Correspondence: Imrich Barák, Institute of Molecular Biology, Slovak Academy of Sciences, 845 51 Bratislava 45, Slovakia. Tel.: +421 2 5930 7418; fax: +421 2 5930 7416; e-mail: email@example.com
The process of cell division has been intensively studied at the molecular level for decades but some basic questions remain unanswered. The mechanisms of cell division are probably best characterized in the rod-shaped bacteria Escherichia coli and Bacillus subtilis. Many of the key players are known, but detailed descriptions of the molecular mechanisms which determine where, how and when cells form the division septum are lacking. Different models have been proposed to account for the high precision with which the septum is constructed at the midcell and these models have been evaluated and refined against new data emerging from the fast improving methodologies of cell biology. This review summarizes important advances in our understanding of how the cell positions the division septum, whether it be vegetative or asymmetric. It also describes how the asymmetric septum forms and how this septation event is linked to chromosome segregation and subsequent asymmetric gene expression during spore formation in B. subtilis.
The earliest apparent event in cell division in bacteria is the formation of an FtsZ ring at the future septum site. FtsZ is a highly conserved GTPase with similarity to the tubulins, a family of eukaryotic cytoskeletal proteins. During vegetative growth, the Z-ring forms at midcell where the cells subsequently divide. Two distinct mechanisms contribute to accurate placement of the division machinery: the Min system and nucleoid occlusion. The Min system functions mainly to eliminate the possibility of division at the cell poles. In Escherichia coli, it achieves this by an extraordinary protein oscillation from pole to pole in dividing cells. Although models have been proposed to explain this oscillation phenomenon, it is by no means fully understood.
Bacillus subtilis, like E. coli and a variety of other prokaryotes, has Min system homologues, and these are important for the prevention of asymmetric septation during vegetative growth. However, B. subtilis lacks one component of the Min system, MinE, and oscillation of the other components has not been observed. In B. subtilis, DivIVA assumes the role played by MinE in E. coli, positioning the MinC division inhibitor close to the cell poles although the two proteins clearly function differently. The DivIVA/MinCD division system appears to have no direct role in the initiation of Z-ring formation at the midcell site, and instead it inhibits division at the polar sites. Possibly, the more important factor in midcell division site selection in B. subtilis is the position of the nucleoid with the midcell site appearing and disappearing cyclically during vegetative growth with rounds of chromosome replication and segregation.
A general problem in developmental biology concerns the process by which cells differentiate. Despite their genetic identity, daughter cells arising from cell division often differ morphologically as well as physiologically and have different fates. As one of the simplest differentiation processes, B. subtilis sporulation represents an excellent model system for studying cell differentiation. Sporulation is closely linked to the cell cycle and in B. subtilis is associated with an asymmetric cell division. The beginning of sporulation is marked by remarkable alterations not only in chromosome partitioning but also in cell division. The asymmetrically positioned sporulation septum bisects the axial filament leaving only one-third of one chromosome in the forespore and creating a transient genetic asymmetry. The remaining two-thirds of the chromosome is then transferred from the larger mother cell into the smaller forespore via a conjugation-like mechanism directed by the partitioning protein, SpoIIIE.
It is our intention in this review to summarize important advances in our understanding of how the cell positions the division septum, be it vegetative or asymmetric. We also intend to describe how the asymmetric septum forms and how this septation event is linked to chromosome segregation and subsequent asymmetric gene expression during spore formation. Aspects of these processes have been described in recent in-depth reviews (Margolin, 2000; Harry, 2001; Errington et al., 2003; Barak & Wilkinson, 2005; Yudkin & Clarkson, 2005).
History of bacterial cell division studies
Bacterial cell division is a complex process in which the division septum is formed at a specific location within the cell in a process that must be temporally and topologically coordinated with other cellular events such as chromosome replication and nucleoid segregation. For many years the cell division process was difficult to study because mutations in cell division genes tend to be lethal and there were no tools for localizing division proteins in living cells. Genetic approaches to the study of cell division were introduced in the late 1960s when large numbers of E. coli conditional mutant strains were identified that failed to septate when grown at the nonpermissive temperature (Hirota et al., 1968). These mutant strains were divided into two general groups. The first group included the filamentation temperature-sensistive (fts) mutants, which formed filaments containing nucleoids that appeared to be regularly distributed along the length of the cell. The second group of mutant strains were characterized by aberrations in the distribution of DNA within the cells and a tendency to form anucleate cells. These were termed partition (par) mutants because they were thought to be defective in chromosome segregation.
The early investigations of the localization of division proteins in the cell used techniques such as cryoimmunotransmission electron microscopy and immunofluorescence microscopy. A breakthrough was made with the establishment of technologies for studying cell division using green fluorescent protein (GFP) and its yellow (YFP), cyan (CFP), blue (BFP) and red (RFP) fluorescent variants [reviewed in (Phillips, 2001)]. Gene fusions using gfp can be constructed with relative ease and their use heightens the sensitivity of detection of proteins in the cell and allows monitoring of protein localization in real time. GFP fusion can disrupt the normal function of proteins, however, and inferences from GFP fusion experiments should be corroborated by other data. Nevertheless, the application of this methodology has led to the observation of such phenomena as protein oscillation and chromosome segregation in bacterial cells. As a result, studies using GFP fusions have strengthened our understanding not only of cell division but also of diverse cellular processes including chromosome replication and partitioning, sporulation, development and signal transduction, as a result changing our view of the organization of the bacterial cell in general.
The rod-shaped E. coli and B. subtilis, which are examples of Gram-negative and Gram-positive bacteria, respectively, have been model organisms for the study of bacterial cell division. Although many of the component cell division proteins are common to the two organisms, there are mechanistic differences in how the septation site is determined and recognized, and in how the septum is constructed. It has generally been assumed that other bacteria follow cell division strategies similar to one or other of these two microorganisms. This assumption can be straightforwardly reassessed by looking for homologues of cell division genes in the many completed bacterial genome sequences.
