Sculpting the Bacterial Cell

Abstract

Prokaryotes come in a wide variety of shapes, determined largely by natural selection, physical constraints, and patterns of cell growth and division. Because of their relative simplicity, bacterial cells are excellent models for how genes and proteins can directly determine morphology. Recent advances in cytological methods for bacteria have shown that distinct cytoskeletal filaments composed of actin and tubulin homologs are important for guiding growth patterns of the cell wall in bacteria, and that the glycan strands that constitute the wall are generally perpendicular to the direction of growth. This cytoskeleton-directed cell wall patterning is strikingly reminiscent of how plant cell wall growth is regulated by microtubules. In rod-shaped bacilli, helical cables of actin-like MreB protein stretch along the cell length and orchestrate elongation of the cell wall, whereas the tubulin-like FtsZ protein directs formation of the division septum and the resulting cell poles. The overlap and interplay between these two systems and the peptidoglycan-synthesizing enzymes they recruit are the major driving forces of cylindrical shapes. Round cocci, on the other hand, have lost their MreBcables and instead mustgrowmainly via their division septum, giving them their characteristic round or ovoid shapes. Other bacteria that lack MreB homologs or even cell walls usedistinct cytoskeletal systemsto maintain their distinct shapes. Here I review what is known about the mechanisms that determine the shape of prokaryotic cells.

Introduction

The control of cell shape is ultimately an epigenetic process, with molecules exerting their effects on various physical constraints. Bacteria like Escherichia coli offer an excellent opportunity to understand shape determination at the molecular level. The cells are relatively simple, the genomes and proteomes manageable in size, and there are many mutants in a single gene that give rise to altered shape. Like plant cells, most bacteria have walls, and the shape of their cells is largely governed by how the wall grows. For most bacteria, then, understanding how the wall grows will be crucial for obtaining a complete picture of cell shape determination. Because the cell wall or murein (also called the sacculus) is one large macromolecule, both the building and the turnover of this large structure help to determine cell shape. A number of recent reviews cover the biochemistry and cell biology of bacterial murein in detail [1-4].

A Great Diversity of Shapes

Despite usually being constrained by a cell wall, the shapes of bacteria are highly diverse, reflecting their large phylogenetic range. For example, of the relatively straightforward shapes, E. coli and Bacillus subtilis are straight rods; Vibrio cholerae is a curved rod; the Borrelia burgdorferi spirochete is a flat wave, an elongated and iterated version of the curved rod; Spiroplasma species are helix-shaped; Mycobacterium tuberculosis is a pleomorphic rod that often branches or swells at the poles; Staphylococcus aureus is a sphere, and Streptococcus pneumoniae is ovoid, with pointed polar caps. Then there are even more interesting ones, including Streptomyces, which form fungi-like mycelial mats and aerial hyphae; Caulobacter crescentus, which forms a curved rod like V. cholerae but with tapered ends and a polar stalk at one end. Archaea also have diverse shapes. While they generally have similar shapes and sizes as bacteria, either round or rod-like, one extreme thermophile (Haloquadratum walsbyi) forms flat squares. Some bacteria are tiny rods, such as Bdellovibrio bacteriovorus, which hunts and invades other bacteria and grows in their cytoplasm. Others are hundreds of microns long and over 20 microns wide, such as Epulopiscium fishelsoni. All this diversity does not even take into account the multicellular structures that many bacteria can form, such as the long chains of Anabaena or the flower-shaped fruiting bodies of Stigmatella aurantiaca. Even bacteria without walls, such as the mycoplasmas, have their own distinctive shapes organized by an internal cytoskeletal scaffold. More detailed information about these species and their shapes can be found in a recent comprehensive review [5].

Why have such morphological diversity? The answer is partly because of selective pressures and in part from physical constraints [6]. One important selective pressure is competition for nutrients, which can be optimized by increasing the surface area relative to total cell volume. The C. crescentus stalk is an extreme example of increasing surface area for nutrient foraging in nutrient-poor environments [7]. Other selective pressures on shape include motility and resistance to predation [8].

One key physical constraint is diffusion limitation. Without obvious cytoplasmic transport mechanisms, the scope of cell sizes for most bacteria is quite narrow, sufficiently small to accommodate diffusion-limited processes but large enough to house the necessary macromolecules, such as ribosomes and chromosomal DNA, to maintain and duplicate the cell. Another physical constraint is turgor pressure. Because of relatively low water activity in the cytoplasm relative to the outside environment, bacteria are essentially pressure vessels, and the cell wall must be oriented to provide sufficient strength to counteract turgor forces of several hundred kilopascals in Gram-negative bacteria and as high as 3 megapascals in Gram-positive bacteria [9,10]. This places limitations on the varieties of shapes bacterial cells can have. However, there are examples where turgor pressure is probably low, such as in extreme halophiles where the molar levels of salt inside the cell are similar to those outside [11]. This, along with a putative internal cytoskeleton, might explain why some species of halophilic archaea can assume triangular and square shapes with sharp corners that are not seen in other species.

