Chapter 101

Cell Division

JOE LUTKENHAUS and AMIT MUKHERJEE

INTRODUCTION

Cell division involves the partitioning of the cytoplasm into two compartments, each containing a copy of the cell’s genetic information. In gram-negative bacteria such as Escherichia coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) the division event, also referred to as septation or constriction, involves the circumferential invagination of the three layers of the cell envelope between the segregated chromosomes. Although many experiments indicate that septation and the processes of chromosome replication and segregation are not strictly coupled, during balanced growth they are well coordinated such that very few DNA-less cells are formed (70). In addition to this temporal regulation, mechanisms must exist to ensure that division is quantitatively and spatially regulated so that it only takes place once per cell cycle between the segregated chromosomes.

The geometrical and physical parameters pertaining to E. coli growth and cell division have been extensively reviewed by Donachie and Robinson (53). In a more recent review, Donachie (51) has emphasized the importance of attaining a critical cell size on the activation of the major periodic events of the cell cycle. Reviews by Nanninga (112) and Ayala et al. (6) emphasize peptidoglycan biosynthesis. The role of periseptal annuli in cell division has been reviewed by de Boer et al. (44), and the genetics of bacterial cell division have been reviewed by Bi and Lutkenhaus (22). The role of FtsZ in cell division has also been reviewed (22, 94).

CELL DIVISION GENES

Genetics of Cell Division

The identification of essential genes required for cell division involves screening conditional lethal mutants of E. coli as well as other bacteria for filamentous morphology at the nonpermissive temperature. This morphology results from continued increase in cell mass in the absence of septation. Such mutants are classified as having a primary defect in DNA segregation (Par, for defect in partitioning) or septation (Fts, for filamenting temperature sensitive) (71). Par mutants are characterized by a large mass of DNA at the filament center or several unevenly distributed masses of DNA. The par genes include DNA replication genes and genes required to topologically separate the newly replicated chromosomes. In contrast, cells of mutants exhibiting an Fts phenotype are filamentous with regularly distributed nucleoids. Filamentation in the absence of an observable effect on DNA segregation suggests that the fts genes are specifically required for septum formation. Table 1 summarizes properties of the known cell division genes. For each, except ftsW, the essential nature of the gene and the filamentous phenotype due to its loss has been confirmed by construction of conditional null alleles (32, 38, 41, 64, 66).

Table 1Genes required for cell division in E. coli

Roles of the Cell Division Gene Products

FtsZ.

Available evidence indicates that ftsZ acts early in septation before other well-studied cell division genes (93). This conclusion stems from the smooth morphology of filaments resulting from loss of ftsZ function (128). In contrast, filaments formed due to a loss of ftsA or ftsI function have indentations assumed to arise from abortive septation events. This order of gene action is also supported by the morphology of fts rodA double mutants at the nonpermissive temperature (12). rodA (or pbpA) mutants, which lack the elongation mode of peptidoglycan synthesis, have a spherical shape that amplifies the influence of septation on morphology. The ftsZ rodA cells are swollen without any sign of constriction, whereas combining rodA with an ftsA, ftsI, or ftsQ mutation leads to swollen cells with partial constrictions.

Immunolocalization studies show that FtsZ is localized to the division site in a pattern designated the FtsZ ring (20) (Fig. 1). During the cell cycle, FtsZ accumulates at the cytoplasmic membrane at midcell before there is visible invagination of the septum. Throughout septation, FtsZ remains at the leading edge of the septum; however, upon completion of the process, FtsZ is not retained at the new cell pole. On the basis of this dynamic behavior, the stability and abundance of FtsZ (about 10,000 to 20,000 molecules per cell), and genetics indicating FtsZ self-interaction, it was suggested that FtsZ forms a cytoskeletal element that mediates septation (20). This hypothesis, still unproven, forms the basis for ongoing experimentation regarding the structure, function, and localization of FtsZ.

Fig. 1Localization of FtsZ at the leading edge of the invaginating septum by immunoelectron microscopy.

Source: Reprinted from reference 94. See also reference 20.

So far, all evidence indicates that the ftsZ gene, which is highly conserved among bacteria (8, 37), is required for all bacterial cell division events. In E. coli, polar localized septation events, which result in minicell formation, involve the FtsZ ring (23). In the gram-positive organism Bacillus subtilis, ftsZ is essential both for vegetative division and for the asymmetric division occurring during sporulation (10). In this organism as well as several gram-positive cocci, FtsZ is localized to the leading edge of the septum (94, 145). In the filamentous organism Streptomyces coelicolor, ftsZ is not essential for viability, but no septation is detected in its absence (101). This surprising result suggests that this organism can grow in the absence of septation but adds support to the hypothesis that ftsZ is required for septation in all eubacteria. More recently, ftsZ has also been cloned from a species of Mycoplasma and an archaebacterium (X. Wang and J. Lutkenhaus, unpublished data). Since these bacteria lack peptidoglycan, this finding has implications for the function of FtsZ (see below).

The study of FtsZ was stimulated by reports that it is a GTPase (43, 103, 117). Although FtsZ does not contain the three motifs that are present in the GTPase superfamily, it does contain a highly conserved sequence motif, GGGTGTG, that is similar to the tubulin signature motif (G/A)GGTG(S/A)G. The importance of this sequence in the ability of FtsZ to hydrolyze GTP is supported by the study of two mutant FtsZ proteins. Both FtsZ3 (GGGAGTG) and FtsZ84 (SGGTGTG) show dramatically reduced GTP binding and GTPase activities (43, 103, 117). Furthermore, the ftsZ84 mutation alters the substrate specificity of FtsZ, allowing it to hydrolyze ATP (118). Even though FtsZ and tubulin share this signature motif, little additional sequence similarity exists. A comparison of a family of FtsZs with members of the three families of eukaryotic tubulins reveals only limited similarities (105). Additional experimentation is necessary to determine if these limited similarities comprise the GTP binding sites in these proteins.