Vegetative cell division in E. coli and in B. subtilis
The basic process of cell division is conceptually similar in eukaryotic and prokaryotic cells and is characterized by the creation of a division septum between the recently duplicated chromosomes. The principal advantage of studying cell division in prokaryotes is its relative simplicity, even though recent advances point to greater complexity than was previously expected. Cell division, often called septation, consists of invagination of the cytoplasmic membrane and peptidoglycan synthesis. Although many of the players in this process have been identified and characterized, the nature and the mechanism of their interactions in forming the division septum are quite poorly understood. Probably the most enduring question in cell division of rod-shaped bacteria is how the position of the division site is determined.
Division site selection
Cell division in E. coli is a remarkably accurate process with respect to localization of the midcell division plane. The earliest event in division is the assembly of FtsZ at the midcell site into a structure called the Z-ring and indeed the Z-ring is considered to be an accurate marker for the position of the division site. The midcell Z-ring in wild-type E. coli and B. subtilis is defined with a high degree of precision with the standard deviation of its location from the cell centre having been determined by immunofluorescence microscopy and FtsZ-GFP fusion protein localization to be 2.6% and 2.2%, respectively (Yu & Margolin, 1999; Sun & Margolin, 2001; Migocki et al., 2002). It is becoming apparent that two distinct mechanisms are responsible for midcell site selection in E. coli and B. subtilis, the Min system and nucleoid occlusion (Fig. 1) [reviewed in (Margolin, 2000; Harry, 2001)]. These two independent mechanisms block division at the poles and in the vicinity of the nucleoid and both are required for precise division site selection.
The nucleoid occlusion model
The positional correlation between the site of Z-ring formation and the zone of separation of the bacterial nucleoid has led to the long-standing hypothesis that the bacterial nucleoid is a determinant of the timing and placement of the division septum. This ‘nucleoid occlusion’ model, originally proposed by Woldringh et al.(1990), states that the nucleoid inhibits division wherever it occupies space in the cell. Thus, the midcell site appears and disappears cyclically during vegetative growth with rounds of chromosome replication (Errington et al., 2003). According to the model, Z-ring formation over the nucleoid is precluded, so that its assembly at the midcell must await the vacation of this area by the nucleoid upon segregation (Fig. 1). Thus, although cell division can in principle take place at any point along the length of a cell, early in the cell cycle the unsegregated nucleoid prevents septation at midcell. Z-ring formation in the nucleoid-free space adjacent to the cell poles is prevented by the presence of the division inhibitors MinCD as discussed below. The nucleoid occlusion model is attractive though rather poorly defined, and moreover it does not appear to have the necessary precision to account for the observed accuracy of midcell site selection in E. coli (Koch, 1990; Margolin, 2000). In E. coli ftsA, dnaAts double mutant strains, there is no correlation between the location of the septum and the position of the nucleoid (Cook & Rothfield, 1999). These cells form filaments with large nucleoid-free zones. When cell division is restored at the permissive temperature, septa are formed at a constant distance, c. 5 μm, from the poles, irrespective of the distance to the nearest nucleoid.
In B. subtilis, the position of the nucleoid seems to be an even more important factor in midcell division site selection. Outgrowing spores have proved to be a particularly fruitful system for studying cell division. This is because (1) in the germinating spore, which is initially spherical, the cell pole is not established by a preceding cell division/Z-ring formation event and (2) the first round of replication and cell division occurs reasonably synchronously within a cell population. In this system, midcell Z-rings have been observed over nucleoids in thymine auxotrophs when DNA replication is blocked by thymine starvation (Harry et al., 1999; Regamey et al., 2000). Under these circumstances, initiation of DNA replication is permitted but elongation of the replication fork is prevented. One explanation is that Z-ring formation and septation can occur in less dense regions of the nucleoid and that medial Z-ring formation occurs after degradation of DNA adjacent to the origin of replication (Regamey et al., 2000). It has also been suggested that a unique, midcell site in B. subtilis, designated NS (nucleation site) is initially masked by its occupation by the stationary replication complex [reviewed in Harry, 2001]. These observations indicate that although DNA replication is not essential for Z-ring assembly, correct positioning of the Z-ring at midcell does require initiation of DNA replication though not elongation of the replication fork. Because replication initiation is required for proper ring localization, it is tempting to speculate that the same machinery that recruits the origin of replication and the polymerase machinery to midcell also helps to unmask the preferred FtsZ nucleation site at this position. The position of the Z-ring and cytokinesis can in certain circumstances coincide with that of the nucleoid in wild-type cells. One example is during sporulation in B. subtilis when the nucleoid is bisected by the asymmetrically positioned septum and the DNA is subsequently pumped across the septum into the forespore (Bath et al., 2000). In addition, Z-rings can form directly over the nucleoid in cells with certain mutations (Sun et al., 1998; Gullbrand & Nordstrom, 2000).
Meanwhile, a specific effector responsible for nucleoid occlusion in B. subtilis, Noc (YyaA), has been identified as an inhibitor of division that is also a nonspecific DNA binding protein (Wu & Errington, 2004). Under various conditions in which the cell cycle is perturbed, Noc prevents the division machinery from assembling in the vicinity of the nucleoid, although the underlying molecular mechanism is not known. Similar to B. subtilis Noc, SlmA protein has been characterized in E. coli as a nucleoid-associated division inhibitor capable of mediating nucleoid occlusion (Bernhardt & de Boer, 2005). SlmA was found by screening for mutations synthetically lethal with a defective Min system. The rationale behind the search was that the Min system, which is not itself essential, would become so in cells impaired in nucleoid occlusion. Interestingly, even though the two proteins probably perform the same function in the cell, Noc and SlmA belong to completely different families of DNA binding proteins and they have no primary sequence similarity.
