Themes and variations in prokaryotic cell division


  • William Margolin

    Corresponding author
    1. Department of Microbiology and Molecular Genetics, University of Texas-Houston Medical School, 6431 Fannin, Houston, Texas 77030, USA
    Search for more papers by this author

*Tel.: +1 (713) 500-5452; Fax: +1 (713) 500-5499, E-mail address:


Perhaps the biggest single task facing a bacterial cell is to divide into daughter cells that contain the normal complement of chromosomes. Recent technical and conceptual breakthroughs in bacterial cell biology, combined with the flood of genome sequence information and the excellent genetic tools in several model systems, have shed new light on the mechanism of prokaryotic cell division. There is good evidence that in most species, a molecular machine, organized by the tubulin-like FtsZ protein, assembles at the site of division and orchestrates the splitting of the cell. The determinants that target the machine to the right place at the right time are beginning to be understood in the model systems, but it is still a mystery how the machine actually generates the constrictive force necessary for cytokinesis. Moreover, although some cell division determinants such as FtsZ are present in a broad spectrum of prokaryotic species, the lack of FtsZ in some species and different profiles of cell division proteins in different families suggests that there are diverse mechanisms for regulating cell division.


In order to proliferate vegetatively, all cells first duplicate their chromosomes into separate subcellular compartments, then split by dividing their cytoplasms somewhere between the chromosomes to yield progeny cells. This basic process of cell division is conceptually similar in eukaryotic and prokaryotic cells. There are several advantages of studying cell division in prokaryotes: the process is likely to be simpler, there are several outstanding model systems for study such as Escherichia coli, Caulobacter crescentus and Bacillus subtilis, and a greater understanding of cell division in bacteria may lead to novel therapeutic antimicrobial compounds. Over the last decade, advances in cytological and genomic technologies have greatly increased our understanding of cell division in prokaryotes, particularly in the model systems. It is likely, however, that we will find a diversity of cell division mechanisms that mirror the diversity of microbial life. The purpose of this review is to summarize the common themes of cell division as well as the likely variations among the vast prokaryotic world.

2Cell division: a major developmental event

In many rod-shaped bacteria such as E. coli and B. subtilis, cell division involves the synthesis of a septum, with some constriction also occurring in addition to septation in E. coli[1]. In others, such as C. crescentus, cell division appears to occur exclusively by simultaneous constriction of the entire cell envelope, resulting in tapered poles [2]. In these cases and in others, including cocci and in species such as cyanobacteria that form chains of cells, it is clear that cell division is a major developmental event, arguably the predominant cellular event in the vegetative life cycle of a bacterial cell [3]. To redirect cell wall growth in a new direction (formation of the septum) and/or to provide the constrictive force at a single location in the cell, particularly against several atmospheres of turgor pressure, it is easy to imagine that a fairly complex molecular machine must be required. Recent evidence indicates that a protein machine dedicated to the process of cell division is assembled between segregated chromosomes at the proper time [4–6]. The key to this machine's assembly is FtsZ [7–9].

3FtsZ, the keystone of the cell division apparatus

FtsZ is by far the most highly conserved of the known cell division proteins. It is present in most species of prokaryotes examined to date (Table 1). Interestingly, it is not present in the obligately intracellular chlamydiae that divide by binary fission within a host inclusion. It is also not present in two free-living species whose genomes have been fully sequenced: Aeropyrum pernix, a crenarchaeon, and Ureaplasma urealyticum, a mycoplasma species. Despite these and probably other exceptions, FtsZ is present in lower and higher plants and appears to be important for chloroplast division (see Section 15).

Table 1.  Cell division proteins across species
Nmeβ-proteobacteriaXX XXX X XXX
Rprα-proteobacteriaXX XXx X    
Cje?-proteobacteriaXX XX  X  X 
Hpy?-proteobacteriaXX XX  X xXX
Ctrchlamydiae   XX  X    
Cpnchlamydiae   XX  X    
TpaspirochaetesXX XXx X  X 
BbuspirochaetesXX XXx X  X 
Mtu+high GCX  XXx X  X 
Mga+low GCX           
Mpn+low GCX           
Uur+low GC   X        
Sspcyano/chlorXx  Xx X XXX
DraDeino/ThermusXX XXx X  X 
AaeAquificalesXX  X  X XX 
TmaThermotogalesXX  X  X  X 
MjaeuryarchaeotaX         X 
MtheuryarchaeotaX         X 
AfueuryarchaeotaX         X 
PhoeuryarchaeotaX         X 
PabeuryarchaeotaX         X 
Bacteria are listed at the top and archaea at the bottom. All species shown are fully sequenced; all have been published except for Uur. Deino/thermus represents the Deinococcus/Thermus family. FtsW entries include the E. coli mrdE and B. subtilis spoVE families. FtsK entries include the SpoIIIE homologs. Large Xs represent strong similarity, and small xs reflect weak similarity. Small xs in the FtsQ column represent weak sequence similarity with E. coli FtsQ but stronger similarity among other FtsQs. DivIVA has some functional similarity with MinE but no sequence similarity.
Eco: E. coli; Hin: H. influenzae; Nme: Neisseria meningitidis; Rpr: Rickettsia prowazekii; Cje: Campylobacter jejuni; Hpy: Helicobacter pylori; Ctr: Chlamydia trachomatis; Cpn: Chlamydia pneumoniae; Tpa: Treponema pallidum; Bbu: Borrelia burgdorferi; Mtu: Mycobacterium tuberculosis; Bsu: B. subtilis; Mga: Mycoplasma genitalium; Mpn: Mycoplasma pneumoniae; Uur: U. urealyticum; Ssp: Synechocystis sp.; Dra: D. radiodurans; Aae: Aquifex aeolicus; Tma: Thermotoga maritima; Mja: Methanococcus jannaschii; Mth: Methanobacterium thermoautotrophicum; Afu: Archaeoglobus fulgidus; Pho: Pyrococcus horikoshii; Pab: Pyrococcus abyssi; Ape: A. pernix.

In E. coli, FtsZ appears to act at the earliest known step in cell division. Conditional mutants of ftsZ in E. coli fail to divide, yielding long filamentous cells that replicate and segregate their chromosomes but have no sign of any division septa or cellular constrictions (thus the term ‘fts’, for ‘filamentous temperature sensitive’). FtsZ is also the target of SulA protein, synthesis of which is induced upon DNA damage [10]. SulA transiently prevents FtsZ from functioning in cell division, thus inhibiting unwanted cell divisions until the damage to the chromosome can be repaired [11–13].

