Bacterial Cell Wall Synthesis: New Insights from Localization Studies
Dirk-Jan Scheffers1,* and Mariana G. Pinho2,*
Department of Molecular Microbiology, Institute of Molecular Cell Biology, Vrije Universiteit, Amsterdam, The Netherlands,1 Microbial Pathogenesis and Cell Biology Laboratory, Instituto de Tecnologia Química e Biológica, Oeiras, Portugal2
*Corresponding author. Mailing address for Dirk-Jan Scheffers: Department of Molecular Microbiology, Institute of Molecular Cell Biology, Vrije Universiteit, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands. Phone: 31 20 5986248. Fax: 31 20 5986979. E-mail: email@example.com. Mailing address for Mariana G. Pinho: Microbial Pathogenesis and Cell Biology Laboratory, Instituto de Technologia Química e Biológica, Avenida da República (EAN), 2781-901, Oeiras, Portugal. Phone: 351 21 4469541. E-mail: tp.lnu.bqti@ohnipgm.
Author information ►Copyright and License information ►
Copyright © 2005, American Society for Microbiology
This article has been cited by other articles in PMC.
In order to maintain shape and withstand intracellular pressure, most bacteria are surrounded by a cell wall that consists mainly of the cross-linked polymer peptidoglycan (PG). The importance of PG for the maintenance of bacterial cell shape is underscored by the fact that, for various bacteria, several mutations affecting PG synthesis are associated with cell shape defects. In recent years, the application of fluorescence microscopy to the field of PG synthesis has led to an enormous increase in data on the relationship between cell wall synthesis and bacterial cell shape. First, a novel staining method enabled the visualization of PG precursor incorporation in live cells. Second, penicillin-binding proteins (PBPs), which mediate the final stages of PG synthesis, have been localized in various model organisms by means of immunofluorescence microscopy or green fluorescent protein fusions. In this review, we integrate the knowledge on the last stages of PG synthesis obtained in previous studies with the new data available on localization of PG synthesis and PBPs, in both rod-shaped and coccoid cells. We discuss a model in which, at least for a subset of PBPs, the presence of substrate is a major factor in determining PBP localization.
The cell wall is the principal stress-bearing and shape-maintaining element in bacteria, and its integrity is of critical importance to cell viability. In both gram-positive and gram-negative bacteria, the scaffold of the cell wall consists of the cross-linked polymer peptidoglycan (PG). Many studies have addressed the relationship between PG synthesis and bacterial growth and cell shape by looking at changes in cell shape in mutants that lack one or several enzymes involved in the synthesis of PG or other cell wall components or by looking at the incorporation of labeled PG precursors into the cell wall (see 41, 57, 80, 159). Recent developments have prompted a renewed effort to understand cell wall growth and shape determination. First, the application of fluorescence microscopy to bacteria has made it possible to study the localization of enzymes involved in PG synthesis in growing cells, as well as to look at localization of newly incorporated PG in live cells. Second, the discovery of an actin-like cytoskeleton involved in bacterial cell shape determination has raised the question of how structural information from inside the cell is translated to the cell wall. In this review, we discuss the recent data on localization of PG-synthesizing enzymes in the light of what is known about PG synthesis from previous studies, and we discuss the role of bacterial cytoskeletal proteins in organizing the cell wall synthesis process.
We focus this review not only on the usual model organisms, the rod-shaped bacteria Bacillus subtilis and Escherichia coli, but also on two cocci, namely, Staphylococcus aureus and Streptococcus pneumoniae, both of which are clinically relevant pathogens. Rod-shaped bacteria always divide through the same medial plane and are thought to have two modes of cell wall synthesis: one responsible for the elongation of the cell and one responsible for the formation of the division septum (Fig. 1). The two modes of synthesis appear to be catalyzed by different protein complexes. Coccoid bacteria like S. aureus divide using three different perpendicular planes in three consecutive cycles of cell division and seem to have only one mode of cell wall synthesis at the septum. S. pneumoniae cells are not “true” cocci, as their shape is not totally round, but instead have the shape of a rugby ball and synthesize cell wall not only at the septum but also at the so called “equatorial rings” (Fig. 1). These differences in the mode of division and sites for cell wall synthesis reflect some of the diversity existing in bacteria, a fundamental aspect of bacterial cell biology.
Incorporation of new cell wall in differently shaped bacteria. Rod-shaped bacteria such as B. subtilis or E. coli have two modes of cell wall synthesis: new peptidoglycan is inserted along a helical path (A), leading to elongation of the lateral wall,...
PEPTIDOGLYCAN STRUCTURE AND COMPOSITION
Peptidoglycan, also called murein, is a polymer that consists of long glycan chains that are cross-linked via flexible peptide bridges to form a strong but elastic structure that protects the underlying protoplast from lysing due to the high internal osmotic pressure (57, 80, 128, 175). The basic PG architecture is shared between all eubacteria except Mycoplasma and a few other species that lack cell wall. The glycan chain is built up of alternating, β-1,4-linked N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) subunits. The chemistry of the glycan chains varies only slightly between different bacteria. However, there is considerable variation in the composition of stem peptides which are linked to the carboxyl group of MurNAc (for an extensive review, see reference 175). Since it is beyond the scope of this review to discuss PG chemistry in great detail, we will discuss only some aspects of PG structure in the model organisms Escherichia coli, Bacillus subtilis, and Staphylococcus aureus.
The length distribution of glycan chains is usually very broad, with a mean of 25 to 35 disaccharide units in E. coli (72) but ranging up to over 80 disaccharide units in E. coli (80) and 100 in B. subtilis (201). In S. aureus the majority of chains have a length of 3 to 10 disaccharide units, with a maximum length of at least 23 to 26 units (14). The stem peptides are synthesized as pentapeptide chains, containing l- and d-amino acids and one dibasic amino acid, which allows the formation of the peptide cross bridge. The dibasic amino acid is often meso-diamoinopimelic acid (m-A2pm), which is present in most gram-negative bacteria and some gram-positive bacteria such as some Bacillus species, or l-lysine, which is present in most gram-positive bacteria. The most common stem peptide found in unprocessed PG in E. coli and B. subtilis is l-Ala(1)-d-Glu(2)-m-A2pm(3)-d-Ala(4)-d-Ala(5), with l-Ala(1) attached to the MurNac (57, 80, 175) (Fig. 2). The peptide cross bridge is formed by the action of a transpeptidase (see below) that links d-Ala(4) from one stem peptide to the free amino group of m-A2pm(3) on another stem peptide (Fig. 2B). In some organisms, the cross-linking of the PG is made via an amino acid bridge, which in the case of S. aureus is made up of five glycines (Fig. 2).
Building blocks and synthesis reactions of the peptidoglycan. (A) The basic unit of the peptidoglycan is a disaccharide-pentapeptide composed of the amino sugars N-acetylglucosamine and N-acetylmuramic acid, which are linked together by β-1,4...
The disaccharide units on the glycan strands may form a helical structure with the stem peptides protruding in all directions and forming angles to one another of about 90° (94, 99) so that cross bridges can be formed in all directions (80). In a two-dimensional PG layer, only every second peptide is in the same plane and therefore sufficiently close to be cross-linked in that layer. Depending on the strain and growth conditions, 44 to 60% and 56 to 63% of the stem peptides are part of cross-links in E. coli and B. subtilis, respectively (5, 63). The long and flexible pentaglycine cross bridge characteristic of S. aureus is able to span the distance between stem peptides from different PG layers which would otherwise be too distant to be cross-linked (102, 103). This permits the very high degree of cross-linking, up to 90%, observed in the staphylococcal PG (207). When a stem peptide is not cross-linked, both terminal d-Ala residues are usually cleaved off by carboxypeptidases (see below). However, in S. aureus these residues are not cleaved due to the lack, or low activity, of dd- and ld-carboxypeptidases (212).
Although the basic structure of the PG is very similar in gram-positive and gram-negative bacteria, the thickness of the PG layer is very different: the gram-positive wall is at least 10 to 20 layers thick, whereas the gram-negative wall is only 1 to 3 layers thick. This means that in gram-negative bacteria, a very thin layer of PG is sufficient to maintain the cell's mechanical stability. Neutron small-angle scattering studies have indicated that in E. coli up to 80% of the PG consists of a monolayer, with the rest of the PG being triple layered (100). As a consequence, the insertion of new wall material and the removal of wall material from the single PG layer must be tightly controlled by enzymes involved in PG synthesis and turnover, since otherwise the mechanical stability of PG would be compromised.
In gram-negative bacteria, the PG is covalently attached to the outer membrane via lipoprotein (Lpp) (Braun's lipoprotein) (17). Gram-positive bacteria, which lack an outer membrane, have a thick cell wall which contains covalently linked, charged polymers, such as teichoic acid and teichuronic acid, as well as proteins that are anchored (covalently or noncovalently) to the cell wall.