The MinCDE system in E. coli
Cells lacking MinCDE often form Z-rings near the cell poles, which can lead to formation of DNA-less ‘minicells’ (Fig. 1b). By contrast, overexpression of MinCD can lead to filamentous cell growth. Thus, the principal function of the Min system, consisting in E. coli of the three proteins MinC, MinD and MinE, is to prevent division at the cell poles. MinC is an inhibitor of division that is activated by MinD, a membrane-associated ATPase that plays several roles (de Boer, 1991). First, it associates with MinC to form a MinC–MinD complex that inhibits FtsZ polymerization and cell division in the absence of other regulators (Raskin & de Boer, 1997). Secondly, it confers on the division inhibitor (MinC) sensitivity to MinE (de Boer et al., 1990). Finally, MinD is required to localize MinE at midcell (Bernhardt & de Boer, 2005).
MinE was shown to be a topological factor that relieves division inhibition in the central region of the cell (de Boer et al., 1992). The relief of division inhibition and the topological specificity functions of MinE reside in different domains of this 88-amino-acid residue protein. The N-terminal domain is responsible for the anti-MinCD function, as shown by the ability of MinE1 – 34 to prevent MinCD-induced filamentation (Pichoff et al., 1995). Topological specificity was proposed to be accomplished by the binding of the C-terminal domain of MinE to a putative topological target molecule at the new division site at midcell (Rothfield & Zhao, 1996). However, this target molecule has not been found.
The mechanism of MinCDE action has been studied by fluorescent microscopy localization of GFP chimaeras. The first studies showed that MinE-GFP localizes as a ring-like structure adjacent to the midcell (Fig. 2a), and that this localization pattern requires the simultaneous expression of minD (Raskin & de Boer, 1997). More recent localization experiments revealed that MinD oscillates from pole to pole in dividing cells (Fig. 2b) with a period of 10–20 s (Pichoff et al., 1995; Rothfield & Zhao, 1996). The division inhibitor MinC binds to MinD and co-oscillates with it (Hu & Lutkenhaus, 1999; Raskin & de Boer, 1999a, b). Similarly, other studies have shown that in living cells MinE also undergoes rapid oscillation coupled to that of MinD (Fu et al., 2001). This oscillating movement of MinD and MinE is codependent because MinE deletion causes uniform distribution of MinD around the cytoplasmic membrane (Rowland et al., 2000).
The consequence of this extraordinary protein oscillation is that the time-integrated concentration of the cell division inhibitor MinC is lowest at the midcell and highest at the cell poles. Thus, the oscillation process prevents division at the cell poles but not at the midcell site. This view of Min protein dynamics has now been refined by the work of Shih et al.(2003a), who have shown in new images with improved resolution that MinCD does not coat the whole cell membrane near one pole but instead forms a long winding filament (Fig. 2c). Similar extended helical arrangements have been observed for the E. coli cytoskeletal protein MreB (Shih et al., 2003b). The MreB and MinCDE coiled arrays do not appear to be identical and recently it was shown that the Min proteins are capable of end-to-end oscillation even in ΔmreB cells, which are spherical and lacking in apparent asymmetry (Shih et al., 2003b). This implies that detectable asymmetric geometry is not required for establishing the axis of Min oscillation. It is more likely that placement of the division site and the development of cell asymmetry are a consequence of polarized Min oscillation. These results also suggest that functionally distinct cytoskeleton-like elements are present in E. coli and that structures of this type can support dynamic changes that play important roles in division site placement and other aspects of the life of the cell. As no candidate proteins have been identified, they also raise the possibility that regions of altered lipid composition in the cytoplasmic membrane provide the matrix for the binding and migration of the Min proteins and other complexes specifically involved in cytokinesis. It is also possible that the underlying matrix for binding of the dispersed clusters of Min proteins, which are visible via fluorescence microscopy, might simply be the inner surface of the cytoplasmic membrane. In either case, the observed dynamics may be a property of the Min proteins themselves or of the postulated matrix or both.
Interestingly, a study by Hu & Lutkenhaus, (2001) has shown that MinE in the presence of phospholipid vesicles stimulates the ATPase activity of purified MinD several-fold. Moreover, for MinE mutants defective in this stimulatory effect, there was a good correlation with loss of function in vivo, and more importantly, with the abolition or increase in the period of the cellular oscillation of MinD. These results clearly link the hydrolysis of ATP by MinD to its oscillatory behaviour. In addition, the C-terminal portion of MinD, which is conserved in most bacteria, as well as in chloroplasts, is essential for membrane localization through direct interaction with membrane lipids (Szeto et al., 2002; Hu & Lutkenhaus, 2003). In fact, the MinD phospholipid-binding region is predicted to be an amphipathic α-helix, one face of which is hydrophobic while the other face is positively charged. The heterogeneous phospholipid composition of the membrane may be important not only for self-assembly of MinD but also for modulation, both spatially and temporally, of the oscillatory characteristics of the Min system [reviewed in (Mileykovskaya & Dowhan, 2005)]. Regardless, these results raise the important question of how oscillation works.
The crystal structures of MinC and MinD have been determined, the latter from three different thermophilic bacteria. MinC has a two-domain architecture in which a C-terminal dimerization domain is connected via a flexible linker to an N-terminal FtsZ binding domain (Fig. 3a) (Cordell et al., 2001). MinD is a single domain monomeric protein with a classical nucleotide-binding fold (Fig. 3b) (Cordell & Lowe, 2001; Hayashi et al., 2001; Sakai et al., 2001). The structures reveal that ATP binding and hydrolysis is not associated with significant conformational changes but that residues implicated in MinC binding map around the nucleotide-binding site. This indicates that ATP binding and hydrolysis regulate MinC–MinD interactions. The structure of a 55-residue fragment of MinE that encompasses the topological specificity domain has been determined by nuclear magnetic resonance spectroscopy. Two MinE fragments are associated in an antiparallel manner in an unusual α/β sandwich (King et al., 2000). Residues critical for topological function are restricted to the centre of the helical pair (Fig. 3c). These structures are an important resource and provide a framework for interpreting mutagenesis data but they provide few insights into how the Min proteins interact or into the mechanism underlying their unusual oscillation behaviour.
