), therefore supporting the idea that glycan strands are likely to be inserted into the sacculus as cross-linked pairs, or higher order polymers.

The recent demonstration that proper bacterial morphogenesis depends on the dynamics of cytoskeletal proteins as MreB spirals and FtsZ rings (Shih & Rothfield, 2006) supports the concept of insertion sites as wandering structures that travel around the sacculus while promoting the synthesis and insertion of peptidogycan strands (Den Blaauwen et al., 2003; Daniel & Errington, 2003; Varma & Young, 2004; Kruse et al., 2005; Varma et al., 2007). Actually, the proposal of insertion sites as dynamic entities is anything but new. It was first proposed more than 20 years ago (Burman & Park, 1984), and was also an intrinsic assumption in the ‘three for one’ model for growth of the murein sacculus (Holtje, 1998).

That insertion of new material may occur at any place of the lateral wall with equal probability does not imply that new and old material is fully intermixed. As a matter of fact, the (sparse) evidence available rather supports generation of mosaic structures made up of interconnected microdomains of all-new and all-old peptidoglycan (De Pedro et al., 2003a; Koch & De Pedro, 2006; Varma et al., 2007).

In conclusion, the present view is that insertion of precursors into the lateral wall of the sacculus occurs at a limited number of discrete sites, that are highly mobile and distributed over the surface of the cell in an MreB-associated pattern.

Topology of precursor insertion sites in septal-polar peptidoglycan

Cell poles are the product of cell division, and by extension polar peptidoglycan is the product of septation. Division of the sacculus requires a change in the mode of synthesis of macromolecular peptidoglycan. At periodic intervals, insertion of precursors instead of promoting elongation starts to generate either an invagination of progressively smaller diameters, or an inwards-growing thick transversal septum in Gram-negative and Gram-positive bacilli, respectively (Spratt, 1975; Holtje, 1998; Nanninga, 1998; Vollmer & Holtje, 2001; Errington et al., 2003). This change has been associated with activation of zonal, that is highly localized, insertion since the earliest studies on cell wall growth, first as a rationalization of the lytic phenotypes observed in response to antibiotics as penicillin (Schwarz et al., 1969; de Pedro et al., 2002), and second by high-resolution autoradiography (Ryter et al., 1973; Burman et al., 1983a; Wientjes & Nanninga, 1989; Mulder & Woldringh, 1991; Nanninga, 1991) and peptidoglycan segregation studies (de Pedro et al., 1997, 2001). The latter study exploited the ability of a, still uncharacterized, periplasmic ld-transpeptidase activity (Caparros et al., 1992) to mediate the exchange of an externally added d-amino acid (d-cysteine) for the d-Ala normally present at the C-terminal position of muropeptides. After isolation of the sacculi, the d-Cys are biotinylated and labeled with biotin-recognizing gold-labeled secondary antibodies. When the growth of the cells in the presence of d-Cys was followed by a chase for one or two mass doublings in the absence of d-Cys, the number of gold particles on the cell poles was still the same compared with unchased samples, whereas the number of particles in the cylindrical part of the cells was diluted by growth (Fig. 5). From these results, it was confirmed that the cell poles are constructed from new peptidoglycan and that they become inert after their synthesis (Burman et al., 1983a; de Pedro et al., 1997).

Figure 5.

 Schematic interpretation of murein synthesis in Escherichia coli cells subjected to a d-Cys label-and-chase experiment. Cells incubated in the presence of d-Cys incorporate the amino acid homogeneously over the entire surface of the sacculus. When cells are transferred to d-Cys-free medium, newly inserted precursors are devoid of d-Cys, and can be differentiated from pre-existing material by immuno-microscopy techniques. The figure depicts the evolution of a newly born cell through two successive generations. At the beginning of the chase period (0 Mass Doubling, MD), the cell wall is heavily and uniformly labeled with d-Cys (dark gray). As the cell elongates, insertion of new material causes dilution of d-Cys-labeled peptidoglycan in the lateral wall (medium gray) but not into the polar regions. Shortly before the cell starts to constrict, incorporation of precursors in the central region is exacerbated, which results in the creation of a band-like region of all-new d-Cys-free murein (white) at the potential division site. As division progresses two new cells are born with a heavily labeled old pole and a new pole devoid of d-Cys (1 MD). The next cell cycle results in a gradual reduction of label in the lateral wall (light gray) and the generation of cells with two unlabeled poles and with one labeled and one unlabeled pole (2 MD). The labeled pole, derived from the original cell, retains a dense labelling and can be traced for at least five generations. Inhibition of cell division proteins results in filaments characterized by the presence of two densely labeled poles (dark gray) and a homogeneously diluted cylindrical wall that presents a series of regularly spaced bands of unlabeled peptidoglycan at the potential division sites when FtsQ is impaired [FtsQ(Ts)].