The E. coli FtsZ GTPase has been characterized in three independent studies (43, 103, 117), and it was observed that the kinetics of GTP hydrolysis vary remarkably depending on the method of FtsZ purification. In one study (117), the purified FtsZ had bound GDP and the GTPase activity did not display a lag. In the other two studies (43, 103), the FtsZ had no bound GDP and the GTPase displayed a lag that was inversely proportional to the FtsZ concentration. The lag could also be suppressed by preincubation with GTP or treatments likely to affect the conformation of the protein. In another study, FtsZ isolated from B. subtilis did not have a lag but the GTPase activity did display a dramatic dependence on the protein concentration, with little activity below 100 μg/ml (144). These studies suggested that the GTPase activity is dependent on interaction of FtsZ molecules and raised the possibility that FtsZ used GTP to form the FtsZ ring (43, 103). This latter possibility is supported by the division defects of FtsZ3 and FtsZ84, which are deficient in interaction with GTP (18, 95). A speculative model has been proposed for FtsZ localization on the basis of analogy with tubulin polymerization (94). In this model, it is suggested that the FtsZ ring is a cytoskeletal element whose formation is dependent on a GTP-GDP cycle. Furthermore, it is proposed that the assembly of FtsZ to form this structure is due to the appearance of a nucleation signal at midcell that is under cell cycle control. At present, nothing is known about the nature of this hypothetical signal.

Support for the suggestion that FtsZ functions as a cytoskeletal element comes from two in vitro studies that demonstrate that FtsZ can polymerize into filaments in the presence of guanine nucleotides (28, 105) (Fig. 2). In the first study, polymerization was observed by electron microscopy by two different methods (105). In one approach polymerized FtsZ was observed by rotary shadowing electron microscopy. The assembly of the filaments required GTP, and the diameter of the observed filaments suggested a linear polymer of FtsZ monomers: a protofilament. In a second approach, the polycation DEAE dextran, which has been used to enhance polymerization of purified tubulin (56), was found to stimulate FtsZ polymerization. Negative-stain electron microscopy revealed microtubular-like structures that might arise from the association of protofilaments. Polymerization required GTP or GDP and occurred without GTP hydrolysis. In addition, two mutant proteins were examined. FtsZ3, which is unable to support cell growth and shows a marked deficiency in interaction with GTP, was found not to polymerize. In contrast, FtsZ2, which binds GTP, displays reduced hydrolysis, and is able to support cell growth, was found to polymerize. In the second study (28), polymerization into microtubular-like structures was observed without the addition of DEAE dextran. Polymerization occurred in the presence of high concentrations of GTP, and FtsZ could be cycled through rounds of assembly and disassembly. Also, the FtsZ84 protein, previously shown to be deficient in GTP hydrolysis, showed a reduced, albeit significant, efficiency to polymerize. Although this latter study suggested that assembly occurred quite readily into microtubular-like structures in the absence of DEAE dextran, it is now apparent that the pH had been inadvertently lowered and that assembly is enhanced at a nonphysiological, acid pH (D. Bramhill, personal communication). Thus, the protofilament may represent the physiologically relevant form of FtsZ. It should be cautioned, however, that proteins such as EF-Tu have been shown to polymerize in vitro although they are unlikely to do so in vivo (11), and filaments corresponding to FtsZ have not been observed in electron micrographs of thin sections of E. coli. However, if protofilaments represent the assembled form of FtsZ, they would be almost impossible to detect in vivo.

Fig. 2Filaments of FtsZ.

Source: Reprinted from reference 105.

Until recently, few mutations existed in the ftsZ gene that might help to further elucidate its function. The classic ftsZ84(Ts) mutant is unable to divide and form the FtsZ ring at the nonpermissive temperature (E. Bi and J. Lutkenhaus, unpublished data), consistent with its reduced ability to polymerize in vitro (28). The ftsZ26(Ts) mutant is also unable to localize FtsZ at the nonpermissive temperature; however, at the permissive temperature, the cells have an altered polar morphology that correlates with an altered geometry of the FtsZ ring, not always perpendicular to the long axis of the cell (21). This result suggests that the geometry of the FtsZ ring is a property of FtsZ itself and suggests that the FtsZ ring directly determines septal growth. Thus, the function of the FtsZ ring may be analogous to that of the contractile ring in animal cells, and the invagination of the septum may be driven by motor proteins, thus far unidentified, acting on the FtsZ ring (28). In addition, the the occurrence of FtsZ in both wall-less bacteria and an archaebacterium, which lack peptidoglycan (Wang and Lutkenhaus, unpublished data), raises the possibility that FtsZ’s primary function is to cause invagination of the cytoplasmic membrane, which may indirectly stimulate septal peptidoglycan synthesis.

Several studies have concluded that the level of FtsZ is rate limiting for cell division, perhaps reflecting a critical concentration required for formation of the FtsZ ring. This conclusion stems from observations on altering the levels of FtsZ. A small increase in the level of FtsZ leads to a hyperdivision activity, seen as a minicell phenotype (54), whereas lowering the level leads to a block to division (39). In several studies, a relatively small decrease (30 to 50%) appeared to be sufficient to block division (38, 58). However, in another study in which an antisense RNA was used to limit ftsZ expression, a 70% decrease resulted in elongated cells that were still able to achieve steady-state growth (130, 131). The increase in cell length was due to a delay in septation which correlated with a delay in nucleoid segregation. These results suggest FtsZ controls the timing of septation. It should also be noted that excess FtsZ (beyond the amount found to induce minicell formation) leads to the same phenotype as loss of FtsZ, i.e., filamentation (146). This phenomenon of the same phenotype with too little or an excess of a protein is seen with some other proteins, for example, the RepA protein of pSC101 or P1 (42, 79, 147). In these cases, it is thought that overproduction shifts the equilibrium from an active monomer to an inactive aggregate. Perhaps the overproduction of FtsZ leads to inappropriate aggregation that reduces activity.