MinD–MinE oscillation has been analysed using simple reaction – diffusion mathematical models (Howard et al., 2001; Meinhardt & de Boer, 2001; Kruse, 2002; Howard & Rutenberg, 2003). The principal conclusion from this modelling effort is that oscillation is a spontaneous self-organized phenomenon arising from an intrinsic instability of the reaction – diffusion dynamics. A vital element of this instability is the varying diffusion constants of the MinCDE proteins: high in the cytoplasm, low when bound to the membrane. In addition, the oscillatory dynamics do not require any pre-existing topological markers to distinguish locations within the cell, as had previously been supposed. These mathematical models were further scrutinized and improved by including in vitro interaction data on MinD and MinE (Hu et al., 2002; Huang et al., 2003). The studies conclude that the oscillations are driven by a cycle in which MinD:ATP first associates with the membrane and then binds MinE. MinE activates ATP hydrolysis after which both MinE and MinD:ADP are released into the cytoplasm. The most recent mathematical model (Drew et al., 2005) is also able to explain the observed ordered cycles of assembly and disassembly of MinD polymers – these filamentous structures have been observed to form in vitro in the presence of ATP or GTP in one investigation (Suefuji et al., 2002), and in the presence of ATP and phospholipids vesicles in another (Hu et al., 2002).
It is clear that the Min system has a crucial role in the precise placement of the division site in E. coli. It has been proposed that the Min system serves as a ‘measuring stick’, capable of determining the distance to each pole using the oscillating behaviour of the Min proteins along their spiral trajectory (Addinall & Holland, 2002). Koch (1990) proposed that the nucleoid, localized to each end of the cell by the newly replicated origins, forms a symmetrical measuring device with its termini at the cell centre. This model implies the existence of specific sites at each pole for fixing the oriC regions of the two chromosomes. A variation of this model proposes that the cell poles periodically produce electrical or chemical signals, which are transmitted along the membrane or through the cytoplasm, thus defining the midcell site (Cook & Rothfield, 1999). As indicated above, such models simply alter the problem from one of delineating a specific site at midcell to one of defining a unique site at the cell poles from which these signals are generated. Although previously thought to be rather unlikely, such mechanisms probably operate even if we do not understand their molecular basis. The oscillation behaviour of the Min proteins establishes the idea that protein concentration gradients can in principle be the measuring stick or ‘measuring spiral’ sought after for many decades.
The MinCD–DivIVA system in B. subtilis
Bacillus subtilis, like E. coli and a wide variety of other prokaryotes, has MinC and MinD homologues, which are important for the prevention of asymmetric septation during vegetative growth. However, B. subtilis lacks MinE, and MinD oscillation is not observed (Marston et al., 1998; Marston & Errington, 1999b). The function of MinE in the topological control of MinCD activity is provided by DivIVA (Cha & Stewart, 1997; Edwards & Errington, 1997). DivIVA has no sequence similarity to MinE, it has a different quaternary structure (Zhang et al., 1998; Muchováet al., 2002; Stahlberg et al., 2004) and it clearly functions differently to MinE. DivIVA oligomers serve as building blocks in the formation of higher order assemblies, giving rise to two-dimensional lattices in a time-dependent manner (Stahlberg et al., 2004). DivIVA is stably associated with the cell poles, to which it recruits MinCD, probably by direct interaction with MinD (Marston et al., 1998; El Karoui & Errington, 2001). The zone occupied by MinD appears to radiate out from the pole to a greater extent than that of DivIVA, suggesting a cooperative self-interaction of MinD, which is perhaps nucleated at the pole by DivIVA. DivIVA requires FtsZ and other cell division proteins for its localization to the division sites late in their maturation during vegetative growth (Edwards & Errington, 1997; Marston et al., 1998). Unlike most division proteins, it is retained at the completed cell poles. In outgrowing spores, however, polar localization of DivIVA is independent of FtsZ localization (Hamoen & Errington, 2003; Harry & Lewis, 2003). Therefore, the requirement of FtsZ, and a second factor, penicillin-binding protein 2B (PBP 2B), for polar targeting of DivIVA during vegetative growth may be an indirect consequence of the role of these factors in the formation of the new cell pole. The polar localization of DivIVA is also independent of the cytoskeletal proteins Mbl and MreB (Hamoen & Errington, 2003). It seems therefore that, in contrast to midcell localization, polar localization of DivIVA does not need any known cell division or cytoskeletal protein. It is possible that DivIVA recognizes a general feature of cell poles as B. subtilis DivIVA can target poles in both E. coli and Schizosaccharomyces pombe, neither of which have DivIVA-like proteins (Edwards et al., 2000). These observations suggest an intriguing function for DivIVA-like proteins as morphogens capable of creating the cell polarity needed for sporulation in B. subtilis. It has been hypothesized that the two-dimensional DivIVA networks specifically recognize a characteristic of the cell's morphology, for example by developing a curvature that precisely matches that of the cell poles (Stahlberg et al., 2004). Subsequently, DivIVA sequesters other specific proteins to these sites.
The MinCD complex is also recruited to division sites, partly independently of DivIVA, though DivIVA is needed to bind the complex at the poles and block asymmetric division in the newly formed daughter cells (Marston et al., 1998; Marston & Errington, 1999b). As in E. coli, the Z-ring in B. subtilis is positioned at midcell with a high degree of precision. However, in B. subtilis, this is achieved by a Min system-independent mechanism (Migocki et al., 2002). These results support the idea that the primary role of the Min system during vegetative division in B. subtilis is to block Z-ring formation at the cell poles and that nucleoid occlusion ensures that cell division occurs precisely at midcell. Comparison of the Min systems of E. coli and B. subtilis with those of other bacteria raises interesting evolutionary questions concerning how the problem of precise division site selection is resolved. According to whole-genome sequence analysis, certain Clostridia species contain homologues of both MinE and DivIVA (Stragier, 2000). It is not known, however, if both are expressed and how or whether they function together. By contrast, other bacterial species, in particular Gram-positive cocci, lack the Min system altogether (Margolin, 2001).