Polar peptidoglycan has a very long life, and seems to remain unchanged for several generations. Neither synthesis nor degradation of peptidoglycan takes place in the poles, which can therefore be considered as stable structures (de Pedro et al., 1997). No chemical or structural features clearly accountable for the high stability of polar peptidoglycan have been detected in sacculi. The relatively minor modifications detected are in general smaller than variations due to growth rate or state of growth (Pisabarro et al., 1985; Glauner & Holtje, 1990; Romeis et al., 1991; Ishidate et al., 1998; Holtje & Heidrich, 2001). When cells are grown in the presence of a fluorescently labeled tripeptide, this material is specifically inserted via the peptidoglycan-recycling pathway (Barreteau et al., 2008; Vollmer et al., 2008a) in to the cylindrical part of the peptidoglycan layer but not in the new cell poles (N. Oldrich, T. den Blaauwen et al., unpublished data). An indication that some modification occurs after synthesis of the new cell poles is that the old cell poles do contain these fluorescent-recycled peptidoglycan muropeptides, whereas the newly synthesized poles are devoid of these peptides. Deficiencies in PBP5, the main dd-carboxypeptidase activity of E. coli, have a dramatic influence on the topology of inert peptidoglycan, in particular when combined with mutations on other low-molecular-weight PBPs such as PBP7 (Nelson & Young, 2000, 2001). These strains often show morphological alterations as branches, buds and kinks, which are always and every time associated with regions of inert peptidoglycan (de Pedro et al., 2003b). How these ectopic regions are generated is under study, but evidence indicates interference with normal septation as aberrant poles are a common feature of such strains (de Pedro et al., 2003b) (M.A. de Pedro, unpublished data). An alternative possibility is that the dynamics of cytoskeletal proteins preclude biosynthetic complexes from polar regions, as suggested by the influence of FtsZ on peptidoglycan synthesis (Varma & Young, 2004; Aaron et al., 2007), and in particular on the extension of polar inert peptidoglycan regions (Varma et al., 2007). Nevertheless, experimental evidence is insufficient to actually settle the issue.

The detailed architecture and distribution of insertion sites involved in septal peptidoglycan synthesis are still poorly known. Activation of septal peptidoglycan synthesis seems to be a multistep process both regarding assembly of division complexes (Aarsman et al., 2005) and insertion of precursors into the central region of the sacculus (Wientjes & Nanninga, 1989). Septal synthesis can be divided into at least two stages related to the involvement of cell division proteins: an early stage, which requires FtsZ as the only division-specific protein, and a later stage, which requires the full complement of division proteins (de Pedro et al., 1997). Activation of zonal insertion at the potential division site requires assembly of the FtsZ ring at the cell center, but can proceed for a significant period of time in the absence of later division proteins as FtsQ and FtsI, leading to the synthesis of an annulus of all-new murein at the potential division site (Fig. 5).

Once a particular, still unknown, check-point is reached in FtsQ and FtsI filaments, septal synthesis apparently stops and the peptidoglycan biosynthetic potential becomes redirected towards lateral wall growth, leaving behind an annulus of inert peptidoglycan (de Pedro et al., 1997). When division is allowed to proceed, septal synthesis goes on promoting the progressive constriction of the sacculus. Incorporation of precursors in this phase seems to take place exclusively at the leading edge of the inwards-growing constriction, in a PBP3-dependent manner (Wientjes & Nanninga, 1989). The nature of the leading edge is still unknown; however, evidence from E. coli mutants defective in N-acetyl-muramyl-l-alanine amidases supports the idea of the leading edge as a differentiated domain in septal peptidoglycan. N-acetyl-muramyl-l-alanine amidases are required at the terminal stages of septation (Holtje & Heidrich, 2001; Heidrich et al., 2002). Defective amiABC mutants form chains of cells separated by incomplete septa (Heidrich et al., 2001, 2002; Priyadarshini et al., 2007). Electron microscopy analysis of sacculi from such strains revealed the existence of an electron-dense ring structure (septal peptidoglycan or SP-ring) tagging the inwards edge of partially constricted division sites (Heidrich et al., 2001; Priyadarshini et al., 2006, 2007) (Fig. 6). Furthermore, SP rings are associated with areas of inert peptidoglycan, suggesting that polar peptidoglycan becomes metabolically inert since the very moment of insertion (Priyadarshini et al., 2006). Apparently, SP rings are able to progress down to a minimum diameter in amiABC mutants, and then stall. Cytokinesis, however, proceeds till the end and cells become independent entities separated by the respective cytoplasmic membranes (Heidrich et al., 2002; Priyadarshini et al., 2007). Therefore, a potential third (late) stage in septation could be marked by the specific requirement for N-acetyl-muramyl-l-alanine amidases activities to allow the closure of the inwards-growing septum and terminate the synthesis of polar peptidoglycan.

Figure 6.

 SP-rings in purified sacculi from a Salmonella thyphimurium SL1344 amiABC triple-deletion mutant. Sacculi purified from exponentially growing cultures were processed for electron microscopy. Sacculi spread on carbon-pioloform-coated copper grids were stained with 1% (w/v) uranyl acetate and observed under the EM either after carbon-platinum shadowing (a and b) to enhance surface detail or directly after staining (c).