For a long time, these selective and physical forces were thought to be the main determinants of bacterial shape, in part because bacteria were not supposed to have cytoskeletons. The genetic and cytoskeletal determinants of bacterial cell size and shape have only been appreciated relatively recently, largely because of the sequencing of multiple genomes and fluorescent protein tags for visualizing protein localization in living cells.

For simplicity, there are three main types of bacterial cell shape to consider: cylindrical, typified by E. coli or B. subtilis, but alsoincluding curved rods such as C. crescentus; ovococcal, typified by S. pneumoniae; and coccal, typified by S. aureus (Figure 1). Cylindrical cells growmainlyby extending the length of the cylinder, and new cell poles are synthesized during a relatively short time window at cell division. This shape has its advantages over a round shape, but requires more sophisticated controls. Ovococci and the cocci both grow mainly via their division septa, but ovococci undergo some length extension, whereas cocci grow exclusively via division septa [12]. As a result, growing cocci are usually present as diplococci because they must divide in order to grow, and their cell wall synthetic machinery localizes to the division septum. Because ovococci grow to some extent independently of their division septa, they need to switch growth modes and place new division septa at the cell midpoint, similar to rod shaped cells. If cell separation is inefficient, then these ovococci form cell chains which are typical for these species. Cocci, on the other hand, are spheroidal and do not form chains, because their division planes alternate in each generation. As a result, many cocci such as Deinococcus radiodurans or S. aureus formclusters or packets. In particular, Neisseria species divide in two alternating planes, whereas S. aureus divides in three alternating planes [12].

Figure 1.

Basic growth modes of six representative species.

Shown are three species that contain MreB (A) and three species that lack MreB (B), with modes of growth summarized below for each. For each species, the arrow refers to the transition between a newborn cell and one at the final stages of division prior to cell separation and new pole formation. For each stage, areas of the cell probably not engaged in significant peptidoglycan synthesis are outlined in blue; areas actively synthesizing peptidoglycan are outlined in other colours. Areas of MreB-dependent wall growth are shown in red or magenta; the magenta in C. crescentus indicates slower growth relative to the red, because of the inhibitory effects of crescentin. Areas of FtsZ-dependent wall growth are shown in green. Solid green outlines indicate septal wall synthesis, and green dots indicate probable locations of active FtsZ-directed sidewall synthesis preceding cell division. Areas of DivIVAdependent wall growth are shown in orange. In species with septal growth but no constriction, proper formation of the new pole requires splitting of the septum and turgordependent reshaping.

How to Build a Rod: Two Distinct Modes of Growth

Growth and division of rod shaped bacteria such as E. coli can be readily explained by sequential switching between two modes of growth. In newborn cells that have just divided, peptidoglycan is synthesized along the sidewall, resulting in the elongation of the cell to ultimately twice the length of the newborn cell. At the time of cell division, the synthesis apparatus switches from sidewall peptidoglycan synthesis to division septum synthesis. These two modes of growth probably compete with one another [13].

Work over many years has shown that peptidoglycan insertion is evenly distributed over many sites throughout the wall, and a number of models have been proposed to explain this type of insertion of new peptidoglycan [2]. Recent cryo-electron tomography and atomic force microscopy data support the model in which individual glycan strands are coiled together into larger fibers, and once inserted into the existing peptidoglycan structure (the sacculus) under turgor pressure, stretch to assume an orientation roughly perpendicular to the long axis of the cell [14]. In contrast, the peptide bridges connecting glycan strands are oriented parallel to the long axis. The end result, at least in B. subtilis, is that the glycan strands appear like a tightly packed array of parallel telephone cords [15].

One attractive model postulates that a complex of cell wall synthases and hydrolases inserts newpeptidoglycan strands between old strands that act as templates to allow long-axis extension without changing cell width [16]. However, such template-directed synthesis would seem to require an ordered lattice of glycan strands, whereas the evidence supports a more disordered pattern [14,17]. This suggests that existing cell wall cannot reliably provide a template for new wall synthesis, and that a cytoskeletal template is needed to maintain growth in the same pattern. Indeed, recent breakthrough work has demonstrated that cytoskeletal cables, not the chemical composition of the wall per se, spatially direct both sidewall and septal wall synthesis.

Bacterial Actin Cables Drive Cell Elongation

Rod-shaped E. coli cells can be converted to roughly spheroidal shape by treatment with beta-lactam antibiotics such as mecillinam, which inhibit PBP2, or by inactivating the function of several genes, including the genes for PBP2 or RodA [18,19]. These round cells expand and lyse under many growth conditions, but with slow growth or increased activity of the cell division protein FtsZ (see below), they can grow and divide asymmetrically via a cleavage-furrow like mechanism, forming heart-shaped dividing cells (Figure 2). They also can revert back to rod shape once the inactivation is relieved. Strikingly, round E. coli cells divide in alternating perpendicular planes like other cocci [20], perhaps because their division plane is specified by the asymmetric geometry of the parent cell [21]. Gram-positive rods such as B. subtilis also have a RodA protein that is required for proper cell shape [22]. Even an ovococcus such as Streptococcus thermophilus requires RodA to maintain ovococcal shape and prevent cells from becoming spherical [23].

