Cytoskeletal elements critically contribute to subcellular organization in all domains of life, acting as dynamic tracks that guide transport processes, as platforms for the assembly of proteins complexes, or as scaffolds that regulate and stabilize cell shape. The prototypical cytoskeletal proteins identified in eukaryotes are tubulins, actins, and intermediate filament (IF) proteins. Members of the first two groups assemble into highly dynamic polymers whose turnover is regulated by nucleotide binding and hydrolysis [Desai and Mitchison, 1997; Pollard et al., 2000]. IFs, by contrast, comprise rigid scaffolds that form spontaneously in a nucleotide-independent manner [Herrmann et al., 2007]. While the composition and regulation of the eukaryotic cytoskeleton has been studied intensively for years, the role of polymeric protein networks in the function of bacterial cells has only recently been appreciated.
The first evidence for the existence of bacterial cytoskeletons came from the discovery that the highly conserved cell division protein FtsZ assembles into dynamic filaments that coalesce into a ring-like structure at the division site [Bi and Lutkenhaus, 1991; de Boer et al., 1992]. Subsequent crystallographic studies revealed that FtsZ, despite the lack of conspicuous sequence similarity, is a structural homolog of tubulin [Löwe and Amos, 1998]. Similarly, the morphogenetic protein MreB was found to share striking structural similarity to actin, forming membrane-associated polymers that govern the activity of cell wall biosynthetic enzymes [Jones et al., 2001; van den Ent et al., 2001]. Meanwhile, the repertoire of bacterial tubulin and actin homologs has expanded significantly. Interestingly, most of the new additions are encoded on phages or low-copy number plasmids, acting as the driving components of DNA segregation systems [Gerdes et al., 2010]. Although all of these proteins consistently require nucleotides for polymerization, their filament architectures, assembly dynamics, and modes of regulation are highly diverse. This variability in both function and biochemical properties indicates a long evolutionary history of tubulins and actins in bacteria, lending support to the view that modern eukaryotic cytoskeletons could have evolved from bacterial predecessors [Wickstead and Gull, 2011]. This hypothesis is further corroborated by the recent identification of a bacterial protein, crescentin, that shows the typical domain organization and polymerization properties of eukaryotic IF proteins [Ausmees et al., 2003]. Thus, all major classes of eukaroytic cytoskeletal elements are also present in bacteria.
In addition to the canonical cytoskeletons, bacteria contain a variety of polymer-forming proteins that are generally not found in eukaryotic cells. Among them are the ParA-like Walker ATPases, a widespread group of proteins that use ATP binding and hydrolysis to form dynamic structures driving the partitioning of DNA and protein complexes during cell division [Gerdes et al., 2010]. Apart from that, there are a variety of bacteria-specific cytoskeletal proteins that polymerize in a nucleotide-independent manner. Many of them have been identified only recently, including coiled-coil-rich proteins (CCRPs), the polar scaffolding proteins DivIVA and PopZ, and bactofilins. Despite the lack of evolutionary relationship, they all assemble into stable macromolecular scaffolds, albeit with distinct polymerization dynamics, subcellular localization patterns, and biological functions.
The biology of bacterial actin and tubulin homologs and of ParA-like ATPases has recently been covered by several excellent reviews [Michie and Lowe, 2006; Graumann, 2007; Pogliano, 2008; Cabeen and Jacobs-Wagner, 2010; Gerdes et al., 2010; Ingerson-Mahar and Gitai, 2012]. This article will specifically focus on nucleotide-independent cytoskeletal elements in bacteria, including the IF protein crescentin as well as the aforementioned bacteria-specific representatives of this group (Table I). We will give an overview of the different protein families, summarize the current knowledge of their structure and polymerization properties, and outline their functional significance in the cell.
Intermediate filaments represent a heterogeneous group of cytoskeletal elements that are abundant in metazoan cells [Godsel et al., 2008]. They form extensive polymer networks in many distinct cellular locations, including the nucleus, the interior of the cytoplasm, and the inner face of the plasma membrane. Different IF networks are interconnected and closely associated with other cellular structures, thereby organizing cytoskeletal function and cell architecture [Herrmann et al., 2009; Starr and Fridolfsson, 2010]. Although existing in large variety, IF proteins share a characteristic domain organization (Fig. 1A). Their middle part is composed of four α-helical coiled-coil domains (designated 1A, 1B, 2A, and 2B) that are separated by short linker sequences, with the fourth domain interrupted by a conserved phasing discontinuity (stutter). This structurally conserved core is flanked by an N-terminal head domain and a C-terminal tail domain, whose sequences are highly variable across different IF protein families [Herrmann and Aebi, 2004]. During polymerization, the coiled-coil regions of two polypeptide chains assemble into a rigid, rod-like structure, giving rise to dumbbell-shaped dimers. These intermediates then further polymerize into stable, yet elastic fibres that finally interact to form higher-order polymeric meshworks. Despite their nucleotide-independent assembly and intrinsic stability, IFs can rapidly reorganize in vivo. These dynamics are regulated by various interacting proteins that control the nucleation, organization, and turnover of IF networks as well as by post-translational modification of individual IF subunits, in particular, phosphorylation and O-linked glycosylation of residues in the head and tail domains [Omary et al., 2006; Godsel et al., 2008].
Recently, a large family of polymerizing bacterial proteins has been identified as IF-like, based (i) on a high content of coiled-coil regions flanked by variable terminal domains and (ii) the ability to self-assemble into stable filaments in the absence of nucleotide cofactors [Ausmees et al., 2003; Bagchi et al., 2008; Fiuza et al., 2010; Walshaw et al., 2010]. Most of these IF-like elements, also known as CCRPs, were shown to contribute to bacterial morphogenesis by mechanically regulating or stabilizing cell shape, reminiscent of the morphogenetic role of many IF proteins in eukaryotic cells.