Rate of peptidoglycan synthesis

The rate of peptidoglycan synthesis has been investigated by pulsed incorporation of radioactive peptidoglycan precursors into synchronous growing cells and asynchronous growing cells (Olijhoek et al., 1982; Wientjes & Nanninga, 1989) (see also ‘Topology of insertion sites in the lateral wall sacculi’). In both cultures, the rate of peptidoglycan synthesis was more or less exponential. However, when the incorporation of radioactive meso-diaminopimelic acid (A2pm, see Vollmer, 2008) was visualized by autoradiography and electron microscopy, an increase in the rate of peptidoglycan synthesis at the constriction site at the expense of a reduction of 40% in the rate of peptidoglycan synthesis in the cylindrical part of the cell was observed (Woldringh et al., 1987; Wientjes & Nanninga, 1989). The incorporated radioactive A2pm was visible after autoradiography as black grains on the cell. In the cylindrical part of the cells, the grains were dispersedly distributed whereas they were clustered at the site of division. Even in cells that showed no visible constriction, but that were about to initiate division, an increase in the number of grains at mid cell was observed. This increase in mid cell peptidoglycan synthesis is dependent on the presence of FtsZ (Woldringh et al., 1987) and coincides with the moment at which the Z-ring is assembled (Den Blaauwen et al., 1999). Remarkably, the increase in mid cell peptidoglycan synthesis is not dependent on the presence or the function of PBP3 (Wientjes & Nanninga, 1989) or probably any of the cell division proteins that are recruited to the Z-ring in the second step of divisome maturation (Aarsman et al., 2005; Varma et al., 2007). As a consequence, these proteins cannot cause the increased rate of peptidoglycan synthesis at mid cell and the accompanying reduction in lateral peptidoglycan synthesis rate. Similar observations have been made with respect to the two modes of peptidoglycan synthesis (Fig. 5). The polar mode of peptidoglycan synthesis is dependent on FtsZ, but not on PBP3 or FtsQ (de Pedro et al., 1997). Therefore, it seems acceptable to assume that the polar mode of peptidoglycan synthesis and the increase in the rate of peptidoglycan synthesis coincide in time and space.

The rate of A2pm incorporation can almost be completely inhibited either by blocking PBP1A/B by their specific inhibitor cefsulodin or by blocking PBP2 and PBP3 simultaneously by mecillinam and cephalexin (Wientjes & Nanninga, 1991). This suggests that the PBP1s and the transpeptidases act together in the synthesis of the cell envelope. Inhibition of PBP2 in synchronized cells reduced the rate of A2pm incorporation by 60% during the entire cell cycle, whereas inhibition of PBP3 by cephalexin reduced the rate of incorporation especially during the constriction period by 35% (Fig. 7) (Wientjes & Nanninga, 1991). This shows that lateral peptidoglycan synthesis continues during the constriction period, albeit at a lower rate as the number of A2pm grains in the lateral wall was reduced by 40%. Because the PBP2 activity was more or less equally inhibited in nonconstricting cells as in constricting cells (Fig. 7), it can be concluded that PBP2 assist in the synthesis of the new cell poles (see also ‘Are specific proteins responsible for the local insertion and increase in peptidoglycan synthesis at the initiation of cell division?’ and ‘Is the cellular localization of PBPs related to their function?’) whereas PBP3 seems to be restricted to new cell pole synthesis.

Figure 7.

 Contribution of PBP2 and PBP3 to peptidoglycan synthesis in relation to the cell division cycle. In the upper panel, the cell number (N), the A450 nm and the percentage of constricting cells in the synhronously growing cell culture of a wild-type Escherichia coli strain are shown. In the lower panel, the percentage of radioactive Dap incorporation into the sacculi is shown in the presence of either the PBP2 inhibitor mecillinam or the PBP3 inhibitor cephalexin (reproduced from Wientjes & Nanninga, 1991 with permission from Elsevier Limited).

The average number of autoradiography A2pm grains at the division site did not diminish in cells that had progressed in polar cap synthesis compared with division-initiating cells (Wientjes & Nanninga, 1989). These observations suggest that peptidoglycan synthesis at the leading edge of the constriction site occurs by more or less the same set of peptidoglycan-synthesizing protein complexes as those that were recruited during the initiation of cell division. These protein complexes have to become more clustered during the constriction process when the FtsZ-ring diminishes in diameter. As a consequence of the constant rate of polar peptidoglycan synthesis, the Z-ring diameter decreases with a rate to the power of two. This is in agreement with the relatively small fraction of deeply constricting cells in a population.

Are specific proteins responsible for the local insertion and increase in peptidoglycan synthesis at the initiation of cell division?

The local increase in peptidoglycan synthesis is dependent on the presence of FtsZ as was discussed above, but not on the proteins that assemble onto the ring at a later stage (Aarsman et al., 2005). It is accompanied by a lengthening of the glycan stands (Ishidate et al., 1998) and the synthesis of a band of inert peptidoglycan at the future site of division (de Pedro et al., 1997). Curiously, this initiation of constriction, which coincided with Z-ring assembly, could not be inhibited by penicillins such as cefsoludin, mecillinam, furazlocilin, cephalexin and moenomycin (Wientjes & Nanninga, 1991). The presence of a penicillin-insensitive peptidoglycan synthetase (PIPS) that interacts with FtsZ was postulated (Nanninga, 1991) to be responsible for the initiation of polar peptidoglycan synthesis. Could any of the other known peptidoglycan-synthesizing proteins be part of PIPS? The third class A PBP of E. coli, PBP1C, is insensitive to the majority of the penicillins and after immobilization on sepharose it was able to retain PBP1B, PBP3 and MltA (Schiffer & Holtje, 1999). It could also retain PBP2, provided that the periplasmic fraction was added to the solubilized membranes (von Rechenberg et al., 1996; Schiffer & Holtje, 1999; Vollmer et al., 1999). This suggests that PBP1C is at least part of the complex responsible for polar peptidoglycan synthesis and that it could represent PIPS. Because PBP1C is reported to bind to moenomycin-coupled agarose, it cannot completely represent PIPS. The Mgt of E. coli is insensitive to moenomycin (Di Berardino et al., 1996) and interacts in a bacterial two-hybrid system with PBP3, FtsW and FtsN (Derouaux et al., 2007). The combination of PBP1C and Mgt could provide all the enzymatic activities for PIPS. However, deletion of either or both proteins simultaneously has no effect on the morphology of the cells (Schiffer & Holtje, 1999). An alternative candidate for the penicillin-insensitive transpeptidase reaction could be provided by the ld-transpeptidase of E. coli (Caparros et al., 1992) involved in incorporation of d-Cys or the one involved in A2pm–A2pm crosslinking (Obermann & Holtje, 1994). These two proteins might be identical, but are most likely not the same as the ld-transpeptidase involved in the attachment of the Braun Lipoprotein (Magnet et al., 2007).