The classical view of PG architecture is that the glycan strands are arranged parallel to the membrane, although to date there are no experimental techniques that can confirm this model (195). Recently, this view has been challenged by a new model, the so-called scaffold model, that proposes that the glycan strands extend perpendicularly to the cytoplasmic membrane, growing outward in a linear rather than a layered fashion (45, 46). In an extensive review, Vollmer and Höltje have discussed both models, pointing out that for the E. coli cell wall, when the scaffold model is applied, the amount of PG present is not sufficient to cover the entire cell surface (195). Therefore, for the purposes of this review we will assume that the PG is in an orientation parallel to the cytoplasmic membrane.
BIOCHEMICAL REACTIONS FOR PEPTIDOGLYCAN SYNTHESIS
The biosynthesis of PG can be divided into three different stages (reviewed in references 162, 190, and 191). The first stage occurs in the cytoplasm and leads to the synthesis of the nucleotide sugar-linked precursors UDP-N-acetylmuramyl-pentapeptide (UDP-MurNAc-pentapeptide) and UDP-N-acetylglucosamine (UDP-GlcNAc). In the second stage, which takes place at the cytoplasmic membrane, precursor lipid intermediates are synthesized. The phospho-MurNAc-pentapeptide moiety of UDP-MurNAc-pentapeptide is transferred to the membrane acceptor bactoprenol, yielding lipid I [MurNAc-(pentapeptide)-pyrophosphoryl-undecaprenol]. Then, GlcNAc from UDP-GlcNAc is added to lipid I, yielding lipid II [GlcNAc-β-(1,4)-MurNAc-(pentapeptide)-pyrophosphoryl-undecaprenol], which is the substrate for the polymerization reactions in bacteria that have directly cross-linked PG. The use of a lipophilic molecule such as bactoprenol enables the cell to transport hydrophilic precursors from the aqueous environment of the cytoplasm, through the hydrophobic membrane, and to the externally situated sites of incorporation into the growing PG. It has been suggested that the translocation of the lipid-linked precursor from the cytoplasmic side to outer side of the membrane, at a high rate that matches the rate of PG synthesis, is catalyzed by a specific translocase or flippase. Although no biochemical evidence has been obtained, FtsW and RodA homologues are good candidates for this function. These proteins are members of the SEDS (shape, elongation, division, and sporulation) family, and homologues have been found in many bacteria that contain a cell wall but not in the wall-less Mycoplasma genitalium or the archaeon Methanococcus janasschii (78). E. coli rodA(Ts) mutants grow as spheres, and depletion of RodA in B. subtilis also leads to the conversion from rod-shaped to spherical cells, implicating RodA in growth of the lateral cell wall (8, 78). Also, in E. coli rodA and pbpA are in an operon, as are ftsW and ftsI (86, 115). In B. subtilis both rodA and ftsW are not found in operons with pbp genes, but a third homologue, spoVE is part of the mur operon, which also contains the upstream transpeptidase gene spoVD (30, 77). The facts that RodA and FtsW homologues are involved in cell wall synthesis, that their genes are often associated in operons with a specific pbp gene, and that the proteins are integral membrane proteins generally predicted to have 10 transmembrane helices are in accordance with the suggestion that they may channel the lipid precursors to their cognate penicillin-binding proteins (PBPs) (50, 87, 88).
The third and final stage of PG biosynthesis, which takes place at the outer side of the cytoplasmic membrane, involves the polymerization of the newly synthesized disaccharide-peptide units and incorporation into the growing PG. This is achieved mainly through the action of the so-called penicillin-binding proteins, which catalyze the transglycosylation and transpeptidation reactions responsible for the formation of the glycosidic and peptide bonds of the PG, respectively (Fig. 2B).
In the transglycosylation reaction, for the formation of glycan strands, the reducing end of the MurNAc of the nascent lipid-linked PG strand is likely transferred onto the C-4 carbon of the glucosamine residue of the lipid-linked PG precursor, with concomitant release of undecaprenyl-pyrophosphate (Fig. 2B). The undecaprenyl-pyrophosphate is then dephosphorylated to yield the lipid carrier bactoprenol, which becomes available for a second round of synthesis. It is still not clear how or whether the growing strand is released from the enzyme and what determines the length of the PG strand.
During transpeptidation, the d-Ala-d-Ala bond of one stem peptide is first cleaved and an enzyme-substrate intermediate is formed, with the concomitant release of the terminal d-Ala. The cleavage reaction provides the energy necessary for the transpeptidation reaction, which occurs outside the cytoplasmic membrane in the absence of energy donors such as ATP. A second step involves the transfer of the peptidyl moiety to an acceptor. This acceptor is the non-alpha amino group of the dibasic amino acid in a second stem peptide (in bacteria with direct cross-linking) or the last amino acid of the cross bridge when it exists. The reaction results in the formation of a new peptide bond between the penultimate d-alanine of a donor peptide and an amino group of the cross bridge of an acceptor peptide (Fig. 2B).
ENZYMES INVOLVED IN THE LAST STAGES OF PEPTIDOGLYCAN SYNTHESIS
Penicillin-binding proteins belong to the family of acyl serine transferases, which includes high-molecular-weight (HMW) PBPs, low-molecular-weight (LMW) PBPs, and β-lactamases (58).
HMW PBPs are enzymes that are composed of two modules located on the outer surface of the cytoplasm membrane and anchored to the cytoplasmic membrane by an N-terminal, noncleavable signal peptide. The C-terminal module is the penicillin-binding domain, which catalyzes the cross-linking of the PG peptides. Depending on the primary structure and on the catalytic activity of the N-terminal domain, HMW PBPs can be divided in two major classes: A and B (58, 65). The N-terminal module of HMW class A PBPs, of which PBP1A and PBP1B of E. coli are the prototypes, has transglycosylase activity. This makes class A PBPs capable of both the elongation of glycan strands (transglycosylation) and the formation of cross-links between the peptides (transpeptidation) of PG. Transglycosylation may also be done by monofunctional glycosyl transferases (43, 139). Class B HMW PBPs, of which PBP2 and PBP3 from E. coli are representatives, have an N-terminal non-penicillin-binding domain whose function is unknown but which has been suggested to be a morphogenetic determinant module (65). At least in E. coli PBP3, the N-terminal domain is required for the folding and stability of the penicillin-binding module, functioning as an intramolecular chaperone (64), and it may also provide recognition sites for interaction with other cell division proteins (132).
The alignment of several class A and class B HMW PBPs showed that there are distinct conserved motifs characteristic of the non-penicillin-binding domains of each class, which may point to a conserved role of these domains (65). The conserved motifs of the transpeptidase domains are better studied and not only are common to these two classes but also constitute the unique signature of all penicillin-interacting proteins: SXXK (which contains the active-site serine), (S/Y)XN, and (K/H)(T/S)G. These motifs are always present in the same order with similar spacing in the primary protein structure, forming the active site in the tertiary structure by folding of the polypeptide chain (58, 65, 114).
LMW PBPs are monofunctional dd-peptidases. Most of these act as dd-carboxypeptidases (although some exhibit transpeptidase or endopeptidase activity) that help to control the extent of PG cross-linking through hydrolysis of the carboxy-terminal d-alanyl-d-alanine peptide bond of a stem peptide, which prevents cross-linking of that peptide (58, 114).
Although the biochemical function of the PBPs is known, their specific role in the cell has remained elusive in most cases, due mainly to the fact that most organisms have redundant PBPs. This is illustrated by the fact that E. coli cells remains viable even after the inactivation of at least 8 of the 12 PBPs present in this organism (37). Nevertheless, some insight has been gained through the analysis of mutants with mutations in the different PBPs, mainly of the model organisms E. coli and B. subtilis. An overview of PBPs from these organisms and from S. aureus and S. pneumoniae and of their proposed functions is presented in Table 1.
Overview of penicillin-binding proteins from B. subtilis, E. coli, S. aureus, and S. pneumoniae
The crystal structures of several high- and low-molecular-weight PBPs from various organisms have been determined (33, 89, 110, 112, 121, 133, 137, 167, 168). Acyl-PBP complexes formed with antibiotics have generated structural insight into PBP-substrate binding and provided understanding of the development of antibiotic resistance as well as a possible starting point for the development of novel antibiotics. Although a detailed discussion of these structures is beyond the scope of this review, it is interesting to note that the recently solved crystal structure of a soluble form of S. pneumoniae PBP1b revealed a conformational change upon ligand binding (112). The active site of the transpeptidase domain of PBP1b was found to exist in an “open” conformation and a “closed” conformation, and the open conformation was dependent on the presence of ligand whereas the closed conformation showed blocked substrate accessibility. The difference between the structures suggests a possible activation mechanism for PBPs (112).