Figure 2.

Roles for the actin (MreB) and tubulin (FtsZ) cytoskeletons in shaping E. coli cells.

Depletion (vertical triangles) of only MreB (A) or only FtsZ (B) during growth disrupts each cytoskeleton and causes cells to either slowly become spheroidal (A) or form cylindrical filaments (B) over time (red downward arrows). Predivisional cells with normal levels of MreB (red) and FtsZ (green) have a Z ring surrounded by MreB rings (top rows in A and B). After cell growth and division over time (rightward black arrows), FtsZ and MreB relocalize into helical patterns, and prepare to divide at the next division site (yellow dashes). During MreB depletion (A), cells will form Z arcs (green) and divide, often asymmetrically, at those arcs to form a diplococcus, which then over time (rightward black arrows) separates to form a single round cell capable of dividing again in a perpendicular plane (yellow dashed line). These MreB-lacking cells will continue to grow and divide as spheroids. In contrast, FtsZ-depleted cells (B) continue to grow their sidewall and elongate their MreB cables but fail to divide, thus forming long filaments. The MreB double helix is shown as a single helix for simplicity. Micrographs representing some of the steps are shown next to the appropriate diagram.

Another gene required to maintain rod shape is mreB [24,25]. MreB and RodA work together to help synthesize the glycan strands of peptidoglycan [26]. A number of observations on MreB together led to a breakthrough in our understanding of bacterial cell shape: first, that MreB is homologous to actin [27]; second, that MreB forms cables in a helical pattern that extended much of the length of the cell [28]; third, that these helical patterns are similar to the pattern of new peptidoglycan insertion, defined by labeling with fluorescent vancomycin [29]; and fourth, that MreB is conserved among many bacteria, but only those that are rod shaped.

These results suggested that MreB cables either direct the pattern of new wall growth or provide physical support to continuously maintain a cylindrical wall. Support for the first model came from studies with an MreB-specific drug, A22 [30]. When A22 is added to rod-shaped cells, it rapidly disrupts the MreB cables, yet the cell shape is unaltered for the short term [31]. Eventually, after additional time, the wall grows abnormally, creating round cells. The lack of cell shape change immediately after complete disruption of MreB cables strongly indicates that MreB cables do not provide an essential physical support but instead direct the pattern of new wall growth. This is remarkably analogous to microtubule-directed cellulose synthesis in plant cells [32].

How do the MreB cables specify the growth pattern? To answer this, we must understand how the MreB cable pattern itself is generated and what the proteins recruited by them do. In E. coli, MreB cables form a double helix with a wide pitch that extend the length of the cell during growth [33]. At the time of cell division, the cables then collapse to form two MreB rings that surround the cytokinetic Z ring (Figure 2; see also below). After cell division, these MreB rings lead to the assembly of new MreB cables in the daughter cells to continue the MreB pattern [34]. MreB undergoes similar redistribution in C. crescentus [35]. Because formation of the division septum switches cell wall growth from an axial to centripetal mode, the redistribution pattern is consistent with MreB being needed to trigger new peptidoglycan synthesis at the required time and place. In support of this idea, MurG, which provides lipid II precursors in the last cytoplasmic step in peptidoglycan synthesis, localizes to both the sidewall and division septum in E. coli and thus is part of both sidewall and septum synthesis machines [36].

Unlike most bacteria, B. subtilis has three MreB paralogs. One of these, called Mbl for ‘MreB-like’, is essential for cylindrical growth and forms a long double helix similar to that formed by E. coli MreB that colocalizes with MreB helices [37]. MreB interacts with Mbl but forms shorter helices. MreC interacts with Mbl but not MreB, consistent with Mbl function being analogous to E. coli MreB [37]. Finally, MreBH also forms helical cables and recruits a cell wall hydrolase, LytE, to those cables, suggesting that MreBH specializes in regulating cell wall turnover [38]. Mutations in any of the paralogs lead to shape abnormalities, indicating that all three contribute to B. subtilis rod shape.

