Cell division in bacteria is mediated by the septal ring, a collection of about a dozen (known) proteins that localize to the division site, where they direct assembly of the division septum. The foundation of the septal ring is a polymer of the tubulin-like protein FtsZ. Recently, experiments using fluorescence recovery after photobleaching have revealed that the Z ring is extremely dynamic. FtsZ subunits exchange in and out of the ring on a time scale of seconds even while the overall morphology of the ring appears static. These findings, together with in vitro studies of purified FtsZ, suggest that the rate-limiting step in turnover of FtsZ polymers is GTP hydrolysis. Another component of the septal ring, FtsK, is involved in coordinating chromosome segregation with cell division. Recent studies have revealed that FtsK is a DNA translocase that facilitates decatenation of sister chromosomes by TopIV and resolution of chromosome dimers by the XerCD recombinase. Finally, two murein hydrolases, AmiC and EnvC, have been shown to localize to the septal ring of Escherichia coli, where they play an important role in separation of daughter cells.
This MicroReview opens with a whirlwind tour of cell division in Escherichia coli. To keep it short and to the point, I have made some oversimplifications and relied almost exclusively on citations to a collection of review articles rather than the original studies. The remainder of the review is devoted to a more detailed and critical look at a few questions that have been the focus of several recent studies. (i) How do the proteins that comprise the septal ring work together during cytokinesis? One step towards answering this question is to identify protein–protein interactions among the division proteins. I will summarize what is known about this topic and suggest reasons why progress has been slow. (ii) How does FtsZ drive cytokinesis? Although the answer remains elusive, recent experiments suggest the rate-limiting step in turnover of the FtsZ polymer is GTP hydrolysis. I will speculate on the implications that this has for the ability of FtsZ to generate force. (iii) How is septum assembly coordinated with partitioning of chromosomes to daughter cells? At least three mechanisms by which FtsK prevents the septum from closing on the chromosomes like a guillotine have been elucidated. I will describe these activities, and briefly compare FtsK with a similar protein, SpoIIIE, involved in sporulation in Bacillus subtilis. (iv) How are daughter cells separated? Two peptidoglycan hydrolases, AmiC and EnvC, have recently been shown to localize to the septal ring and play an important role in this process in E. coli. I will summarize some of what is known about AmiC and note parallels to the Atl murein hydrolase of Staphylococcus aureus.
Cell division explained in 500 words
In E. coli, cell division is mediated by a collection of proteins that localize to the division site, where they appear to assemble into a multiprotein complex called the septal ring (Fig. 1) (for recent reviews, see Rothfield et al., 1999; Margolin, 2000; Errington et al., 2003; Ryan and Shapiro, 2003). The process starts with polymerization of the tubulin-like protein FtsZ into the Z ring. The Z ring is the heart of the division apparatus – it serves as a landing pad for recruitment of other proteins to the division site, and might also use energy from GTP hydrolysis to drive cytokinesis. The other proteins that comprise the septal ring fall into several functional classes: (i) modulating the assembly state of FtsZ (FtsA, ZipA, ZapA), (ii) connecting the Z ring to the cytoplasmic membrane (FtsA, ZipA), (iii) coordinating septation with chromosome segregation (FtsK), (iv) synthesis of peptidoglycan cell wall (FtsI, FtsW) and (v) hydrolysis of peptidoglycan to separate daughter cells (AmiC, EnvC). The septal ring also contains many proteins of essentially unknown function [FtsEX, FtsQ, FtsL, FtsB (formerly called YgbQ) and FtsN].
Cell division proceeds by the concerted inward growth of all three layers of the cell envelope – the cytoplasmic membrane, peptidoglycan wall and outer membrane. A plausible model for this process, first put forward by Hale and de Boer (1997), can be updated as follows. Constriction of the Z ring pulls the cytoplasmic membrane inward. Proteins involved in peptidoglycan synthesis, such as FtsI and FtsW, reside in the cytoplasmic membrane, so when the membrane invaginates, peptidoglycan synthesis follows. Even as the peptidoglycan layer grows inward, hydrolases like AmiC are at work splitting the septal murein to separate daughter cells. Finally, the peptidoglycan layer is connected to the outer membrane by a variety of bridging proteins such as Lpp (Braun's lipoprotein). Thus, the outer membrane is expected to follow the peptidoglycan passively as new bridging proteins are incorporated in the wake of the peptidoglycan synthases.