The IF-Like Protein Crescentin
The first IF-like protein to be identified and characterized in bacteria was crescentin from the stalked Gram-negative species Caulobacter crescentus [Ausmees et al., 2003]. Crescentin was found to be required for establishing the characteristic curved morphology of this organism, [Ausmees et al., 2003; Cabeen et al., 2009], with deletion of its gene (creS) leading to complete straightening of the cells. Both in vivo and in vitro evidence supports the hypothesis that crescentin indeed is a bona fide IF protein. It exhibits the typical composition of four coiled-coil domains separated by three short linker regions and possesses a stutter interrupting the fourth domain (Fig. 1B). When localized within the cell (Fig. 2A, right panel), crescentin forms a filamentous structure at the inner cell curvature lining the inner face of the cytoplasmic membrane [Ausmees et al., 2003]. However, after detachment from the membrane, the filament collapses and assumes a helical shape, suggesting that it functions as an elastic spring that exerts a mechanical force on the cell envelope [Cabeen et al., 2009]. Based on this observation, it was proposed that crescentin affects cell curvature mechanically by producing a localized strain on the peptidoglycan cell wall that, in turn, leads to differential peptidoglycan biosynthesis and, thus, uneven longitudinal growth (Fig. 2B) [Cabeen et al., 2009].
In line with a mechanical role, crescentin polymers do not show any appreciable turnover in vivo [Charbon et al., 2009; Esue et al., 2010]. They appear to lack any overall polarity, as new subunits are added along the entire length of the filament. While increasing in size, the polymer grows bidirectionally. Its extension is a slow process that occurs gradually over the course of the cell cycle. The mechanism that coordinates crescentin polymerization with cell elongation is still unclear. However, crescentin assembly was shown to cease in regions of extreme cell curvature [Charbon et al., 2009]. Thus, polymerization may continue until the tips of the filament touch the cell poles. Further elongation would induce the formation of U-turns, leading to a distortion in the structure of the polymer that makes the addition of new subunits energetically unfavorable. During cytokinesis, the crescentin filament is locally disassembled at the cell division site, allowing its equal distribution to the daughter cells. This splitting process fails when crescentin is detached from the cell envelope, suggesting the existence of a specific, membrane (divisome)-associated machinery that locally disrupts the crescentin filament during cell constriction [Cabeen et al., 2010].
Consistent with its IF-like behavior in vivo, crescentin shares many biochemical properties of IF proteins. Most importantly, it can self-assemble into stable filaments (Fig. 2A, left panel) in the absence of nucleotides or any other cofactors [Ausmees et al., 2003; Esue et al., 2010], with its polymerization being stimulated by divalent cations [Esue et al., 2010]. Moreover, the mechanical properties of these filaments, as determined by quantitative rheology, are similar to those of IFs: they show similar phase angles (i.e., flow under mechanical stress) and assemble into solid-like networks that are elastic and can recover a large part of their elasticity after shear [Esue et al., 2010].
The precise mechanism underlying the morphogenetic activity of crescentin is still incompletely understood. In particular, it remains to be clarified how the cytoplasmic crescentin filament can stably attach to the cell envelope such as to exert a mechanical force on the periplasmic cell wall. A possible link between the two structures is the actin-like cytoskeletal protein MreB, which forms membrane-associated polymers that interact, directly or indirectly, with components of the cell wall biosynthetic machinery. In support of this idea, crescentin is detached from the cell envelope upon treatment of cells with the MreB inhibitor A22 and copurifies with MreB in coimmunoprecipitation experiments [Charbon et al., 2009]. Moreover, it is able to induce cell curvature even when expressed heterologously in Escherichia coli, indicating its interaction with a generic cellular component, such as MreB. Thus, the association between the actin and IF cytoskeletons observed in eukaroytes [Chang and Goldman, 2004] may also be conserved in bacteria. However, detachment of crescentin was also observed upon antibiotic-mediated inhibition of cell wall biosynthetic enzymes [Cabeen et al., 2009]. The requirement for MreB could, therefore, also be more indirect and rely on its regulatory role in cell wall formation.
Mutational analysis revealed that the positively charged N-terminal region of crescentin (aa 1–27; Fig. 1B) is required for cell envelope attachment [Cabeen et al., 2009], potentially facilitating the alignment of crescentin filaments with the negatively charged phospholipid membrane. The rest of the head domain (aa 28–79) and the coiled-coil-rich rod domain, by contrast, have a key role in filament assembly, with their deletion resulting in complete delocalization of crescentin in vivo [Cabeen et al., 2011]. Removal of linker region L1 still allows for polymerization, but the polymers are short and aggregate into a single polar focus in the cell. A crescentin derivative lacking the stutter, on the other hand, was only able to form short filamentous patches in vivo, while assembling into massive paracrystalline polymers in vitro, suggesting a role of the stutter in the regulation of polymer bundling [Cabeen et al., 2011]. The C-terminal tail of crescentin only has a minor role in filament assembly, with its deletion resulting in both wild-type and spotty localization patterns. Overall, the domain organization of crescentin and the functions of the different domains are highly reminiscent of eukaryotic IF proteins [Herrmann and Aebi, 2004]. Crescentin thus appears to be a genuine IF protein, presenting a readily amenable bacterial model for the study of IF assembly and function. This finding raises the question of whether crescentin is derived from a eukaroytic IF protein that has been acquired through horizontal gene transfer or whether IF proteins had already evolved before the advent of the first eukaryotic cells.