Because PBP2 is involved in cell pole synthesis (Den Blaauwen et al., 2003), it is possible that PBP2 is able to initiate cell division when either PBP3 or PBP1A/B is inhibited. If this would be the case, the peptidoglycan synthetases do not provide any specificity to explain the switch from lateral to polar peptidoglycan synthesis. Specificity would have to be found in, for instance, the activity of the hydrolases such as the carboxypeptidases or in the compartmentalization that the cytoskeleton of the FtsZ-ring and the MreB helix somehow seems to cause. The occurrence of a switch in the mode of peptidoglycan synthesis that leads to new cell pole synthesis is firmly established, but its nature, as is outlined above, is very poorly understood. Future studies might need to include not only the proteins involved in the synthesis of the peptidoglycan layer but also the physical arrangement of the biosynthetic complexes as a function of the cell cycle and the metabolic and signaling state of the cells.

Is the cellular localization of PBPs related to their function?

The PBPs (Sauvage et al., 2008) are responsible for the periplasmic stage of peptidoglycan synthesis by inserting the lipid II peptidoglycan precursors into the existing peptidoglycan layer during length growth or by synthesizing new peptidoglycan during cell division (See ‘Rate and topology of peptidoglycan synthesis during the cell cycle’). Although some of the PBPs have a unique function, many are redundantly present in most bacterial species investigated thus far. Their specific function, apart from their enzymatic activity, is not very well known. The exceptions are the class B high–molecular-weight PBPs that have transpeptidase activity. In E. coli, these are PBP2 and PBP3, which could easily be characterized as essential for length growth and division (Spratt, 1975; Botta & Park, 1981; Popham & Young, 2003) because their inhibition has such a dramatic effect on cell morphology (i.e. conversion to spheres or filaments, respectively). Deletion of the genes encoding for PBPs has not provided clear-cut information on their function, as PBPs of the same class are often capable of replacing the deleted one. The class A PBP1A and PBP1B in E. coli that have transglycosylase as well as transpeptidase activity can each, but not simultaneously, be deleted without any morphological consequences and are apparently not that specific (Yousif et al., 1985). Surprisingly, all of the four class A PBPs (PBP1, PBP2C, PBP2D and PBP4) of B. subtilis can be simultaneously deleted, resulting only in a reduction of growth rate and minor morphological effects (McPherson & Popham, 2003). Analysis of the muropeptide composition of the quadruple-class A mutant strain indicates that a novel unidentified enzyme must perform the glycosyltransferase activity required for peptidoglycan synthesis (McPherson & Popham, 2003).

An alternative approach to elucidate the function of the various PBPs is to determine where they localize in the bacterial cell using either fluorescent protein fusions or antibodies, and fluorescence microscopy. Promising results have been obtained over the last few years [see for a comprehensive list of localized PBPs; (Scheffers & Pinho, 2005)]. Localization of PBPs has been studied most extensively in B. subtilis where 13 out of the 16 PBPs were localized using strains in which chromosomally encoded PBPs were replaced by GFP–PBP fusions (Scheffers & Errington, 2004; Scheffers et al., 2004). Three localization patterns could be observed: (1) localization in the cylindrical envelope and at the septum, (2) septal localization only and (3) localization in the form of a collection of foci that might be organized in arc-like or helical structures (Scheffers et al., 2004). The last pattern belonged to PBPs that were either expressed less or were presumably needed for specific growth or sporulation (PBP3, PBP4A, and PBP4*). The class A PBP1B and the class B PBP2B localized at the septum. The class A PBP4, the class B PBP2A and the major carboxypeptidase PBP5 (Lawrence & Strominger, 1970) localized as well in the cylindrical wall as at the septum.

The lateral growth responsible class B PBP2 of E. coli was shown to be present in the cylindrical part of the envelope as well as at the constriction site using a functional GFP-PBP2 (Den Blaauwen et al., 2003). As expected, the class B PBP3 responsible for septal peptidoglycan synthesis localizes specifically at the constriction (Weiss et al., 1999). In the case of PBP2, it was shown that its localization at the constriction was required to maintain the diameter of the new cell poles (Den Blaauwen et al., 2003). Therefore, it seems likely that the new cell poles of rod-shaped bacteria are synthesized with the lateral mode of peptidoglycan synthesis as well as the septal mode of synthesis, whereas the cylindrical part is only synthesized in the lateral mode. This hypothesis is supported by the observation that some cocci change their shape or diameter after addition of the, for PBP2-specific, antibiotic mecillinam or after deletion of genes encoding for proteins that are part of their putative elongation machinery (Pinho et al., 2000; Thibessard et al., 2002; Zapun et al., 2008).