Spatial and temporal regulation of cell division is accomplished primarily at the level of Z ring assembly. Two mechanisms are particularly important: inhibition of Z ring assembly at the midcell by nucleoid occlusion and inhibition of Z ring assembly at the poles by the Min proteins. Nucleoid occlusion refers to the observation that Z ring assembly is inhibited in the vicinity of the nucleoid, for reasons that are obscure. The MinC and MinD proteins form a complex, MinCD, which binds to FtsZ and prevents Z ring formation. In E. coli, MinCD has the astonishing property of oscillating from pole to pole with a period of ≈30 s, a process that requires a third protein, MinE. Oscillation involves redistribution of the Min proteins among coiled polymers that extend along the inner surface of the cytoplasmic membrane (Shih et al., 2003). In a variation on this theme, MinCD of B. subtilis remains statically associated with the poles; this requires a protein called DivIVA that is not homologous to MinE of E. coli. The net effect of MinCD oscillation in E. coli and polar sequestration in B. subtilis is a time-averaged concentration minimum at the midcell (Meinhardt and de Boer, 2001; Howard et al., 2001; Howard, 2004). Thus, nucleoid occlusion and the Min system work together to ensure that the only permissive site for Z ring assembly is the DNA-free region that opens up at the midcell as chromosomes begin to segregate to incipient daughter cells.
Assembly of the septal ring: a model for protein–protein interactions?
Identifying and characterizing interactions among the proteins that constitute the septal ring is important, as this information should help us to understand how these proteins work together during cytokinesis. Studies of protein localization in various E. coli mutant backgrounds have revealed a set of dependencies that imply the various division proteins localize to the septal ring in a defined order (Fig. 1) (Buddelmeijer and Beckwith, 2002; Errington et al., 2003). The process starts with polymerization of FtsZ into a contractile ring structure at the inner face of the cytoplasmic membrane. FtsA and ZipA bind directly to FtsZ, and localize next, presumably as the Z ring is assembling. Z rings can assemble in mutants that lack FtsA or ZipA, but Z rings are not observed in the absence of both proteins (Pichoff and Lutkenhaus, 2002). The ZapA protein, recently discovered in B. subtilis, also binds directly to FtsZ, so one would expect it to localize as the Z ring is assembling, but this has not been studied yet. Once both FtsA and ZipA are in place, the remaining proteins are recruited in the following order: [FtsE + FtsX] (probably an ABC transporter complex) → FtsK → FtsQ →[FtsL + FtsB] (probably a heterodimer) → FtsW → FtsI → FtsN → AmiC. This order of recruitment is generally interpreted to reflect the assembly pathway for a multiprotein complex, and thus makes predictions about which proteins interact.
Several negative regulators of Z ring assembly have been shown to interact with FtsZ directly, including SulA, MinC and EzrA (Hu et al., 1999; Cordell et al., 2003; Haeusser et al., 2004). SulA and MinC are not part of the septal ring, but the distritubtion for EzrA is more complex. EzrA is a negative regulator of Z ring assembly found throughout the low-GC Gram-positive bacteria. In non-dividing B. subtilis cells, EzA is distributed around the cytoplasmic membrane, where it helps prevent Z rings from assembling at inappropriate sites such as the poles. But in dividing cells a significant fraction of the EzrA pool is recruited to the septal ring in an FtsZ-dependent manner (Levin et al., 1999).