A structural role for FtsZ was initially suggested by its abundance, about 15 000 monomers per average E. coli cell [14], and by its localization by immunogold labeling to a ring structure at the future site of division [15]. Located at the cytoplasmic membrane, the Z ring, as it is called, appears to contract dynamically along with the membrane as it invaginates during formation of the division septum. The Z ring has also been detected by light microscopy in whole E. coli cells and in other bacteria and archaea by using either immunofluorescence or green fluorescent protein (GFP) fusions to FtsZ [16–20]. These methods confirmed the previous findings, demonstrated that Z rings occur in diverse prokaryotic species, and the GFP studies directly demonstrated the dynamic properties of the Z ring during cell growth and division of E. coli[21].

The localization of FtsZ to a ring structure at the division site of E. coli (Figs. 1A and 2A) suggests that FtsZ protein forms some type of cytoskeletal structure [5]. Unfortunately, an actual FtsZ structure has not yet been visualized in thin sections of E. coli cells. However, much has been learned about the properties of FtsZ protein that can help formulate a model about its structural role in cell division. There is now overwhelming evidence in favor of the idea that FtsZ is a homolog of tubulin, the ubiquitous eukaryotic cytoskeletal protein involved in many essential cellular processes including mitosis [7]. Despite only limited primary sequence homology centered around a GTP binding motif termed the ‘tubulin signature sequence’[22–25], the recently solved crystal structures of FtsZ and tubulin show extensive structural homology throughout the proteins [26]. In addition, FtsZ, like tubulin, binds and hydrolyzes GTP and assembles into protofilaments that have structures similar to those within microtubules [27,28]. This assembly is GTP-dependent [29,30] and disassembly occurs when the GTP is exhausted, suggesting that FtsZ polymers, like microtubules, are dynamically unstable [31]. FtsZ and tubulin also share similar responses to hydrophobic dyes: while bis-anilino-naphthalenesulfonate (bis-ANS) inhibits polymerization of both proteins, the related dye ANS has no effect on either [32]. Another link between FtsZ and tubulin in vivo is that they can be made to coalign as polymers in mammalian cells in the presence of vinblastine, a microtubule-destabilizing drug [33].

Figure 1.

Models for the assembly of the cell division apparatus in E. coli. (A) A possible time course of events. Nucleoids are shown as red ovals, the cell envelope as large black ovals, and the Z ring and its associated components as colored rings. Different ring colors represent its putative different states of assembly. Early in the cell cycle, MinE forms a ring near the future division site, keeping the MinCD oscillating inhibitor away and clearing the way for Z ring assembly. DNA replication results in the duplication and rapid segregation of origins toward the poles [169]; this is followed by termination of replication, further nucleoid alignment and decatenation. Approximately when replication termination occurs, the Z ring assembles at the division site. The MinE ring may still persist but for simplicity is not shown after this point. Then the other essential cell division proteins are recruited in a linear order. This may occur all at once or may occur sequentially over a significant time period; for clarity, a sequential type of recruitment is shown. Finally, the ring contracts, the septum is laid down and the cells separate. (B) A possible cross-sectional view of a small section of the E. coli division machine at the cytoplasmic membrane. About 100 monomers of FtsZ, comprising one protofilament, are in the pictured part of the ring. In this speculative model, the other proteins are found in well-distributed clusters, with FtsA being central for the recruitment of the cluster. ZipA is shown in a separate location from the cluster, but there is no evidence for or against this idea other than FtsA and ZipA are recruited independently.

Figure 2.

Micrographs of bacterial cells. A–C show immunofluorescence images of Z rings in an E. coli ftsK mutant (A), wild-type B. subtilis (B) and an E. coli spherical (rodA) mutant (C). Nucleoids are stained with DAPI in (A) and (B) and appear red in pseudocolor. D–E show phase contrast images of E. coli cells before (D) and after (E) induction of synthesis of a C-terminally truncated FtsA, which causes the cells to bend dramatically [52].

FtsZ and the Z ring are essential for cell division [34]. Thermoinactivation of a mutant FtsZ or depletion of native FtsZ results in the rapid disappearance of rings and the failure of cells to divide, yielding long filamentous cells lacking Z rings or septa [35]. Because FtsZ is rate-limiting for cell division [36,37], one possible trigger for assembly of the ring is a critical concentration of FtsZ monomers in the cell. In E. coli, the small variation in FtsZ levels during the cell cycle [38] is unlikely to be sufficient to trigger assembly. The average in vivo concentration is estimated to be about 10 μM per average cell, about 5–10 times higher than the critical concentration for assembly in vitro [14,29,39]. Therefore, FtsZ may always be in excess in the cell and ring assembly may be subject to negative regulation by FtsZ assembly inhibitors (see Section 16). The specific cell cycle signal that triggers Z ring assembly at the proper time remains elusive, as is the signal that triggers constriction of the ring once it is assembled. One clue to the latter puzzle is Era, a highly conserved and essential GTPase. Depletion of Era causes a cell cycle delay, resulting in cells containing four nucleoids and an unconstricted Z ring [40]. Whether Era localizes to the Z ring and how it might be involved in regulating its activity awaits further investigation.

What is the function of the Z ring? Does it provide the contractile force? Or does it lack motor properties itself but instead serves as a cellular organizer that recruits other proteins into a molecular machine? To address this question, it is necessary to review what is known about other proteins essential for cell division in E. coli.

4FtsA and Z-interacting protein A (ZipA), FtsZ-interacting proteins

FtsA was found by the original fts screen, except that ftsA mutant filaments exhibited regular constrictions instead of no constrictions observed in ftsZ mutant filaments. This suggested that FtsA acted later in the division process [41,42]. FtsA is the only essential cell division protein other than FtsZ that is predicted to reside solely in the cytoplasm (Fig. 2B). FtsA is a member of the ATPase superfamily that includes actin and HSP70 [43]. It is well-conserved in many, though not all, bacteria, but it is notably absent in mycobacteria, cyanobacteria and mycoplasmas. It is also absent in archaea. The ftsA gene is usually present immediately upstream of ftsZ within the cell division–cell wall biosynthesis gene cluster (dcw) in a variety of diverse species, suggesting that its product may interact with FtsZ. Direct evidence for such an interaction comes from yeast two-hybrid screens and colocalization experiments with GFP-tagged FtsAs. Cytological experiments demonstrated that FtsA localizes to the Z ring and its localization depends on FtsZ, whereas Z ring localization does not depend on FtsA [17,18,44,45]. This is consistent with the later cell division defect apparent in the ftsA (ts) mutant. However, inactivation of FtsA also prevents the recruitment of a number of other cell division proteins (see Sections 5 and 6). This indicates that FtsA may act relatively early in division, and that the numerous ftsA (ts) mutants that have been isolated may all have residual activity, allowing septum formation to initiate but then abort.