The combination of a class A and a class B PBP is in principle sufficient for bulk cell wall synthesis. Based on the localization studies in B. subtillis, one could envision two separate peptidoglycan-synthesizing machineries as already proposed in the 1980s by Higgins & Shockman (1971): one for lateral peptidoglycan synthesis and one for septal peptidoglycan synthesis, where the latter is assisted by the lateral synthesizing machinery during synthesis of the new cell poles. However, the observation that the Class A PBP1B localizes in the lateral wall as well as at the septum, where it directly associates with the class B PBP3 (Denome et al., 1999; Bertsche et al., 2006), which localizes exclusively at the division site, suggests that a rearrangement of the lateral peptidoglycan machinery occurs to included PBPs such as PBP3 that are specialized in septal peptidoglycan synthesis. The functionally for PBP1B inter exchangeable PBP1A is expected to localize as well in the cylindrical part of the cell as at the constriction site.

The low-molecular-weight PBPs modify the peptidoglycan by making the peptide side chains inaccessible for cross-linking or by cleaving crosslinks (see Mainardi et al., 2008; Vollmer et al., 2008b). The morphological defects of deletion of different combinations of the LMW PBPs of E. coli produce cells with random shapes, length and diameters (Denome et al., 1999; Vega & Ayala, 2006). Therefore, the majority of these PBPs are presumably important for fine-tuning of the peptidoglycan synthesis (Nilsen et al., 2004). Especially, the dd-carboxypeptidase PBP5 of E. coli seems to be important for the maintenance of cell diameter, surface uniformity and overall topology of the peptidoglycan layer (Nelson & Young, 2000). Therefore, it is not surprising that E. coli PBP5 localizes like B. subtilis PBP5 in the lateral wall and at the septum (A. Karczmarek, T. den Blaauwen et al., unpublished data). When cell division is inhibited by the PBP3 inhibitor aztreonam, PBP5 shows a random and disperse membrane localization, whereas it localizes almost exclusively at the septum in spherical cells obtained by growth in the presence of the PBP2-specific inhibitor mecillinam (A. Karczmarek, T. den Blaauwen et al., unpublished data). This indicates that PBP5 appears to be where the highest peptidoglycan-synthesizing activity occurs. This would agree with the high carboxypeptidase activity reported during division in synchronized cell cultures (Mirelman et al., 1977, 1978). In contrast, the carboxypeptidase PBP3 of Streptococcus pneumoniae localizes everywhere but at the site of septal peptidoglycan synthesis from which it was concluded that by making the peptide side chains of the glycan strands unavailable for crosslinking, the synthesizing PBPs would not be able to localize and insert new cell wall material (Morlot et al. 2004; Zapun et al., 2008). Interestingly, overproduction of PBP5 in E. coli results in spherical growth (Markiewicz et al., 1982). It is not clear whether the spherical shape is the result of specific inhibition of the elongation or whether it is caused by general inhibition of the transpeptidation in these cells. In the latter case, the spherical morphology could be due to insufficient peptide crosslinks to support the rod structure.

Elevated levels of PBP5 or six can restore cell division in ftsI23(Ts) mutants at the restrictive temperature. Larger amounts of dd-carboxypeptidase increase the amount of tetrapeptide and tripeptide acceptors preferentially used by PBP3. A similar correction of the mutation can be performed by the addition at a low concentration of d-cycloserine, which inhibits the d-Ala–d-Ala ligase and thus increases the amount of tripeptide. The level of ld-carboxypeptidase that hydrolyzes the tetrapeptide into tripeptide fluctuates during the cell cycle with a sharp increase at the time of division (Beck & Park, 1976). The proportion of penta-, tetra- and tripeptide might thus be an important factor in determining the cell shape by controlling the relative rate of division and elongation (Begg et al., 1990).

Attachment of the outer membrane (OM) to peptidoglycan and consequences for elongation and division

Linkage between the OM and the petidoglycan

The peptidoglycan layer is connected to the OM via a number of proteins. Of these, the Braun's lipoprotein is the only known protein that is covalently bound to the peptidoglycan while inserted into the OM by its acyl chain (Braun & Rehn, 1969). Lipoprotein is the most abundant OM protein and exists in a free (60–70%) and a peptidoglycan-bound (30–40%) form (Inouye et al., 1972; Braun, 1975). The l,d-transpeptidase YbiS, and to a lesser extent the proteins ErfK and YcfS, are responsible for the linkage of the α-carbonyl of A2pm to the side-chain amide of the l-Lys residue located at the C-terminus of lipoprotein (Magnet et al., 2007). The free form might stabilize the peptidoglycan-bound form by forming trimers (Choi et al., 1986). Lipoprotein is homogenously distributed along the cell envelope (Hiemstra et al., 1986, 1987) and by its sheer numbers, 100 000 copies μm−2 (Nikaido, 1996), it is in general essential for the maintenance of contacts between the OM and the peptidoglycan layer.

Other proteins like the peptidoglycan-associated lipoprotein (PAL) and OM proteins A, C and F (OmpA, OmpC and OmpF) interact noncovalently with the sacculus (Leduc et al., 1992; Koebnik, 1995). Deletion of either lipoprotein or Pal causes loss in the integrity of the OM, release of periplasmic proteins and OM vesicle formation or blebbing (Suzuki et al., 1978; Yem & Wu, 1978). Overexpression of Pal can compensate for the loss of lipoprotein in an lpp deletion strain, but overexpression of lipoprotein cannot rescue a pal deletion strain (Cascales et al., 2002), suggesting that Pal has a more specific function than lipoprotein.