Although crescentin is the most studied IF-like protein in bacteria, many aspects of its function have remained obscure. The dynamics of crescentin polymers as well as the mechanism coordinating filament growth with cell elongation and division are still unknown. Moreover, there is currently no information on the quaternary structure of crescentin and the structural features that give rise to the helical shape of detached crescentin filaments. Finally, it will be interesting to determine how the mechanical force exerted by crescentin influences cell wall biosynthesis at the molecular level. Thus, more research is required to fully unravel the biology of this intriguing cytoskeletal protein.
Other Bacterial CCRPs
While crescentin is unique to C. crescentus, many bacteria contain functionally similar proteins that are rich in coiled-coil domains but lack the typical domain architecture of eukaryotic IF proteins [Bagchi et al., 2008; Walshaw et al., 2010]. The characterization of these proteins has started only recently, and in most cases, our knowledge about their precise biological functions and their mechanisms of action is still very limited.
Many CCRPs were shown to form intracellular scaffolds that are involved in the maintenance of cell shape. The spiral-shaped human pathogen Helicobacter pylori, for instance, contains a total of four CCRPs (Ccrp58, Ccrp59, Ccrp1143, and Ccrp1142; Fig. 1B). Each of these proteins is able to polymerize spontaneously in the absence of any cofactor in vitro [Waidner et al., 2009; Specht et al., 2011]. Deletion of ccrp59 results in the complete loss of spiral shape, while inactivation of other CCRPs affects cell morphology to a lesser extent and in varying ways, depending on the strain background [Waidner et al., 2009; Specht et al., 2011]. Ccrp59 was shown to localize in a punctate/patchy pattern, with several static foci distributed along the long-axis of the cell, suggestive of a helical filament arrangement [Waidner et al., 2009]. Similar to crescentin, the H. pylori CCRP cytoskeleton may thus serve to mechanically reduce the rate of peptidogylcan biosynthesis within a region that winds helically around the cell cylinder, thus leading to spiral-like deformation of the peptidoglycan sacculus. Alternatively, it could modulate cell shape more directly by controlling the positioning of cell wall biosynthetic enzymes, but information on potential interaction partners is still missing. Notably, recent work has shown that its helical shape is critical for H. pylori to successfully colonize the gastric epithelium [Sycuro et al., 2010], indicating that CCRPs may be important pathogenicity factors.
Another CCRP involved in cell shape determination is RsmP (Fig. 1B) from the rod-shaped organism Corynebacterium glutamicum. RsmP is essential for viability, and its depletion leads to rounding up of the cells [Fiuza et al., 2010]. Consistent with its ability to polymerize spontaneously in vitro, RsmP assembles into massive, long filaments upon overproduction in C. glutamicum. Although the native localization pattern still remains to be determined, it may thus form an envelope-associated cytoskeleton that helps to control cell wall biosynthesis. Consistent with this idea, corynebacteria are among the few rod-shaped species that lack MreB homologs and elongate by polar growth [Letek et al., 2008], indicating that there must be other factors such as RsmP (and DivIVA, see below) that take the lead in the regulation of cell morphogenesis. Interestingly, RsmP was shown to be phosphorylated in the head and tail domains by the serine/threonine protein kinase PknA in vitro, and phosphomimetic mutations in the three target sites greatly reduced its tendency to polymerize in vivo [Fiuza et al., 2010]. Thus, post-translational modification may be an important mechanism to regulate the morphogenetic function of at least some CCRPs in response to environmental or cell cycle cues. A similar situation has been reported for eukaryotic IFs, whose assembly properties and cellular functions are controlled by phosphorylation of the head and tail domains [Omary et al., 2006].
A third CCRP whose function has been analyzed in more detail is FilP (Fig. 1B) from the actinomycete Streptomyces coelicolor [Bagchi et al., 2008]. Similar to its relative C. glutamicum, S. coelicolor elongates by polar growth, but cells form long branched filaments that develop into extended hyphal networks. A strain lacking FilP shows growth and germination defects. Moreover, mutant hyphae have an irregular, crooked morphology associated with reduced cellular stiffness and elasticity [Bagchi et al., 2008]. FilP may, therefore, have a regulatory role in polar cell wall biogenesis or mechanically stabilize cell shape. Consistent with this notion, FilP was recently shown to form extensive filament networks that line the cell envelope (Fig. 2C, right panel). During growth, most of these polymers accumulate at the hyphal tips, with their abundance decreasing in a gradient-like fashion with increasing distance from the pole (Fig. 2D) [Fuchino et al., 2013]. Stationary cells, by contrast, show a rather homogeneous distribution of FilP polymers within the hyphal filaments. Notably, the localization of FilP to polar growth zones is mediated through interaction with the polar landmark protein DivIVA (see below) [Fuchino et al., 2013]. In vitro, FilP and two of its homologs from related actinomycete species were shown to polymerize spontaneously into profilament bundles. For FilP, various types of polymer architectures have been observed. When polymerizing slowly in physiological buffers, FilP assembles into extensive filament networks that may be related to the network-like structures identified in vivo (Fig. 2C, left panel). However, it also forms striated polymer bundles that resemble the structures formed by the eukaryotic IF protein lamin under similar in vitro conditions [Stuurman et al., 1998; Bagchi et al., 2008]. Analysis of the domain organization revealed that FilP displays two distinct coiled-coil regions: a short N-terminal region containing the canonical heptad repeats characteristic of the coiled-coil segments in eukaryotic IF proteins and a long C-terminal region that is composed of an unusual 51-residue (penindaenad) repeat [Walshaw et al., 2010]. However, the biological significance of this structural alteration still remains to be determined.