Despite the genetic and cytological evidence, little is known about the biochemical properties of FtsA protein, and it is completely unknown how FtsA helps FtsZ to divide the cell. Purified FtsA can be phosphorylated and bind ATP, and the viability of a mutant that cannot bind ATP suggests that ATP binding and possible hydrolysis may not be essential for FtsA function [46]. Additional clues to FtsA function may emerge from mutants of FtsZ that can form Z rings but apparently can no longer recruit FtsA; these mutants have residue changes within a short conserved region of the FtsZ C-terminus which may represent a protein–protein interaction domain [47,48]. Another clue is the importance of the ratio between FtsA and FtsZ in the cell, which is estimated to be about 1:100. Altering this ratio by overproducing FtsA, for example, inactivates septation, but cell division can be restored under these conditions by proportionally increasing levels of FtsZ [49,50]. FtsA and FtsZ appear to interact directly in bacteria, as described above, and in the yeast two-hybrid system [48,51]. Taken together, the evidence suggests that FtsA binds to FtsZ monomers within the ring, but because of the 1:100 ratio, FtsA must bind only to a small subset of these FtsZ monomers. Because the other cell division proteins with periplasmic components (see Sections 5 and 6) are probably less abundant in the cell than FtsA, an important function of FtsA may be to limit the number of assembly sites for these proteins (Fig. 1B). If FtsA indeed distributes itself among a limited number of FtsZ monomers within the ring, it is not clear how this would be accomplished.

One unusual property of FtsA is that overproduction of a C-terminally truncated FtsA causes E. coli cells to become markedly curved [52] (Fig. 1D,E). This effect does not require functional FtsZ or FtsI [53]. One speculative explanation of this phenomenon is that high levels of the truncated FtsA protein can assemble at a single site on the cytoplasmic membrane, independent of the Z ring, and recruit but not properly orient a subset of cell division proteins. This would stimulate abnormal septum-like wall synthesis only on one side of the cell, resulting in a curved cell morphology.

ZipA was not found in the screen for fts mutants. Instead, it was found in a biochemical screen for proteins that interacted with purified FtsZ [54]. ZipA is predicted to be an integral membrane protein, with an unusual N-terminal membrane anchor and a C-terminal cytoplasmic domain. It is not particularly conserved, with homologs only in other γ-proteobacteria (Table 1). Nevertheless, in E. coli, ZipA is essential for cell division, requires the Z ring for recruitment, and is recruited to the Z ring independently of FtsA [55]. Z rings are present when ZipA is depleted from the cell, although in some cases the number of rings is significantly reduced. This finding, as well as recent biochemical evidence [56], suggests that ZipA may function to stabilize the Z ring immediately after it is assembled. The topology of ZipA implies that this stabilization may be accomplished by ZipA-mediated anchoring of the ring to the cytoplasmic membrane (Fig. 2B). Interestingly, recent evidence indicates that the conserved extreme C-terminus of FtsZ may be involved in recruitment of ZipA as well as FtsA [47,57]. Assuming that this domain is directly responsible for recruitment, it seems at first that it would not be possible for one small protein domain to recruit two different proteins. However, the predicted large excess of FtsZ relative to FtsA and ZipA might easily allow recruitment of both proteins to separate segments of the ring (Fig. 2B). It is notable that overproduction of ZipA can inactivate cell division [54], which is analogous to the effect of FtsA overproduction. The mechanism of this inhibition is unknown. It may occur in part because ZipA or FtsA, when overproduced, may compete for accessible FtsZ sites and therefore prevent recruitment of the other.

5FtsK and FtsW, polytopic integral membrane proteins

FtsK and FtsW are both essential cell division proteins that are predicted to span the membrane multiple times [58–60] (Fig. 2B). FtsK is recruited to the Z ring and requires both FtsZ and FtsA for recruitment, but not FtsI or FtsQ [61,62]. FtsW may have a ZipA-like role, as depletion of FtsW results in a decrease in Z rings; however, nothing else is known about how FtsW functions in cell division or how it is recruited to the ring [63]. One attractive idea is that FtsW serves to integrate signals between the cytoplasmic components of the machine (FtsZ and FtsA) and the periplasmic components including FtsQ (see Section 6).

FtsK is one of the largest proteins in E. coli and appears to have multiple functions. Its N-terminal 15% is sufficient for its targeting to the Z ring and its function in cell division [61,64]. Interestingly, the requirement for FtsK in cell division can be bypassed by deletion of the gene for penicillin binding protein 5 (PBP5), a carboxypeptidase that removes a dipeptide from the pentapeptide side chain of peptidoglycan to make a tripeptide. Overproduction of FtsN (see Section 6) can also bypass the FtsK requirement [64]. The mechanism behind the bypass effects is not clear. In the case of PBP5, shifting the balance toward precursors with pentapeptide side chains, which are normally used in cell elongation, may compensate for the loss of FtsK. This suggests that FtsK-mediated septum closure may require different peptidoglycan precursors than the tripeptide precursors normally used for septum synthesis [65]. Interestingly, overproduction of FtsK, as with ZipA and FtsA, results in inhibition of cell division; in the case of FtsK, it appears that the major effect is by inhibiting the assembly of Z rings [64]. The mechanism of this inhibition is unknown but may prove to be generally significant for our understanding of protein–protein interactions within the division complex.

Whereas the N-terminal domain of FtsK is essential for cell division, the C-terminal domain appears to have a role in coupling cell division with chromosome segregation [66,67]. This domain is highly similar to the C-terminus of SpoIIIE, which is required in B. subtilis to transport chromosomes through septa in which they have been trapped [68]. In E. coli, deletions of the C-terminal domain of FtsK result in abnormal chromosome positioning and chromosome bisection by closing septa. These deletions also result in failure to resolve chromosome dimers at the dif site, suggesting that many of the positioning defects may be caused by dimer resolution problems [69]. The current working model for FtsK proposes that it helps to keep chromosomes away from the contracting cell division complex so as to prevent chromosome scission. It does so presumably by actively transporting DNA away from the closing septum while simultaneously achieving the last step in septum closure. The viability of ftsK C-terminal deletions [70] indicates that not all chromosomes get trapped and that backup systems exist. However, such deletion strains have defects in cell–cell separation, stationary phase survival and adaptation to stress [70]. Moreover, combining an ftsK C-terminal deletion with a null mutation in mukB, which is probably involved in chromosome condensation, yields non-viable cells; this suggests that decondensed chromosomes are more susceptible to being trapped [66]. A similar synthetic phenotype between SMC (a probable functional homolog of mukB) and spoIIIE was found in B. subtilis[71]. FtsK is also induced upon SOS-mediated DNA damage, which is consistent with its important role in DNA dynamics [62].