GROWTH OF THE PEPTIDOGLYCAN: HYPOTHETICAL MULTIENZYME COMPLEXES FORMED BY PENICILLIN-BINDING PROTEINS
To grow and divide, bacteria need not only to synthesize new PG but also to break the covalent bonds of the existing PG sacculus that involves the cell, in order to enable the insertion of new material. It is accepted that PG hydrolases are therefore essential for the growth of the cell wall, together with the PBPs.
Specific hydrolases exist for almost every covalent ligation of the PG, and these enzymes can be classified as muramidases, glucosaminidases, amidases, endopeptidases, and carboxypeptidases, depending on the specific bond of the PG which they cleave. Besides a role in growth of the cell wall, PG hydrolases have been proposed to be involved in, for example, cell separation after division, cell wall turnover, and muropeptide recycling or sporulation (82, 177, 180).
Several models have tried to answer the question of how the cell controls the activity of hydrolases during PG synthesis in order to avoid autolysis (24, 80, 95). Different models have been proposed for gram-positive and gram-negative bacteria, since the former have thick, multilayered cell walls while the latter have essentially a monolayered cell wall (100).
Koch has proposed the so-called “inside-to-outside” growth model for gram-positive bacteria. New material is inserted on the inner face of the wall, adjacent to where the PBPs are. According to the surface stress theory, the wall which is laid down immediately outside the cytoplasmic membrane is in an unextended conformation. As subsequent additions of PG occur, the wall moves outward, becomes stretched, and bears the stress due to hydrostatic pressure (95). Autolytic activity would be expected to be greater in the more stressed (external) layers, since the stress provides a lowering of the reaction activation energy (93, 95). Under this growth mode, the wall is never weakened because new covalently closed layers of the sacculus are formed before the older and outer ones are breached (95).
A more elaborate growth model was proposed for PG synthesis in E. coli. It is called “three-for-one” because it states that for every three new PG strands inserted in the cell wall, one old strand is removed; three glycan strands cross-linked to each other are covalently linked by transpeptidases to the free amino groups in the cross bridges on both sides of one old strand, called the docking strand. Specific removal of the docking strand leads to the insertion of the three new strands into the stress-bearing sacculus (80, 81) (Fig. 3A). This is in accordance with the “make-before-break” strategy proposed by Koch and Doyle (95), since also in this case the new material is made and inserted in the cell wall before the old material is removed, whereby the risk of autolysis is avoided. It is also in accordance with the high rates of PG turnover observed in E. coli during growth and division (66, 138), which may correspond to the release of the docking strands. The three-for-one mechanism requires the coordination, in time and space, of several different enzymes. Therefore the existence of multienzyme complexes in E. coli was proposed. These complexes would combine the activities of PG synthases, namely, transpeptidases and transglycosylases, and PG hydrolases, in a macromolecular PG-synthesizing machinery (80, 81) (Fig. 3B). This hypothesis is supported by affinity chromatography studies that revealed protein-protein interactions between bifunctional transpeptidase-transglycosylase PBPs, monofunctional transpeptidases, lytic transglycosylases, dd-endopeptidases, and structural proteins (163, 196, 197). Supporting evidence was obtained in vivo with Haemophilus influenzae, using cross-linking agents to identify two penicillin-binding multienzyme complexes, each containing several PBPs that interact via salt bridges (3). This work looked only into the PBP components of the isolated complexes and did not go further in identifying other proteins that may be present. It is nevertheless interesting that one of the complexes contained PBP2 (which is essential for the formation of rods), while the other contained PBP3 (which is implicated in cell division), in agreement with the idea that there are two different complexes for cell elongation and for cell constriction (80, 166). Work from the same group also identified two PBP complexes in E. coli (11) and B. subtilis (178) by cross-linking, but in these cases, both elongation- and cell division-specific PBPs were present in both complexes. Recently, coimmunoprecipitation performed using Caulobacter crescentus membranes and anti-PBP2 antibodies revealed that PBP2 interacts with PBP1a and PBP3a (54), providing further support for the existence of multienzyme complexes formed byPBPs.
Proposed mode of insertion of new peptidoglycan in gram-negative bacteria. (A) The three-for-one growth mechanism suggests that three newly synthesized, cross-linked glycan chains in a relaxed state (white circles) are covalently attached to the free...
LOCALIZATION OF PEPTIDOGLYCAN SYNTHESIS
Rod-shaped bacteria are considered to have two modes of cell wall synthesis: one associated with elongation and one associated with formation of the division septum, which, after division, becomes the “new pole” of both daughter cells (Fig. 1). This notion stems from the observation that, in both E. coli and B. subtilis, elongation- and division-specific PBPs have been identified (170, 183, 202, 215), as has an elongation-specific function for RodA, the putative lipid II translocase (8, 38, 78). The two modes of synthesis can be dissociated in mutants, and therefore results on the localization of PG synthesis are commonly interpreted as activities associated either with synthesis of the lateral wall (elongation) or with synthesis of the septum (division). Some cocci seem to display only septal PG synthesis, whereas streptococci synthesize PG both at the septum and at the equatorial rings (Fig. 1).
Incorporation of Labeled Peptidoglycan Precursors
Classical experiments on the localization of PG synthesis and turnover involved the use of labeled PG precursors (an overview of methods is shown in Fig. 4). These studies led to various important insights about the mode of PG synthesis in rod-shaped organisms. The insertion of cell wall material in the E. coli lateral wall was found to occur in a diffuse fashion (25, 29, 206, 210) (an example of [3H]diaminopimelic acid-labeled E. coli sacculi is shown in Fig. 4A). Later labeling studies, where the incorporation of d-Cys combined with immunodetection was used (an example is shown in Fig. 4B), revealed that PG is inserted in patches, bands, and hoops in the lateral wall of E. coli (40, 41). Studies with B. subtilis revealed localized insertion of wall material, with old material segregating along with DNA and new material inserting into new patches (152, 173, 174). In this type of pulse-chase experiment, the incorporated labeled material is followed over time after a chase with nonlabeled precursor. The resolution of the method is not high enough to exclude the possibility that new, unlabeled wall material is inserted at sites where labeled material is also present. Studies with another rod-shaped organism, Bacillus megaterium, also point towards diffuse intercalation of PG precursors along the lateral wall (34).
Different methods to visualize the incorporation of PG precursors. (A) Autoradiogram of sacculi prepared from steady-state E. coli cells grown at 28°C in the presence of [3H]diaminopimelic acid. (Adapted from reference 210 with permission of...
When the synthesis of the division septum occurs, the rate of PG synthesis notably increases compared to that of lateral wall synthesis (29, 210). Interestingly, localized insertion of PG at future division sites was also observed in filaments formed by an E. coli dnaX(Ts) strain (which contain a central nucleoid and long DNA-free cell ends) in areas that are not occupied by the nucleoid (123) but was not observed in filaments of a temperature-sensitive ftsZ strain (210). This indicates that this localized synthesis activity is dependent on activation of the cell division cascade by FtsZ and takes place at nucleoid-free sites that allow formation of an FtsZ ring. d-Cys labeling studies confirmed and refined the notions outlined above: the occurrence of diffuse incorporation of lateral wall material, localized synthesis at future cell division sites dependent on FtsZ but independent of other cell division proteins such as the cell division-specific transpeptidase PBP3, and heightened PG synthesis at the septum (40).
The material inserted at the septum will form one of the two cell poles of each of the daughter cells, after separation. The cell wall at the poles is “inert”; i.e., no new material is inserted at these sites, and polar PG is not subject to degradation/turnover as is material in the lateral wall, both in E. coli (25, 40) and in B. subtilis (119). The differentiation of PG to inert PG takes place at the cell division site, before septation is complete (40), but whether this differentiation includes structural changes to PG composition or other changes is not known.
A special sort of pole is observed during branch formation by E. coli. Branches originate randomly from side walls and are thought to form from (small) asymmetries in the side wall (67). In E. coli strains defective for PBP5, a high frequency of branching is observed. Like at cell poles, the PG present at the tips of these branches is inert (42) (Fig. 4C). Areas of inert PG at the poles were often split by areas of active synthesis, which explains branching from the cell poles. Areas of inert PG in the side wall were also observed and are thought to act as de novo poles around which new cell wall material is incorrectly oriented, resulting in branch formation (42). Interestingly, the poles of rod-shaped cells not only are metabolically inert for PG but also constitute an area of restricted mobility for periplasmic proteins (55) and outer membrane proteins (39, 42).