The above examples demonstrate that both the molecular architecture and the cellular function of bacterial CCRPs are highly diverse. CCRPs vary greatly in the number and lengths of their coiled-coil regions, often lack the typical nonhelical head and tail domains of eukaryotic IF proteins and do frequently not contain a stutter. Their functional diversity is illustrated by the fact that there are not only CCRPs with (varying) roles in cell shape determination, such as FilP, RsmP, and the H. pylori Ccrp proteins but also several representatives lacking any obvious morphogenetic function. In the predatory bacterium Bdellovibrio bacteriovorus, for instance, inactivation of Ccrp leaves cell shape unaffected but provokes cellular indentations under conditions of osmotic stress [Fenton et al., 2010], suggesting that this protein forms a stabilizing scaffold that confers rigidity to the cell. The protein CfpA from the spirochete Treponema phagedenis, on the other hand, assembles into membrane-associated ribbon-like structures, in which neighboring CfpA filaments are interconnected by bridging proteins. Inactivation of the protein leads to a cell division defect that probably results from chromosome missegregation and hyper-condensation [Izard, 2006], suggesting a role of CfpA in nucleoid architecture or DNA transport. A similar function may also be attributed to the CCRP Scc from the spirochete Leptospira biflexa. Scc can form extensive polymer bundles in vitro and upon heterologous expression in E. coli cells. These polymers can interact nonspecifically with nucleic acids and rearrange into regular, rod-shaped structures when complexed with DNA molecules. Cells lacking Scc display normal cell shape and viability, but they give rise to considerably smaller colonies on solid medium [Mazouni et al., 2006], indicating that Scc must have a significant physiological role.
There are more filament-forming CCRPs with a potential cytoskeletal role, such as AglZ and FrzS from the social bacterium Myxococcus xanthus, which are required for single-cell and social motility of this organism, respectively [Ward et al., 2000; Yang et al., 2004]. However, it is so far unclear whether and how their ability to polymerize contributes to their cellular functions. Interestingly, bioinformatic analyses showed that proteins with extensive coiled-coil regions are wide-spread among bacteria [Bagchi et al., 2008; Walshaw et al., 2010]. It will be interesting to see whether the candidate proteins identified in silico in fact assemble into cytoskeletal elements in their host organisms.
Cytoskeletal-Like Scaffolding Proteins
In addition to CCRPs, bacteria also contain proteins unrelated to known eukaryotic cytoskeletal elements that are able to assemble into higher-order polymeric structures. Some of them, such as DivIVA and PopZ, lack the ability to form extended filaments but rather assemble into compact two- or three-dimensional scaffolds that line the tips of the cell, acting as polar landmarks in a variety of cellular pathways.
Polar Scaffolds in Gram-Positive Bacteria: DivIVA
A prototypical example of a bacteria-specific polar scaffolding protein is DivIVA, a factor widely conserved among Gram-positive bacteria. DivIVA is composed of two central coiled-coil regions that are separated by a flexible linker and flanked by short terminal noncoiled segments [Oliva et al., 2010]. Some homologs have longer C-terminal extensions, but generally the conservation of DivIVA in both domain structure and sequence is relatively high. Although reminiscent of CCRPs in its domain structure, DivIVA displays distinct polymerization properties. Its assembly initiates with the alignment of two chains into a parallel coiled coil [Oliva et al., 2010]. Electron microscopic studies on the Bacillus subtilis DivIVA homolog suggest that these dimeric complexes further oligomerize into bone-shaped structures, which in turn interact end-to-end and laterally to form extended two-dimensional lattices (Fig. 3A, left panel) [Stahlberg et al., 2004]. Owing to exposed hydrophobic and positively charged residues and, possibly, specific structural features, these assemblies preferentially attach to negatively curved membrane regions, as usually observed at the cell poles and the site of cell division [Carballido-Lopez, 2006; Lenarcic et al., 2009; Ramamurthi and Losick, 2009; Oliva et al., 2010].
In B. subtilis, newly synthesized DivIVA molecules initially localize to the growing septum, where they form a stable, ring-like structure (Fig. 3B) [Edwards and Errington, 1997]. After cell separation, this structure collapses into smaller fragments that remain associated with the new pole [Eswaramoorthy et al., 2011]. The septal DivIVA ring recruits the MinCDJ division inhibitor complex [Harry and Lewis, 2003; Bramkamp et al., 2008; Patrick and Kearns, 2008] to prevent premature reinitiation of cytokinesis at the new cell poles after the cells have separated [Cha and Stewart, 1997; Gregory et al., 2008]. Moreover, during sporulation, polar DivIVA assemblies interact with the DNA-binding protein RacA to facilitate attachment of the chromosomal origin region to the prespore pole [Ben-Yehuda et al., 2003; Wu and Errington, 2003]. Finally, DivIVA also appears to promote natural competence by interacting with the competence-activated division inhibitor Maf [Briley et al., 2011] and the post-transcriptional regulator ComN [dos Santos et al., 2012], but the underlying mechanisms still remain to be clarified. It will be interesting to analyze the regulation of DivIVA function and determine how this protein can organize different biological processes in distinct regions of the cell.