Pal is anchored to the OM by an N-diacyl glyceride moiety at its second serine residue (Mizuno, 1979; Gennity & Inouye, 1991) and interacts strongly with the peptide moiety of the peptidoglycan layer by its carboxy-terminal region (Lazzaroni & Portalier, 1992; Parsons et al., 2006). TolB competes with the peptidoglycan peptide side chain for the same region on Pal (Ray et al., 2000). TolB is part of the Tol–Pal system that connects the inner membrane with the OM (Fig. 8). It consists of at least the cytoplasmic membrane proteins TolA, TolQ and TolR, which interact by their transmembrane helices (Journet et al., 1999), and the periplasmic TolB and Pal protein anchored to the OM. TolQ is an integral membrane protein with three membrane helices whereas TolR and TolA are bitopic membrane proteins with two domains located in the periplasm (see for a review Lazzaroni et al., 2002). The C-terminal domain of TolA interacts with the N terminal domain of TonB (Dubuisson et al., 2002; Walburger et al., 2002). The crystal structure of TolB reveals a two-domain structure (Abergel et al., 1999; Carr et al., 2000) with a mixed α helix plus sheet of five β strands N-terminal domain connected to a six-bladed β-propeller C-terminal domain. This propeller domain interacts with Pal. In addition to the interaction of Pal with TolB and peptidoglycan, it interacts independently with OmpA and with TolA (Cascales & Lloubes, 2004). The interaction of TolA with Pal is lost after dissipation of the electrochemical gradient of ions or protons (Cascales et al., 2000; Cascales & Lloubes, 2004). Based on sequence comparisons and mutagenesis studies, TolQR have been proposed to form a proton or an ion channel, which converts the proton motive force (pmf) to mechanical energy to drive a conformational change in the TolA protein (Goemaere et al., 2007a). The channel is probably formed by 14 transmembrane helices derived from a 4 : 2 stoichiometry of TolQ to TolR (Cascales et al., 2001). The C-terminal periplasmic domain of TolR seems to have a role in the regulation of the opening and closing of the channel entrance (Goemaere et al., 2007b). Whether it regulates the ion flow by forming a penetrating loop, a plug or a surface structure at the periplasmic side of the channel is not yet clear.

Figure 8.

 Schematic representation of the Tol–Pal system. The cytoplasmic membrane proteins TolQ (purple with channel-closing domains) and TolR (dark blue) probably form a channel and interact with TolA (overflowing from purple to green) in the periplasm. Pal (light green) is anchored to the OM and interacts strongly with the peptide moiety of the peptidoglycan layer by its carboxy-terminal region. TolB (two domains represented by a circle and a square in sea green) competes with the peptidoglycan peptide side chain for the same region on Pal. In addition to the interaction of Pal with TolB and peptidoglycan (purple railway), it interacts independently with OmpA (pink) and with TolA.

All five proteins of the Tol–Pal system accumulate at the division site during the constriction process and mutants lacking this system exhibit a delay in OM invagination and contain large membrane blebs at the constriction site and the cell poles (Gerding et al., 2007). The localization of the Tol–Pal at mid cell is dependent on the presence of FtsN and therefore, probably on a mature and functioning divisome. TolA, which normally requires FtsN for its recruitment to the divisome, localizes in an ftsN null strain, suggesting that FtsN does not recruit the Tol–Pal complex directly (Bernard et al., 2007). Gerding et al. (2007) propose that the trans-envelope Tol–Pal system constitutes a subcomplex of the division machinery in Gram-negative bacteria that is specifically used to ensure proper invagination of the OM at the constriction site. The Tol–Pal proteins appear to be uniformly distributed in the bacterial envelope of young cells, to accumulate at mid cell and follow the leading edge of constriction in dividing cells. This dynamic behavior is only possible if the transperiplasmic connection is of a transient nature. The competition of TonB and peptidoglycan for the same binding site on Pal, and perhaps of TolB and TolA for Pal, might allow a discontinuous connection that enables the Pal and Ton proteins to migrate to mid cell. As the same time, the connection is sufficiently continuous to force the OM to follow the invaginating cytoplasmic membrane and peptidoglycan layer and even to be resistant to plasmolysis (MacAllister et al., 1987).

The pmf dependence of the Tol–Pal interaction might also be related to the frequent occurrence of cell division protein mutants such as FtsZ84(TS) that are sensitive to low or high osmotic strength. Tol–Pal mutants themselves form chains of cells under osmotic stress conditions (Dubuisson et al., 2005; Gerding et al., 2007), indicating that they are deficient in cell division. The cell division proteins FtsE/X have been characterized as conditional osmoremedial essential proteins, and deletion of their genes could be compensated for by overproduction of FtsQAZ or FtsN (Reddy, 2007). Reddy (2007) suggested that the assembly of the divisome is possibly intrinsically osmosensitive. This osmosensitivity might be caused by the pmf dependence of the Tol–Pal system.