It should be noted that there are other indications that FtsZ occupies a critical position in the cell division pathway. Most notably, FtsZ is the apparent target of the cell division inhibitors SulA, produced during the SOS response (18, 85, 92), and MinCD, a cell division inhibitor involved in placement of the division site (17, 47). Both of these inhibitors block division by preventing formation of the FtsZ ring (23). The negative effect of these inhibitors on cell division can be overcome by increasing ftsZ expression, either through increased gene dosage (47) or through an increase in the activity of positive activators such as sdiA (144). The inhibitors can also be suppressed by mutations in ftsZ, originally referred to as sulB or sfiB mutations (18, 92). These mutations appear to result in a form of FtsZ that is resistant to SulA.

FtsZs from a wide variety of eubacteria including gram-negative and gram-positive organisms (8, 72, 98, 152) are rich in glycine and alanine, with these residues accounting for almost 50% of the invariant residues. Also, all of the FtsZs are rich in acidic amino acid residues; the isoelectric point of the E. coli protein is 4.5, which makes it one of the most acidic proteins in E. coli (152). Comparison of the FtsZs indicates that the protein can be divided into two domains (Wang and Lutkenhaus, unpublished data): an N-terminal, GTP binding domain comprising about 80% of the protein that is highly conserved (30% invariant residues) and a C-terminal domain comprising about 20% of the protein that is highly variable except for a short segment at the extreme carboxyl end (Wang and Lutkenhaus, unpublished data). Despite the high degree of conservation, the ftsZ of B. subtilis cannot substitute for the E. coli ftsZ; instead, it expresses a dominant negative interference leading to filamentation (8).

FtsA.

As noted above, cells lacking FtsA function form indented filaments resulting from a block after septation has initiated, suggesting that they are blocked at a step after formation of the FtsZ ring. FtsA is present at about 200 molecules per cell, and about half of the molecules are associated with the cytoplasmic membrane (115). FtsA has sequence homology to the ATP binding domain of a number of ATPases, including actin, DnaK, and sugar kinases. Similar homology is also observed between these ATPases and MreB, which is involved in cell shape determination and was previously shown to have limited homology to FtsA and DnaK (99). The homology of FtsA to this actin/DnaK/hexokinase family has been used to build a structural model (24, 122). In support of this model, FtsA has been shown to bind to an ATP affinity column and to undergo phosphorylation at a threonine that corresponds to a threonine that is phosphorylated in DnaK (122). One possibility is that FtsA acts directly on FtsZ, possibly preventing nonproductive aggregation. This suggestion stems from the observation that the ratio of FtsZ to FtsA is important for cell division to proceed (39, 50).

The location of a portion of FtsA on the cytoplasmic membrane and its importance in FtsZ function raise the possibility that FtsA acts to link the FtsZ ring and septal specific peptidoglycan biosynthesis. It has been suggested that FtsA acts at the septum, since it has been observed that abortive septa formed in the presence of FtsA3 (which is irreversibly inactivated at 42°C) are not immediately used upon return to permissive conditions (133). In contrast, abortive septa formed in the presence of other mutant FtsA proteins are readily used upon return to permissive conditions. One interpretation is that the inactivated FtsA3 protein is retained at the aborted septa, blocking newly synthesized protein from acting. On the other hand, it is possible that at the nonpermissive temperature, the FtsA3 protein acts enzymatically to produced damaged septa. FtsA may interact with penicillin-binding protein 3 (PBP3) because FtsA3 prevents the in vitro labeling of PBP3 with a radioactive penicillin derivative (132).

FtsI (PBP3).

E. coli contains four high-molecular-weight PBPs (PBP1a, PBP1b, PBP2, and PBP3) which are thought to have transpeptidase and transglycosylase activities (55). Of these, the ftsI gene product, PBP3, is specifically required for septal peptidoglycan biosynthesis (123). This unique requirement for PBP3 has been demonstrated by the use of both PBP3-specific antibiotics (25) and conditional lethal mutations that map to the ftsI gene (125). Both ways of blocking PBP3 activity specifically block cell division and result in the formation of indented filaments with well-separated nucleoids. This result indicates that PBP3 activity is required only for septation and may be activated only during septation.

PBP3, like the other high-molecular-weight PBPs, has a single transmembrane domain that fuses a short cytoplasmic tail to the large periplasmic domain (27). It was noted that PBP3 has a potential lipoprotein modification sequence; however, biochemical and genetic studies argue that it is not a lipoprotein (68). It has also been demonstrated that PBP3 undergoes a posttranslational C-terminal cleavage that removes 11 amino acids (110), which raises the possibility that posttranslational modification is involved in regulation of the activity of PBP3. However, the functional significance of this cleavage is unclear. Constructs in which the C-terminal portion is removed function quite well, and a mutant that is unable to carry out the cleavage grows and divides normally (65, 67). Thus, a crucial question remaining is how the activity of PBP3 is topologically and temporally regulated.

FtsQ, FtsL, and FtsN.

Three additional cell division genes have been identified and designated ftsQ, ftsL (mraR), and ftsN (13, 41, 64, 136). Each of the genes encodes a low-abundance membrane protein (∼50 molecules per cell) that has a simple transmembrane topology. In each case, a short N-terminal cytoplasmic tail is fused through a hydrophobic transmembrane domain to a larger periplasmic domain. Even though the gene products are relatively low in abundance, the cell division process is relatively insensitive to their overproduction. Increases of 50-fold in the level of these gene products do not appear to affect the division process (13, 41, 64). This result suggests that if these gene products are directly involved in septation, they need to be activated locally.

Even though these genes have been characterized, almost nothing is known about their biochemical activities. FtsL has a potential leucine zipper motif in its periplasmic domain, raising the possibility that it could dimerize (64). In addition to a conditional null mutation, two temperature-sensitive mutations have been characterized in ftsL (80, 136). One mutation altering the start codon results in filamentation, similar to the null phenotype, whereas a second missense mutation results in a temperature-dependent lysis phenotype. These results indicate that lack of FtsL blocks cell division but that a defective product affects the integrity of the septation process.