Notably, in actinomycetes such as S. coelicolor and C. glutamicum, the function of DivIVA has diverged considerably (Fig. 3B). There, tip-associated DivIVA complexes serve as assembly platforms for the peptidoglycan biosynthetic machinery directing the characteristic apical growth of these organisms [Flärdh, 2003; Ramos et al., 2003; Kang et al., 2005; Letek et al., 2008]. In S. coelicolor, these complexes are associated with another coiled-coil protein, called Scy [Holmes et al., 2013]. In its absence, hyphae display numerous irregularly shaped and mislocalized DivIVA clusters along the length of the filaments, accompanied by tip splitting and elevated branching frequencies. Overproduction of Scy, by contrast, induces copious abortive branch formation, leading to striking fractal growth patterns. These findings suggest that Scy acts as a regulatory factor that specifies the sites of DivIVA assembly and, potentially, controls the size of the resulting clusters. In addition to Scy, S. coelicolor DivIVA interacts with the CCRP FilP [Holmes et al., 2013] and with a cellulose synthase-like protein involved in hyphal tip growth [Xu et al., 2008]. The DivIVA scaffold thus appears to be the foundation of a large tip-organizing center, termed polarisome [Hempel et al., 2012] or TIPOC [Holmes et al., 2013], that governs polar hyphal extension, similar to the Spitzenkörper in filamentous fungi [Sudbery, 2011]. Apart from regulating hyphal growth, actinomycete DivIVA has an important role in chromosome dynamics. Reminiscent of the DivIVA-RacA interaction in sporulating B. subtilis cells, the C. glutamicum homolog (Fig. 3A, right panel) associates with the chromosome partitioning protein ParB, thereby immobilizing the chromosomal origin regions at the cell poles [Donovan et al., 2012]. Interestingly, a similar interaction was also observed for the corresponding S. coelicolor proteins [Donovan et al., 2012]. Moreover, the DivIVA-associated protein Scy was found to interact with the chromosome segregation ATPase ParA [Ditkowski et al., 2013], which interacts with ParB to drive origin segregation. These findings suggest that the S. coelicolor TIPOC may have a role in positioning chromosomal DNA within the hyphal filaments. However, since in streptomycete species, the chromosomal origin regions are usually found at pole-distal positions, the underlying mechanistic details remain unclear.
The regulation of DivIVA function is still poorly understood. In several species, DivIVA is phosphorylated by eukaryotic-type serine/threonine protein kinases, and its modification was shown to be critical for proper cluster formation and cell morphology [Kang et al., 2005; Manteca et al., 2011; Fleurie et al., 2012; Hempel et al., 2012]. In S. coelicolor, for instance, multiple residues within the C-terminal region of DivIVA are targeted by the kinase AfsK, which colocalizes with DivIVA at the hyphal tips [Hempel et al., 2012]. Whereas inactivation of AfsK results in a reduced branching frequency, synthesis of a constitutively active variant leads to disintegration of the polar DivIVA clusters, hyper-branching, and irregular cell shape. Notably, strong phosphorylation of DivIVA was induced by treatment of cells with antibiotics blocking cell wall biogenesis [Hempel et al., 2012], suggesting that AfsK could serve to direct growth of the mycelium away from adverse environments.
Polar Scaffolds in Gram-Negative Bacteria: PopZ
Although DivIVA is restricted to Gram-positive bacteria, many Gram-negative species (the alpha-proteobacteria) produce a functionally similar protein, called PopZ [Bowman et al., 2008; Ebersbach et al., 2008]. Evolutionarily unrelated to DivIVA, PopZ homologs do not contain any coiled-coil regions. They are structured into three distinct domains, including (1) a conserved N-terminal domain largely composed of a short α-helix, (2) an intervening nonstructured region that is poorly conserved in sequence but characterized by a high content of proline and acidic residues, and (3) a conserved C-terminal domain containing three predicted α-helical segments [Laloux and Jacobs-Wagner, 2013]. PopZ assembles into filamentous oligomers that further interact to form extensive polymeric networks, frequently connected by three-way junctions (Fig. 3C, left panel). The resulting matrix has gel-like properties and is permeable to small proteins but not to macromolecules, as apparent by the existence of conspicuous ribosome-free zones at the subcellular locations of PopZ in vivo [Bowman et al., 2008; Ebersbach et al., 2008; Bowman et al., 2010].
In C. crescentus, PopZ localizes dynamically to the poles of the cell and, similar to DivIVA, functions in multiple cellular pathways. In G1 phase, it only forms a single complex at the old cell pole, but upon initiation of DNA replication, a second complex appears at the opposite end of the cell (Fig. 3C, right panel). PopZ captures and immobilizes the chromosomal origin regions through interaction with the origin-bound chromosome partitioning protein ParB [Bowman et al., 2008; Ebersbach et al., 2008]. Upon entry into S-phase, the old PopZ complex releases ParB, possibly to facilitate replication initiation, whereas the PopZ complex formed at the opposite end captures and immobilizes the origin copy that is partitioned to the new cell pole (Fig. 3D). In addition to its role in origin attachment, PopZ interacts with the chromosome partitioning ATPase ParA and thereby helps ensure the directionality of the DNA segregation process [Schofield et al., 2010]. Finally, the old-pole PopZ complex was shown to mediate, directly or indirectly, the polar localization of several proteins involved in cell cycle control [Ebersbach et al., 2008; Bowman et al., 2010]. Consistent with the central importance of proper origin and protein localization in the regulation of cell division and development in C. crescentus [Thanbichler and Shapiro, 2006; Kiekebusch et al., 2012; Kirkpatrick and Viollier, 2012; Tsokos and Laub, 2012], inactivation of PopZ strongly disturbs division site placement and cell cycle progression. It thus acts as a central hub that temporally and spatially coordinates key processes in the cell.