A leap forward in the visualization of nascent PG was achieved recently by work of Daniel and Errington, who developed a novel high-resolution staining method to label nascent PG in gram-positive bacteria by using a fluorescent derivative of the antibiotic vancomycin (Van-FL) (31) (Fig. 4D). Vancomycin binds the terminal d-Ala-d-Ala of PG precursors. Control experiments with B. subtilis showed that Van-FL binds to externalized but unincorporated lipid-linked PG precursors and to the recently inserted lipid-linked subunit at the growing end of a glycan strand and can therefore be used as a marker for nascent PG synthesis. Older PG is not recognized by Van-FL, as most of the terminal d-Ala-d-Ala is either removed by transpeptidation reactions or cleaved off by carboxypeptidases (31). Van-FL staining of growing B. subtilis cells showed prominent staining at an area around the division site, corresponding with the areas of high PG synthesis activity observed with radiolabeled PG precursors, and no staining at the cell poles. The staining at the lateral wall, although less bright, could be resolved in a helical pattern, which turned out to be dependent on the cytoskeletal protein Mbl, providing a clear indication for the role of the bacterial actin homologues, the MreB-like proteins, in growth of the lateral wall in rod-shaped organisms (discussed in more detail below) (31) (Fig. 4D). The same staining procedure using Van-FL allowed visualization of PG synthesis in other organisms: Streptomyces coelicolor, a gram-positive organism that grows as hyphae, revealed polar growth, whereas staining of Corynebacterium glutamicum, which lacks a homologue of the MreB protein family, revealed growth at the division site and at the poles, which are normally inert in most rod-shaped organisms. These results illustrate the different growth modes of rod-shaped organisms that either contain or lack MreB-like proteins.
For coccoid bacteria, which lack MreB homologues (90), incorporation of new PG has been recently studied in S. aureus. Due to the low carboxypeptidase activity present in this organisms, a considerable percentage of muropeptides containing the terminal d-Ala-d-Ala, which is recognized by vancomycin, is present even in “old” PG. For these reasons, specific staining of nascent PG using Van-FL requires a previous incubation of the cells with an excess of d-serine, which is incorporated in the last position of the pentapeptide, followed by a short pulse of d-Ala, which is then recognized by vancomycin. This procedure showed that new cell wall synthesis seems to occur mainly, if not only, at the division site (Fig. 4E). This would imply that S. aureus divides by synthesizing new wall material specifically at cell division sites, in the form of a flat circular plate that is subsequently cleaved and remodeled to produce the new hemispherical poles of the daughter cells, in accordance with earlier studies which state that at least some cocci have only a septal mode of cell wall synthesis (60, 79) but contrary to the suggestion that the spherical form could result from diffuse growth over the entire cell surface (96).
In one of the earliest studies in the field, Cole and Hahn used fluorescently labeled globulin to study bacterial cell wall growth in Streptococcus pyogenes and found that streptococci display equatorial growth to synthesize lateral wall combined with septal growth associated with division (28). By following the segregation of [3H]choline-labeledteichoic acid in the cell wall, evidence for equatorial cell wall growth was found in S. pneumoniae (18). In a different study, Tomasz and colleagues took advantage of the fact that choline-containing cell wall is sensitive to autolysin whereas ethanolamine-containing cell wall is not. Autolysin treatment of choline-pulsed ethanolamine-grown cells confirmed that new cell wall is inserted in the central portion of dividing cells (187, 188). These results were recently supported by Van-FL staining of S. pneumoniae, which clearly indicated PG synthesis activity at the septal and (stained less brightly) the equatorial rings (31).
Localization of PBPs
A different way to identify the places where PG is synthesized is by looking at the localization of the main PG-synthesizing enzymes, the PBPs. Until recently, limited information on the localization of the cell division-specific PBPs in E. coli and B. subtilis was available from immunofluorescence (IF) studies (32, 141, 200, 205). The use of green fluorescent protein (GFP) fusions to determine protein localization within the bacterial cell (for an overview, see, e.g., reference 109) has, in the past 3 years, led to an enormous increase of localization data on PBPs in different live bacteria (summarized in Table 1). Until now, the localization of the non-penicillin-binding monofunctional transglycosylases has not been addressed.
PBP localization in B. subtilis.
PBP localization in rod-shaped organisms has been studied most extensively in B. subtilis. Thirteen out of a total of 16 genes coding for PBPs, encompassing all biochemical activities, were replaced on the chromosome by gfp-pbp gene fusions, and the localization of the GFP-PBP fusion proteins during both vegetative growth and sporulation was studied (169, 171).
During vegetative growth, three main localization patterns were observed (Fig. 5A). First, a disperse localization pattern at the peripheral wall and division site was observed for PBP4, -2c, and -2d; PBP2a, -H, and -4b; and PBP5 (169, 171) (Table 1). PBP2c, -2d, and -4b are active only during sporulation (116, 203), so the relevance of their localization during vegetative growth when expressed from an inducible promoter can be questioned. It is interesting to notice that PBP2a and -H, which are mutually redundant for lateral wall synthesis in B. subtilis (202), did not localize exclusively to the lateral wall but also localized to the division site (171).
Localization of PBPs in different organisms. (A) Three different patterns, i.e., disperse, septal, and spotty, are observed in B. subtilis. Shown are GFP-PbpH (i), GFP-PBP1 (ii), and GFP-PBP3 (iii). Bar, 5 μm. (Adapted from reference 171 with...
A second localization pattern was observed for PBPs localizing specifically to the site of cell division. Division site localization, previously shown with IF for PBP1 and the cell division-specific PBP2b (32, 141), was reproduced with GFP-PBP fusions, and the same pattern was observed for PbpX (171) (Table 1).
The third localization pattern was that of PBPs localizing in distinct spots at the cell periphery, which sometimes resolved in short arcs, as was observed for PBP3, -4a, and -4*. These short arcs were even more clear in three-dimensional image reconstructions when GFP-PBP3 and GFP-PBP4a fusions were combined in one strain to increase the GFP signal (169, 171).
The PBP1 localization at the septum is consistent with the notion that PBP1 functions in cell division (170). A PBP1knockout strain is viable but has a reduced growth rate, slightly elongated cells, and less efficient sporulation (158), as well as abnormal septal structures (141). By showing that in a PBP1 knockout strain the sporulation deficiency is the result of a defect in the formation of the asymmetric sporulation septum and that GFP-PBP1 depends on other membrane-associated cell division proteins for its localization (all membrane-bound cell division proteins in B. subtilis are interdependent in terms of localization ), the role of PBP1 as a nonessential component of the cell division machinery was confirmed (170).
The localization patterns of the GFP-PBP fusion proteins were also monitored during spore formation. Sporulation in B. subtilis starts with the formation of two polar division sites, one of which develops into the sporulation septum. After asymmetric division, the prespore develops in a mature spore through the coordinated action of dedicated genes that are expressed either in the larger mother cell or in the prespore (51). GFP-PbpX, normally located at the mid-cell division site, appears to spiral out from the medial division site to both asymmetric division sites during the switch from medial to asymmetric cell division at the start of sporulation (169). This spiraling resembles patterns described for FtsZ, FtsA, and EzrA (9). The division-specific PBP1 and -2b also localize to the asymmetric division site, but only to the one that has been committed to form the sporulation septum. In contrast, PbpX goes to both asymmetric division sites and so follows some of the early cell division proteins, before commitment to formation of the sporulation septum has taken place (169). Whether PbpX is a true component of the cell division machinery remains to be established, since a pbpX knockout strain shows no cell division or shape defects in vegetative growth or sporulation (169).
Also during sporulation, PBP2c and -2d were found to be redistributed from the peripheral wall to the sporulation septum. From the septum, these proteins followed the mother cell membrane during engulfment and ended up in the outer prespore membrane (169). Localization to the prespore is consistent with the role of PBP2c and -2d in sporulation, although it has to be noted that normally these proteins are expressed predominantly in the prespore and not in the mother cell (116, 143, 156). When GFP-PBP2c was expressed in the mother cell from a promoter that is activated only after the sporulation septum has been closed, the same targeting to the prespore septum and engulfing membrane was observed. This suggests that PBP2c is targeted to the prespore membrane by diffusion and capture, a mechanism whereby a protein is inserted into the membrane in a “random” fashion and then diffuses to its destination, where it is “captured” by other proteins with which it interacts (164, 169). Interestingly, two recent papers describe the targeting of proteins to the sporulation septum through interactions of the extracellular domains of two proteins across the septum (13, 47). A complex network of proteins with overlapping anchoring capacity is proposed to play a role in anchoring membrane proteins in the prespore septum, with a crucial role for the prespore protein SpoIIQ (13, 47). It will be interesting to see whether the localization of PBP2c and -2d also depends on SpoIIQ and other network partners.
PBP localization in E. coli.