FtsN was isolated as a multicopy suppressor of a temperature-sensitive ftsA mutation (41). Increased levels of FtsN also suppress a temperature-sensitive ftsI mutation, and to some extent an ftsQ mutation, but not a variety of other temperaturesensitive mutations, including ftsZ84(Ts). The mechanism of suppression is unknown. However, it is surprising that multicopy ftsN can suppress both a mutant protein in the cytoplasm, FtsA, and one in which the mutant domain is present in the periplasmic space, PBP3. Possibly FtsN acts in a complex with FtsA and PBP3 at the septum and the increased FtsN favors the formation of complexes destabilized by mutant proteins.

Products of Other fts Genes.

Several other genes with the fts designation have been described. Among these, ftsW encodes a protein that is homologous to RodA, a protein that functions with PBP2 to maintain cell shape (81, 126). By analogy, it has been suggested that FtsW functions in concert with PBP3 during septum formation (78). However, morphological studies suggest that ftsW acts early, before PBP3 (88). A mutation in ftsE, located at 76 min in an operon with ftsY and ftsX, leads to filamentation (62). Since this filamentation is medium dependent, it has been questioned whether ftsE represents a true cell division gene (128). The FtsE protein has homology to a family of proteins involved in diverse transport processes, but its specific role is unknown (61). Initially, ftsH was classified as a cell division gene. However, the original mutant had two mutations, and the one responsible for filamentation was in the gene encoding PBP3 (ftsI) (15). It is possible that FtsH has a role in PBP3 assembly into the membrane.

Divisome, Septalsome, and Septator

The biosynthesis of the septum is surely a complex process. It must involve a subset of the cell’s peptidoglycan biosynthetic machinery, mechanisms for the localized action of this machinery and for the circumferential invagination of the nascent septum. Because of the number of gene products involved, the existence of a macromolecular complex that functions to form the septum has been suggested. Several very reasonable names have been suggested for this hypothetical complex, including septalsome (73), divisome (112), and septator (139). However, the evidence to date does not allow one to distinguish between division proteins acting in sequence or as an actual complex. The evidence consists of inhibitory effects of altering the levels of one of the gene products on division, especially in mutant backgrounds (39). Examples are the inhibitory effect of increasing ftsQ, ftsL, or ftsI in various mutant backgrounds (ftsZ, ftsA, and ftsI) but not in the wild type and the already noted inhibitory effects of varying the levels of FtsZ and FtsA and how they can counteract each other (39, 86).

Gene Organization within the Large 2-Min Cluster

Many of the mutations that have been isolated that affect cell division map at 2 min within the large mra cluster. This region has been completely sequenced, and genetic and biochemical studies have delineated the functions of many of these genes (Fig. 3). Within this cluster, most of the genes function either in the biosynthesis of the precursor for peptidoglycan (UDP-N-acetylmuramyl pentapeptide) or in its transport across the membrane or have the fts designation, indicating that their primary role is in cell division. An exception is the envA gene located at the 3' end of this cluster. envA has long been considered to have a role in cell separation (53); however, it has recently been shown to encode a gene required for an early step in lipopolysaccharide biosynthesis (116). In addition, upstream of ftsL there are two open reading frames of unknown function (6). The tight juxtaposition of genes in this cluster, all oriented in the same direction, indicates that these genes may be cotranscribed. This possibility is consistent with the location of the only transcriptional terminator in this region, just downstream of the envA gene (9). Thus, transcription beginning anywhere within the cluster could be expected to continue to this terminator. Despite this simple organization, it is clear that many promoters are scattered throughout the region, providing possibilities for differential expression. It is likely that regulation of this region, only beginning to be addressed, is quite complex.

Fig. 3Gene organization within the 2-min region of the E. coli chromosome. The complete nucleotide sequence of this region has been assembled by J. A. Ayala (GenBank accession number X55034). All genes are transcribed left to right. The 3? end of the gene cluster is a terminator just beyond the envA gene (indicated by the hairpin structure) (9). The 5? boundary is not well defined. The locations of promoters that impact ftsQAZ expression are indicated, although evidence suggests that additional upstream promoters contribute to the express of ftsZ. The putative antisense RNA for ftsZ would initiate at PstfZ and end at the terminator (indicated by the hairpin structure( located between the ftsA and ftsZ genes. This terminator only affects transcription proceeding from right to left.

Source:

Regulation of the ftsQAZ Cluster.

Increasing the gene dosage of the ftsQAZ cluster leads to an increase in the level of these gene products, resulting in a hyperdivision activity expressed as a minicell phenotype (146). This observation, along with other observations that raising the level of ftsA or ftsZ alone is deleterious, suggests that the level of these proteins normally has to be maintained within a narrow range (39).

On the basis of complementation, the use of promoter detection vehicles, and primer extension experiments, the promoters in the ftsQAZ region have been identified (Fig. 3) (5, 119, 120, 127, 153). Two promoters, P1_ftsQ and P2_ftsQ , are located upstream of ftsQ within the ddl gene and transcribe ftsQ, ftsA, ftsZ, and envA (5). Although these genes are coordinately expressed from these promoters, the levels of the gene products vary as a result of different efficiencies of translation (104) and additional promoters located downstream. The two ftsQ promoters are differentially regulated. Expression from P2_ftsQ occurs throughout growth, and expression from P1_ftsQ , a so-called gearbox promoter, increases as the growth rate declines (5). Gearbox promoters are defined as those exhibiting a level of expression that is inversely proportional to the growth rate of the culture (5). Expression from P1_ftsQ is dependent on rpoS, and increased expression from this promoter may be responsible for increasing ftsQAZ expression as the growth rate declines. This increased rate of expression could in part be responsible for an increased rate of division that results in smaller cells as cultures enter stationary phase (91). Expression from P2_ftsQ is controlled by sdiA, a member of the luxR family of transcriptional regulators (57, 144). Cloning sdiA on a multicopy plasmid leads to increased expression from P2_ftsQ , which results in a minicell phenotype and suppression of the block to septation due to overexpression of minCD (144). The physiological role of sdiA is unknown, however, as inactivation of sdiA does not affect the pattern of division under standard laboratory conditions. Nonetheless, it seems likely that these promoters allow the cell to coordinately adjust the level of ftsQAZ expression in response to changes in growth rate and other physiological conditions and thereby adjust cell size. For example, it is likely that the rpoS-dependent reduction in cell length that occurs during entry into stationary phase is due in part to rpoS acting directly or indirectly at P1_ftsQ (91). It is also possible that the presumed increased expression of ftsZ due to increased amounts of rcsB or rcsF occurs through either P1_ftsQ or P2_ftsQ (59, 60). This needs to be further clarified.