FtsZ is the most highly conserved of the cell division proteins. It is present in most bacteria, archaea (Rothfield et al., 1999; Margolin, 2000) and it has an important role in plastid division in plants (Osteryoung et al., 1998). FtsZ is absent from most mitochondria despite the prokaryotic origin of these organelles. Z-ring formation establishes the location of the future division site and is an integral part of the temporal regulation of cytokinesis. The initiation of Z-ring assembly must be tightly controlled both temporally and spatially to prevent aberrant septation. In addition to the Min system and nucleoid occlusion, which have already been discussed, Z-ring formation can be influenced by the FtsZ concentration, and the initiation of DNA replication [reviewed in (Addinall & Holland, 2002; Romberg & Levin, 2003)].
Originally it was believed that the cellular FtsZ concentration could determine the timing of Z-ring formation because FtsZ assembles in vitro in a concentration-dependent manner and both E. coli and B. subtilis cells can modulate transcription of the ftsZ operon (Aldea et al., 1990; Joseleau-Petit et al., 1999; Weart & Levin, 2003). However, it was shown that the FtsZ to cell-mass ratio remains constant during a complete cell cycle, indicating that Z-ring assembly is not driven by changes in the concentration of this protein (Rueda et al., 2003). Artificially lowering the levels of ftsZ expression does not change the timing of medial ring formation, but it can cause slight increases in cell size (Palacios et al., 1996) while raising the expression levels can increase the incidence of polar Z-ring formation (Ward & Lutkenhaus, 1985). These data suggest that although high FtsZ expression levels do not affect medial Z-ring formation, they can overcome the action of MinC in preventing assembly of the division apparatus at the cell poles. Overall, therefore, fluctuations in FtsZ levels seem not to play an important role in the growth rate-dependent regulation of Z-ring formation in either E. coli or B. subtilis.
The third factor important for positioning the Z-ring at midcell is the initiation of DNA replication. Although blocking the initiation of DNA replication does not prevent Z-ring formation, it does seem to influence the positioning of the Z-ring (Sun et al., 1998; Sun & Margolin, 2001). Because Z-rings form proximal to the previous location of the origin of replication, it is possible that the same machinery that recruits the origin of replication and the DNA polymerase complex to midcell also helps to unmask the preferred site for FtsZ-ring assembly at this site.
A structural role for FtsZ was initially proposed based on its abundance in the cell and its localization by immunogold labelling to a ring structure at the future site of division (Bi & Lutkenhaus, 1991). FtsZ is a homologue of tubulin, the eukaryotic cytoskeletal protein involved in many essential processes including mitosis (Erickson, 1997). The crystal structure of FtsZ from Methanococcus janaschii (MjFtsZ; Fig. 3) confirmed the close relationship to tubulin and revealed a two-domain molecule with an N-terminal Ras-type GTPase domain and a C-terminal α/β domain, the two domains being arranged around a central helix (Fig. 3d) (Lowe & Amos, 1998). FtsZ polymerizes in a GTP-dependent manner into protofilaments that have structures that were assumed to be similar to those formed by tubulin (Fig. 4a). From nucleotide-free FtsZ obtained following an unfolding/refolding procedure, Löwe and coworkers were able to crystallize the protein as a semicontinuous protofilament closely resembling polymerized tubulin (Oliva et al., 2004). One of the molecular interfaces is especially tight and associated with a subunit spacing of 43 Å, which is similar to the spacing observed in electron micrographs of polymerized MjFtsZ. After soaking of GTP into these crystals, it is apparent that the GTP binding site in the N-domain of one subunit is completed by two key residues from the C-domain of the neighbouring subunit, which are positioned to polarize water for the GTPase reaction (Oliva et al., 2004). This work thus provides a model for the structure of the FtsZ polymer and illustrates how GTP binding and hydrolysis might be coupled to the assembly and disassembly of protofilaments.
What is the nature of the cell cycle signals that regulate FtsZ localization and its polymerization dynamics? There is, as yet, no clear answer to this question. The processes of FtsZ localization to the midcell site, its nucleation and polymerization at this site, and the subsequent constriction of the Z-ring are affected by the direct or indirect interactions of at least eight different proteins (MinC, FtsA, ZipA, EzrA, ZapA, Noc, SlmA and SulA). These proteins can inhibit polymerization of FtsZ, they can affect the position of the Z-ring, and they can change the dynamics of the Z-ring structure. Some of these proteins are present only in a subset of bacteria and it is likely that other, as yet unidentified, proteins influence Z-ring dynamics.
Vegetative septum formation in E. coli and in B. subtilis
In E. coli, cell division is a constriction process, involving three envelope layers – the outer membrane, the peptidoglycan layer and the cell membrane. After the selection of the septation site, the cell division process can be divided into four general steps: (1) assembly of an FtsZ–FtsA–ZipA complex; (2) interaction of the FtsZ–FtsA–ZipA complex with division proteins that provide a linkage to the cell membrane and cell envelope layers; (3) assembly of proteins with major extracellular domains such as PBPs, which control the synthesis of the cell wall and outer membrane material that will constitute the new cell poles; and (4) constriction and septum closure.
FtsA, an actin-like protein, localizes to and stabilizes the Z-ring (Ma & Margolin, 1999; van den Ent & Lowe, 2000). The interactions of FtsA and FtsZ have parallels with those of actin and tubulin, which are involved in cytokinesis in eukaryotes. Although lacking transmembrane domains, it has been shown that FtsA serves as the principal membrane anchor for the Z-ring (Pichoff & Lutkenhaus, 2005). Specifically, the conserved C-terminal amphipathic helix of FtsA is required for membrane targeting. FtsA is an ATPase and a key component in the sequential recruitment of other components of the divisome (Rothfield et al., 1999; Margolin, 2000). Unlike FtsZ and FtsA, the other cell division proteins have clear membrane-spanning sequences. The cellular FtsZ to FtsA ratio of 100 : 1 is important for correct division (Dai & Lutkenhaus, 1992; Dewar et al., 1992).