6The bitopic cell division proteins

Four known essential cell division proteins in E. coli, FtsQ, FtsL, FtsI and FtsN, possess a short N-terminal cytoplasmic anchor, a transmembrane segment and a relatively large periplasmic domain [72–74]. FtsI, also known as PBP3, has been relatively well-characterized. There is good evidence that it is a transpeptidase and functions in one of the key enzymatic steps in septal peptidoglycan synthesis [75]. The β-lactam drugs cephalexin and aztreonam preferentially inhibit PBP3 activity without significantly inhibiting the activity of other PBPs that synthesize lateral wall peptidoglycan, such as the PBP1 class. These drugs have been useful for specifically blocking cell division, and they are currently the only known drugs that directly target the cell division apparatus.

All four of these proteins localize to the Z ring (Fig. 2A,B). By combining cytological techniques with fts mutants, enough evidence has accumulated to propose a model for their recruitment. For example, FtsQ tagged at its cytoplasmic N-terminus with GFP localizes to the Z ring independently of FtsL and FtsI but requires FtsZ and FtsA [76]; a similar GFP–FtsI fusion protein fails to localize in the absence of FtsZ, FtsA, FtsQ or FtsL [77], consistent with immunofluorescence results with FtsI [63]. FtsL requires FtsZ, FtsA and FtsQ but not FtsI to localize [78]. FtsN is recruited last, and depends on FtsI [79]. In summary, these proteins appear to be recruited in a linear series after FtsZ, FtsA, ZipA and FtsK (and perhaps FtsW) are localized (Fig. 2A).

The functions of FtsQ, FtsL and FtsN are unknown, but their tripartite domain structure has been exploited in domain swapping studies. These studies indicate that replacing domains becomes increasingly deleterious the later the protein is recruited. For example, all three domains of FtsI, the periplasmic and cytoplasmic domains of FtsL, but only the periplasmic domain of FtsQ are essential for function and cannot be replaced [72,76,77,80]. The periplasmic portion of FtsL has a coiled coil structure and is essential for its localization and function; recent results suggest that this domain multimerizes and that this multimerization is important for function [80]. FtsQ and FtsL are moderately conserved among bacteria, but FtsN is only found in E. coli and Haemophilus influenzae.

The role of FtsN is puzzling. Despite its late recruitment, FtsN when overproduced can suppress mutations in ftsA and ftsK[64,81], implicating FtsN in some type of global process. Consistent with this, depletion of FtsN results in an ‘early’ cell division defect, with filaments lacking indentations. This suggests that most or all of the essential cell division proteins are already in place early during the septum formation process. One way that FtsN might influence the cell division machinery globally is by regulating the availability of substrates or modulating the ability of the cell envelope to invaginate. It is also possible that FtsN does not need to localize to the ring in order to be active.

7A conserved septum/cell wall synthesis cluster

Many of the essential cell division genes, including ftsI, ftsW, ftsQ, ftsA and ftsZ, are found together with genes for peptidoglycan synthesis within the dcw cluster. The dcw genes not exclusively involved in cell division include the mur (murein), ddl (alanine ligase) and mra gene families. Despite the omission or addition of specific members, this cluster and the gene order within it are highly conserved in a number of bacteria. In E. coli, the transcriptional organization of the dcw cluster is complex, but all transcription proceeds in one direction [82]. Genes ftsZ and envA, a gene involved in a step in lipid biosynthesis that is also important for cell separation [83], lie at the 3′ end. It is likely that ftsZ is the most heavily transcribed of the genes, and there are additional upstream promoters throughout the cluster that are responsible for high-level ftsZ expression [84]. Regulation of ftsZ gene transcription does not appear to play a significant role in activating cell division in E. coli, as cells in which ftsZ is expressed ectopically from an inducible promoter in the absence of a functional native ftsZ grow and divide normally. However, several proteins regulate transcription of dcw genes, particularly ftsA and ftsZ, and it remains to be seen how this regulation affects and is affected by cell physiology [4,85,86].

The genes at the termini of the cluster, including ftsZ at the downstream end and mraZ-mraW at the upstream end, appear to be the most conserved among bacteria, whereas the arrangement of internal genes within the cluster is more variable among different species. This variability amidst the common core of dcw genes may correlate with the likely differences in cell division mechanisms in different species. For example, many of the usual dcw genes are present in Streptococcus pyogenes but are distributed around the genome in three non-contiguous clusters [87]. In mycoplasmas that contain ftsZ, the gene order is mraZ-mraW-MG223-ftsZ, indicating that the termini of the cluster in this ‘minimal genome’ are still conserved but that all of the other intervening dcw genes are replaced by a single gene of unknown function, MG223. In archaea, mraZ-mraW are missing, and ftsZ is the only recognizable dcw representative.

8Other genes and factors that affect cell division

The discussion so far has been limited to nine genes known to be essential for E. coli cell division. All but zipA were found in genetic screens for effects on cell division. The recent availability of genomic sequences of a variety of microbial species suggests that other highly conserved genes may be required for cell division. The mraZ-mraW genes are examples of highly conserved genes that do not seem to have a cell division phenotype. However, yihA, which has GTPase motifs, has a cell division phenotype when inactivated but it is not clear if the effect is direct or indirect [88]. It is likely that further functional genomic analysis of potential cell division genes will lead to additional essential cell division genes.

In many cases, cell division can be inhibited by mutations or overexpression of genes known to be involved in other cellular processes, indicating that there are many indirect ways to affect cell division. One class of inhibitory effects includes mutations that perturb chromosome replication, structure or segregation; inhibition of cell division results from either SOS-mediated induction of SulA or a topological veto by the nucleoid (see Section 16). Another class includes direct effects on cell division proteins. For example, altered synthesis of DnaK, trigger factor, and the immunophilin homolog slyD results in filament formation [89–91]. In these and other cases, overproduction of FtsZ often can suppress the inhibition, suggesting that the inhibitory effects are caused by a decrease in the level of FtsZ activity such as decreased ftsZ gene expression or improper FtsZ folding. Deficiencies in factors as diverse as S-adenosyl methionine [92] or specific membrane phospholipids [93] result in specific cell division defects. Constitutive activation of the cpxA periplasmic stress pathway does not affect the cell's ability to divide, but division often occurs at abnormal sites [94]. Dissecting the mechanisms by which these and other proteins affect cell division will be a challenge because of the multiple roles of many of these proteins in cell physiology.