The mechanisms that recruit PopZ to the cell poles and control its localization dynamics are still incompletely understood. It was shown that oligomerization of PopZ, mediated by the two C-terminal α-helices H3 and H4, is critical for the assembly of polar PopZ clusters [Laloux and Jacobs-Wagner, 2013]. However, unlike DivIVA, these clusters do not appear to have an intrinsic affinity for curved membranes. Instead, they localize preferentially to subcellular regions devoid of chromosomal DNA, suggesting a role of the nucleoid in PopZ positioning [Ebersbach et al., 2008; Saberi and Emberly, 2010]. Recent work demonstrated that the characteristic cell cycle-dependent localizaton dynamics of PopZ are achieved by coupling the formation of new clusters to the segregation of the chromosomal origin regions [Laloux and Jacobs-Wagner, 2013]. The two processes are linked through the chromosome partitioning protein ParA, which acts as a nucleator of PopZ assembly. ParA gradually accumulates at the new cell pole as separation of the newly replicated origin regions proceeds. Its condensation is thought to increase the local concentration of PopZ (which interacts with ParA) until it reaches a certain threshold that is required to initiate cluster formation. Thus, the assembly of a new cluster is promoted by a cell cycle-controlled skew in the distribution of its subunits. It will be interesting to dissect the underlying assembly mechanism and the precise architecture of PopZ scaffolds. Notably, analysis of eukaryotic cytoskeletal and signaling systems showed that multivalent proteins can produce gel-like molecular networks that establish distinct, nonmembrane-bounded compartments within the cell [Li et al., 2012]. The protein scaffolds formed by PopZ may result from a similar phase separation event, suggesting that sol-gel transitions could have evolved as a general mechanism to create micro-scale organization from nano-scale components.
Recently, a new class of cytoskeletal proteins, termed bactofilins, has been identified in bacteria [Kühn et al., 2010]. Its representatives are widespread across the bacterial phylogeny, with many genomes encoding multiple paralogs (Fig. 4) [Kühn et al., 2010]. Bactofilins are characterized by a conserved DUF583 domain (also known as bactofilin domain) [Punta et al., 2012], which is typically flanked by short, often nonstructured terminal regions. Although generally small in size (∼20 kDa) and predicted to be soluble, some DUF583-containing proteins display long N- or C-terminal extensions, some of which include potential transmembrane segments (Fig. 5). Similar to IF-like proteins, bactofilins are capable of self-assembling into stable filaments in the absence of any cofactor [Kühn et al., 2010; Koch et al., 2011]. In most organisms, the precise cellular function of these structures is still unknown. However, current evidence suggests that bactofilins may be involved in a range of different pathways that vary among species.
Bactofilins were initially discovered in a protein localization screen aimed at identifying factors with a potential role in the biogenesis of the C. crescentus stalk [Kühn et al., 2010]. The two bactofilin paralogs revealed in this species (BacA and BacB) are soluble proteins that carry a central DUF583 domain bordered by proline-rich, charged terminal peptides. Both of them spontaneously polymerize into filament bundles when heterologously produced in E. coli (Fig. 6A, left panel). These polymers lack obvious dynamics and are insensitive to chelating agents, changes in pH and dilution, indicating high intrinsic stability. BacA and BacB are produced constitutively over the course of the cell cycle. While being evenly distributed in stalkless G1 cells, they both condense at the stalked pole during at the onset of S phase, when stalk formation is initiated (Fig. 6A, right panel). In doing so, the two proteins copolymerize into sheet-like structures that line the inner face of the polar cytoplasmic membrane [Kühn et al., 2010]. These assemblies in turn recruit the cell wall biosynthetic enzyme PbpC, a bitopic membrane protein with a proline-rich cytoplasmic domain that interacts with the BacAB sheet (Fig. 6B). C. crescentus mutants lacking the two bactofilins or PbpC still form stalks, but stalk length is significantly reduced [Kühn et al., 2010], suggesting that this system has an accessory role, for example, in cell wall modification. The mechanism triggering the cell cycle-dependent polar accumulation of BacAB is still unclear. However, clues may come from the observation that overexpression of the two proteins causes expansion of the polar bactofilin patches into elongated, ribbon-like structures that impose strong positive curvature on the cell envelope regions they are associated with [Kühn et al., 2010]. Reminiscent of DivIVA [Lenarcic et al., 2009; Ramamurthi and Losick, 2009], BacAB sheets could thus be intrinsically curved and, as a consequence, preferentially associate with the positively curved membrane regions established at the base of the growing stalk. However, further studies are required to test this hypothesis and see whether membrane topology is a common cue for targeting polymeric protein scaffolds to the cell poles.
Whereas the two bactofilins found in C. crescentus copolymerize and share a common function, many species synthesize multiple paralogs with clearly distinct roles. The chromosome of the social Gram-negative bacterium Myxococcus xanthus, for instance, harbors four bactofilin genes. Each of the encoded proteins (BacM-P) is able to polymerize into stable filament bundles when overproduced heterologously in E. coli [Kühn et al., 2010]. Moreover, recent work showed that fibers of BacM can be readily purified from the insoluble fraction of M. xanthus cells [Koch et al., 2011]. Similar to the heterologously produced polymers, these BacM fibres are resistant to high-salt or detergent treatment [Kühn et al., 2010; Koch et al., 2011]. Notably, they largely consist of an N-terminally truncated version of the protein. Since both the full-length and truncated forms of BacM are detectable in cell lysates, BacM may, therefore, be proteolytically cleaved before or immediately after incorporation into the polymer [Koch et al., 2011]. It is tempting to speculate that this processing reaction could regulate the polymerization dynamics or function of BacM fibres in the cell. The precise subcellular localization of BacM is not entirely clear, because its function is compromised by fluorescent protein tags. Current data suggest that BacM forms helical or rod-like filaments that show some preference for the polar regions of the cell but can also extend throughout the entire cell body [Kühn et al., 2010; Koch et al., 2011]. Cells lacking BacM have an uneven, kinked morphology when grown in liquid medium. Moreover, they show increased sensitivity toward antibiotics interfering with peptidoglycan biosynthesis [Koch et al., 2011]. Thus, unlike in C. crescentus, where bactofilins have a specialized function in stalk formation, BacM appears to be required for maintaining general cell shape. The underlying mechanism still remains to be determined. However, it is conceivable that similar to BacAB in C. crescentus, BacM filaments may interact with parts of the cell wall biosynthetic machinery to temporally or spatially control their activity. The function of the three other bactofilin homologs (BacN-P), whose genes are organized in an putative operon, still remains to be clarified. Localization studies suggest that they also form filaments in vivo [Kühn et al., 2010]. However, their inactivation does not affect the localization of BacM nor does it cause any obvious morphological defects [Koch et al., 2011], indicating functional redundancy or diversification.