The constriction of the Z-ring requires that it be tightly bound to the membrane. An additional candidate for attaching the Z-ring to the membrane is ZipA (FtsZ interacting protein), which has been shown to interact with FtsZ and to stabilize the Z-ring immediately after it is assembled (Hale & de Boer, 1997). The structure of the C-terminal FtsZ-binding domain of ZipA in complex with a C-terminal peptide fragment of FtsZ has been solved (Mosyak et al., 2000; Moy et al., 2000), revealing that ZipA has a six-stranded β-sheet, one face of which packs against three α-helices, while the opposite face forms a shallow hydrophobic groove in which the FtsZ peptide is bound. ZipA localizes to the Z-ring in an FtsA-independent manner (Hale & de Boer, 1999; Liu et al., 1999). Interestingly, Z-rings can form in the absence of either ZipA or FtsA but not in the absence of both (Pichoff & Lutkenhaus, 2002). The C-terminal tail of FtsZ forms part of the binding site for both ZipA and FtsA; however, because the number of FtsZ molecules in the cell vastly exceeds the number of ZipA and FtsA molecules, it is unlikely that the latter pair compete for FtsZ binding sites. The assembly of the Z-ring thus depends on either or both of FtsA and ZipA, while the localization of the latter pair of proteins depends on FtsZ. Although ZipA and FtsA play partially redundant roles in Z-ring assembly, both are needed for septal constriction. The Z-ring may play a role in both the assembly and the constriction of the septal ring (Erickson et al., 1996). In E. coli, the division proteins are recruited into the ring according to a sequential and almost linear pathway as follows: (FtsE, FtsX), FtsK, FtsQ, (FtsB, FtsL), FtsW, FtsI, FtsN, AmiC and EnvC, where the proteins in parentheses assemble simultaneously (Vicente & Rico, 2006). Thus, the first of these proteins, FtsE and FtsX, do not require any downstream proteins to assemble at the Z-ring. The localization of the last protein from this set, EnvC, depends on all of the other proteins. This hierarchical localization of division proteins in E. coli is likely to reflect a sequence of protein–protein interactions that leads to the assembly of a protein complex(es) of the divisome.
Bacillus subtilis has homologues of most of the E. coli division proteins, including FtsZ, FtsA, ZipA (a possible functional homologue of B. subtilis EzrA), FtsL, FtsB (DivIC in B. subtilis), FtsQ (DivIB in B. subtilis), FtsW (YlaO in B. subtilis) and Pbp3 (Pbp2B in B. subtilis) [reviewed in Errington et al., 2003]. An additional protein, SepF, is present only in B. subtilis (Hamoen et al., 2006). Recently it was shown that FtsZ and FtsA associate prior to Z-ring formation at the midcell site and that FtsA is required for efficient Z-ring formation in B. subtilis (Jensen et al., 2005). There are two additional major differences in vegetative septum formation in these two microorganisms. First, the E. coli divisome comprises several additional proteins. Higher complexity could be a consequence of the fact that in E. coli cytokinesis requires constriction of peptidoglycan layers, the outer membrane and the inner membrane, in comparison with B. subtilis which has only one membrane and a thick peptidoglycan layer. Secondly, in E. coli the recruitment of the different components seems to occur in an ordered manner, so that the depletion of one cell division protein leads to failure to recruit the downstream proteins in the interaction pathway. This is in contrast to B. subtilis in which the equivalent division proteins are recruited in a more concerted manner [reviewed in Errington et al., 2003]. DivIB, DivIC, FtsL, Pbp2B and probably YlaO are all completely interdependent in their assembly at the division site and depletion of FtsA, DivIC, FtsL or Pbp2B abolishes the positioning of the other cell division proteins at midcell.
Asymmetric cell division during sporulation of B. subtilis
In addition to midcell septation, a completely different type of cell division is observed in rod-shaped Bacillus and Clostridia species during sporulation. Sporulation is one of three distinct cell division processes in the life cycle of these organisms, the other two occurring during vegetative growth and spore germination (Fig. 5). In nutritionally rich media, B. subtilis undergoes longitudinal growth and midcell division as described above. Upon starvation, one of several adaptations available to B. subtilis is to form a spore.
Upon entry into sporulation, the replication origin regions of the two daughter chromosomes in the predivisional cell migrate to opposite poles (Fig. 5b) (Glaser et al., 1997; Lin et al., 1997). Electron microscopy of thin cell sections at this stage shows axial filament formation by the chromatin. The first clear morphological step in sporulation is the formation of an asymmetric septum (Fig. 5c). This septum divides the cell into a larger cell, called the mother cell, and a much smaller cell, called the prespore or forespore. The forespore is next engulfed by the mother cell in a process resembling eukaryotic phagocytosis (Fig. 5e). In the later stages, the forespore develops into a very resistant spore and the mother cell lyses (Fig. 5f). The spore can remain dormant indefinitely and germinate and resume vegetative growth when living conditions improve (Fig. 5g).
Sporulation is activated by complex regulatory circuits and begins when a threshold concentration of phosphorylated Spo0A is reached. Spo0A phosphorylation is under the control of an expanded two-component sensory signalling system termed a phosphorelay (Hoch, 1998), in which environmental signals trigger the ATP-dependent autophosphorylation of one or more of five sensor kinases (Stephenson & Hoch, 2002). The phosphate group is relayed via Spo0F∼P and Spo0B∼P intermediates onto Spo0A. Phosphorylation leads to the formation of Spo0A dimers that can activate the transcription of scores of genes (Asayama et al., 1995; Lewis et al., 2002). Activated Spo0A is required for the switch from symmetric to asymmetric septation at the onset of sporulation (see below). Through its regulation of expression of specific sets of genes, Spo0A also regulates chromosome partitioning, helps to establish differential gene expression in the forespore and, as was shown recently, becomes a compartment-specific transcription factor in the mother cell (Fujita & Losick, 2003). Spo0A is therefore an extraordinarily versatile transcriptional regulator.