Despite some recent progress towards identifying interactions among the division proteins, it is remarkable that so many of the expected interactions have yet to be observed. To some extent this situation may reflect technical challenges and a lack of effort, but several observations suggest that interactions among the division proteins may be more complex than implied by the linear order of recruitment observed in E. coli. (i) The FtsQ–FtsL–FtsB complex mentioned above was detected even in cells depleted of FtsK (Buddelmeijer and Beckwith, 2004), which is required for localization of FtsQ and downstream proteins to the septal ring. Apparently FtsQ, FtsL and FtsB are recruited to the septal ring as a preformed complex, despite the fact that FtsQ can localize to the septal ring without FtsL or FtsB. (ii) Overproduction of several proteins, including FtsI and FtsQ, by ≈ 50-fold does not interfere with division, even though much of the ‘extra’ protein is found delocalized around the membrane or mislocalized to the poles (Guzman et al., 1997; Weiss et al., 1999; Boyd et al., 2000). This observation is difficult to reconcile with high-affinity pair-wise interactions, because that should result in sequestration of downstream division proteins. Rather, it seems likely that some division proteins only interact strongly in the context of the septal ring, perhaps because each protein binds weakly to two or more proteins. (iii) Most of the septal ring proteins are associated with the cytoplasmic membrane. Colocalization of proteins to the membrane increases their local concentration and might allow weak protein–protein interactions to drive assembly of the septal ring. However, such weak interactions might be difficult to detect in assay systems where the proteins are not tethered in close proximity. (iv) The septal ring is a transient structure. It disassembles by the end of constriction, and at least some of its components undergo constant turnover even when the ring is present (see below). The cooperativity that results from a network of interactions would seem well suited for providing both high affinity and facile disassembly. (v) Most of the same division proteins exist in B. subtilis, but exhibit considerable interdependence, as if assembly of the septal ring is a highly concerted process in that organism (reviewed in Errington et al., 2003). (vi) It is not yet known whether the proteins in the current model represent a complete set. Six new proteins (FtsB, ZapA, AmiC, FtsE, FtsX and EnvC) have been added to the picture just within the last 3 years (Buddelmeijer et al., 2002; Gueiros-Filho and Losick, 2002; Bernhardt and de Boer, 2003; Schmidt et al., 2004; Bernhardt and de Boer, 2004).
The assertion that most interactions among the late proteins have yet to be demonstrated needs to be qualified, as there is a recent report of numerous interactions – some expected, some not – using a bacterial two-hybrid system based on reconstitution of a phage repressor in E. coli (Di Lallo et al. 2003). One concern regarding the interpretation of these experiments is that fusions were made to full-length division proteins. Thus, the assay might be detecting co-assembly of proteins into the septal ring rather than simple pair-wise interactions. This caveat aside, reconstitution of phage repressor function could reflect direct interactions and implies at the very least that the respective division proteins are in very close proximity, which is more than could be inferred from observations of colocalization by fluorescence microscopy.
Structure and function of FtsZ
Of the dozen or so proteins that comprise the septal ring, the most highly conserved is FtsZ. Homologues of FtsZ exist in almost all bacteria, many archaea, some chloroplasts and a few primitive mitochondria (Vaughan et al., 2004). FtsZ is a prokaryotic homologue of tubulin, one of several major cytoskeletal proteins in eukaryotic organisms. Like tubulin, purified FtsZ exhibits GTPase activity and undergoes reversible, GTP-dependent polymerization into filaments. In vivo, FtsZ forms a contractile ring at the division site (Bi and Lutkenhaus, 1991). As division proceeds, the Z ring constricts, so as to remain at the leading edge of the developing septum.
FtsZ of E. coli appears to comprise four domains (Vaughan et al., 2004). There is a short and poorly conserved N-terminal leader (≈15 residues), a highly conserved domain (≈300 residues) that is structurally and functionally homologous to tubulin, a poorly conserved linker (≈50 residues) and a well-conserved C-terminal tail (≈15 residues). The C-terminal tail is the binding site for two division proteins, FtsA and ZipA (Din et al., 1998; Ma and Margolin, 1999; Hale et al., 2000; Yan et al., 2000; Haney et al., 2001). As noted above under Assembly of the septal ring, FtsA and ZipA are in turn required for recruitment of downstream division proteins. Recently, a mutant form of FtsA that can support division in the absence of ZipA has been described (Geissler et al., 2003). The existence of this bypass mutant may simplify efforts to study septal ring assembly and function in vitro and might also explain why most bacteria lack a clear ZipA homologue.