Another example of a bactofilin involved in morphogenesis has recently been reported for Helicobacter pylori, a human pathogen notorious for causing peptic ulcers. H. pylori carries a single bactofilin gene, named ccmA. Its deletion abolishes the typical helical cell shape of this species and instead produces straight or slightly curved morphologies [Sycuro et al., 2010]. Importantly, this morphological defect strongly impairs the ability of cells to colonize the gastric mucus, identifying bactofilins as important pathogenicity factors. It has been proposed that CcmA, together with three periplasmic LytM-type endopeptidases (Csd1–3), mediates the hydrolysis of peptidoglycan crosslinks at specific points along the long-axis of the cell, thereby inducing slight rearrangments in the cell wall that result in torsion of the cell body [Sycuro et al., 2010]. Although the in vivo localization of CcmA is still unknown, the polymer-forming properties of other bactofilin homologs suggest that it could assemble into membrane-associated longitudinal filaments that, directly or indirectly, position the Csd1–3 hydrolases in the cell.
Taken together, the functional studies performed to date suggest that bactofilin polymers in many cases serve to position cell wall biosynthetic enzymes, thereby regulating various aspects of cell morphogenesis. However, several bactofilin homologs have also been implicated in motility. The swarming bacterium Proteus mirabilis, for example, contains a single bactofilin gene, ccmA, that gives rise to two different products: the full-length protein, which carries two N-terminal transmembrane helices, and an N-terminally truncated form that remains peripherally associated with the cytoplasmic membrane [Hay et al., 1999]. Both variants are highly expressed at the swarm stage, and deletion of ccmA leads to a reduction in swarming speed and a high frequency of cells that are noticeably curved. Synthesis of a C-terminally truncated derivative yields a more pronounced phenotype, with cells showing strong bending, uneven width and a severely reduced swarming efficiency [Hay et al., 1999]. In agreement with a role of bactofilins in motility, a different study showed that a bactofilin homolog, together with a LytM endopeptidase, is strongly upregulated during surface attachment and swarming in Vibrio parahaemolyticus [Gode-Potratz et al., 2011]. Moreover, two DUF583-containing proteins from B. subtilis were found to be required for proper swimming motility [Rajagopala et al., 2007]. The involvement of bactofilins in locomotion thus appears to be conserved in evolutionarily distinct lineages. However, in all of the above cases, the underlying mechanisms are still unknown. Considering that several bactofilin homologs are involved in cellular morphogenesis, the observed motility phenotypes may be caused by morphological defects that prevent mutant cells from aligning into multicellular rafts (as required for swarming motility) or maintaining the direction of movement. Conversely, bactofilins could also act more directly by facilitating the proper positioning of flagella or other motility-related proteins.
Evidence for such a direct mechanism has been obtained in M. xanthus, a species that uses two complementary mechanisms—single-cell and social motility—to move on solid surfaces. Social motility depends on type IV pili that are positioned at the leading cell pole, pulling the cell forward through repeated cycles of extension and retraction. M. xanthus erratically changes the direction of movement, in part by relocating a subset of pilus-associated proteins from one cell pole to the other. Proper positioning of these mobile proteins depends on the protein SofG, a small, Ras-like GTPase that forms a single, subpolar cluster within the cell [Bulyha et al., 2013]. The unusual localization of SofG was recently shown to rely on the bactofilin homolog BacP (see above), which forms filamentous structures at both cell poles (Figs. 6C and 6D). SofG associates with one of these structures and then shuttles back and forth along the BacP polymer, dependent on its GTPase activity [Bulyha et al., 2013]. The molecular mechanism by which SofG mediates the polar localization of pili-associated proteins is still poorly understood. It may either actively transport interacting proteins toward the pole by translocating along the BacP filament or, as a gatekeeper, regulate their BacP-dependent association with the cell pole. This sytem shows striking analogies to the regulation of cytoskeletal function by small GTPases in eukaryotes [Sit and Manser, 2011] and provides another example for the fact that many basic concepts in cell biology also apply to bacteria.
Although our knowledge on bactofilins is still very limited, the above examples suggest that DUF583 domain-containing proteins share the ability to assemble into polymeric structures that serve as localization factors for other proteins. However, the nature of their interaction partners and the biological systems they are involved in appear to vary considerably. Consistent with this functional diversity, bactofilins display a relatively low level of sequence conservation in the regions bordering the conserved DUF583 domain. It will be interesting to see whether all of their representatives form membrane-associated polymers, as demonstrated for BacAB from C. crescentus [Kühn et al., 2010] and BacM from M. xanthus [Koch et al., 2011]. Moreover, structural studies will be required to reveal the precise architecture of bactofilin polymers and their mechanism of assembly.
Conclusions and Perspectives
In recent years, an ever-increasing number of polymer-forming proteins has been identified in bacteria. Some of them are highly conserved, whereas others are only found in certain lineages or some closely related species. This diversity is likely explained by the long evolutionary history of bacteria, which has provided ample time for the development of new cytoskeletons and their divergence in assembly properties and function. Bacterial cells inhabit a wide range of environments and, as a consequence, display a variety of different morphologies and life cycles. Therefore, it does not come as a surprise that they have developed a range of different solutions to control their cellular architecture, tailored precisely to their specific needs.