In addition to the upstream promoters that transcribe ftsZ coordinately with ftsQ and ftsA, ftsZ is also expressed from promoters located within the ftsA gene (96, 119, 127). Originally four transcriptional start points were identified within ftsA by primer extension analysis (5). However, one of these ends, corresponding to P1_ftsZ , is generated by RNA processing (C. A. M. Kaymevang and J.-P. Bouche, personal communication). Of the remaining three promoters, P2_ftsZ accounts for most of ftsZ expression arising from promoters within the ftsA gene (58). Recent studies from Vicente’s laboratory (58) using a PCR transcription titration assay to quantitate promoter activity at the resident chromosomal location demonstrate that the P2_ftsZ promoter undergoes periodic activation during the cell cycle coincident with the initiation of DNA replication. In contrast, Zhou and Helmstetter (154), using a quantitative S1 hybridization assay, found that all promoters for ftsZ, including those upstream of ftsQ, exhibited cell cycle periodicity. They concluded that the periodicity resulted from an inhibiton of transcription as replication forks pass the 2-min region. This aspect of ftsZ regulation needs to be further clarified because it may be critical for the observed coordination between DNA replication and cell division (130). One attractive candidate for temporal regulation of ftsZ, dnaA, does not appear to play such a role (58). There is some evidence that ppGpp may activate ftsZ expression and by a process of elimination operate at P2_ftsZ ; it does not operate at the four upstream promoters (142).

The regulation of expression of ftsZ may be even more complex, since it is affected by natural antisense RNAs. The first antisense RNA for ftsZ, designated dicF, is part of a defective prophage present in some, but not all, E. coli strains as well as other bacteria (26). No physiological condition that leads to expression of this antisense RNA is known, although it has been used experimentally to manipulate the level of FtsZ (130, 131). More recently, indirect evidence for a second putative antisense RNA has been presented. This antisense RNA, designated stfZ, is produced by antisense transcription of the 5' end of ftsZ. The antisense RNA initiates at a promoter that must lie just inside the 5' end of the ftsZ gene and ends at a terminator, which is active only in the antisense direction, located between ftsZ and ftsA (49). This RNA inhibits division when the gene for stfZ (but not the entire ftsZ gene) is present at high copy number and at high temperature. This same requirement for high temperature for effective inhibition of division is also seen with dicF (130). Whether stfZ plays a physiological role is unknown.

Expression of ftsI (pbpB) and ftsL.

The ftsI gene is located near the 5' end of the 2-min cluster just downstream of ftsL (mraR) (111). Both of these genes are weakly expressed, resulting in approximately 50 molecules per cell (64, 124). Two additional open reading frames of unknown function lie just upstream of ftsL, and their proximity to ftsL makes it likely that they are cotranscribed with ftsL and ftsI. In fact, available evidence suggests that expression of ftsI is dependent on sequences far upstream of ftsL (66). The level of PBP3, as monitored by β-lactam binding, is constant during the cell cycle, suggesting that PBP3 is activated during septation (149). It is known that the level of PBP3 is affected by the mre locus, which is involved in shape determination (143). A mutation in the mreB gene leads to an increase in the level of PBP3, and spherical morphology and deletion of the entire mre locus lead to even further increases in PBP3. The regulatory mechanism for this effect is unknown. It is also unknown if the spherical morphology is due to the increase in PBP3 or to other effects of inactivation of the mre locus.

INFLUENCE OF DNA REPLICATION AND SEGREGATION ON DIVISION

SulA-Mediated Inhibition of Cell Division

In E. coli, blocking DNA replication results in inhibition of cell division by at least two mechanisms (31). The first and best understood mechanism involves induction of SulA (SfiA), a component of the SOS response (75, 76). A rapid increase in SulA occurs following DNA damage, leading to a rapid block to cell division. This block is transient because as DNA damage is repaired and the SOS response is again repressed, SulA disappears through rapid degradation by the lon protease (103). Genetic and biochemical evidence indicates that FtsZ is the target of SulA (18, 85, 92), and immunoelectron microscopy indicates that SulA inhibits formation of the FtsZ ring (23), thus blocking cell division at an early step. Once SulA is removed by proteolysis, the FtsZ ring is able to form and cells divide, even without additional FtsZ synthesis (23, 97). Thus, SulA is an effective reversible inhibitor, in part because it is an unstable protein that is readily degraded by the lon protease and in part because SulA does not irreversibly damage the division apparatus. An attractive possibility is that SulA blocks FtsZ polymerization.

Since the SulA-mediated inhibition of cell division is lethal when enhanced in a lon background, suppressors can be readily selected. Most suppressor mutations that do not inactivate SulA map to the ftsZ gene (63). A number of these mutations have been characterized and found to be scattered throughout the ftsZ gene (18). One common property of the corresponding mutant proteins is that they display a reduced GTPase activity (40). However, other mutant proteins with reduced GTPase activity (43, 117), such as FtsZ84, are not resistant to SulA (18), suggesting that selection for SulA^r yields a special class of mutations. One suggestion is that such mutations alter the FtsZ protein, yielding a form of the protein which is resistant to interference by SulA.