9Cell division in other model systems: B. subtilis

Cell division in B. subtilis differs from that of E. coli in several ways. First, cytokinesis in the Gram-positive B. subtilis is achieved by formation of a thick septum without significant constriction of the outer cell envelope, whereas cytokinesis in E. coli appears to occur by simultaneous formation of a septum and cell envelope constriction. Second, septation in B. subtilis normally occurs at the cell midpoint during vegetative growth. However, upon starvation, B. subtilis cells initiate a highly regulated developmental program that culminates in the formation of a terminal endospore. One of the critical steps in this program is the switch from medial cell division to polar cell division, which results in the formation of a septum that separates the mother cell from the prespore. The asymmetric compartments generated by this spore septum are important for the creation of two cells with complete genomes but with different developmental fates [95,96], although different-sized compartments are not absolutely required for this to occur [97]. The alteration of Z ring localization from midcell in vegetatively growing cells [98] (Fig. 1B) to the poles is controlled by the master response regulator Spo0A [16], but it is not yet known what factors directly cause the change in Z ring localization. Identification of factors involved more directly in the switch should contribute important insights into the spatial control of the division plane.

As might be expected from their different cell envelopes, there are some differences in the proteins involved in B. subtilis cell division with respect to those of E. coli. Normal B. subtilis cell division requires FtsZ, FtsA, FtsL (a distant relative of E. coli FtsL), the FtsQ homolog DivIB and the FtsI homolog PBP2b [99–103,176]. The B. subtilis homolog of FtsK, SpoIIIE, is essential for segregating the prespore chromosome into the prespore and is involved in fusing the inner and outer membranes of the spore upon its engulfment by the mother cell [104]. However, SpoIIIE is not required for vegetative division despite its recruitment to the septum [105], and detectable homologs of the other proteins essential for E. coli division are not present or are not essential for division in B. subtilis. Conversely, DivIC is an essential cell division protein in B. subtilis that is not present in E. coli or most other species [106]. Like DivIB, DivIC is highly abundant, bitopic in structure and acts late, requiring FtsL for localization [99,107–109]. Cytological evidence suggests a linear order of recruitment to the Z ring, with FtsZ and FtsA arriving first, followed by FtsL, DivIB and PBP2b. This is somewhat different from E. coli in that FtsL recruitment in E. coli requires FtsQ(DivIB); however, B. subtilis FtsL shares little sequence homology with E. coli FtsL and may also not be functionally homologous. In addition, some factors involved in specifying the position of the Z ring are different and appear to act by a different mechanism (see Section 16).

10Cell division in other model systems: C. crescentus

The third well-characterized system for bacterial cell division is C. crescentus. One of the reasons C. crescentus is attractive for the study of cell division and the cell cycle is that like B. subtilis, C. crescentus undergoes an asymmetric division that results in daughter cells having different developmental fates. The two distinct cell types are swarmer and stalked cells, which arise initially by the development of asymmetry within the mother cell prior to cell division. The cell cycle appears to be regulated much more tightly than in the other bacterial model systems and thus is reminiscent of an eukaryotic cell cycle, in which each step is dependent on and regulated by a previous step. For example, DNA replication initiates only in stalk cells and not in swarmer cells. Moreover, cell division is regulated in ways that appear to be distinct from those of E. coli or B. subtilis. First, on a structural level, cell division appears to occur by constriction or a cleavage furrow and not by formation of a septum [2]. As a result, the poles are tapered, which may facilitate the biogenesis of the polar stalk. Whether this reflects the plasticity of the cell envelope or a different mechanism of cell wall synthesis is not known. Second, unlike in E. coli, C. crescentus FtsZ levels are tightly regulated by transcriptional and posttranslational control [110,111]. The result is that FtsZ is only present in cells during a short cell cycle window. Expression of ftsZ is repressed by the global regulator CtrA in swarmer cells, which do not divide, but is derepressed in stalked cells [111]. In addition, FtsZ is proteolyzed specifically in cells undergoing division, effectively removing FtsZ from both daughter cells. Interestingly, ftsZ is expressed first, followed by ftsA and ftsQ; this timing of expression reflects the timing of function of the gene products [112]. It will be instructive to find out how the placement of the Z ring is controlled in this organism, given the apparent lack of some negative regulatory controls which are present in other species (see Section 16).


Very little is known about the molecular aspects of cell division in other bacteria. Nevertheless, one common thread likely to emerge is the role of FtsZ as a cellular organizer. In most standard rod-shaped cells, FtsZ probably forms a ring at the cell midpoint to divide the cell. Given the variable conservation of the other cell division proteins and the examples of B. subtilis and E. coli, it is likely that the Z ring recruits a different set of proteins to the division complex in different species. Such plasticity may reflect different cell wall biosynthetic pathways, PBPs, lipid composition, and/or mechanisms of chromosome organization and partitioning.

The problem becomes more complex in cocci, which have an infinite number of theoretical cell division planes that can give rise to two equal daughter cells, as opposed to rods, which have essentially one division site that fulfills this criterion. Moreover, the larger diameter of most cocci relative to rods means that the Z ring must be significantly larger and therefore may need to regulate assembly of the division machine differently.

Enterococci apparently deal with the problem of identification of the division plane by zonal cell wall growth. New wall growth begins at a structure called a wall band, which is represented by a circumferential hump-like structure on the outside of the cell. Wall material then grows outward bidirectionally, duplicating the band to form a notch and forming a Y-shaped structure. The two bands push apart until they are located at the midcell point of the two daughter cells. A new notch is then formed at each band by duplication, and the process repeats [113]. Thus, the division site is differentiated by duplication of the wall band and division culminates in the synthesis of a septum at the base of the wall band. Just as FtsZ seems to be responsible for the switch from elongation to septal synthesis mode in E. coli, it may also be responsible for the duplication of the wall band. In this case the division site seems to be an epigenetic structure.

In other cocci such as Deinococcus, however, this model does not seem to hold. Deinococcus radiodurans and the Gram-negative cocci represented by Neisseria, among others, appear to divide in alternating planes [114–116]. Staphylococcus aureus mutants defective in cell separation can also divide this way [117]. While the enterococcal system provides an explanation for the selection of the division plane, it is not at all clear how alternating perpendicular planes are specified and how the switch in direction is made. It is reasonable to propose that in rods, a major function of FtsZ is to switch the direction of cell wall synthesis by 90°, and perhaps a similar directional switch functions in spheres. Interestingly, spherical mutants of E. coli divide in alternating planes [118], and studying Z ring assembly and placement in such mutants (Fig. 2C) may be able to address these issues in a more tractable system. It is tempting to speculate that the perpendicular switching by FtsZ may be evolutionarily related to the two perpendicular structures of the animal cell centriole, which plays an important role in organizing the microtubule cytoskeleton.