Interestingly, while nucleotide-dependent actin and tubulin cytoskeletons are widely conserved in both eukaryotic and prokaryotic cells, nucleotide-independent cytoskeletal proteins appear to be more variable and limited in their phylogenetic distribution. In particular, there is a great variety of bacterial CCRPs, which despite comparable polymer architectures and mechanical properties hardly share any structural similarity with each other. Similarly, a number of poorly conserved CCRPs with proven or hypothetical cytoskeletal roles has been identified in metazoans and other eukaryotic lineages, none of which show any apparent similarity to bacterial CCRPs [Rose and Meier, 2004; Ciska et al., 2013]. It remains to be determined whether these proteins have diverged from a common IF-like ancestor or whether they are analogs that have evolved independently in each of the different lineages. Notably, there is precedent for the horizontal acquisition of bacterial cytoskeletal proteins (such as the Prosthecobacter tubulins BtubA/B) from ancient eukaryotic cells [Schlieper et al., 2005; Martin-Galiano et al., 2011; Pilhofer et al., 2011], suggesting that a similar process could have taken place in the case of bacterial CCRPs. However, it is difficult to unequivocally identify cytoskeletal CCRPs and their homologs by sequence analysis, thus preventing an in-depth analysis of their distribution and evolutionary relationship. Similar to their history, the full range of functions adopted by CCRPs is still unclear. The representatives investigated so far are generally thought to act as rigid scaffolds that support or mechanically modulate cell shape. However, it is conceivable that CCRPs may recruit proteins involved in cell wall biogenesis to complement their mechanical function. Moreover, it will be interesting to see whether they have also adopted cellular functions unrelated to cell morphogenesis, for instance, by acting as general protein localization factors that facilitate the assembly or positioning of protein complexes.
Similar to CCRPs, bactofilins polymerize in a nucleotide-independent manner to form stable cytoskeletal scaffolds. However, they do not have a mechanical function but rather serve as spatial landmarks within the cell. Bactofilins are almost universally conserved among bacteria. Their genes are often highly expressed and present in several copies per genome. They are, therefore, likely to perform vital functions in the organisms that produce them. However, these functions, such as the modulation of cell shape and motility, may not be essential under laboratory conditions, explaining why this group of proteins has not been identified earlier. Bactofilin polymers resemble IFs in their stability and inertness in vitro. However, the cell cycle-dependent formation of BacA sheets in C. crescentus indicates that their assembly may be dynamic and tightly controlled in vivo. It will be a challenge for future research to investigate the precise mode of bactofilin polymerization and to study the mechanisms that temporally and spatially regulate this process in the cell. Moreover, it will be interesting to determine the diversity of biological systems that bactofilins are involved in.
Considering that only a small number of bacterial species are currently amenable to cell biological studies, the cytoskeletal elements known to date may only represent the tip of the iceberg [Briegel et al., 2006]. Moreover, our knowledge on the dynamics of nucleotide-independent cytoskeletons is probably far from complete. For instance, recent work has revealed that mollicutes of the genus Spiroplasma display fibril bundles that traverse the spiral-shaped cell bodies along their long axis, mediating complex cell shape changes that propel the bacteria forward in liquid medium [Kurner et al., 2005; Trachtenberg, 2006; Trachtenberg et al., 2008]. These polymers are highly stable in vitro and thus probably lack nucleotide-dependent polymerization dynamics. Nevertheless, they are able to alter their shape in a controlled manner, based on conformational changes in their subunits [Cohen-Krausz et al., 2011]. The fact that a protein polymer is inherently stable may, therefore, not necessarily mean that it is unable to drive dynamic processes in the cell. Future studies will show whether such polymerization-independent mechanisms for controlling cytoskeletal rearrangements are a more common theme among bacteria.
Nucleotide-independent cytoskeletal scaffolds have emerged as key players in the organization of bacterial cells. We have only started to understand their diversity and function, and it will be exciting to see how they cooperate with other cytoskeletal structures to control the temporal and spatial organization of bacterial cells.
We thank Christine Jacobs-Wagner, Marc Bramkamp, Nora Ausmees, Grant Bowman, and Lotte Søgaard-Andersen for providing images, Daniela Kiekebusch for critical reading of the manuscript, and the anonymous reviewers for constructive criticism. The work from the authors' laboratory discussed in this review was supported by the Max Planck Society, the DFG Graduate School “Intra- and intercellular transport and communication” (GRK 1216), and a Young Investigator Grant (RGY69/2008-C and RGY0076/2013) from the Human Frontier Science Program.
Table 1. Typical Properties of Nucleotide-Independent Cytoskeletal Proteins in Bacteria
The functions of the proteins in the table have only been investigated in a limited number of model organisms. Additional biological roles may, therefore, be identified in the future.
In these groups, the existence of CCRPs has been experimentally verified. Additional candidates have been discovered in Bacillales and Planctomycetes by sequence analysis [Bagchi et al., 2008]. However, CCRPs are difficult to identify by homology searches, and detailed structure-based analyses of their phylogenetic distribution are still missing, suggesting these proteins may be more widespread than currently known.
Extended coiled-coil regions flanked by variable head and tail domains
Stable long filaments
Cell shape maintenance
Widespread, e.g., in Proteobacteria, Actinobacteria, and Spirochaetesb
Two short coiled-coil regions connected by a flexible linker and flanked by short terminal regions
Targeting of proteins to septa or cell poles
Highly conserved in Gram-positive bacteria
Central proline-rich region flanked by one N-terminal and three C-terminal α-helices
Targeting of proteins to cell poles
Highly conserved in alpha-proteobacteria
Central bactofilin (DUF583) domain flanked by non-structured N- and C-terminal regions