In E. coli, two different transpeptidases are involved in the synthesis of cell wall material during elongation (PBP2) and division (PBP3, also called FtsI) (183). The difference between the activities of these proteins isthought to be caused by substrate specificity, with PBP3 exhibiting a preference for tripeptide side chains and PBP2 for pentapeptide side chains (16, 148). The subcellular localization of both these PBPs has been determined using IF and GFP fusions and is in accordance with their proposed roles.
The elongation-specific transpeptidase, PBP2, localizes both in a spot-like pattern in the lateral wall and at the division site, not dissimilarly to PBP4 and -5 from B. subtilis (36) (Fig. 5B). The presence of PBP2 at the division site prompted the question as to whether PBP2 is part of the E. coli division machinery, but those authors showed that this is not the case (36). A role for PBP2 in maintenance of the correct diameter of the cell was proposed, since cells grown in the presence of the PBP2 inhibitor amdinocillin (also known as mecillinam) show an increased pole diameter at newly formed cell poles (36).
The division-specific transpeptidase, PBP3, was shown to localize specifically to the division septum (200, 205). This localization was shown to be dependent on the PBP3 transmembrane domain and cell division proteins FtsZ, -A, -Q, -L, and -W (118, 204). Recently it was shown that the transmembrane domain of PBP3 alone localizes to the division septum, with residues on one side of the helix being essential for localization, forming a proposed helix-helix interaction with another division protein (144, 209). Using a two-hybrid analysis that is suitable for membrane proteins, Karimova et al. identified interactions of PBP3 with several other cell division proteins in E. coli (FtsA, FtsL, FtsN, FtsQ, and FtsW) and an uncharacterized but potential cell division protein (YmgF) (91). This confirmed and extended findings obtained with another two-hybrid analysis of the E. coli cell division machinery (44). Interestingly, all interactions except the PBP3-FtsQ interaction required significant portions of the periplasmic domain of PBP3, suggesting that the localization of the transmembrane helix of PBP3 to the division septum (209) is mediated by its interaction with the FtsQ transmembrane helix (91). It should be noted that in the two-hybrid studies both interacting proteins are overexpressed to 10 or more times their normal cellular levels, which may lead to the observation of interactions that do not take place within a growing cell. Previous domain swap analysis experiments showed that the cytoplasmic domain and transmembrane helix of PBP3 are essential for its function but that the FtsQ transmembrane helix can be swapped for a transmembrane helix from an unrelated protein (68). This means that even though the FtsQ and PBP3 transmembrane helices may be interacting, PBP3 localization is not dependent on the FtsQ transmembrane helix. This suggests that PBP3 is stabilized at the division site by multiple interactions with cell division proteins that all (partially) contribute to targeting of PBP3.
PBP localization in S. aureus.
The coccoid bacterium S. aureus has only four PBPs and, as mentioned above, has mainly one place where cell wall synthesis occurs, the septum (146), although some inside-to-outside growth may occur (59). In accordance with this idea, PBP2, the only PBP in S. aureus with both transpeptidase and transglycosylase activities, localizes at the septum (146, 147) (Fig. 5C). Fluorescence microscopy of a GFP-PBP2 fusion revealed that at the beginning of septum formation, PBP2 localizes in a ring around the future division plane. As the septum closes, the fluorescence signal from PBP2 changes from two spots, corresponding to a ring, to a line across the cell, corresponding to a disk colocalized with the division septum (147). Interestingly, septal localization of PBP2 is lost in the presence of the beta-lactam antibiotic oxacillin, indicating that substrate recognition may be important for PBP2 localization (147). Beta-lactams are structurally analogous to the d-alanyl-d-alanine terminus of the muropeptides, i.e., to the substrate of PBPs, and bind in a nonreversible manner to the active site of PBPs, thereby preventing further binding of PBPs to their natural substrates.
Beta-lactam-resistant strains of S. aureus have an extra PBP, PBP2A, which has low affinity for the antibiotic and is therefore capable of transpeptidation even in the presence of high concentrations of beta-lactams (71, 161). In these resistant strains, localization of PBP2 is maintained in the presence of oxacillin, suggesting that PBP2A is able to maintain PBP2 at the septum via (direct or indirect) protein-protein interaction (147).
PBP localization in S. pneumoniae.
A comprehensive study of the localization of all HMW PBPs in S. pneumoniae was done by Morlot and colleagues, using immunofluorescence (122), and three different patterns were found. PBP1a and PBP2x (Table 1) both have septal localization and follow FtsZ localization (with a delay of approximately 5 min under the conditions used in the study) (Fig. 5D). PBP2b and PBP2a both follow the localization of the duplicated equatorial rings (Fig. 5D). S. pneumoniae cells are encircled by an outgrowth of the cell wall called the equatorial ring. An initial ring is duplicated, and the two resultant rings are progressively separated, marking the future division sites of the daughter cells (186) (Fig. 1). The third localization pattern was observed only for PBP1b, which exhibits septal localization (in about 60% of the cells) and equatorial localization (in about 40% of the cells), although simultaneous septal and equatorial localization was not observed in an individual cell. These results confirm that PBP1a and PBP2x are the HMW class A and B PBPs, respectively, involved in cell division, while PBP2a and PBP2b are the HMW class A and B PBPs, respectively, associated with peripheral cell wall synthesis. The fact that two modes of cell wall synthesis, catalyzed by two sets of PBPs, seem to exist in S. pneumoniae implies a similarity with rod-shaped bacteria and a difference from “true” cocci such as S. aureus.
The same study by Morlot et al. found that constriction of the FtsZ ring precedes the shrinking of septal localization of the PBPs; i.e., FtsZ-driven membrane invagination is uncoupled from PBP-catalyzed PG synthesis. This suggests that although interaction with cell division proteins has a role in recruiting PBPs to the septum, the cell wall synthetic complex may subsequently dissociate from other division proteins (122). Similar uncoupling of membrane invagination from cell wall ingrowth was reported for B. subtilis upon depletion of the cell division-specific transpeptidase PBP2b (32) and for an E. coli mutant lacking several cell wall hydrolysases (74).
In many organisms there is a certain degree of redundancy between the existing PBPs. In S. pneumoniae none of the three class A PBPs was found to be essential because single mutants had no obvious phenotype, but a double mutant with mutations in PBP1a and PBP2a could not be obtained (84, 136). Interestingly, both PBP1a and PBP2a are able to modify their localization in the absence of the other class A PBPs. PBP1a displays both septal and equatorial localizations, although never simultaneously, when it is the only HMW class A PBP left in the cell, and the same happens for PBP2a (122). The fact that one PBP can take over the normal localization of a different PBP from the same class may explain the redundancy of these proteins.
S. pneumoniae PBP3, a LMW dd-carboxypeptidase that degrades the substrate of HMW PBPs, is evenly distributed on bothhemispheres and is absent from the future division site (120). This results in the substrate of HMW PBPs being present only at the division site. When PBP3 is absent from the cell, the substrate for HMW PBPs is no longer restricted to the division site (but should exist over the entire surface of the cell), and HMW PBPs loose their normal colocalization with FtsZ rings. This implies that, in wild-type cells, localization of HMW PBPs at midcell depends on the availability of substrate exclusively at that place, similarly to what was found for S. aureus (147).
PBP localization in other bacteria.
Another noncoccoid bacterium for which PBP localization was studied is C. crescentus. Using IF, PBP2, involved in elongation, was found to localize in stripes along the length of the cell in a pattern that resembled that found for the morphogenetic protein MreB (Fig. 5E) (54).
Fluorescein-labeled β-lactam antibiotics have been used to label PBPs in Streptomyces griseus, which grows as hyphae and is capable of sporulation. The labeling revealed high activity of PBPs in sporulation septa within sporulating hyphae and identified an 85-kDa PBP as a possible sporulation-specific PBP in this organism (70).
Localization of Peptidoglycan-Degrading Enzymes
PG hydrolases are enzymes that catalyze the turnover or degradation of PG in bacteria. Among other roles, these enzymes are proposed to participate in cell wall growth and turnover and in cell separation.