Other systems pose additional challenging questions. For example, chlamydia species lack FtsZ (Table 1), yet clearly undergo binary fission as reticulate bodies within the chlamydial inclusion [119]. Having minimal genomes, do chlamydiae co-opt host dynamin, which is involved in pinching off membranes during endocytosis, or microtubules for their own cell division [120]? If so, how do they regulate the placement of the division plane? If they use a chlamydial protein, what is it? Recently, a protein of unknown function has been immunolocalized to the equator of dividing reticulate bodies that may be a candidate for a cell division protein [121]. It is intriguing that despite the lack of FtsZ, chlamydiae contain several other homologs of essential cell division genes including ftsK, ftsW and ftsI. Assuming these genes have functional roles in division, how do their gene products work if there is no FtsZ to recruit them?

13Mycoplasma and L-forms

Mycoplasma are bacteria that lack cell walls and cause a variety of infections. They also contain small genomes that are good models for a minimal set of genes. As with organelles and archaea (see below), the only recognizable essential cell division homolog in mycoplasmas is ftsZ (Table 1). The absence of the other genes such as ftsA, ftsQ, etc. suggests that FtsZ is sufficient for cell division and that the other proteins serve to coordinate cell wall and septum synthesis with FtsZ-mediated cytokinesis. It is tempting to speculate, based on this genomic evidence, that FtsZ itself can provide the constrictive force necessary to split cells. However, mycoplasmas also have unusual cytoskeletal-like proteins that are involved in cytadherence; interestingly, perturbing the gene encoding one such protein results in a morphological defect that may be a result of cell division inhibition [122]. More work needs to be done to demonstrate if this or other large attachment proteins are involved in cell division in mycoplasmas. Mycoplasmas appear to have divergent division patterns, with some species dividing mainly by binary fission and others by budding [123]. Most intriguing is the lack of FtsZ in U. urealyticum, indicating that some other means must be used to divide these cells.

Bacteria other than mycoplasmas can be stripped of their walls and cultivated. These wall-less variants, called L-forms, can be made in many species including E. coli[124]. The fact that E. coli L-forms can grow and divide indicates that the cell wall is not essential for the division machinery to function. The tantalizing question here is whether the cell division proteins other than FtsZ are necessary for L-form division; if not, then E. coli L-forms may be a viable model system for studying the sufficiency of FtsZ for cell division.


Despite having many eukaryotic characteristics, such as eukaryotic-based transcription and DNA replication systems, archaea look like bacteria [125]. It was therefore exciting but not a complete surprise when it was found that archaea contain FtsZ homologs and that FtsZ forms a ring at the midcell division site [20,126,127]. However, several surprises have emerged from genomic studies of archaeal cell division genes. First, as with mycoplasmas, no other known essential cell division homologs are found in archaea despite the presence of cell walls. Second, many archaeal species harbor two distinct paralogs of FtsZ, and a compilation of many archaeal genome sequences indicates that these paralogs form two separate groups [128]. Essentially nothing is known about archaeal cell division, including the different functions, if any, of the dual FtsZs. It should be emphasized that additional copies of ftsZ are also found in plants (see Section 15), and there is one case in bacteria (Sinorhizobium meliloti) of two distinct ftsZ homologs [129]. This suggests that ftsZ has a tendency to form gene families and potentially multiple isotypes, as does tubulin. Whereas the FtsZ paralogs in plants may have distinct functions (see Section 15), the differential function of FtsZ paralogs in prokaryotes has yet to be explored.

The third, and biggest, surprise revealed by archaeal genomes is the complete absence of ftsZ to date in a subset of the archaea known as the crenarchaea. This group includes the recently sequenced A. pernix[130]. Whereas the absence of ftsZ in chlamydia can be rationalized because chlamydiae are obligately intracellular and depend on the host for many cellular functions, the absence of ftsZ in free-living bacteria that divide is difficult to understand. Sulfolobus species, crenarchaea that divide by budding and grow under fairly normal conditions, have the potential to be a good model system for cell cycle studies [131]. In particular, this organism also may serve as a good FtsZ-free system in which to study FtsZ-based bacterial cell division and potentially to assemble the division apparatus de novo inside a cell.


It is currently accepted that organelles arose from bacterial endosymbionts: chloroplasts came from cyanobacteria while mitochondria came from the α-proteobacteria [132]. As organelles need to divide in order to proliferate and be maintained within their dividing eukaryotic hosts, it is natural to ask whether bacterial division mechanisms have been conserved. Recent evidence strongly indicates that they have been [133]. Nuclear-encoded FtsZ homologs have been found in a number of photosynthetic eukaryotes, ranging from protists to monocotyledonous and dicotyledonous plants. Not surprisingly, these FtsZs are most highly related phylogenetically to cyanobacterial FtsZs. In two different plant species, including the model system Arabidopsis thaliana, inhibition of FtsZ by antisense or by knockouts causes chloroplasts to enlarge and stop dividing [134,135]. Interestingly, A. thaliana contains at least two FtsZ homologs, one having a chloroplast import sequence and the other lacking it. This evidence suggests the possibility that two Z rings may assemble, one inside the organelle and the other outside. This would nicely fit with the microscopic observation of two concentric electron-dense rings at the site of chloroplast division [136]. However, this raises the question of how targeting of two FtsZs on either side of the chloroplast might be coordinated.

The role of FtsZ in mitochondrial division is less clear, but has been illuminated in some recent studies. Several completely sequenced animal and fungal genomes, including nematodes and budding yeast, lack obvious FtsZ homologs. This indicates that FtsZ is not universally required for mitochondrial division. Recently, however, a mitochondrial FtsZ homolog from a chromophyte alga has been isolated [137]. This homolog is most related to the FtsZs of the α-proteobacteria, further supporting the idea that mitochondria descended from this family of prokaryotes. The absence of FtsZ homologs in animal and fungal species so far suggests that mitochondria of most species either do not need to divide regularly and can segregate by other mechanisms, or that another cytoskeletal protein such as dynamin may act to divide organelles in these species [138].

16Factors involved in the specification of the division plane

16.1The Min system

As stated previously, cell division is a major developmental event. The primary question that arises is how the cell division plane is identified. Clearly, we now know that bacteria are not just bags of enzymes and DNA, and they have within them distinct subcellular addresses for protein assembly. In rod-shaped bacilli, the correct division site is usually the midpoint between the two poles. How does FtsZ find this site?