The three-dimensional structures of FtsZ from Methanococcus jannaschii and the αβ-tubulin heterodimer from beef brain were reported simultaneously, and offered important insights into how these proteins function (for reviews, see Nogales et al., 1998; van den Ent et al., 2001; Romberg and Levin, 2003). Although FtsZ and tubulin have < 20% amino acid identity, they adopt very similar folds, especially in the vicinity of the GTP binding site, where not only the fold but also several critical amino acid residues are highly conserved. GTP binds at one end of the FtsZ monomer, designated the ‘plus’ end by convention. Addition of another monomer completes the GTPase catalytic site; it is the minus end of the second monomer that interacts with GTP and the preceding subunit. This binding geometry can be used to rationalize why GTP promotes polymerization of FtsZ and why hydrolysis of GTP is associated with depolymerization. Moreover, the fact that amino acids involved in GTP hydrolysis are derived from both subunits explains why FtsZ assembly is necessary for GTP hydrolysis. Interestingly, the minus end of FtsZ is bound by the division inhibitor SulA, which prevents FtsZ polymerization by blocking FtsZ–FtsZ interaction (Cordell et al., 2003).
The FtsZ polymer in the Z ring has yet to be observed in electron micrographs. There are several potential reasons for this – it may be too small, the bacterial cytoplasm may be too dense to provide contrast, and/or the Z ring might not have a repetitive and ordered structure that would make it visually distinctive. In the absence of a telltale micrograph, conjecture as to the structure of the FtsZ polymer in the Z ring has been guided by studies of FtsZ polymers formed in vitro (Romberg and Levin, 2003). Purified FtsZ assembles in a GTP-dependent fashion into protofilaments, linear arrays of FtsZ molecules stacked end to end. Protofilaments formed of GTP-bound FtsZ are straight, whereas those of GDP-bound FtsZ are curved (Lu et al., 2000). Because FtsZ protofilaments are observed under a variety of conditions, and because similar protofilaments are formed by tubulin, there is wide agreement that protofilaments are the fundamental building block of the Z ring. About one-third of the cellular FtsZ pool is present in the Z ring, making it six to eight protofilaments wide in some E. coli strains (Stricker et al., 2002). How the protofilaments are arranged is very much an open question. Because the lateral surfaces of FtsZ are different from those of tubulin, FtsZ is unlikely to form structures similar to microtubules (Nogales et al., 1998). Nevertheless, protofilaments associate laterally into a variety of higher-order structures in vitro, including pairs of protofilaments that form tight spirals, sheets of parallel and anti-parallel protofilaments, and bundles of protofilaments that are arranged in complex ways. Formation of these higher-order structures is promoted by cations like Ca2+, Mg2+ and DEAE-dextran. Interestingly, two division proteins, ZipA and ZapA, have been observed to promote bundling in vitro, fuelling speculation that they do so in vivo as well (RayChaudhuri, 1999; Hale et al., 2000; Gueiros-Filho and Losick, 2002). Thus, FtsZ bundles might be stabilized by lateral FtsZ–FtsZ interactions, by cross-linking proteins, or a combination of the two. With so many higher-order structures to choose from, it is difficult to say which, if any, is physiologically relevant. The recent report of a mutant form of FtsZ that fails to support division and can assemble in vitro into protofilaments but not Ca2+-induced bundles strongly suggests that a higher-order FtsZ structure is involved in cell division (Koppelman et al., 2004).
There has been some controversy as to whether protofilaments assembled in vitro consist primarily of FtsZ-GTP or FtsZ-GDP (Mingorance et al., 2001; Scheffers and Driessen, 2002). A recent study has concluded that FtsZ-GTP predominates (Romberg and Mitchison, 2004). Assuming that this result holds up to further analysis, why might it be interesting? As noted by Romberg and Levin (2003), a polymer composed of FtsZ-GTP is stable, whereas a hypothetical FtsZ-GDP polymer is likely to be metastable – recall that this is the form that favours depolymerization. In other words, an FtsZ-GDP polymer would be like a loaded spring that could disassemble rapidly, releasing the energy from many GTP hydrolysis events all at once. But if protofilaments consist primarily of FtsZ-GTP, they would presumably release energy in small quanta and are not well suited for generating large forces. This is different from microtubules, which are a metastable assemblage of tubulin-GDP with a tubulin-GTP ‘cap’ at the plus end. Perhaps microtubules need to generate larger forces because they operate on a greater geometric scale than does the Z ring.