In S. aureus the major autolysin is encoded by the atl gene. The gene product is a precursor protein which is exported from the cytoplasm and undergoes two cleavage events to generate a mature amidase and glucosaminidase (135). The cell surface localization of the two enzymes was studied by scanning and transmission immunoelectron microscopy (213). The atl gene products form a ring structure on the cell surface at the septal region for the next cell division site, which is in agreement with the proposed function of Atl in the hydrolysis of the PG for the separation of daughter cells after division. Targeting of the amidase and glucosaminidase to a specific site within the bacterial envelope is directed by repeat domains (R1, R2, and R3) located at the center of the Atl precursor protein (6
The peptidoglycan (PG) cell wall is a defining feature of the bacteria. It emerged very early in evolution and must have contributed significantly to the success of these organisms. The wall features prominently in our thinking about bacterial cell function, and its synthesis involves the action of several dozen proteins that are normally essential for viability. Surprisingly, it turns out to be relatively simple to generate bacterial genetic variants called L-forms that completely lack PG. They grow robustly provided that lack of the cell wall is compensated for by an osmoprotective growth medium. Although their existence has been noted and studied on and off for many decades, it is only recently that modern molecular and cellular methods have been applied to L-forms. We used Bacillus subtilis as an experimental model to understand the molecular basis for the L-form switch. Key findings included the discovery that L-forms use an unusual blebbing, or tubulation and scission mechanism to proliferate. This mechanism is completely independent of the normal FtsZ-based division machinery and seems to require only an increased rate of membrane synthesis, leading to an increased surface area-to-volume ratio. Antibiotics that block cell wall precursor synthesis, such as phosphomycin, efficiently induce the L-form switch without the need for genetic change. The same antibiotics turned out to induce a similar L-form switch in a wide range of bacteria, including Escherichia coli, in which we showed that proliferation was again FtsZ-independent. Aside from further basic science, future work on L-forms is likely to focus on their possible role in chronic or recurrent infections, their use as a model in studies of the origins of life, and possibly, biotechnological applications.
- cell proliferation
- cell wall
The bacterial cell wall
The cell wall is an almost ubiquitous feature of the domain Bacteria. The major component of the cell wall is called peptidoglycan (PG), which comprises long glycan strands cross-linked by short peptide bridges . The PG forms an elastic meshwork that covers the whole surface of the cell and which serves to protect the cell from damage, resist the outward turgor pressure due to the high osmolarity of the cytoplasm, and confer shape. The precursor molecule for wall synthesis, called lipid II, is made inside the cytosol. It is composed of a disaccharide of amino sugars, N-acetylglucosamine and N-acetylmuramic acid, carrying a short peptide side chain containing unusual d-amino acids. The precursor is linked to a special isoprenoid lipid carrier called bactoprenol to form lipid II. This is then flipped to the outer surface of the cytoplasmic membrane, where enzymes, called glycosyltransferases, polymerise the disaccharide moieties to form the glycan strands and transpeptidases, called penicillin-binding proteins, make bridges between adjacent strands . A new family of putative glycosyltransferases, the RodA/FtsW (also called SEDS) proteins, was recently described [3,4] (see also ).
The shape of the cell is, to an extent, dictated by the shape of the PG layer. In many rod-shaped bacteria, such as Bacillus subtilis and Escherichia coli, an actin-like family of proteins, called MreBs, form polymers at the inner surface of the cytoplasmic membrane, where they are thought to organise the cell wall synthetic enzymes and provide spatial direction to cell wall synthesis, thereby governing cell shape . Another important ‘cytoskeletal’ protein, FtsZ, of the tubulin superfamily plays a similar role in governing PG synthesis during cell division [7,8]. Some rod-shaped bacteria lack MreB proteins and use a different growth strategy in which PG synthesis and remodelling occur at the tip of the rod [9,10]. Coccoid bacteria have yet other strategies for controlling their shape and wall synthesis, but virtually all bacteria have PG as their main cell wall shape-determining component .
Applying modern molecular cell biology methods to the L-form problem
In the mid-2000s, my laboratory became interested in an old problem relating to the apparent existence of bacterial variants that are capable of living in a cell wall-free state, called the L-form (or L-phase). There was an extensive literature on L-forms, going back to the 1930s [13,14], largely based on hospital case histories of patients with infections refractory to treatment with β-lactam antibiotics, or challenge studies in which animals were infected with walled or L-form cells, and progression of infection or clearance of the bacterial cells was followed. In the early days, there was confusion about whether L-forms should be distinguished from pleuropneumonia-like organisms (PPLOs). PPLOs are now called mycoplasmas, which are wall-deficient bacteria that have undergone millions of years of evolution to adapt to the wall-deficient state. L-forms, in contrast, are now usually assumed to be closely related to walled bacteria and often are able to switch back to the walled state.
L-forms are pleomorphic and osmotically sensitive because of their cell wall defect. However, they are also completely resistant to a range of antibiotics that work on cell walls, and there were sporadic reports of L-forms being involved in a wide range of often chronic or recurrent infections (reviewed in refs [15,16]). The literature on L-forms was quite extensive, but it peaked around the late 1970s, just before the advent of DNA sequencing and other methods that would have made molecular studies of the L-form state more tractable. In ca. 2004–2005, we decided to revisit the L-form problem using modern molecular cell biology and genomic methods in our favourite laboratory organism, B. subtilis.
B. subtilis had been reported to be able to enter the L-form state in earlier laboratory work , as well as in environmental studies of plant–microbe interactions . Richard Daniel, then a postdoc in my laboratory, acquired an environmental L-form isolate of B. subtilis from a laboratory in Aberdeen (that of Eunice Allan; ) and began investigating its properties. Working with the strain was frustrating because it was tricky to grow (e.g. requiring osmotically supportive medium) but also because our attempts to introduce fluorescent (GFP) markers or other genetic changes that would help us to study its properties could not be achieved by our standard genetic manipulation methods. The classic laboratory strain of B. subtilis is attractive as a model because it is extremely amenable to genetic transformation, but other environmental isolates are often not so tractable. Nevertheless, imaging of the ‘naked’ L-forms revealed a startling degree of morphological complexity, including long strands of cytoplasm joining adjacent pleomorphic cells, so we were encouraged to continue with the project. A couple of years later, after my laboratory had moved from Oxford to Newcastle University, a finishing PhD student, Mark Leaver, wished to stay on for another year to carry out some high-risk, high-reward experiments and became interested in the L-form project. With Richard, he spent a few frustrating months trying to work out how to generate L-forms from B. subtilis. L-forms are very slow growing and are rapidly outgrown by walled cells. So, classical L-form protocols often rely on the presence of antibiotics such as penicillin to select for L-forms and prevent the growth of walled cells. However, it turned out that, at least for B. subtilis, the presence of penicillin actually blocks the initial generation of L-forms, for reasons that are only becoming clear now (Kawai et al., in preparation). Mark eventually succeeded in forcing B. subtilis to make the L-form switch . He took advantage of a strain that Richard had made in which the genes for cell wall precursor formation could be turned on or off depending on the presence of an inducer, xylose (Pxyl-murE). Repression of wall synthesis in this strain, by withholding xylose, forced B. subtilis into a wall-deficient state, provided that they also had an osmoprotectant (in this case sucrose) to prevent cell lysis. The key to the protocol was to select with penicillin later, after the cells had the chance to switch into the L-form state, following which they appeared to be able to grow indefinitely. We also played around with some genetic tricks, such as having a second copy of the xylose repressor gene in the cells, to prevent mutants capable of making cell wall in the absence of xylose from emerging and taking over the plates. Once this protocol had been developed, Mark found that he could select for L-form growth in any of our genetically manipulated strains .
It was clear from the frequency at which the L-forms emerged that at least one mutation (in addition to repression of Pxyl-murE) was needed to enable the cells to grow. Attempts to map that mutation were frustrating because the L-forms were not easy to manipulate. However, we were fortunate that the timing of the project coincided with the emergence of whole genome sequencing. Newcastle had just set up a facility to do this and Jonathan Coxhead helped us to obtain the sequence of an L-form variant. After a considerable amount of bioinformatics, Mark was able to identify a single-point mutation that was present in the L-form compared with its parent strain. The mutation lay in a gene called ispA, encoding geranyl-geranyl pyrophosphate synthase, which is conserved from bacteria to man . It is required for synthesis of polyprenoid lipids. The mutation was a single base substitution, but it altered a residue that had been shown in the rat enzyme to virtually eliminate function. Mark showed that the mutation was sufficient to enable L-form growth when cell wall synthesis was shut down, but it was several years before we worked out precisely what the mutation did. In fact, at the time, we did little to follow up on this result because it seemed obvious from metabolic maps that ispA should be required for making the carrier molecule, bactoprenol, on which PG precursors are assembled. Since we were blocking another (later) step in precursor synthesis, we assumed that the ispA mutation prevented accumulation of a toxic intermediate or compensated for some kind of metabolic imbalance that occurs when PG precursor synthesis is shut down. However, this turned out not to be the whole story (see below).