We have recently developed a model that invokes negative regulation. In this model, the two main determinants of division site placement are (i) the Min system and (ii) the nucleoid and its associated components [139]. The Min system is composed of three proteins, MinC, D and E which are encoded by the minCDE operon of E. coli[140]. In E. coli, MinC and D act together to negatively regulate the assembly of the Z ring, while MinE acts to negatively regulate the action of MinC and D [141]. GFP fusions to the Min proteins have revealed much about how they might function as regulators [142]. MinE–GFP localizes as a ring-like structure near the cell midpoint (Fig. 2A), although the ring is often off-center [143]. In contrast, the Z ring is precisely centered [139,144]. Localization of MinE is independent of FtsZ, consistent with the role of MinE as a specificity determinant for FtsZ placement [143]. GFP fusions to MinC and MinD, on the other hand, oscillate from pole to pole with a periodicity on the order of 10–60 s [145–147]. The membrane localization and oscillation of MinC depend upon MinD, and both MinC and D appear to oscillate as a complex [146]. Interestingly, the oscillation of MinCD depends on MinE, and the midcell localization of MinE depends on MinD. The movement of MinCD may function to average the distance between the two poles, while the MinE ring may act as a molecular slingshot to repel MinCD during its cellular traverse and to allow Z rings to assemble near the protection of MinE. Away from MinE, MinCD may act as a sweeper to keep Z rings from assembling at poles, one half-cell at a time (Fig. 3). Purified MinC is sufficient to inhibit FtsZ polymer assembly in a dose-dependent manner [147]. MinD, therefore, probably enhances MinC action by recruiting MinC to the membrane where its local concentration is significantly increased.

Figure 3.

A model for negative regulation of division site placement in E. coli and B. subtilis. Cells at successive points in the cell cycle, from early in the replication cycle to initiation of septation, are shown from top to bottom for each species. Points above the dotted line represent intervals of less than a minute, in order to highlight the MinCD oscillation phenomenon. Blue ovals represent nucleoids, and yellow bands near midcell in E. coli represent MinE rings. MinE rings are shown to move about the center in order to explain why they are often found off-center, but there is no evidence that they actually move in this way. Dark shaded areas within cells represent the MinCD division inhibitor and the degree of shading illustrates the putative concentration gradient with respect to the poles. Red dots denote FtsZ molecules with potential to multimerize. Orange dots denote FtsZ molecules that have successfully nucleated and polymerized into a Z ring. The green squares denote nucleation sites at midcell that are revealed once chromosome segregation begins. Black lines above the cells indicate regions of the cell covered by the nucleoid that appear to inhibit multimerization of FtsZ.

What happens in the absence of the Min system in E. coli? Cells with inactivated minC or D divide either at their midpoint or at a pole, resulting in a mixture of short filaments, nucleoid-free minicells and nucleoid-free rods [139,148]. These mutations are not lethal because enough cells still divide medially. Z rings in a ΔminCDE strain form promiscuously in all gaps between nucleoids and often form closely packed double and triple rings [139]. This result indicates that Z rings require the presence of the Min system in order to be properly directed, and suggests that MinCD inhibits Z ring assembly in all parts of the cell, not just at midcell and at the poles.

In addition to the defect in division site placement, min mutants also have defects in chromosome partitioning [148,149]. This is of interest as MinD is homologous to the ParA family of plasmid and chromosomal partitioning proteins. One of these, Soj, oscillates from pole to pole in B. subtilis and interacts with the centromere binding protein Spo0J [150,151]. Soj is non-essential for growth, cell division or chromosome segregation. It remains to be seen what specific role MinD plays in E. coli chromosome segregation.

The Min system of B. subtilis is also involved in preventing unwanted polar divisions [152], but there are important distinctions in both the players involved, their localization and their importance for division site selection. As in E. coli, MinCD acts as a division inhibitor, preventing assembly of Z rings [153]. However, there is no MinE homolog in B. subtilis. Instead, a protein called DivIVA acts as the MinE-equivalent, antagonizing MinCD inhibition [154]. Another major difference is that MinC and D have not been observed to oscillate. Instead, in contrast to the E. coli system, MinCD and DivIVA in B. subtilis are recruited to the Z ring after it forms, and stay bound to the ring until after cell division [153–155]. DivIVA helps to keep MinCD at the poles of the daughter cells, so that Z ring assembly is specifically inhibited at the poles (Fig. 3). One of the conclusions to be made from these observations is that MinCD prevents assembly of new rings but not existing rings. Moreover, there must be factors other than MinCD that help to regulate localization of Z rings to the medial site in B. subtilis, because MinCD is recruited by FtsZ and most cells lacking MinCD still contain properly placed medial Z rings [156]. A B. subtilis-specific additional negative regulator of Z rings, EzrA, has recently been isolated. When ezrA is inactivated, extra rings appear at the poles [157]. EzrA is recruited to the Z ring like MinCD and DivIVA, but EzrA is not retained later at the poles. This suggests that unlike MinCD, EzrA may serve to destabilize preassembled FtsZ polymers.

In some ways, the B. subtilis division site system seems more streamlined than the E. coli system. By using the existing ring to recruit the MinCD inhibitor and retain it at the incipient poles where it acts (Fig. 3), B. subtilis Min proteins do not need to be able to identify the division site. In E. coli, on the other hand, FtsZ is not able to identify specific sites as readily and needs MinE to do this. It can be postulated that FtsZ of E. coli can assemble constitutively and is negatively regulated, whereas FtsZ of B. subtilis either is capable of identifying the right site on its own or uses other, yet unknown, cues. The factors that determine FtsZ targeting to the proper subcellular sites in B. subtilis are still unknown.

MinD is well-conserved among many bacteria and archaea, and is also present in a number of eukaryotes that carry chloroplasts [158]. This is not as surprising as it seems, because recognizable Min homologs are present in cyanobacteria. It will be interesting to see if MinD is required for proper Z ring placement in archaea and organelles, or if it acts in some other pathway. Judging whether a species has MinD is complicated by (i) the presence of MinD-like proteins such as ParA and Soj that may oscillate but probably do not influence division site placement, and (ii) incorrect sequence annotations citing homology to MinD when little similarity actually exists. Nevertheless, obvious homologs of MinD as well as MinC appear to be absent in many species. MinE is present in only a few characterized species. The B. subtilis system and comparative genomic data suggest that (i) many species may use other factors that interact in novel ways with MinD to regulate division site placement, and (ii) that some species may not use an analogous site placement system. H. influenzae, for example, is highly related to E. coli but lacks the Min proteins. So do many (but not all) cocci. One rationale for this is that short rods, such as H. influenzae, do not need to prevent polar divisions because there are no nucleoid-free areas at the poles. The rationale for most cocci is that most of them have other mechanisms for identifying the division plane. Interestingly, Neisseria and Deinococcus species both have MinD homologs and Neisseria also has MinC and MinE; as pointed out above, these species divide in alternating perpendicular planes. This may not be coincidental, and it is possible that this type of division system may use the negative regulation by Min proteins as an extra level of topological control. It will be interesting to see if min knockouts in these species result in randomized division planes, and whether toporegulation by the Min system conforms to one of the known models.