Because the studies of what nucleotide is bound to FtsZ in protofilaments have been carried out with purified FtsZ in vitro, it must be noted that FtsZ polymers in vivo could behave differently. In particular, they might consist of FtsZ-GDP if the protofilaments engage in lateral interactions that stabilize FtsZ-GDP assemblies, as is the case for tubulin-GDP in microtubules. Alternatively, Z ring-associated proteins might change the relative rates of the various steps in the GTPase cycle. Despite these caveats, studies of FtsZ dynamics in vivo support the notion that GTP hydrolysis is the rate-limiting step in the cycle of FtsZ assembly and disassembly in the Z ring (see below).
Several mechanisms have been suggested for how FtsZ might use GTP hydrolysis to power constriction of the Z ring during septum assembly (reviewed in Ryan and Shapiro, 2003). These are: (i) depolymerization, (ii) sliding of stable FtsZ filaments against each other by a hypothetical motor protein functioning as a ratchet and (iii) curvature of the polymer upon hydrolysis of GTP. While each of these models is reasonable, it is worth noting that it has yet to be established whether, let alone how, FtsZ powers cytokinesis. It is widely assumed that the septal ring must constrict against turgor pressure, which exerts an outward force of approximately three atmospheres on the cytoplasmic membrane of E. coli (Cayley et al. 2000). However, cell division is accomplished by synthesis of new cell envelope material, so it is not clear to what extent turgor pressure resists invagination of the cell envelope. Another reason for suspecting that FtsZ provides energy for constriction is the analogy to eukaryotic cytoskeletal elements such as actomyosin and microtubules that have been shown unambiguously to do work in animal cells. But while the eukaryotic proteins are perhaps best known for their capacity to generate force, this is not their only use (Desai and Mitchison, 1997). Microtubules, for instance, take advantage of GTP hydrolysis-driven cycles of polymerization and depolymerization to probe the cytosol as they attempt to capture chromosomes during the early stages of mitosis. By analogy, the ultimate purpose of GTP hydrolysis by FtsZ might be to allow facile assembly and disassembly rather than exert force on the cell envelope. Of course, these functions of GTP hydrolysis are not mutually exclusive.
A related issue is whether energy might indeed be needed for constriction, but something other than FtsZ is the source. The only alternative that has received significant attention is the idea that inward synthesis of the peptidoglycan cell wall might ‘push’ cytokinesis. At present, this seems unlikely because invagination of the cytoplasmic membrane has been observed to continue in E. coli and B. subtilis even after synthesis of septal peptidoglycan has been blocked (Daniel et al., 2000; Heidrich et al., 2002). Moreover, some bacteria lack a cell wall, but use a Z ring for cell division (Wang and Lutkenhaus, 1996).
For the Z ring to exert force on the cell envelope, it needs a solid connection to the cytoplasmic membrane. There appear to be redundant mechanisms for this. ZipA is an integral membrane protein and binds directly to FtsZ (Hale and de Boer, 1997), but is only found in a subset of the proteobacteria and even there the requirement for ZipA can be bypassed by a mutation in ftsA (Geissler et al., 2003). FtsA also binds directly to FtsZ. Although FtsA is nominally a soluble protein, some of it fractionates with cytoplasmic membrane in E. coli, suggesting that FtsA is a peripheral membrane protein (Sanchez et al., 1994). Finally, Koppelman et al. (2004) recently demonstrated that FtsZ can bind to inverted vesicles derived from E. coli cells, indicating that FtsZ itself has affinity for the membrane or a protein(s) in the membrane.
The Z ring undergoes rapid turnover
The Z ring is like a duck in the water – calm above the surface, paddling like mad underneath. The ‘calm surface’ is a cycle of assembly, persistence, constriction, and disassembly that was revealed by numerous studies using FtsZ-GFP or immunofluorescence microscopy (e.g. Addinall et al., 1996; Sun and Margolin, 1998). Interestingly, such studies also revealed that Z rings assemble as much as 20 min before they constrict when growth is slow (doubling time 85 min) (Den Blaauwen et al., 1999). The significant lag between Z ring formation and constriction probably reflects the time needed to complete assembly of the septal ring, but also suggests that assembly and constriction are regulated separately. If so, it will be important to identify the signal that initiates constriction.