A great deal is known about the regulation of gene expression during sporulation in B. subtilis. Spo0A brings about global changes in gene expression in the cell (Fawcett et al., 2000; Molle et al., 2003). Spo0A∼P represses transcription of many vegetative genes and initiates the transcription of early sporulation-specific genes. Gene expression is subsequently orchestrated through the activity of four compartment-specific sigma factors (σF, σE, σG and σK), activated either in the forespore or in the mother cell. Morphological changes and the programme of gene expression are closely linked throughout sporulation to ensure that genes are expressed at the proper time, at appropriate levels and in the right compartment. Each of the four compartment-specific σ factors (σF, σE, σG and σK) is inactive at the time of its synthesis and requires subsequent activation [recently reviewed in (Errington, 2003; Hilbert & Piggot, 2004)].
Chromosome segregation during sporulation
Cell division and the faithful segregation of the newly duplicated chromosomes to each daughter cell must be coordinated. This is ensured in vegetatively growing B. subtilis cells by complex mechanisms. The switch to polar cell division during sporulation has an interesting consequence for chromosome segregation. At the beginning of sporulation, instead of segregating, the two chromosomes form an elongated structure known as the axial filament. oriC localization experiments using Lac repressor (targeted to tandem copies of lacO sequences placed near oriC) and the chromosome segregation protein Spo0J (by means of GFP fusions) revealed that axial filament formation is accompanied by migration of the oriC regions to the opposite poles of the cell (Fig. 5b) (Glaser et al., 1997). Interestingly, the chromosome anchoring sites are in a region lying about 150–300 kbp from oriC. A protein candidate for recruiting the chromosomes to the poles is DivIVA (Quisel et al., 1999).
The movement of the two chromosomes is under the control of the phosphorelay system (Graumann & Losick, 2001). Three DNA-binding proteins are involved–Spo0J, RacA and Soj. Spo0J colocalizes with the oriC region, and in spo0J mutants the prespore chromosome is misorientated (Sharpe & Errington, 1996). Soj displays a remarkable pattern of subcellular localization. This protein can oscillate from one pole to the other with a period of c. 20 s (Quisel et al., 1999), a movement which is dependent on oriC-bound Spo0J (Marston & Errington, 1999a). The Spo0J concentration increases in the early stages of sporulation. In the absence of Spo0J, Soj localizes to the nucleoid where it preferentially binds to early sporulation promoters (Cervin et al., 1998). In vitro experiments have shown that Soj inhibits Spo0A-activated transcription (Cervin et al., 1998). Thus, the nucleoid-associated form of Soj is expected to prevent expression of Spo0A∼P-regulated genes in vivo. By contrast, it is believed that the oscillating form of Soj does not repress Spo0A-dependent promoters. The Spo0J/Soj system thus ensures that incomplete partitioning of the chromosomes blocks expression from Spo0A-dependent promoters effectively operating as a checkpoint that couples chromosome partitioning to developmental gene expression. RacA also binds to the oriC region of the chromosomes and forms a component of the mechanism that attaches them to the cell poles, possibly through contacts to DivIVA, which is already localized at the cell pole (Fig. 5b) (Ben Yehuda et al., 2003; Wu & Errington, 2003).
The sporulation septum bisects the axial filament leaving only one-third of one chromosome in the forespore and creating a transient genetic asymmetry (Wu & Errington, 1994). The remaining two-thirds of the chromosome is transferred, over a period of 10–20 min, from the mother cell into the forespore by a conjugation-like mechanism directed by the partitioning protein SpoIIIE (Wu & Errington, 1997). SpoIIIE has an ATP-dependent DNA-tracking activity and a hydrophobic amino-terminal domain which targets it to the sporulation septum (Bath et al., 2000). A later function of SpoIIIE in the membrane-fusion process during forespore engulfment has also been suggested (Sharp & Pogliano, 1999).
Asymmetric septum formation
The first observable morphological step in sporulation is the formation of the asymmetric septum (Fig. 5c). In B. subtilis, the cell division septum is normally a thick structure with a substantial amount of peptidoglycan separating the two daughter cells at cytokinesis. By contrast, the asymmetric septum formed during sporulation is much thinner and most of the peptidoglycan separating the two lipid bilayers that comprise the septal structure is removed soon after septation is complete. The pliable septal structure then migrates towards the nearer pole of the cell, so as to engulf the forespore, which matures subsequently into a dormant endospore. Based on electron micrographs of sporulating cells, it has been suggested that localized regions of lipid bilayer fusion occur within the septal structure following removal of the peptidoglycan. This might be significant in relation to communication between the two sporangium compartments (Higgins & Piggot, 1992; Sharp & Pogliano, 1999).
The proper positioning of the sporulation septum is dependent on Spo0A, which is required for the assembly of the cell division proteins near the cell pole instead of at midcell. Spo0A indirectly controls migration of the Z-ring from the midcell, where it is initially assembled, to polar sites through a spiral intermediate (Ben Yehuda & Losick, 2002). This positional switch may be partially triggered by the activity of SpoIIE, which is expressed under Spo0A control (Barak & Youngman, 1996; Feucht et al., 1996; Khvorova et al., 1998). SpoIIE has a central domain which is involved in oligomerization and in interactions with FtsZ (Fig. 4b) (Lucet et al., 2000). Experiments with SpoIIE-GFP fusions in B. subtilis (Arigoni et al., 1995) and Bacillus megaterium (Barak et al., 1996) using fluorescence microscopy showed localization of SpoIIE to the asymmetric sporulation septum in the form of a ring. A number of spoIIE mutants, such as spoIIE20, spoIIE21 and spoIIE60, give rise to thick asymmetric septa reminiscent of vegetative septa. Relative to the wild-type strain, these septa are reduced in numbers, and delayed in their formation (Illing &Errington, 1991; Barak & Youngman, 1996; Feucht et al., 1996). SpoIIE is the only known sporulation-specific protein whose deletion or mutation is associated with changes in the ultrastructural features of the asymmetric septum. These results imply that SpoIIE can specifically recognize the sporulation septation site and initiate the formation of sporulation septa. Otherwise the available evidence suggests that asymmetric septation is a modified form of vegetative septation that uses the same basic cell division machinery.