The prediction, then, is that cables formed by MreB homologs (including Mbl cables of B. subtilis) directly recruit cell wall synthetic enzymes, thus ensuring that cell wall synthesis occurs in a helical pattern. This appears to be the case in B. subtilis, where MreB cables interact directly with PBP1 and several other peptidoglycan synthetic enzymes [39]. However, other interactions are probably indirect. The mreB gene usually is in an operon with two other conserved genes, mreC and mreD. In C. crescentus, MreC, an integral membrane protein, interacts with MreB and peptidoglycan synthetic enzymes and forms helical patterns in the membrane [40-43]. Interestingly, at any given time, the MreB and MreC helices do not colocalize, but instead one helix alternates with the other, with PBP2 colocalizing mostly with the MreC helix [35]. This suggests that whereas MreB cables continuously move by treadmilling [44-46] to specify where the next wall needs to be synthesized, MreC helices may remain relatively fixed at the current site of wall synthesis, bound to peptidoglycan synthetic complexes, and lag behind MreB helices. Less is known about MreD, but it has been shown to interact with MreC [47,48]. Importantly, inhibition of MreB cable formation does not prevent cell wall synthesis, as rod-shaped cells expand their walls uniformly to become round under these conditions. This indicates that the cell wall synthetic complexes are always active to some degree, and need to be constantly constrained in space by the cytoskeleton in order to inhibit the default state of uniform wall expansion. It is not yet known how peptidoglycan synthetic complexes move, but the models for peptidoglycan insertion and replacement suggest that the complexes are probably continuously assembled and disassembled, and not processive like plant cellulose synthases are thought to be. Fluorescence photobleaching analysis of PBP3 in C. crescentus shows a high turnover rate, consistent with this idea [49].

If the MreB cable localization pattern directs the pattern of wall growth, then what directs the MreB cable pattern? In somewhat of a surprise, it turns out that inactivating the transmembrane proteins MreC, MreD or RodA can perturb the cytoplasmic MreB helix. This suggests that the localization of at least some of these proteins is interdependent and that the integrity of the transmembrane proteins stabilise MreB cables [47,50]. As MreB lacks an obvious membrane-anchoring domain, the other proteins may at the very least help to tether MreB cables to the membrane. Many membrane-associated proteins localize in a helical pattern [51-54], and even lipids localize in such patterns [55]. This indicates that the tendency of protein polymers such as FtsZ or MreB to form helices inside cells [56] cannot be the sole explanation for the patterns.

The helical pattern of labeling by vancomycin or ramoplanin, which both label nascent glycan strands [57], indicates that synthesis of new cell wall is helical, supporting the original prescient model of Mendelson [58]. As a result, insertion of new cell wall may result in a multiple-start helix, such that everything that localizes to the membrane, even the membrane itself, is influenced by the underlying helical architecture. In support of this idea, even outer membrane proteins and lipopolysaccharide of E. coli localize in a helix [59,60]. If the helical ‘grooves’ are relatively imprecise, which is consistent with the direct imaging of the peptidoglycan strands described above, then the patterns will localize with different pitches.

Protein factors also regulate MreB assembly. For example, the tubulin homolog FtsZ (see below) is required for the collapse of MreB cables from extended helices to rings near the future division site [34,41], although the mechanism for this FtsZ-dependent relocalization is not known. Recently, another conserved bacterial shape protein called RodZ was discovered which interacts with MreB and is crucial for proper assembly of MreB cables [61-63]. Like MreB, RodZ usually localizes as a helix, and its localization pattern requires MreB, indicating that RodZ and MreB localization are dependent on each other. As RodZ crosses the membrane once, its cytoplasmic domain probably contacts MreB, whereas its extracytoplasmic domain may contact peptidoglycan synthesis enzymes. Inactivation of RodZ in C. crescentus or E. coli results in huge, misshapen cells. Overproduction of RodZ causes similar shape abnormalities, indicating that RodZ levels need to be in an optimal range. Overproduction of MreB also leads to a dominant-negative effect on cell shape, but overproducing both MreB and RodZ mostly restores normal rod shape, suggesting that the MreB:RodZ ratio is crucial. Future studies will be needed to understand the mechanism by which RodZ regulates MreB cable assembly and cell shape.

The bacterial tubulin cytoskeleton coordinates growth, cell division, and cell size. FtsZ, the bacterial tubulin homolog [64], assembles into a cytokinetic ring called the Z ring at the site of cell division [65,66]. The Z ring is used for cytokinesis in most bacteria, although some families such as the Planctomycetes and Chlamydia lack FtsZ and therefore must divide using another as yet unknown mechanism [67]. Although FtsZ primarily localizes to the division septum, when a cylindrical cell such as E. coli or B. subtilis is not dividing or preparing to divide, FtsZ localizes as a dynamic coil in the cytoplasm [68,69] (Figure 2). In C. crescentus, FtsZ relocalizes from a focus at one cell pole to midcell upon chromosome segregation [70]. Just as MreB relocalizes from sidewall to septum depending on the temporal and spatial requirements for peptidoglycan-synthesis, FtsZ relocalization may serve the same function. In support of this, FtsZ participates in both septal synthesis and a subset of sidewall synthesis by recruiting peptidoglycan-synthesis machinery via PBP 2 [71-73] (Figure 1). However, FtsZ cannot direct cylindrical growth without MreB, probably because only MreB can properly organize the growth pattern. Interestingly, both FtsZ and MreB help to distribute the peptidoglycan synthetic machinery evenly, because E. coli or S. aureus mutants that lack both proteins synthesize their cell wall in large patches [74,75]. Therefore, both actin and tubulin cables are important for tethering the peptidoglycan synthetic machinery to ensure a properly dispersed, nonrandom distribution of new wall material. In cocci that lack MreB and grow by their division septum, FtsZ is likely the sole means of recruitment and tethering of the peptidoglycan-synthetic machinery.