Proliferation without a division machine
The L-forms had, as expected, the highly pleiomorphic shapes described in earlier literature and seen in our earlier experiments with environmental L-forms. They also had a huge range of sizes. Part of Mark's motivation for developing L-forms had been to ask a fundamental question about the function of the central player in bacterial cell division, FtsZ. FtsZ forms a ring-like structure at the site of impending cell division, where it also recruits various proteins required for cell wall synthesis . We did not know whether the Z-ring worked directly to drive constriction of the cell membrane at the division site, or whether it simply recruited division proteins, including cell wall synthases, which contributed the constrictive force. We anticipated being able to answer this question in L-forms because of their lack of cell wall function. With Mark's protocol, we could make L-forms from any of our genetically manipulated strains, so one of the first things that Mark did was to make L-forms from a strain bearing an FtsZ–GFP fusion, so that we could look at the Z-rings. It turned out to be quite difficult to visualise the GFP fusion because of the heterogeneous size and shape of the L-forms, their fragility, and our inability to immobilise them, so that they would stay in a focal plane. Nevertheless, the experiments suggested that L-forms rarely assemble the regular ring-like FtsZ structures of walled cells. We then started to wonder whether the L-forms used FtsZ at all to divide, so Mark built an L-form strain in which we could shut down expression of the ftsZ gene. In walled calls, this brings about a lethal cell division defect. The cells elongate without dividing, then become unstable and lyse. Remarkably, it seemed that repression of ftsZ expression made no difference in the viability or growth rate of our L-forms! One of the reviewers of our first L-form paper thought that this was such an important result that we needed to demonstrate that we could delete the ftsZ gene to prove that it was non-essential. This was technically challenging because of the essential nature of ftsZ in walled cells, but Mark was eventually able to build an L-form strain with a complete deletion of ftsZ (and the adjacent ftsA division gene for good measure), conclusively showing that the L-forms did not require the normal division machine . We were astonished by this result because FtsZ was widely conserved across the bacterial domain, and essential for viability virtually everywhere it had been tested (Streptomyces being one notable exception; ). This was our first hint that the study of L-forms might turn out to be much more interesting and important than we had anticipated.
An unexpected bizarre mode of proliferation
The ftsZ result raised an important question about the nature of L-form proliferation. Long-term time lapse imaging of the L-forms turned out to be difficult for various technical reasons. Nevertheless, one Saturday morning I got an excited email from Mark. He told me that he had got the time lapse imaging to work and had seen proliferative events, but that they did not fit with any of our models. I excitedly waited for the movies to download and was amazed at what I saw. In fact, I went running around the house trying to find someone else to show — sadly, my daughter (then aged about 17) was not as excited as I was!. The most prominent event captured in the movie (Figure 1A) was a cell, more or less round, which, over several hours, grew in size before elaborating a protrusion, which grew into a long tube that then resolved into a chain of what appeared to be progeny cells, which appeared to remain connected by tiny tubular connections. Another larger L-form showed a somewhat different behaviour (Figure 1B). To begin with, it had a more or less spherical shape but then, again over a period of hours, surface features, bulges, and dimples appeared at multiple places on the surface. This was followed by the eruption of multiple progeny across at least three different sites on the cell. Thus, L-forms clearly did not follow the binary fission process that typifies almost all cells that have been described. The startling new findings on loss of requirement for FtsZ and the bizarre mode of proliferation, which we termed ‘extrusion resolution’, were published in a full article in Nature, which coincided with the celebration of Darwin 200 . I later mused that Darwin would have been interested in the identification of a possible early intermediate step in the evolution of life.
Numbers refer to time (min) of observation (from ref. ). (A) An event we called extrusion–resolution. A spherical L-form increases in size, then a tubular protrusion emerges which breaks down into a chain of connected progeny cells. (B) A larger L-form again starts as a sphere, then undergoes pulsating changes in shape before multiple small progeny cells erupt from at least three different places on the cell surface.
Nothing is new under the sun
A little while after the paper was published I received a typewritten letter from Gertrud and Otto Kandler. The letter was generally complimentary but it contained a wry sentence. ‘We are delighted to see that the L-forms still replicate by the mechanism described in our paper of 1954’. We rushed off to find a copy of the paper  and found that it did indeed contain wonderful phase contrast images similar to the ones we had just published. The paper had been written in German, did not contain the key word ‘L-forms’, and had only been cited a handful of times, finally in the 1970s, so we did not feel too bad about having missed it. We published a corrigendum in Nature pointing to the earlier paper. Of course, in 1954, nothing was known about the FtsZ ring machinery, or even about the structure of the cell wall, and the nature or origins of the organisms described by the Kandlers were not well understood. Nevertheless, the story makes clear that it is very difficult to be completely original in science!
Patri Domínguez-Cuevas and Romain Mercier joined the laboratory, respectively, just before and just after Mark Leaver moved on, and they took up the mantle of the L-form work. I also received an ERC Advanced Grant to study the problem further. Patri made new L-form strains and discovered one that switched efficiently back and forth between the L-form and walled state . By sequencing this organism, we discovered that the emergence of L-forms from walled cells requires, at least under some conditions, e.g. penicillin selection, mutations that facilitate escape of the membrane-bounded L-form from the cylindrical cell wall. One regulatory mutation probably works by counteracting penicillin-triggered defence mechanisms that prevent or limit the degradation of the wall, and which would lead, under non-osmotically protective conditions, to cell lysis. The second mutation affected a cell division gene sepF, mutation of which leads to malformed division septa  that can apparently fracture, leading to L-form escape .
Membrane fluidity and its importance in L-form progeny scission
In these early days, we were convinced that the amazing extrusion resolution process must be driven by proteins. Figure 2 shows an early model in which we assumed that cytoskeletal fibrils (red lines), potentially of MreB, would drive the shape changes responsible for division. In Figure 2B, we imagined alternative models in which mechanisms responsible for segregation of chromosomes (which still remain unclear) might drive formation of the membrane protrusions. This class of model had the advantage that it would ensure that L-form progeny efficiently acquire the genetic information needed to propagate. Cytoskeletal proteins, such as mreB, seemed good candidates for proteins capable of driving the protrusions. However, extensive candidate gene knockout studies failed to identify factors needed for L-form growth . These included making L-forms from a triple knockout of all three mreB paralogues of B. subtilis , as well as genes such as divIVA, encoding a versatile protein required in different ways for polar morphogenesis in a range of Gram-positive bacteria , or chromosome segregation genes (in case segregation drove blebbing or tubulation). However, availability of a mutant that could readily switch between states enabled us carry out an unbiased genetic screen . Patri and Romain eventually found a mutant that could grow normally in the walled state but was completely unable to grow as an L-form. Genome sequencing of the mutant revealed that it had a point mutation probably inactivating a gene called bkd that was required for branched-chain fatty acid synthesis. Romain showed that the mutation probably worked directly on enzyme activity because he could restore growth to the mutant by providing the branched-chain fatty acid precursors that would be lacking in the mutant. This provided the first hint that properties of the cytoplasmic membrane might be important for L-form growth. Romain showed that the primary effect of the mutation was to reduce anteiso-branched-chain fatty acids, which increase membrane fluidity relative to the closely related iso-forms. Morphological characterisation of the mutant showed that the L-forms could grow for a significant period of time, and undergo changes in shape, but they did not resolve into separate progeny. We concluded that the mutant was affected in the final step of division, which we called scission, and that a relatively high level of membrane fluidity was required for this process .
(A) The central path illustrates that L-forms have a wide range of sizes. After a period of growth, proliferation can take any one of many forms, from essentially binary fission to the proliferative events exemplified in Figure 1. Red lines illustrate hypothetical cytoskeletal filaments that could be involved in driving shape changes leading to proliferation. (B) Model for proliferation based on the idea that active segregation of chromosomes (illustrated again by putative cytoskeletal or motive filaments) could drive shape changes leading to proliferation.
A simple biophysical mechanism for L-form proliferation
Romain Mercier teamed up with another geneticist, Dr Yoshi Kawai, to carry out an extensive genetic analysis of L-form growth . All of the experiments we had done so far were based on repression of cell wall precursor synthesis (Richard Daniel's Pxyl-murE construction). They wished to find out whether there were other ways to trigger L-form growth. They therefore took a strain with an ispA mutation and developed a way to screen for L-form colonies in this background. They also wished to find mutations that did not simply abolish wall precursor synthesis (like the Pxyl-murE construct), so they looked for L-forms that could resume growth as walled cells. Among the mutations they found, one stood out — a mutation just upstream of the accDA operon, which encodes the catalytic subunit of acetyl-CoA-carboxylate synthase. This enzyme catalyses the key committing step in fatty acid synthesis. Yoshi showed that the mutation worked by increasing synthesis of the proteins and that it could be mimicked by a construction that enabled overproduction of the proteins via a second copy of accDA controlled by Pxyl, elsewhere in the chromosome. Induction of Pxyl-accDA in the presence of an ispA mutation induced L-form growth just as efficiently as repression of Pxyl-murE. From previous work on fatty acid synthesis in B. subtilis , it was anticipated that increased acetyl-CoA-carboxylase synthase would increase the accumulation of malonyl-CoA, which would in turn induce the expression of various genes required for fatty acid synthesis, leading to increased membrane lipid accumulation. Furthermore, repression of PG precursor synthesis (as in our original L-form experiments with Pxyl-murE) also led indirectly to increased membrane synthesis, by a mechanism that is still unclear, as did various other mutations that emerged from our genetic screens. These results led us, originally with a certain degree of scepticism, to the notion that L-form growth might simply be promoted by excess membrane synthesis .