Mobility of OM proteins is constrained at cell poles

The strong and complex interaction between OM and murein sacculus has a particularly dramatic manifestation in the behavior of OM proteins at the cell poles. Label and chase experiments designed to follow the segregation and mobility of OM proteins showed that proteins at the poles were essentially immobile, while in other locations they became progressively intermixed with newly synthesized proteins during the course of cell elongation (de Pedro et al., 2004). In vivo studies on the dynamics of the OM protein LamB also demonstrated coexistence of mobile and static populations of the protein (Gibbs et al., 2004). A likely reason for such unexpected behavior lies in the inert nature of the peptidoglycan (IPG) making up the polar caps of the sacculus (de Pedro et al., 1997). Because peptidoglycan subunits in the polar regions are locked at their positions, the mobility of any other (macro)molecules interacting with these affixed subunits would be constrained. In the case of covalently bound proteins such as Braun's lipoprotein (Braun, 1975; Magnet et al., 2007), or strongly interacting with the sacculus as OmpA (Wang, 2002) or the Tol–Pal complex (Gerding et al., 2007), the proteins become effectively anchored to the subjacent IPG. Because these proteins do interact with the OM, and are present in large numbers, they can in turn affect the mobility of other OM proteins and even the physico chemical properties of the polar area as such. Indeed, differentiation of ectopic or abnormal IPG regions in sacculi of certain mutants is systematically associated with a parallel development of ‘low protein mobility’ areas in the OM (de Pedro et al., 2001, 2003b), thus reinforcing a causal relationship. Whether every single protein species in the OM is similarly affected in the polar area is unknown. Applied methods preferentially labeled a set of proteins accessible to reagents in the extracellular space (de Pedro et al., 2004; Gibbs et al., 2004). Therefore, free movement of nonreacting protein species in and out of the polar regions cannot be excluded. Polar IPG regions might not only influence OM protein mobility but also secretion. Fluorescently labeled proteins in the poles show a very slow decline in absolute signal intensity with time, indicating a similarly slow rate of dilution with newly made, unlabeled proteins (de Pedro et al., 2004). Although still a controversial matter (Foley et al., 1989; Mullineaux et al., 2006), diffusion of periplasmic materials could also be influenced by the temporal stability of the OM-sacculus complex in the polar regions.

The constraint of OM protein mobility at the cell poles is likely to play important roles in a number of cellular functions, such as the generation of asymmetries in the distribution of cell envelope proteins (Lybarger & Maddock, 2001), polar targeting of specific proteins as IcsA (Charles et al., 2001; Janakiraman & Goldberg, 2004), secretion of autotransported proteins (Jain et al., 2006), twitching mobility (Skerker & Berg, 2001) and possibly others.


The definition of what should be called an L-form has been discussed since 1939, when it was shown that many bacterial species gave rise to forms similar to the L1 culture (L from Lister Institute) that Klieneberger had isolated from Streptobacillus moniliformis. Since 1942, the methodology for establishing L-forms routinely involves numerous passages on complex hypertonic penicillin plates over an extended period of time. The first growing cells obtained, the unstable L-forms, are spherical and osmosensitive, and they revert to normal morphology in the absence of penicillin. After further passages, often extending over several years, stable (nonreverting) derivatives are obtained, and some authors have suggested that only these should be called L-forms (Klieneberger-Nobel, 1960). Others, however, have presented convincing evidence that stabilization is a secondary event, which simply prevents the reconstitution of a normal cell wall in the absence of penicillin but does not affect L-form growth (Lederberg & St Clair, 1958). What then is an L-form? In the absence of a recognized authority empowered to establish such definitions, the wisest course is to describe clearly the origin and cultivation of the organisms used, whatever name they go by. This, unfortunately, is not always the case in the L-form literature. To avoid confusion, we will call the cefsulodine or penicillin-induced spherical and osmosensitive forms ‘L-form-like’ cells.

Because of the prolonged growth of the L-forms in the presence of penicillin and their mycoplasma-like morphology, they are thought to be devoid of a cell wall. Electron microscopy showed in some cases that there is no visible cell wall in L-forms of various bacterial species, including E. coli (Gumpert & Taubeneck, 1983). Biochemical analyses of cell wall constituents in L-forms have yielded variable results, with numerous reports in which muramic acid, A2pm, d-glutamate or glucosamine was or was not detected in extracts of L-forms of various bacteria, usually with little quantification (Gumpert & Taubeneck, 1975, 1976, 1983). Hence, the authors are unaware of any published data that eliminate the possibility of a residual peptidoglycan in an established L-form.

D'Ari and coworkers (Joseleau-Petit et al., 2007) have found that the growth of L-form-like E. coli cells induced by cefsulodin or penicillin requires residual peptidoglycan synthesis amounting, in the former case, to 7% of that of wild-type cells. This is in some ways reminiscent of the paradox of chlamydial species, which are reputed to have no cell wall and yet seem to require peptidoglycan synthesis, probably for cell division (McCoy & Maurelli, 2006). The amount of peptidoglycan in L-form-like cells is far too little to form a sacculus covering the entire cell (Wientjes et al., 1991). Although techniques are not presently available for locating this peptidoglycan within the cell, the following arguments suggest that it may be at the division site of the spherical L-form-like cells. When, the for the septation process specific, PBP3 is inhibited by piperacillin in cefsulodin-induced L-form-like cells, a rapid block of the viable cell count is observed. This suggests that PBP3 is required for the propagation of the L-form-like cells. The transglycosylase activity of PBP1B is required for rapid growth of the L-form-like cells, and this protein has been shown to interact directly with PBP3 (Bertsche et al., 2006). In addition, PBP1B has been implicated in cell division under certain conditions (Garcia del Portillo et al., 1989). The division protein FtsZ is also required for the propagation of L-form-like cells (Joseleau-Petit et al., 2007). The simplest hypothesis to explain these observations – requirements for peptidoglycan synthesis, FtsZ, PBP3 and PBP1B – is that cell division in cefsulodin-induced L-form-like cells, as in rods, is by means of peptidoglycan synthesized in the division plane and indispensable for cytokinesis (Joseleau-Petit et al., 2007).