FtsZ is important for bacterial shape in two other respects. The first is that its orchestration of septal wall synthesis determines the shape of the cell poles (Figure 1). Cell poles of C. crescentus are tapered because, as mentioned above, FtsZ contributes significantly to cell wall extension near the division site, resulting in mostly constriction instead of septum formation in this species. In contrast, B. subtilis divides mainly by septation without constriction, resulting in its typically squared off poles. As B. subtilis is Gram-positive and thus lacks an outer membrane, there is no need for coordinated constriction. E. coli has round poles because it divides with a combination of septum formation and constriction of inner and outer membranes, which are all coordinated [76]. Once poles are formed during cell division, they generally do not change their shape, and their peptidoglycan is inert (see below). Polar morphology in E. coli can be altered, but usually only by mutations in FtsZ or FtsZ-dependent peptidoglycan-modifying enzymes that perturb the geometry of the septum [77-80].

FtsZ can also control cell size, an important component of shape. For example, E. coli increases cell size — mostly via increased length — as nutrients become more plentiful and growth rate increases [81]. Cell division also requires cells to attain a threshold length. However, even in rich medium, mutants of E. coli or B. subtilis have been discovered that make cells ~20% shorter than normal at the same growth rate. The B. subtilis mutants are in UgtP, a component of a glucolipid pathway for cell wall biosynthesis [82]. In addition to its enzyme activity, UgtP negatively regulates FtsZ assembly, resulting in delayed Z ring formation and larger cells, but only in rich medium. In poor growth medium, cells are small whether or not they have UgtP. But in rich medium, lack of UgtP results in the assembly of more Z rings per cell mass, and shorter than normal cells. Therefore, this pathway uncovers the first known link between bacterial metabolic activity and cell size. In E. coli, a gain-of-function mutation in the actin-like cell division protein FtsA also results in ~20% shorter cells in rich medium, by increasing FtsA–FtsZ interactions and generally enhancing the integrity of the Z ring [83], although no metabolic link has yet been established in this case. In conclusion, regulation of Z ring activity by nutritional and other inputs can determine the cell length at division, which has an obvious impact on cell shape.

Caulobacter crescentus as a Model for Bacterial Shape Control

C. crescentus is genetically tractable and has an obvious asymmetric shape: it is a curved rod with tapered poles, and predivisional cells have a flagellum at one pole and a thin extension of the cell body, called a stalk, at the opposite pole. These properties make C. crescentus an excellent model system to study bacterial cell shape and polarity. A knockout of a gene called creS, encoding crescentin, converts the normally curved rods to straight rods that still have polar stalks and flagella [84]. Although not obvious from its primary sequence, crescentin has a similar domain structure to that of eukaryotic intermediate filaments, and indeed purified crescentin can spontaneously assemble into filaments without nucleotide. A single crescentin filament normally localizes to the membrane at the inner curve of C. crescentus cells. Because a crescentin filament can initiate assembly of a new filament on any part of the cell membrane, its inner curve location is the result of its action, not the cause. The available evidence suggests that a crescentin filament constrains wall growth only on the side of the cell where it is present. Crescentin filaments are stretched, because release from the membrane results in contraction of the filament into a helix.

These results suggest that the stretched crescentin filament may locally inhibit new wall synthesis by locally decreasing strain in the peptide crosslinks of the peptidoglycan, resulting in increased growth of the wall on the opposite side and ultimately curving the cell [85]. In support of this model, exogenously produced crescentin in E. coli induces severe coiling of the cells. The model also predicts that other proteins that localize to the membrane, form polymers, and interact with MreB cables might also be able to distort cell shape. Interestingly, altered expression or mutants of MinE, MreB, FtsZ and FtsA can all cause similar dramatic coiling of E. coli cells under certain conditions [86-89], although there is no evidence so far of direct interaction between MreB and these other proteins. Moreover, when C. crescentus cells are forced to assume unnatural shapes by growing them in a microfabricated chamber, they maintain the shapes when removed [85]. This is consistent with earlier findings suggesting that the shape of the bacterial sacculus is determined by its growth pattern and remains rigid [90,91].

To test if crescentin might interact with MreB, crescentin was exogenously produced in Agrobacterium tumefaciens, which lacks MreB but is rod-shaped. The presence of crescentin had no effect on shape [92]. This and other evidence demonstrates that interaction of crescentin with the cell membrane requires that it bind to MreB cables; if MreB cables are perturbed, crescentin filaments become dislodged from the membrane and lose their shape-changing properties. The potentiation of crescentin function by MreB cables is similar to the organization of intermediate filaments in eukaryotic cells by actin and microtubules [93].