The question of how asymmetric division sites are recognized and established is intriguing. Evidence that cells contain potential division sites at the poles has come primarily from studies of minicell mutants (Rothfield & Zhao, 1996). Experiments in which plasmid and chromosomal DNA sequences were localized very early in a round of replication in E. coli and B. subtilis revealed that these cells have identifiable 1/4 and 3/4 cellular positions (Niki & Hiraga, 1997; Li & Austin, 2002; Lau et al., 2003). It is reasonable to assume that these sites are also used for the placement of the Z-ring at asymmetric positions during sporulation.
What is known about the activation of asymmetric division during sporulation? Spo0A and σH are clearly involved in shifting cell division to the polar sites (Levin & Losick, 1996; Ben Yehuda & Losick, 2002). It was previously suggested that the switch is effected primarily by blocking Z-ring formation at midcell (Levin & Losick, 1996; Barak et al., 1998). Asymmetric division in sporulation occurs at about the time medial division would have occurred had vegetative growth continued. Therefore, at the point at which sporulation is initiated, medial division has to be blocked. In strains harbouring spo0A null mutations, Z-ring relocalization does not take place and division occurs at the midcell site (Levin & Losick, 1996). As mentioned above, the switch from medial to polar Z-rings in wild-type cells is accomplished via spiral-like FtsZ intermediate structures emanating from midcell outward towards the cell poles where they are converted into bipolar rings (Fig. 5b) (Ben Yehuda & Losick, 2002). SpoIIE plays a crucial role in this process, possibly by activating bipolar Z-ring formation and/or by stabilizing them once formed. SpoIIE may, in addition, play a role similar to ZipA in medial division of E. coli by serving as a membrane anchor for the Z-ring.
Following FtsZ and SpoIIE ring formation, the entire division machinery is assembled at the cell poles. An interesting characteristic of this process is that there are two potential division sites, one at each cell pole and indeed both are used for septum formation in disporic mutants, such as strains defective in spoIIE, spoIIAA, spoIIAC (encoding σF), spoIIGA or spoIIGB (encoding σE). It is crucial in wild-type cells that division occurs at only one of these sites and this is ensured by one or more σE-dependent gene products which block the maturation of the second polar division site (Eichenberger et al., 2001).
Although asymmetric cell division during sporulation resembles vegetative division, little is known of the mechanisms involved in accurate placement of the division machinery. First, it does not appear to involve the Min system because mutations in minC and minD have little effect on the sporulation frequency (Li & Austin, 2002; Huang et al., 2003; Drew et al., 2005). It is not possible, however, to exclude a partial role of the Min system during sporulation in light of the fact that in a small proportion of minD mutant cells, a sporulation-like septum is misplaced from its normal polar site (Barak et al., 1998; Thomaides et al., 2001). Secondly, during sporulation nucleoid occlusion is overcome and the polar septum constricts around the nucleoid. The nature of the effector that overcomes or eliminates the spatial veto exerted by the nucleoid is not known.
Although the spatial regulation of vegetative and sporulation cell division differs significantly, both processes use essentially the same protein machinery with the exception of SpoIIE, which is a specific component of the sporulation septum. The hierarchy of assembly of midcell division machinery and sporulation division machinery also appears to be similar. Asymmetric septum formation has additional consequences as it leads to the establishment of differential gene expression in spore-forming bacilli, which enables them to survive unfavourable conditions.
Cell division, as a fundamental cellular process, still holds many secrets that are waiting to be revealed. The major current challenges lie in understanding the assembly and disassembly of protein complexes at the site of division. Unravelling the molecular mechanisms of these processes will require state-of-the-art experimental methods for defining the order of assembly, the stoichiometry and the overall structures of cell division protein complexes, as well as biophysical and cytological methods to explain such phenomena as asymmetry of protein localization, protein oscillation and protein spiral formation. Such experiments should help us to understand the molecular device that allows the cell to recognize with high precision its middle during vegetative growth, or polar division sites during sporulation.
Cell division complexes are by their nature very large macromolecular assemblies, which present formidable challenges to structural characterization. The isolated components of these complexes have also proved to be difficult to work with owing to complexities associated with their overexpression, their stability, and their intrinsic tendency to oligomerize and aggregate. A high proportion of the cell division protein structures determined to date are of orthologues from thermophilic bacteria. Mutagenesis studies guided by these structures confirm their relevance and provide links to the genetic and cytological studies, which have largely been carried out in E. coli and B. subtilis. In this way, the active surfaces on many of these molecules have been mapped and this information is being complemented by crystal structures of binary protein complexes, which reveal the details of molecular associations. Structures of larger assemblies represent the future challenge and it is anticipated that these will combine data from multicomponent crystals and reconstruction using images generated by electron microscopy.
Work in our laboratory is supported by grant NMP4-CT-2004-013523 from the EC 6th Framework, by grant 2/1004/21 from the Slovak Academy of Sciences, by grant APVT-51-027804 from the Ministry of Education of Slovak Republic and by grant ERAS-CT-2003-980409 from the EC 6th Framework as part of the European Science Foundation EUROCORES Programme EuroSCOPE (I.B.) and by grants from the Wellcome Trust, the BBSRC, UK and the EC (A.J.W.).