It was recently reported that an established E. coli L-form isolated nearly 40 years ago has acquired mutations in several genes required for peptidoglycan synthesis and cell division (Siddiqui et al., 2006). From the sequence of 36 kb of L-form DNA, the authors deduced that the FtsA, FtsW and MurG proteins have one or two amino acid changes each, the ftsQ gene has an amber triplet at codon 132 (of 276 codons) and the mraY gene has a frameshift in codon 294 that should produce a protein of 298 amino acids (instead of 360). The functional consequences, in rod-shaped cells, of the missense mutations and of the truncation of MraY are unknown. The truncated FtsQ protein would almost certainly be nonfunctional for cell division in rod-shaped cells, although a low level of amber suppression could provide the 22 molecules of intact FtsQ estimated to be needed for division (Carson et al., 1991). Further characterization of this classical L-form should establish clearly whether or not the cells carry out residual peptidoglycan synthesis and, if they do, whether it is essential for their propagation.

Cell division in normal bacteria is inseparable from cell wall synthesis at the septum, and now (Joseleau-Petit et al., 2007) it has been shown that this requirement is also present for L-form-like cells. The question is why cell division would require peptidoglycan synthesis, and the simplest answer is that macromolecular peptidoglycan is both a substrate and a product of the reaction, and the residual 7% of macromolecular peptidoglycan would provide the required scaffold for assembly of the division leading edge.

Concluding remarks

The elongation of the sacculus is performed by a dynamic protein complex ‘the elongase’, which inserts peptidoglycan precursors at a limited number of discrete sites while using the cytoskeletal MreB helix as a tracking device. The elongase consists of at least MraY, MurG, MreBCD, RodA, PBP1 and PBP2. Based on the average number of MurG foci in the cylindrical part of the cell (Mohammadi et al., 2007) and the number of PBP2 molecules per average cell (Dougherty et al., 1996), it seems unrealistic to expect more than 50 elongases per average cell depending on, the growth rate (Fig. 9). Upon initiation of cell division by positioning of the cytoskeletal Z-ring at mid cell, a switch from the dispersed to a concentrated local peptidoglycan synthesis occurs. From this point on peptidoglycan synthesis is for a large part redirected from the elongating activity to the synthesis of new cell poles that consist of new peptidoglycan. This synthesis is performed by the divisome or septosome. Initially, the synthesis seems to be performed by a premature divisome that contains PIPS (see ‘Are specific proteins responsible for the local insertion and increase in peptidoglycan synthesis at the initiation of cell division?’) to synthesize new peptidoglycan. After about one fifth of a generation time, the divisome matures and might be envisioned as an extended elongase because apart from its basic peptidoglycan synthesis activity specific functions have to be added. These are a change in morphology from a cylinder to a sphere (e.g. FtsK, FtsQLB, FtsW, PBP3), the invagination of the OM (e.g. Tol-Pal-FtsN) and the hydrolases that allow the extension and invagination of the peptidoglycan layer (e.g. MipA, MltA, AmiC, EnvC). Again, based on the number of tightly membrane-bound FtsZ molecules (T. den Blaauwen, unpublished data) and the number of FtsQ and PBP3 molecules present per average cell (Dougherty et al., 1996; D'Ulisse et al., 2007), not more than 50 divisome subassemblies will be present at the division site (Fig. 9). As the rate of peptidoglycan synthesis at the leading edge is constant during cell division, these subassemblies will be active during the progression of the leading edge to the closure of the constriction into a separated daughter cell pole (Fig. 9). The elongase and the divisome are dynamic hyperstructures (Norris et al., 2007) that probably at least share part of their proteins (e.g. PBP1, PBP2, MurG, MraY?, carboxypeptidases). A picture is emerging in which the high-molecular-weight PBPs, on which the focus has been for the major part of the last century, are the proteins that are the least involved in the regulation of the creation of the shape-maintaining sacculus.

Figure 9.

 Schematic representation of the elongation and septal mode of peptidoglycan synthesis in a slice perpendicular to the length axes of the bacterium. Insertion of peptidoglycan precursors in to the existing peptidoglycan layer is accomplished by the elongases that use the MreB cytoskeletal helix as a tracking device to ensure dispersed lateral extension of the sacculus. Based on the abundance of the elongase subunits, between 20 and 100 elongases per cell are expected to be active. After c. 40% of generation time, the FtsZ-ring is formed and a switch occurs to polar peptidoglycan synthesis by a penicillin-insensitive peptidoglycan-synthesizing machinery. About one fifth of a generation time later, the proteins FtsK up to AmiC assemble onto the ring, and constriction of the cell is initiated. The proteins that constitute a mature divisome are present in about 50 complexes (represented here by 12) that are predicted to remain attached to the Z-ring during the whole constriction process.


The authors thank Benoît Wolf for the drawing of Figs 3 and 4 and Elsevier Limited Oxford, UK for the permission to reproduce Fig. 4 (here Fig. 7) from Wientjes & Nanninga (1991). This work was supported in part by the European commission within the ‘EUR-INTAFAR’ LSHM-CT-2004-512138, and ‘COBRA’ LSHM-CT-2003-503335 networks and by a Vernieuwingsimpuls grant 016.001.024 of the Netherlands Organization for Scientific Research (NWO) to T.d.B., by the Belgian State, Prime Minister's Office, Science Policy programming (IAP no. P6/19), the Actions de Recherche Concertées (grant no. 03/08-297), the Fond de la Recherche Fondamentale Collective (contract no. 2.4543.05) to M.N., and BFU2006-04574 from Ministerio de Educación y Ciencia, Spain, to J.A.A. and an institutional grant from Fundación Ramón Areces.