Although difficult to find by amino-acid similarity alone, bacterial intermediate-filament-like proteins are widespread as judged by the presence of central segmented coiled coil domains and the ability to assemble spontaneously into filaments [94]. One such homolog is FilP of Streptomyces coelicolor, which also forms intracellular cytoskeletal filaments. In the absence of FilP, cell growth and morphology were abnormal, and the hyphae became susceptible to mechanical breakage [94]. Therefore, like eukaryotic intermediate filament proteins, at least one bacterial homolog seems to impart mechanical strength to the cell, although the mechanism is unclear. Another, the cytoplasmic filament protein of Treponema spirochetes, spans the entire cell length and seems to be important for cell division [95].

The C. crescentus stalk is also of interest because it acts as a nutrient antenna and can grow considerably longer than the cell body under certain conditions. Inactivation of PBP2, or depletion of MreB or RodA, results in a stalk elongation defect [96]. These results are consistent with the idea that the stalk is a very narrow continuation of the cell body, and that MreB cables most likely act indirectly, possibly via PBP2, at the cell pole to initiate and maintain this small diameter (about 5 times smaller than the cell itself). FtsZ is also required for formation of cross-band structures in the stalk [43].

Phylogeny of Shape

It is likely that rod-shaped bacteria arose first in evolution, and cocci have evolved multiple times from them by losing crucial genes, including those for the cytoskeleton [97,98]. Phylogenetic comparisons are completely consistent with the idea that cocci are the default shape of bacterial cells that have lost the MreB cytoskeleton; MreB is almost never found in cocci. As MreB probably acts to shape the cell wall via MreC, it comes as no surprise that essentially all species that have MreB also have MreC (exceptions include Wolbachia endosymbionts of insects and Thermotoga maritima, which have MreB but no MreC) [61]. MreD is often present in bacteria with MreB and MreC, but is absent in others such as the γ-proteobacteria. RodZ is usually only present in species that have MreB, which is consistent with its role as an MreB assembly factor; however, RodZ is somewhat less conserved than that of MreB, with some species containing MreB but lacking RodZ [61]. The distribution of some cell shape proteins in several divergent species is shown in Figure 3.

Figure 3.

Phylogeny of some shape-determining proteins across representative bacteria.

The species are grouped by family. For each species, typical cell shape is shown (not drawn to scale), along with the presence or absence of the protein as encoded in the genomic sequence. The presence or absence of MreB, MreC, MreD or RodZ in different species as identified by STRING COG was described in [61]. STRING COG was also used to identify DivIVA homologs. DivIVA is absent in all the Gram-negative species listed, but some members of the Gram-negative δ-proteobacteria contain DivIVA orthologs.

There are a couple of surprises from the phylogeny of shape proteins. One is that MreC and MreD are sometimes present in species with no MreB, such as Gram-positive cocci. This suggests that MreB was specifically lost in these species, but that MreC and MreD still retain a role in cell wall synthesis. The other surprise is that a fairly large number of rod-shaped species lack MreB, along with MreC and MreD. Clues are now emerging about how these bacteria maintain their rod shape without MreB cables.

Cell Poles and Cytoskeletal Control of Tip Growth

Some bacteria form branches, grow at their cell poles without the benefit of MreB cables, or both. One example of an MreB-containing species that branches extensively is the filamentous fungus-like Streptomyces, which grow vegetatively as a mycelial filamentous mat and branch extensively [99]. During starvation, Streptomyces then send up aerial hyphal branches to make spores. The branching process requires that new cell poles be initiated along the sidewall of an existing filament, which is not observed in most rodshaped bacteria. The apparent trigger of new pole formation is DivIVA, a protein found in many Gram-positive bacteria [100]. In Streptomyces, DivIVA forms foci along hyphal side-walls at incipient branches, and overproduction of DivIVA can force new zones of cell wall to be assembled [101] (Figure 4A). The importance of DivIVA for growth in Streptomyces is underscored by the dispensability of MreB for cell wall extension in these species, although MreB is essential for spore formation [102].

Figure 4.

Cytoskeletons trigger new shapes in bacteria.

Shown are two examples of new shapes resulting from bacterial cytoskeletal proteins. (A) In Streptomyces, DivIVA (orange balls) forms large assemblies at the poles of hyphae, possibly recognizing a region of sharp membrane curvature, but also forms seemingly random foci along the sidewall. As DivIVA tends to self-assemble, larger foci are easily generated (middle panel). These larger assemblages probably trigger branch formation, because a DivIVA focus usually forms under a new pole prior to visible branch initiation. A DivIVA lattice is subsequently maintained at the new branch tip (green) and is required for continued tip extension. (B) Cells of Mycoplasma pneumoniae lack peptidoglycan, and would be spheroidal without a cytoskeleton (top). Cytoskeletal proteins form the terminal organelle, which is important for gliding motility (red rightward arrow); the resulting cell extension (red) upon movement and subsequent duplication of the organelle at the opposite pole (not shown) gives these wall-less cells a roughly cylindrical shape with narrow tips.