Several studies indicated that the Z ring can assemble and disassemble rapidly, within 1 min or less (Addinall et al., 1997; Sun and Margolin, 1998; Rueda et al., 2003). Nevertheless, it came as a surprise when fluorescence recovery after photobleaching (FRAP) revealed that FtsZ molecules in the ring turn over rapidly (Fig. 2) (Stricker et al., 2002). Briefly, a laser was used to bleach Z rings in cells that expressed ftsZ-gfp, and return of fluorescence to the ring, owing to exchange with unbleached FtsZ-GFP from the cytoplasmic pool, was monitored by time-lapse photography. The initial study found that the half-time for remodelling in E. coli was 30 s, but more recent work, under different experimental conditions, indicates a half-time of about 9 s in both E. coli and B. subtilis (Anderson et al., 2004). Because all of the other division proteins require FtsZ for septal localization, if FtsZ subunits are turning over, the remaining proteins in the septal ring might be too. This expectation was confirmed for ZipA (Stricker et al., 2002).
A number of additional interesting observations have come from these studies. First, the rate-limiting step in turnover is probably GTP hydrolysis. In support of this inference, the FtsZ84(Ts) mutant protein, which has a lesion in the GTP binding site and diminished GTPase activity, exchanges about threefold slower than wild-type FtsZ (Anderson et al., 2004). Moreover, the half-time of ≈ 9 s for turnover of wild-type FtsZ as determined by FRAP means that, on average, each FtsZ molecule cycles into and out of the Z ring approximately five times per minute. This rate is strikingly similar to the rate of GTP hydrolysis – 5–10 GTP per FtsZ per minute – determined in vitro under conditions that support formation of protofilaments (Lu et al., 1998; Romberg and Mitchison, 2004). If the rate-limiting step in FtsZ turnover is in fact GTP hydrolysis, the Z ring consists primarily of FtsZ-GTP and has a limited capacity to generate force. A second intriguing finding is that turnover does not appear to be coupled in any simple way to constriction. The rate of remodelling is the same before and during constriction, and ftsZ84(Ts) mutants divide fairly normally at the permissive temperature despite the diminished GTPase activity and slower turnover of the FtsZ84 protein. These observations caution against using the observed rapid turnover of FtsZ as support for the notion that depolymerization drives cytokinesis. A third remarkable finding is that elimination of ZapA, EzrA or MinCD, all of which are implicated in modulating FtsZ assembly in vivo, had little or no effect on the rate of Z ring remodelling (Anderson et al., 2004). Finally, as noted by Romberg and Levin (2003), rapid turnover has implications for the regulation of cell division. Because the Z ring must be actively maintained, assembly of a Z ring does not commit the cell to division at that site. Instead, it is easy for cells to disassemble an existing septal ring to abort division, as occurs during the SOS response to DNA damage in E. coli (Bi and Lutkenhaus, 1993), or to redeploy FtsZ to another site, as occurs during the switch from medial to polar septation during sporulation in B. subtilis. (Ben-Yehuda and Losick, 2002)
Coordinating septum assembly with nucleoid segregation
As already mentioned, nucleoid occlusion is one mechanism for linking cell division to chromosome segregation. Another mechanism involves FtsK.
FtsK of E. coli is an enormous protein, 1329 amino acids in length, and has several activities. The N-terminal domain (≈200 residues) alone is sufficient to support cell division (Draper et al., 1998; Wang and Lutkenhaus, 1998). It contains four transmembrane helices (Dorazi and Dewar, 2000) and localizes to the septal ring, where it is needed for recruitment of several additional essential division proteins (Wang and Lutkenhaus, 1998; Yu et al., 1998a; Chen and Beckwith, 2001). The essential activity supplied by the N-terminal domain of FtsK is not entirely clear. Its role in recruitment of downstream division proteins implies an essential function in assembly of the septal ring before the onset of cytokinesis, but there is also evidence that the N-terminal domain of FtsK plays additional roles during septum closure (see below).