The main comfort for us was that there turned out be both theoretical and practical support for the model . Thus, Peter Walde's laboratory, in particular, had described experiments in which a lipid vesicle was induced to undergo L-form-like replication simply by ‘feeding’ it with fatty acids. By intercalating into the surface of the vesicle, the fatty acids increase its surface area to volume ratio, which is sufficient to generate ‘baby’ vesicles . An exciting outcome of these experiments was that they showed that L-forms might provide an interesting model for the replication of primitive cells, way back before the evolutionary emergence of the cell wall [11,32].
The L-form as an experimental tool for studying cell wall synthesis and cell division
L-forms are useful as a model for the origins of life, but they also provide powerful experimental systems for studying certain functions that are normally essential but become non-essential in the L-form state, such as the cell wall synthetic system itself and the FtsZ-based division machinery. We took advantage of this to study a long-standing question about whether cellular form requires a pre-existing template. This had long been debated , and it was a feature of certain models for cell wall synthesis, such as the 3-for-1 model of Holtje . In the latter model, an existing glycan strand in the wall is hydrolysed and replaced with a ‘triple pack’ of new strands. Yoshi Kawai tested the formal requirement for a cell wall template by making L-forms in which PG precursor synthesis was prevented, by deletion of the essential murC gene, and then, after a period of growth in the absence of PG, transformed the cells with a plasmid carrying the murC gene, which resulted in restoration of normal growth and rod-shaped form. Thus, we were able to conclusively reject models for cell morphogenesis that require a pre-existing template .
Generalisation of L-form principles to other bacterial groups
By about 2013, we were happy that we had achieved a relatively good understanding of the general principles underlying the walled to L-form transition and L-form growth in B. subtilis. However, it seemed important to explore whether these principles could be extended to other groups of bacteria. We could easily test whether inhibition of cell wall precursor synthesis might elicit the L-form switch by taking advantage of antibiotics, such as phosphomycin and d-cycloserine, that inhibit enzymes in the precursor pathway. Romain Mercier took a range of organisms of different classes and tested whether the inhibitor, in the presence of osmoprotectant, could generate L-forms. He showed that Corynebacterium glutamicum, a high G + C Gram-positive actinobacterium, switched beautifully into an L-form that grew well in liquid culture medium, much like B. subtilis L-forms. Even E. coli, a Gram-negative bacterium, separated by perhaps 1 billion years of evolution from B. subtilis, could grow in an L-form state, albeit requiring a semi-solid matrix or agar plate surface for efficient growth . Earlier work on E. coli had suggested that under at least some conditions, L-form-like cells required a residual level of PG synthesis , suggesting that they might differ, perhaps quite fundamentally, from the B. subtilis L-forms. However, on the basis of our B. subtilis results, we argued that a useful operational definition of the L-form state was ability to grow in the absence of the division machine, and Romain was able to take advantage of the powerful E. coli genetics to generate null mutations in various genes, including those for cell wall precursor synthesis and two different essential cell division genes, ftsZ and ftsK .
There have been a plethora of different kinds of conditions used to generate cells called L-forms and it may well be that, in different organisms, the extent to which they can survive and thrive with reduced levels of cell wall synthesis may vary. We therefore suggest that ability to proliferate in the absence of the normally essential FtsZ-based division machine is a useful operational definition for the L-form. It is clear that, at least for Gram-positive B. subtilis and Gram-negative E. coli, these organisms are intrinsically able to switch readily to a mode of proliferation that is independent of the normally complex and essential FtsZ-based machine.
Interestingly, one of the organisms that Romain examined, Caulobacter crescentus, an α-proteobacterium, resisted his attempts to force growth in the L-form state. We speculate that this may be due to the intricate dependence of the cell cycle of this organism on polar morphogenesis , which is presumably impacted badly by cell wall inhibition.
Solving the ispA conundrum
In the course of carrying out a detailed genetic dissection of the B. subtilis L-form transition, Yoshi and Romain also looked for mutations different from ispA that could support L-form growth when PG precursor synthesis was blocked (e.g. by repression of the Pxyl-murE construct). Most primary mutations of this class, which we termed class II, lay in or near the ispA gene. To avoid this, we introduced a second copy of the ispA gene, so that two mutational hits would be required to eliminate IspA function. The new mutations, which generally gave rise to weaker growth than ispA, mapped to a variety of different genes. However, many of them lay in genes involved in the respiratory chain and oxidative phosphorylation. Others would induce oxidative stress responsive genes, while a third group would affect glycolysis. These findings led us to propose that the mutations work by reducing oxidative damage . In support of this idea, Yoshi showed that inhibition of cell wall synthesis resulted in up-regulation of oxidative stress responsive genes, and that this stress was reduced by the class II mutations, including ispA. We then realised that rather than blocking bactoprenol synthesis (see above), which would in any case be lethal in non-L-form cells, the ispA mutation might also reduce or block the synthesis of menaquinone, another isoprenoid lipid, and a component of the respiratory chain. In further support of a model in which oxidative stress is experienced by wall-deficient B. subtilis, Yoshi showed that growth in the L-form state could be stimulated without an ispA mutation by use of anaerobic conditions or by the presence of exogenous scavengers of reactive oxygen species (ROS) . Similar results were obtained for E. coli, suggesting that the oxidative stress effect is broadly conserved. We currently favour a model in which a block in cell wall synthesis, and thus utilisation of sugar-phosphate intermediates, leads to increased flux through the TCA cycle. ROS are then generated as a by-product of the metabolism of molecular oxygen by the electron transport chain. We are currently investigating how the design of metabolism and intricate connections between PG precursor synthesis, fatty acid synthesis, and other outputs of glycolysis combine to generate the above effects.
The L-form project has ramified over the last 10 years to generate at least four different areas of interest. First, we still do not fully understand the basic biology of the L-form state. It is not completely clear how blocking cell wall precursor synthesis, stripping the cell wall (without blocking synthesis), or overproducing membrane, all elicit oxidative stress. The fact that this occurs in both Gram positives and negatives suggests that there are common principles in the design of metabolism that have been conserved over immense evolutionary time. The generation of oxidative stress by various antibiotics has been a controversial topic over the last few years  and further studies of L-forms may contribute significantly to the understanding of this complex area.
A second major topic of interest lies in the use of L-forms to inform about possible mechanisms for early steps in the evolution of cellular life [11,32]. The bacterial cell wall appears to be very ancient, possibly dating back to the earliest bacterial cells. Indeed, it is plausible that invention of the cell wall was a key step in enabling the bacterial radiation, providing the ability to withstand adverse changes in osmolarity, to achieve a defined shape and an efficient, tightly regulated division process . Comparative studies of L-form proliferation and the replication of simple lipid vesicles are likely to be an interesting and informative area, identifying also the minimal requirements for proliferation.
Minimal cells are also of interest in biotechnology, in principle, providing a way to reduce the metabolic energy that could otherwise be directed towards biosynthesis of commercial products. Elimination of the wall could also provide a way to remove a potential barrier to the secretion of proteins and avoid synthesis of wall fragments that have potentially toxic immunostimulatory effects.
Finally, many key questions still remain in relation to the possible role of L-forms in all kinds of infectious diseases [14,15,41]. Now that we have a much better understanding of the molecular and physiological changes that accompany and promote L-form growth, we are in a strong position to revisit questions about L-forms in disease. We are presently engaged in various collaborations aimed at identifying L-forms or L-form-like cells in various disease states. Ongoing work appears very promising, but the ‘killer’ experiments that will finally generate the data to convince sceptical infectious disease clinicians remain tantalisingly out of reach. Please watch this space!
Early work on L-forms in the laboratory was funded by grants from the UK Biotechnology and Biosciences Research Council. Work since 2009 was funded by successive Advanced Investigator Grants from the European Research Council [250363, OPAL and 670980, ELFBAD] and a Marie-Curie Intra-European Fellowship [255096, CFILP].
The Author declares that there are no competing interests associated with this manuscript.
I am indebted to a wonderful team of dedicated and brilliant researchers, mostly mentioned above, with whom I have shared many amazing, stimulating hours of discussion and who carried out all of the experiments.
Abbreviations: PG, peptidoglycan; PPLOs, pleuropneumonia-like organisms; ROS, reactive oxygen species.