Another related species, Corynebacterium glutamicum, is roughly rod-shaped and yet lacks MreB, MreC and MreD. How can it maintain a rod shape without MreB cables? Like Streptomyces, C. glutamicum uses polar targeting of DivIVA to drive tip growth, which is mediated by PBP1a, PBP1b, and RodA [103]. Staining of C. glutamicum cells with vancomycin showed that young cells stained at the poles while older cells stained at the septum [29]. This is consistent with the idea that DivIVA initiates tip growth, then switches growth to the division septum when cells reach a critical length (Figure 1). This is analogous to the switch from sidewall to septal growth in MreB-containing bacteria. In support of this model, depletion of DivIVA in C. glutamicum results in nearly round shape, whereas overproduction of DivIVA results in swollen poles, indicative of locally increased peptidoglycan synthesis at polar sites of DivIVA localization [104]. The closely related Mycobacterum smegmatis also needs DivIVA to grow as pleomorphic rods [105]. In the unrelated S. pneumoniae, which also lacks MreB but is ovoid, DivIVA localizes at the major site of peptidoglycan synthesis — the division septum — and probably is involved in septal peptidoglycan synthesis [106].

The role of DivIVA as a local organizer of tip growth is reminiscent of the role and localization of the apical body (Spitzenkörper) of filamentous fungi. This function is not as farfetched as it would seem, because DivIVA has properties of a cytoskeletal element. The carboxy-terminal coiled-coil domain of DivIVA has some homology with tropomyosins, which, intriguingly, localize preferentially to fungal Spitzenkörper [107]. Moreover, purified DivIVA protein forms ‘doggy bone’ structures in vitro that probably translate into a membrane-associated lattice in vivo [108]. How this putative DivIVA lattice recruits cell-wall synthesizing enzymes is not yet known, given that no MreC and MreD are present in C. glutamicum and related species. However, these species may use protein kinases to regulate the activity of peptidoglycan synthesizing enzymes [109].

How does DivIVA localize to the cell poles and division septum? Recent evidence suggests that DivIVA recognizes regions of high negative membrane curvature [110]. This is supported by its ability to localize along arcs at cell poles of other species that normally lack DivIVA, such as E. coli and even fission yeast [111]. The DivIVA foci in Streptomyces filaments that trigger formation of a new pole may initially form randomly, and then form a lattice at the higher curvature once the branch has initiated. Alternatively, the foci may trigger local membrane curvature that then stabilizes DivIVA assembly at the curve. This is a potential example of shape inducing and reinforcing a distinct protein localization pattern, basically the converse of the concept of cytoskeletal localization driving growth.

Another example of shape inducing localization in bacteria is the recent discovery that a 26-amino acid B. subtilis membrane-associated protein, SpoVM, localizes specifically to regions of high positive curvature [112], possibly by forming a lattice structure. Positive membrane curvature inside bacterial cells is rare, but is a distinct feature of the outer surface of the developing endospore, where SpoVM localizes and functions in building the spore coat. As mentioned earlier, the future division plane of E. coli may depend on cell shape cues [21]. We will probably soon be learning about more proteins that use similar shape cues for their localization and/or activity, which might help to reinforce and maintain certain types of shapes.

Many species of the a-proteobacteria are rod-shaped, and yet lack MreB, MreC, MreD, RodZ and DivIVA. An unrelated γ-proteobacterium and the cause of tularemia, Francisella tularensis, also lacks these proteins and yet is rod-shaped [61]. Because vancomycin does not penetrate the outer membrane of these Gram-negative cells to be able to bind to its periplasmic target, we do not yet know if they grow by tip extension like corynebacteria. However, some α-proteobacteria lacking the known shape proteins above share a strong tendency with corynebacteria and mycobacteria to form branches instead of filaments when their cell division machinery is perturbed [113-116]. This suggests that cells of these species also expand by tip growth instead of extension along the cell cylinder, and inhibiting cell division forces new poles to accommodate the mass increase. It is likely that these α-proteobacterial species have a DivIVA-like cytoskeletal protein that localizes to cell poles and organizes a putative peptidoglycan-synthesis complex there. Curiously, C. crescentus, which is also an α-proteobacterium but has MreB, synthesizes a protein, TipN, that is conserved in some of the above species. TipN localizes to the division site and persists in the daughter cells at the newly formed cell pole [117,118]. When overproduced in C. crescentus, TipN triggers ectopicpoles to form, similar to the phenotype during recovery from MreB depletion [96], suggesting that TipN may be involved in tip extension in the related MreB-lacking species.

What might be the molecular mechanism for initiation of branches in species that lack a DivIVA-like organizer? Recent work on E. coli offers some hints. The sidewall of E. coli is constantly growing, indicating that peptidoglycan turnover is high. This peptidoglycan is the flexible skin of the cell that can change its shape in response to turgor pressure thanks to autolytic enzymes and cytoskeletal proteins described above. In contrast, the peptidoglycan at the division septum and resulting cell poles exhibits little if any turnover [119], making it relatively inert. This property may be a result of different orientation of the peptidoglycan strands after splitting of the division septum, which discourages autolytic cleavage [10]. The cell wall is therefore a mosaic of all new and all old domains [71,120].