16.2The nucleoid

The other regulatory element hypothesized to be involved in division site selection is the nucleoid. This makes good sense, as the proper replication, packaging and segregation of the chromosome should be a prerequisite for cell division. However, in E. coli, evidence from several approaches suggests that chromosome replication or segregation is not an absolute requirement for cell division to occur. For example, Z ring assembly and septum formation can be triggered at cellular locations far from nucleoids and yet with some spatial precision [139,148,159]. Z rings and septa can also bisect abnormal nucleoids that have not yet segregated [159–161].

Nevertheless, wild-type E. coli cells do not septate on top of or far away from nucleoids. In fact, there is some evidence that the nucleoid acts negatively to inhibit Z ring formation in its vicinity. This nucleoid veto is the basis of the ‘nucleoid occlusion’ model proposed by Woldringh and coworkers [162]. The model suggests that all positions in the cell are potential division sites and that the nucleoid prevents division from occurring at midcell until the veto is released, presumably because of segregation. According to this model, minicells result from increased nucleoid-free zones at poles in min mutants. The bisection of nucleoids in certain mutants can be rationalized by the idea that nucleoids with abnormal structure no longer have effective veto power. However, some evidence clearly contradicts the idea that nucleoid occlusion is sufficient for division site placement. First, Z rings are present at or near midcell in nucleoid-free cells [161]. Second, divisions occur within a fixed distance from a pole in DNA replication-defective mutant filaments far from any nucleoids [159]. Third, the Min system, which must have a role in division site selection, is not included in the nucleoid occlusion model. Therefore, this model is not sufficient to explain division site placement.

16.3A new model for E. coli division site placement

We have formulated a new model that fits the known results and combines roles of both the Min system and the nucleoid in negative regulation of Z ring placement in E. coli (Fig. 3). Essentially, this model proposes that a combination of negative regulatory signals from the nucleoid and the Min system are necessary and perhaps sufficient for correct localization of the ring. Like the nucleoid occlusion model, our model suggests that all locations in the cell are potential division sites, but these sites are masked by (i) the MinCD inhibitor, which sweeps the cell from pole to pole, and (ii) the nucleoid, which occupies much of the cell center. Z rings are kept from forming in polar nucleoid-free regions by MinCD, while the nucleoid inhibits Z ring assembly in the remainder of the cell. It is not known what the molecular mechanism of the proposed nucleoid veto is, but as proposed in the original nucleoid occlusion model, this veto is relieved at some point during nucleoid segregation. This relief may occur early in segregation of normal chromosomes, because Z ring assembly occurs about the same time as replication termination [163] (Fig. 2A). MinCD would normally be able to still prevent Z ring formation except for the presence of MinE at midcell, which protects FtsZ from MinCD. The promiscuous assembly of Z rings in nucleoid-free regions of cells lacking the Min proteins (but not in cells with the Min proteins) prompted us to formulate this model. The bisection of nucleoids by septa under certain conditions, as discussed above, can be rationalized. In these cases, nucleoid structure is perturbed, and intact nucleoid structure may be required for the veto.

What could FtsZ ultimately be sensing? One possibility is that FtsZ is targeted to a membrane domain that also targets replication proteins. For example, SeqA in E. coli and DNA polymerase in B. subtilis also localize to midcell and quarter-cell regions [164,165]. Does an intact, properly positioned oriC complex influence FtsZ localization? Is the nucleoid veto effect observed in other species? One important postulate in our model is that the precise positioning of the cell division site may result from the previous centering of the replication apparatus. It is unknown, however, how either replication proteins or MinE find their midcell locations, and this is fertile ground for future study.

Studies of germinating spores of B. subtilis may address the nucleoid veto mechanism. By being able to synchronize initiation of DNA replication in this system, it may be possible to dissect the effects of oriC structure and nucleoid dynamics on Z ring assembly [166]. It is interesting that the switch from medial to polar septation occurs at the same time as the formation of a highly elongated nucleoid structure called the ‘axial filament’[167]. Whether this filamentous nucleoid structure results in a nucleoid veto across most of the cell, pushing the Z rings to the poles, remains to be seen but is intriguing.

The other major model system, C. crescentus, may also offer some valuable insights into division site placement. Despite the tight control of protein levels and the importance of chromosome integrity, the nucleoid veto effect in C. crescentus appears to be weak. For example, topoisomerase IV mutants of C. crescentus are inhibited in chromosome segregation, but Z rings bisect the unsegregated nucleoids, forming multiple constrictions throughout the cell [168]. These constrictions fail to complete division, suggesting that C. crescentus has a cell cycle checkpoint that acts unusually late, although there are other chromosome-related checkpoints that act earlier [169]. These findings, along with the lack of evidence so far for a Min system in C. crescentus or other α-proteobacteria, suggest that some other system negatively regulates Z ring assembly in these species. S. meliloti and its relatives such as Agrobacterium species also share an unusual response to cell division inhibition: they form branched cells [170,171]. The mechanism behind branching, whether it is in these species or in E. coli, where it occurs much less frequently [172], is completely unknown but should eventually provide important insights into growth and form [173].


Despite our vast knowledge of how gene expression is regulated and how biosynthetic pathways are interrelated, the challenging questions of how any cell grows and divides remain unanswered. The typical reductionist approach used so successfully to tackle other problems, such as DNA replication, becomes much more challenging in light of the size, complexity, membrane association and topological constraints of the cellular division apparatus. The recent emergence of powerful genomic and cytological tools for studying bacteria has significantly enhanced the ability to address the problem, but the hardest part will be bridging the gap between what is observable on the whole cell level and what can be ascertained by biochemical approaches. This gap is where macromolecular assemblies reside. The only way that we can fully understand how bacterial cell division and other major cellular processes work is by being able to characterize, build and quantify these assemblies. Future work in reconstituting the division apparatus will be difficult, but is really the only way to understand and dissect how this machine works. Such studies may be enhanced by the comparative genomics of a number of microorganisms and the potential for the use of species that either do not require cell division for vegetative growth [174,175], or species such as the crenarchaea that use FtsZ-independent mechanisms. The future looks bright indeed for gaining a better understanding of the fundamental problem of cell division.


I gratefully acknowledge support from the National Science Foundation Grant MCB-9513521 and National Institutes of Health Grant 1R55-GM/OD54380-01.