Non-SulA-Mediated Inhibition, the Par Phenotype

Even in the absence of SulA, however, cell division is at least transiently blocked whenever DNA replication or segregation is affected (31). Almost any perturbation of DNA metabolism leads to some degree of filamentation and in some cases minicell formation. When DNA replication or segregation is completely blocked, filamentation occurs, but after a delay some degree of division resumes, away from the DNA mass, producing DNA-less cells (107). Thus, the DNA replication-segregation cycle and the cell division cycle are not obligatorily coupled (16, 114). However, interference with the DNA replication-segregation cycle prevents division from occurring at midcell and causes at least some filamentation before division resumes in DNA-free regions. One well-studied example is thymine starvation, which blocks DNA replication and leads to filamentation and, after a lag, to the production of DNA-less cells (84). In the presence of SulA, cell division is more efficiently blocked and fewer DNA-less cells are produced. Thus, conditional lethal mutations in any of the DNA replication genes will probably lead to filamentation and the production of DNA-less cells at the nonpermissive temperature, especially if SulA is not induced (107). Although the block to division could be a direct consequence of blocking replication, it may be due to the absence of DNA termination and segregation. In addition, inactivation of genes required for topological resolution of the replicated chromosomes results in a Par phenotype. Such genes include those encoding for gyrase (77, 87), topoisomerase IV (1), and the dif system (89; chapters 100 and 105 in this volume).

Other Genes Affecting Segregation and Division

The thinking about segregation of chromosomes has been dominated by the replicon model, which invokes attachment of the chromosomes to the membrane and segregation resulting from zonal growth between the attachment points (82). However, several studies suggest that E. coli has a system for the rapid displacement of replicated chromosomes (52, 69). It should be noted that this conclusion is still controversial, as the methodology for measuring nucleoid positioning has been questioned (138). This topic is further discussed in chapter 105.

A candidate for a protein involved in an active segregation is MukB, which has characteristics of a force-generating protein (113). The mukB gene was identified following an ingenious screen designed to look for mutations producing anucleate cells (70). The phenotype of a mukB null mutant at low temperature includes the production of normal-size anucleate cells (5%) and cells that have twice the normal content of DNA. In addition, some cells are observed in which septation has apparently cleaved the chromosome, a rare phenomenon for E. coli. This phenotype at low temperature is distinct from that of par mutants that have defects in untangling the newly replicated chromosomes described above, since division appears to occur at the normal time and place. However, at higher temperatures, mukB is essential and the phenotype resembles Par, with filamentous cells containing extended nucleoids (113). This phenotype cannot be traced to lack of topoisomerase activity or SOS induction, indicating it may be directly due to a lack of mukB.

Deletion of dnaK also results in a Par phenotype (29, 30). At 30°C, the dnaK gene is not absolutely essential; however, an obvious Par phenotype is apparent. Since the primary role of DnaK is in the protein folding pathway, it is likely that a protein involved in partition requires DnaK. Interestingly, the filamentous morphology of DnaK deletion mutants can be suppressed by increasing the level of FtsZ, suggesting that the segregation defect is also suppressed (29).

There are several observations that appear to link ftsZ to nucleoid segregation. Reducing the level of expression of ftsZ fivefold by expressing the antisense RNA, dicF, results in an increase in cell volume and retarded nucleoid separation (130). This was a surprising result, since it had previously been shown that cell division was quite sensitive to smaller changes in FtsZ levels and that nucleoid segregation appeared quite normal as cells filament in the complete absence of FtsZ function (38). To explain the effect of reduced FtsZ on nucleoid separation, Tetart et al. (130) suggested that FtsZ inhibits nucleoid segregation until a threshold level that is able to carry out division is reached. Upon reaching this threshold level, the FtsZ may take on a new form that is able to initiate division but can no longer inhibit nucleoid segregation. In light of results on FtsZ localization (20) and in vitro assembly of FtsZ (28, 105), this new form could be the FtsZ ring. Some caution must be taken in interpreting these results, however, as dicF could have effects in addition to reducing ftsZ expression. Also, segregation is normal in the ftsZ84(Ts) mutant at the nonpermissive temperature (29, 71) even though it must be close to the threshold for division, since increasing the level of the mutant protein less than twofold can rescue division (144). On the other hand, it is has been observed that the Par phenotypes seen in dnaK and min mutants are reversed by overexpressing FtsZ (19, 29) and that the ftsZ84(Ts) mutation can be suppressed by mutations in gyrB (121).

Four genes have been identified on the basis of resistance to camphor, which is known to cause increased ploidy in eukaryotic cells (134). Mutations in these genes, designated mbrA to mbrD, confer resistance to camphor, temperature sensitivity, and sensitivity to growth on rich media. An examination of the morphology of the mbrA4, mbrC17, and mbrD19 mutants after a shift from minimal to rich medium revealed that the cells filamented and contained ribbons of DNA, indicating retarded segregation. The results suggest that these mutants have difficulty in adjusting the rate of chromosome segregation and cell division to the new growth rate; they cannot keep up with the rate of cell mass increase. Interestingly, segregation and cell division are restored as cells enter stationary phase, where the decrease in growth rate may allow DNA segregation and cell division to catch up. On the basis of similar morphology of a mutation in murI and map location, it has been suggested that mbrC is identical to murI (7), a gene required for the biosynthesis of d-glutamic acid, a specific component of peptidoglycan (54).

LOCALIZATION AND FORMATION OF THE DIVISION SITE

In cells of E. coli or S. typhimurium in balanced growth, cell division occurs at the center of the long axis of the cell; however, the mechanism for positioning the division event at midcell is unknown. Models for positioning the division event encompass two extreme possibilities and suffer from a lack of corroborating biochemical or genetic data. In one model, potential division sites preexist on the cell envelope and division at midcell occurs by selection of the topologically correct site (33, 44). In a second model, no sites exist on the cell envelope and positioning of the division event is determined by replication and segregation of the DNA (150). In any event, the min locus affects placement of the division site, and continued analysis of this locus may lead to the underlying mechanism.