The current model proposes that branches (new cell poles) occur in the sidewall because of a patch of inert peptidoglycan, and that MreB, RodA, PBP2 and other components of the wall elongation machinery normally prevent unwanted patches of inert peptidoglycan from forming [121]. In E. coli lacking the PBP5 carboxypeptidase, which trims side chains from the peptidoglycan and keeps peptidoglycan from becoming inert, cells can assume a variety of bizarre shapes, including multiple branches, which are consistent with growth around the inert peptidoglycan, muchlike a tree grows around a blocking object [122]. The lack of wall turnover at the cell pole may create a more static environment for many membrane proteins to localize specifically to cell poles [123]. Examples of proteins that may localize to cell poles by this mechanism include bacterial chemoreceptors [124] and ActA/IcsA, which are required to nucleate actin tails at one cell pole of Listeria or Shigella species to trigger mobility inside the eukaryotic host and evade host defenses [125,126].

Thinking outside the (Peptidoglycan) Box

What happens in prokaryotes without typical peptidoglycan walls? Membranes by themselves usually assume spheroidal shapes unless they are organized by a cytoskeleton. Nevertheless, there are several examples of cells without walls of any kind that have some semblance of shape. The most notable are the mycoplasmas, which completely lack cell walls, and yet often assume non-spheroidal shapes because of a unique internal cytoskeleton visible by ultrastructural analysis [127,128]. Mycoplasma pneumoniae, for instance, uses cytoskeletal proteins to assemble new terminal organelles to extend its tips when it is gliding along surfaces, and assumes a rod-like cell shape [129,130] (Figure 4B). Therefore, an internal cytoskeleton is required for motility, resulting in a distinctive cell shape. Another wall-less mycoplasma, Spiroplasma citri, is helical in shape and probably also uses a motor to determine its shape. A flat cytoskeletal ribbon, composed mainly of a protein called Fib, runs the shortest distance down the entire cell length [131,132]. Interestingly, S. citri has multiple homologs of MreB, and MreB is one of the components of the ribbon, suggesting that the actin cytoskeleton helps to mold these wallless cells into a helix.

Bacteria such as E. coli or B. subtilis can be forced to lose their wall under selective growth conditions, and these socalled L-forms can grow and divide slowly as spheroids but are often unstable. Very recently, a stable cell line of B. subtilis L-forms was isolated that proliferate, albeit inefficiently, even in the absence of FtsZ, by blebbing off membrane extrusions which contain chromosomal DNA [133]. It is reasonable to speculate that this blebbing process requires a cytoskeleton to provide the force, and because B. subtilis biology is so well established, these cells should be useful in the future in dissecting the minimal requirements for cell division and shape.

Finally, archaea lack typical peptidoglycan walls and most instead have surface layers, or S-layers, consisting of a protein lattice on the outside of the cell [134]. Some bacteria, such as C. crescentus, also have S-layers. One possibility is that S-layers comprise an exoskeleton that determines shape of archaeal cells. However, species with different shapes, such as the spheroidal Haloferax volcanii and the rod-shaped Halobacterium salinarum, have similar S-layer components, suggesting that S-layers do not themselves direct cell shape [135]. This is probably analogous to bacterial peptidoglycan, which passively maintains the shape determined actively by the cytoskeleton [134]. Many putative cytoskeletal proteins are present in archaea, including actin homologs present in some wall-less archaea [136]. It is likely that these and other cytoskeletal proteins are crucial for cell shape determination in this kingdom, although the difficulty in working with archaea will make progress slower.

Conclusions and Outlook

It is clear that the cytoskeleton of bacteria actively organizes cell shape. In walled bacteria, the patterns of synthesis and degradation of the cell wall ultimately determine cell architecture when under turgor pressure. In bacteria lacking walls, other cytoskeletal structures can dynamically change the shape of membranes, much like in protozoan or animal cells. In turn, changes in shape can provide feedback to the bacterial cytoskeleton, and localized mechanical stresses may as well, as they do in plant cells [137]. Actin and tubulin homologs are prominent in the bacterial cytoskeleton, but other proteins such as RodZ, DivIVA, and intermediate filament homologs are emerging as important supporting players. Although we know many of the players and how some of them interact, the mechanisms by which cytoskeletal components dictate cell shape still need to be worked out. While it is generally true that cell architecture is inherited [138], it is also true that rod shaped cells can emerge from round cells or from round dormant spores, which must depend on de novo assembly of cytoskeletal structures. Membrane lipid biosynthesis also must be regulated to match growth of the cytoplasm and wall, and may respond to feedback mechanisms [139,140]. It will require much more work to understand the molecular mechanisms for these processes or how to explain the diversity of shapes. At present, we can see only the tip of the cylinder.