The remainder of FtsK is cytoplasmic. It consists of a proline- and glutamine-rich region (≈500 residues) that might serve as a linker followed by a C-terminal domain (≈500 residues) involved in DNA segregation (Yu et al., 1998b; Steiner et al., 1999). The C-terminal domain belongs to the AAA family of ATPases, a set of proteins associated with a wide variety of cellular activities (Vale, 2000). Some AAA domains catalyse the folding and unfolding of proteins, whereas others disassemble protein complexes or generate unidirectional movement along tracks.
To understand how FtsK facilitates chromosome segregation, it is helpful to know that newly replicated chromosomes are linked, and that these linkages must be resolved for partitioning to go to completion. One way in which chromosomes are linked is that they are catenated. Catenanes are resolved by Topo IV. The C-terminal domain of FtsK interacts directly with Topo IV, recruits it to the midcell and stimulates its decatenase activity (Espeli et al., 2003). The other form of linkage is that chromosomes are sometimes dimeric. Chromosome dimers arise from recA-dependent (homologous) recombination between sister chromosomes, and are resolved by XerCD, a recombinase that acts at a 28 bp chromosomal site near the terminus named dif. The C-terminal domain of FtsK activates the XerCD recombinase and positions the dif sites so that they can form an appropriate synapse (Aussel et al., 2002; Capiaux et al., 2002). Finally, the cytosolic portion of FtsK is an ATP-dependent DNA translocase (Aussel et al., 2002). There are several possibilities, none of which are mutually exclusive, for how the translocase activity might facilitate chromosome segregation. These include aligning dif sites, creating a domain of supercoiled DNA where Topo IV acts and pumping DNA away from the closing septum.
Depending on the growth conditions, the C-terminal domain of FtsK is only essential in the ≈ 10% of cells in a population that contain chromosome catenanes or dimers that need to be resolved. Most of the cells in a culture (≈ 90%) divide well even if the C-terminal domain of FtsK is absent.
FtsK belongs to a large family of proteins implicated in DNA translocation. Another well-studied member of this family is SpoIIIE, a protein required for sporulation in B. subtilis. Sporulation involves an asymmetric division event to create a large mother cell (that ultimately lyses) and a small forespore (that ultimately matures into a spore). The fact that cell division occurs close to one pole means that DNA has to move farther to be properly partitioned to the forespore than to a daughter cell during vegetative growth, when division occurs at the midcell. For this reason, sporulation provides an excellent model system for studying chromosome movement. SpoIIIE localizes to the polar septum, where it appears to function as a unidirectional pump that exports DNA from the mother cell to the nascent forespore (Bath et al., 2000; Sharp and Pogliano, 2002). In addition, the N-terminal membrane anchor domain of SpoIIIE is needed for a membrane fusion event at the completion of engulfment, a phagocytosis-like process in which the cytoplasmic membrane of the mother cell envelops the forespore (Sharp and Pogliano, 2003).
Whether FtsK participates in membrane fusion during cell division in E. coli is not yet known. While an ftsK depletion strain forms smooth filaments, indicating that FtsK is required to initiate cytokinesis, several point and truncation mutants form deep constrictions, as if these forms of FtsK are specifically defective in the final stages of septum closure (Begg et al., 1995; Diez et al., 1997; Wang and Lutkenhaus, 1998; Yu et al., 1998a). These observations are consistent with a defect in membrane fusion, but it has also been suggested that FtsK is needed for the final stages of peptidoglycan synthesis because FtsK deletions can be rescued by deleting dacA, which encodes a carboxypeptidase (Draper et al., 1998). FtsK deletions can also be suppressed by overexpression of FtsN (Draper et al., 1998), the structure of which has revealed a potential peptidoglycan binding site (Yang et al., 2004).
Separating daughter cells
The murein hydrolases, some of which function as autolysins, constitute a diverse set of enzymes that attack different types of bonds in the peptidoglycan sacculus – lytic transglycosylases hydrolyse the glycan backbone, amidases cleave between the glycan backbone and the peptide side-chain, and d, d-endopeptidases cleave the peptide cross-links. It has long been appreciated that murein hydrolases are needed for splitting of the septum to separate daughter cells, but proving this point in E. coli has been difficult because hydrolase mutants usually grow and divide normally. This presumably results from the tremendous redundancy of these enzymes, as there are 18 known peptidoglycan hydrolases in E. coli. The recent construction of mutants deleted for multiple hydrolase genes (up to seven) has provided the missing proof (Heidrich et al., 2002). Septum cleavage is retarded in such mutants, resulting in the formation of chains of cells anywhere from 3 to 100 cells long, depending on the number and identity of the hydrolases inactivated.