The min Locus

The placement of the division site is dramatically affected by mutations at the min locus, which consists of three genes, minC, minD, and minE (46). The classic min mutation of Adler et al. (2) occurs in minD (109) and results in a minicell phenotype, namely, division events at the poles to produce anucleate minicells. Deletion of the entire min locus results in the same phenotype, stressing that min is not required for the process of septation but rather is required for its efficient localization at midcell (46). An early hypothesis for min function was that it inhibited cell division from occurring at the cell poles, the old division sites, thus restricting division to midcell (129). This hypothesis received support from experiments demonstrating that a division inhibitor was encoded by the min locus (46). By analyzing the effects of expressing various combinations of the min genes, it was revealed that minC and minD encode a bipartite division inhibitor. Subsequent work has shown that MinC is the portion of the inhibitor that is proximal to the division apparatus (47, 48) and that MinD is a membrane-bound ATPase that activates MinC (45). It appears that this ATPase activity is required for MinD function.

Importantly, the activity and topological specificity of MinCD are somehow determined by the ratio of MinCD to a third component, MinE, a small polypeptide consisting of 88 amino acids (46). When MinE is in excess, the inhibitory effect of MinCD is canceled and a minicell phenotype is observed, division occurring at midcell or the cell poles (46). When MinCD is in excess, its inhibitory activity is dominant and cell division is blocked at all sites. However, when the ratio of MinE to MinCD is normal, the typical division pattern is observed, with septation only occurring at midcell. How this combination of MinCDE achieves this topological specificity is unknown. The target of the MinCD inhibitor appears to be FtsZ, since increasing the level of FtsZ and certain mutations in ftsZ can suppress the effect of MinCD (17, 47). Also, MinCD, like SulA, can inhibit the formation of the FtsZ ring (23). Interestingly, MinC can also combine with DicB, encoded by a defective prophage, to become an inhibitor without topological specificity (47, 90). Since this inhibitor is nonresponsive to MinE, this finding, along with other evidence, implies that the topological specificity of the MinCDE complex is imparted by MinE.

An additional unexpected feature of the min phenotype is the appearance of a fraction of cells in the population with abnormally segregated nucleoids (3, 83, 106) and branched cells, with the frequency of branching dependent on the genetic background and the medium (4). The observed abnormal DNA segregation raises the issue of whether the primary function of min is associated with chromosomal segregation, the effect on septal positioning being secondary. Although the suspected interaction of min with FtsZ favors a more direct effect on septation, the possibility that FtsZ can also affect segregation (130) emphasizes that more needs to be learned about the activities of all proteins involved.

Periseptal Annuli

In the periseptal annuli model, the placement of the division event is determined by preexisting annuli, circumferential zones of adhesion between the cytoplasmic membrane and the murein-outer membrane layers (44). Periseptal annuli flanking a completed septum were first detected by electron microscopy of a mutant of S. typhimurium that forms chains of cells (100). The existence of periseptal annuli raised the question of whether their formation preceded the initiation of septation or was the result of the septation process. If their formation preceded septation, they were likely to be involved in determining the division site. By determining the positions of plasmolysis bays, used as markers for the boundaries of annuli, it was observed that newborn cells growing in rich media contain incomplete annuli (indicated by the presence of partial plasmolysis bays) at midcell (33). As cells of increasing length are analyzed, the annuli at midcell are complete and additional new annuli appear to arise at midcell and migrate to the one-quarter and three-quarters positions. Following cell division, these latter annuli appear to be retained and become the future periseptal annuli. If the interpretation of these observations is correct, it suggests that the annuli are morphological precursors of the division site.

The relevance of plasmolysis bays in the biogenesis of the division site has been questioned in another study in which it was observed that plasmolysis bays occur primarily at new cell poles but internal bays occur fairly randomly along the cell’s long axis (108). Thus, the positioning of plasmolysis bays may be artifactual and determined mainly by hydrodynamic effects. In addition, the random position of plasmolysis bays in ftsZ mutant filaments did not correlate with the rather discrete division of the filaments upon a return to the permissive temperature.

More recently, Cook and Rothfield have reexamined the positioning of plasmolysis bays in ftsZ and ftsA mutant filaments (35, 36). Significantly, it was observed that plasmolysis bays occurred at regular intervals along the cell length in ftsZ mutant filaments, in contrast to earlier results reporting random positioning (34). The earlier results with the ftsZ mutant were obfuscated by using conditions that did not completely block division at the nonpermissive conditions. The earlier result was interpreted as ftsZ having a primary role in positioning periseptal annuli (44); however, these recent results rule out this role for ftsZ. The discrete positioning of plasmolysis bays in ftsZ mutant filaments, which are smooth cylinders and therefore not subjected to the hydrodynamic effects present in indented filaments, suggests that the bays arise from attachment of the cytoplasmic membrane to the cell wall.

Nucleoid Occlusion

In a second model, the nucleoid occlusion model proposed by Woldringh et al. (150), the location of the division event is determined as a result of DNA replication and segregation. This model arose from observations on the division pattern in DNA replication mutants (107). When DNA replication or segregation is blocked, even in the absence of the SOS response, septation is inhibited in the area of the nucleoids although septation, after some delay, occurs in DNA-free regions of the cell. These division events producing anucleate cells do not occur at a fixed distance from the poles but rather occur randomly in the anucleate regions. As cells resume DNA replication, septation takes place close to the nucleoid, suggesting that the actively replicating nucleoid stimulates septation nearby. In this model, it is suggested that the nucleoid exerts a negative effect on septation but that the replicating nucleoid exerts a positive effect. When nucleoids segregate, the positive effect can locally overcome the negative effect, leading to septation. Furthermore, it is suggested that these positive and negative effects directly influence the topography of peptidoglycan synthesis, leading to septation. Although this model suggests similarities between prokaryotes and eukaryotes, with chromosome segregation playing a critical role in determining the site of cell division, it is a difficult model to test.