Whereas most of the known E. coli murein hydrolases contribute to septum cleavage, two amidases, AmiA and AmiC, appear to be especially important (Heidrich et al., 2001). AmiA and AmiC are exported to the periplasm by the twin-arginine protein transport (Tat) pathway, which probably explains why Tat mutants form chains of cells (Bernhardt and de Boer, 2003; Ize et al., 2003). An AmiA-GFP fusion protein is distributed fairly uniformly throughout the periplasm, but AmiC-GFP localizes to the septal ring during cell division (Bernhardt and de Boer, 2003). AmiC has four functional domains: an N-terminal Tat signal sequence (≈ 30 residues), a targeting domain (≈ 150 residues), a potential linker sequence (≈ 60 residues) and an amidase catalytic domain (≈ 170 residues). Although the Tat signal sequence is required for export, the targeting domain is both necessary and sufficient for localization of GFP to the septal ring (Bernhardt and de Boer, 2003). Localization of AmiC depends on FtsN, making AmiC the last known recruit to the septal ring. More recently, a second murein hydrolase that localizes to the septal ring in E. coli has been discovered (Bernhardt and de Boer, 2004). This enzyme, named EnvC, has homology to lysostaphin and is probably a metal-dependent endopeptidase that cleaves peptide cross-links.
Some Gram-positive bacteria contain only a few murein hydrolases, and in such organisms loss of a single hydrolase has been found to prevent cell separation. Of particular interest here, because of the analogy to AmiC, is the Atl (autolysin) enzyme of S. aureus. The Atl protein localizes to the division site, a structure called the equatorial surface ring, and mutants that lack atl grow as large clusters of cocci (Baba and Schneewind, 1998). Localization to the equatorial surface ring is mediated by the repeat domains R1 to R3, which are not homologous to the targeting domain in AmiC. The actual targets recognized by Atl and AmiC – perhaps a protein, perhaps a modification of the peptidoglycan – have yet to be identified.
Some issues for the future
This MicroReview has touched upon some recent insights into the mechanism of bacterial cell division, especially the astonishingly rapid turnover of the Z ring, the function of FtsK and the identification of proteins responsible for separating daughter cells once septation is complete. Despite this progress, we still know very little about how a mother cell becomes two daughters. Considering only the subset of topics covered here, a number of important questions remain to answered. Does FtsZ drive constriction, and if so, how? Why is the N-terminal domain of FtsK essential? How do the division proteins interact, and how do these interactions enable these proteins to work together during cytokinesis? Beyond these questions lie many others pertaining to issues that this MicroReview has not attempted to discuss in any serious fashion. How is septal peptidoglycan synthesized, and how is this process regulated? Many division proteins are of essentially unknown function – what do they do?
Finally, given that cell division has so far only been studied intensively in E. coli and B. subtilis, one wonders how widely the paradigm will pertain. While genome sequencing has revealed that the cell division proteins known from E. coli and B. subtilis are, to a first approximation, widely conserved among the bacteria, there are important differences (Rothfield et al., 1999; Margolin, 2000). The apparent lack of FtsZ in chlamydiae and mycoplasma, and the recent characterization of a cyanobacterial division protein with no clear homologue in E. coli or B. subtilis (Mazouni et al., 2004), are but two examples. At first the diversity will complicate our picture of cell division, but ultimately themes will emerge, and with these themes will come an understanding of which details are just details and which are of more fundamental importance.
I thank Laura Romberg and Ken Marians for helpful discussions, Piet de Boer, Harold Erickson, Joe Lutkenhaus, Laura Romberg and several anonymous reviewers for comments on the manuscript. Thomas Bernhardt, Piet de Boer, David Anderson and Harold Erickson graciously communicated results before publication. I thank Mark Wissel for help with preparing Fig. 1, and David Anderson and Harold Erickson for supplying Fig. 2. The work in my lab is supported by a grant from the National Institutes of Health (GM59893).