PEPTIDOGLYCAN SYNTHESIS AND SEPTATION

The process of septation involves the invagination of the cytoplasmic membrane, with the accompanying ingrowth of the peptidoglycan layer and outer membrane. The invagination of the outer membrane may be passive and not required for the division of the cytoplasmic contents into two distinct compartments. In contrast, there is considerable evidence that ingrowth of the peptidoglycan layer is critical for the septation event, and it is not clear if the division of the cytoplasm into two compartments can occur without it. Thus, a crucial event in septation may be the activation of septal peptidoglycan synthesis.

Topology of Peptidoglycan Synthesis

The topology of peptidoglycan biosynthesis during the cell cycle has been examined by in situ autoradiography to monitor [meso-³H]diaminopimelic acid incorporation into peptidoglycan (148). In nondividing cells, there is diffuse incorporation along the entire length of the cell, whereas in dividing cells, incorporation occurs mainly at midcell at the leading edge of the constriction. This observation suggesting localized synthetic activity raises questions as to how the biosynthetic activity switches from diffuse to septal biosynthesis and whether there is a difference in composition between the lateral cell wall and the polar caps. Despite intense efforts using current separation techniques, no muropeptide that is specific to the septum has been identified (112), even though septation specifically requires PBP3.

PBP3 is required for septation but not for initiation of this event (112). Filaments that are formed because of lack of PBP3 activity contain blunt constrictions that appear to arise from septation events that initiate but are then aborted because of the lack of PBP3 activity. Examination of the topology of diaminopimelic acid incorporation in such filaments revealed that diaminopimelic acid is incorporated as such blunt constrictions are formed, but once they are formed, there is no further localized incorporation. This result suggests that a penicillin-insensitive step may occur early in septation before PBP3 is involved. This step has been designated penicillin-insensitive peptidoglycan biosynthesis (112).

Substrates for PBP3

One approach to the regulation of PBP3 activity has been to try to identify interacting proteins genetically by looking for extragenic suppressors of a temperature-sensitive mutation in ftsI, the gene encoding PBP3. One such attempt revealed the unexpected finding that an increase in the level of PBP5 (a dd-carboxypeptidase) compensated for the reduced PBP3 activity at the nonpermissive temperature (14). The suppression required the temperature-sensitive protein, suggesting that mutant PBP3 was not being bypassed but that its activity was somehow being increased. Further study revealed that other manipulations that could be expected to increase the level of the tripeptide acceptor for peptidoglycan synthesis also suppressed a temperature-sensitive allele of ftsI (14). This finding led to the hypothesis that PBP3 prefers the tripeptide acceptor as a substrate and suggests that changes in the ratio of the tri- to pentapeptide acceptor may be important in switching cells from an elongation mode of peptidoglycan growth to a septal mode of growth. This result also suggests that the initial goal, identification of interacting proteins, will be difficult to achieve by this approach.

Role of PBP2

For wild-type E. coli to be viable and maintain its rod shape (chapter 68), it requires the activity of a second high-molecular-weight PBP, PBP2, plus RodA, the product of the adjacent rodA gene (123, 125). RodA, an integral membrane protein, appears required for the in vitro enzymatic activity of PBP2, and so the two are thought to function in concert (81). Some mutants of E. coli are able to grow and divide in the absence of PBP2 and exhibit a spherical morphology (141). This bypass of the PBP2 requirement can be accomplished by either increasing the level of ppGpp or increasing the level of FtsZ (142). One possible explanation for these two results is that spherical cells, which have a larger volume, require an increased level of FtsZ and that ppGpp is able to stimulate ftsZ expression (140, 142). If a portion of ftsZ expression is dependent on ppGpp, it may explain the filamentous phenotype of a mutant deleted for spoT and relA which totally lacks ppGpp (151) (see chapter 92).

Cell Separation

As the leading edge of the septum invaginates, the newly synthesized peptidoglycan is cleaved to form the new poles of the two daughter cells. The enzyme(s) responsible for this topologically restricted cleavage of the peptidoglycan is not known. As stated earlier, the envA gene, long a candidate for this role since a mutation in this gene led to growth of the cells in chains, is involved in an early step in lipopolysaccharide biosynthesis and must affect the hydrolytic cleavage indirectly (116). E. coli contains at least 10 hydrolytic activities, including two soluble lytic transglycosylases, a membrane-bound lytic transglycosylase, endopeptidases, and N-acetylmuramyl-l-alanine amidases (74). Which of these enzymes might be responsible for cleaving the septal peptidoglycan is not known. Deletion of the gene encoding the major soluble lytic transglycosylase or inactivation of amiB encoding one of the N-acetylmuramyl-l-alanine amidases does not have a significant effect on cell morphology (135, 137). The lack of a conditional lethal mutant with the expected chain morphology raises the possibility that this activity, which must be well regulated to prevent lysis, is redundant, with more than one enzyme able to carry it out.

CONCLUDING REMARKS

Although progress has been made over the past few years in our understanding of bacterial cell division, this aspect of bacterial physiology remains obscure compared with other cellular processes in E. coli and S. typhimurium. The continued application of techniques based on genetics along with the more recent introduction of biochemical analyses of the gene products that are essential or those that dramatically affect the process should reveal details about the overall process. Thus, the progress made recently in the biochemical analysis of the ftsZ, ftsA, and min gene products is a foundation on which future research can be built. For example, the continued physiological characterization of ftsZ mutants along with the biochemical analyses of an array of FtsZ mutant proteins should help to clarify the role of FtsZ in cell division and its interactions with other proteins. As revealed with FtsZ, the localization of the protein can have important implications in its function. Although other cell division proteins are less abundant and therefore undetectable by immunolocalization, the development of more sensitive methods would extend this important facet of the biology of cell division. Thus, there is hope that the study of cell division, which has long remained refractory to molecular and biochemical analyses, will begin to yield details that will be both fascinating and novel.

ACKNOWLEDGMENTS

We thank colleagues for sharing unpublished work and sending unpublished manuscripts.